Negative Feedback, TIM Distortion, etc.
Randy Yates
I have heard for a very long time the pooh-poohing of negative
feedback in audio circuits (amps, preamps, etc.). Back in the
70s (I believe), it was linked to excessive TIM (transient intermodulation
distortion).
I have never really been able to get from "here" to "there" on this
issue, "here" being the level-headed, unassuming engineer that likes
to see everything demonstrated analytically, "there" being the
distortion that is purported to exist. Can anyone bridge this gap
with absolutely no hand-waving?
--
% Randy Yates % "Maybe one day I'll feel her cold embrace,
%% DIGITAL SOUND LABS % and kiss her interface,
%%% Digital Audio Sig. Proc. % til then, I'll leave her alone."
%%%% <ya...@ieee.org> % 'Yours Truly, 2095', *Time*, ELO
http://www.shadow.net/~yates
Trevor Wilson
Randy Yates <ya...@shadow.net> wrote in message
news:3741FAFA...@shadow.net...
> Hello Group,
>
> I have heard for a very long time the pooh-poohing of negative
> feedback in audio circuits (amps, preamps, etc.). Back in the
> 70s (I believe), it was linked to excessive TIM (transient intermodulation
> distortion).
**Understand a few points, here:
1)ALL amplifiers use NFB. Every single on of them. Without exception. Some
amplifiers use low levels, or no, LOOP NFB. The distinction is VERY
important.
2)AFAIK, there is no recogised measurement for TIM.
>
> I have never really been able to get from "here" to "there" on this
> issue, "here" being the level-headed, unassuming engineer that likes
> to see everything demonstrated analytically, "there" being the
> distortion that is purported to exist. Can anyone bridge this gap
> with absolutely no hand-waving?
**I is my belief (and that of many others), that the effects of LOOP NFB
lead to problems of interaction between the feedback path and the back EMF
from the loudspeaker. This is a rather contentious issue, however. There is
no recogised measurement to quantify the distortion resulting from the use
of loop NFB. It has been known as SID (Speaker Induced Distortion) and may
be partially responsible for the preference of many listeners for tube
amplfiers, which generally use low levels of loop NFB.
Some years ago, I was involved in some DBT's, suing two identical
amplifiers: One utilising a small amount of loop NFB and the other, none.
All conventionally measured specs (THD, IMD, FR, Damping factor, etc) were
below audible limits, for both amps. The zero loop NFB product was preferred
by all listeners, under all conditions.
--
Cheers,
Trevor Wilson
Richard D Pierce
>Hello Group,
>
>I have heard for a very long time the pooh-poohing of negative
>feedback in audio circuits (amps, preamps, etc.). Back in the
>70s (I believe), it was linked to excessive TIM (transient intermodulation
>distortion).
>
>issue, "here" being the level-headed, unassuming engineer that likes
>to see everything demonstrated analytically, "there" being the
>distortion that is purported to exist. Can anyone bridge this gap
>with absolutely no hand-waving?
Much of the handwaving was centered around the fact that many first-
generation transistor designs were based on very narrow-bandwidth
open-loop designs that had little or no gain margin at high frequencies
(that is, often as low as a few hundred Hz!). Look what happens to the
summing junction of the input amp when the rest of the circuit is
literally too slow to get out of its own way: you end up with large error
voltages that then go through the open-loop gain. When that input amp
saturates as a result, obviously the gain of the circuit to all but that
saturated error signal is 0, and intermodulation results.
Add to that the fact that, in the late 60's, few if any amplifiers were
tested into realistic loads.
Further, much of the hoopla was based on utterly unrealistic test
signals that far exceeded what music would do.
No magic, no new discovery, just another example of the audio industry
veing years behind the electronics business in general (this stuff was
described by George Philbrick at least 15 years before Otala burst upon
the scene).
But while all this handwaving was going on, the industry had already
figured it out and solved the problem in most cases.
Otala, unfortunately, obscured the real culprit to a great extent,
launching more of a mythology than anything else.
Yes, it yields quite readily to straightforward analysis, despite the
rapid, strident handwaving of some in the high-end business. And,
regrettably, their uninformed handwaving leads to out-and-out falsehoods
like "feedback is bad," "single-eded triode amplifiers have no feedback"
and other such patent nonsense.
--
| Dick Pierce |
| Professional Audio Development |
| 1-781/826-4953 Voice and FAX |
| DPi...@world.std.com |
Barry Mann
>Hello Group,
>I have heard for a very long time the pooh-poohing of negative feedback
>in audio circuits (amps, preamps, etc.). Back in the 70s (I believe), it
>was linked to excessive TIM (transient intermodulation distortion).
>I have never really been able to get from "here" to "there" on this
>issue, "here" being the level-headed, unassuming engineer that likes to
>see everything demonstrated analytically, "there" being the distortion
>that is purported to exist. Can anyone bridge this gap with absolutely
>no hand-waving?
Go to the Audio Engineering Site and look through the JAES (Journal of the
Audio Engineering Society) archives. In the early 80's there were a bunch
of papers on the subject. In many cases the citations will be more useful
than the papers.
Basically the unwary assumed that, given enough feedback, any trash open
loop amplifier could be turned into something wonderful. Perhaps this was
true for a 1000 Hz steady signal, but at higher frequencies that sad
little open loop amplifier would cause grief.
Even though the problem was officially "discovered" by the audio community
at about 1980 (I don't recall the exact date of Otala's paper), I can
remember a few voices in the wilderness warning of the problem in the mid
1960's. I can't imagine anyone who graduated from a good undergraduate
school after the mid 60's (and who paid attention) was surprised by the
"discovery". I suspect that many of the senior audio designers in the 70's
had never attended a formal course on feedback. In the late 40's and 50's,
the feedback calculations that we can perform on a napkin, were doctorial
stuff and beyond.
-----------------------------------------------------------
nite...@voicenet.com (Barry Mann)
-----------------------------------------------------------
Kurt Strain
: Hello Group,
: I have heard for a very long time the pooh-poohing of negative
: feedback in audio circuits (amps, preamps, etc.). Back in the
: 70s (I believe), it was linked to excessive TIM (transient intermodulation
: distortion).
: I have never really been able to get from "here" to "there" on this
: issue, "here" being the level-headed, unassuming engineer that likes
: to see everything demonstrated analytically, "there" being the
: distortion that is purported to exist. Can anyone bridge this gap
: with absolutely no hand-waving?
One of the best series of postings of the use/misuse of NFB was
written in rec.audio.tubes by Scott Frankland. Deja News might be
of use there.
I learned a lot from Scott, stuff they don't teach you in undergraduate
school for EE, probably just due to having too much to cram in too short
a program - as if 4-5 years isn't enough time.
Some things about NFB that I have gathered:
It has been reported in 1947 by H.F. Olson that pentode distortion is
more objectionable than triode distortion, THD being equal. Hence one
of the earlier proposed weighted THD equations was introduced, whereby the
weight of the harmonic was N/2 times higher, N being the harmonic number.
(This is oversimplified, but a start at attempting to quantify the effect).
In 1953, R.O. Rowlands showed that NFB produces distortion in higher order
harmonics where none existed before, mathematically proven. By bringing
a distorted output back to the input (inverted), "harmonics of
harmonics" are created. Overall THD will be lower, but sometimes the
upper harmonics are increased or generated from zero compared to no NFB.
(You will get a 3rd order harmonic from an amp that distorts with only
a 2nd order harmonic open loop by applying NFB, and then some more in
an infinite series).
Increasing NFB has the effect of increasing the number of harmonics and
IMD products but also has the effect of lowering them all as well. This
has been shown by experiment.
Subjective reports say that mild amounts of NFB can be more detrimental
than no NFB or using max NFB. We might have a clue as to why from the
above. Scott Frankland seems to be a proponent of using max NFB for the
results he desires. I have found it to be wise sometimes, but usually
I have found that I have enjoyed better results with no NFB, or make it
no *intentional* (i.e. minimal possible) NFB in the circuit.
Overall NFB has the additional property of destroying "soft clipping" as
opposed to just local NFB.
What is not known is what level of THD is inaudible. It is virtually
impossible to know this for sure because:
Some harmonics damage more than others, and it is not a simple progression
as reported by Olson.
All equipment used to test this distort, especially the transducers, and their
impact on the test is about impossible to isolate.
You can hear sounds below a random noise floor - experiments show that
it is possible to discern signals almost 30 dB below the noise floor.
And so it is not clear as to how much of the harmonics "buried in the
noise floor" are not buried at all. Distortion below the noise floor
ought to be distinguished as "residual error" as opposed to random noise,
something you cannot "average out".
NFB is not fully understood still as regards to audio
amplifier performance to a subjective listening audience, and how it
impacts people differently and in different situations. Generally, though,
it is agreed that NFB makes bad amps better but good amps that run
open loop sometimes poorer to some people. It is still loaded with
opinion even though we know some facts about it.
Kurt
Hans Beijner
any form of feedback, local or global, there are even some examples comercially
available from Audionote for instance. But of course most of single ended tube
amplifiers have at least some type of global feedback.
BR Hans
Richard D Pierce wrote:
> In article <3741FAFA...@shadow.net>, Randy Yates <ya...@ieee.org> wrote:
> >
> >
>
>
>
> <snip>
Arny Krüger
Randy Yates wrote in message <3741FAFA...@shadow.net>...
>Hello Group,
>
>I have heard for a very long time the pooh-poohing of negative
>feedback in audio circuits (amps, preamps, etc.). Back in the
>70s (I believe), it was linked to excessive TIM (transient
intermodulation
>distortion).
>
>I have never really been able to get from "here" to "there" on this
>issue, "here" being the level-headed, unassuming engineer that likes
>to see everything demonstrated analytically, "there" being the
>distortion that is purported to exist. Can anyone bridge this gap
>with absolutely no hand-waving?
It can't be done. Fear of negative feedback is a religion, not
science or technology.
Arny Krüger
Trevor Wilson wrote in message ...
>**I is my belief (and that of many others), that the effects of LOOP
NFB
>lead to problems of interaction between the feedback path and the
back EMF
>from the loudspeaker.
How does this relate to your former claims about zero loop NFB being
a desirable property for a phono preamp?
>This is a rather contentious issue, however. There is
>no recogised measurement to quantify the distortion resulting from
the use
>of loop NFB.
That's because it is imaginary, and its hard to quantify the
imaginary.
>It has been known as SID (Speaker Induced Distortion) and may
>be partially responsible for the preference of many listeners for
tube
>amplfiers, which generally use low levels of loop NFB.
When you reduce feedback, particularly loop feedback, you raise
output impedance. When you raise output impedance, you increase
frequency response variations due to variations in the impedance of
the speaker with frequency and amplitude. Therefore, it is the
low-feedback amplifiers that have the increased distortion (primarily
linear distortion in this case).
>Some years ago, I was involved in some DBT's, suing two identical
>amplifiers: One utilising a small amount of loop NFB and the other,
none.
>All conventionally measured specs (THD, IMD, FR, Damping factor,
etc) were
>below audible limits, for both amps. The zero loop NFB product was
preferred
>by all listeners, under all conditions.
Actually, Trevor has partially revealed the circumstances of these
tests in the ppast on RAO, and they were hardly DBT's. If I can coax
him into doing it again, the truth will be outed.
Arny Krüger
Hans Beijner wrote in message <3743933A...@nrj.ericsson.se>...
>Actually at least with tube technology it IS possible to make an
amplifier without
>any form of feedback, local or global, there are even some examples
comercially
>available from Audionote for instance. But of course most of single
ended tube
>amplifiers have at least some type of global feedback.
Providing you happen to find some output tubes with infinite plate
resistance... ;-)
Arny Krüger
Kurt Strain wrote in message <7hv9jg$n...@canyon.sr.hp.com>...
>
>Some things about NFB that I have gathered:
>
>It has been reported in 1947 by H.F. Olson that pentode distortion
is
>more objectionable than triode distortion, THD being equal. Hence
one
>of the earlier proposed weighted THD equations was introduced,
whereby the
>weight of the harmonic was N/2 times higher, N being the harmonic
number.
>(This is oversimplified, but a start at attempting to quantify the
effect).
This approach is counter to accepted knowlege about how the ear
works, at least for fundamental tones above 3 or 4 KHz. Sine the
ear's sensitivity is sharply dropping with increased frequency (and
harmonic number) higher order harmonics are less likely to be heard.
The whole concept of THD suggested here seems bogus. THD is not a
property of equipment. It is a means to characterize equipment
properties. THD relates to nonlinear distortion in equipment.
Nonlinear distortion, unlike THD is an equipment property. Nonlinear
distortion, with musical program material as a signal (or complex
tones) produces both harmonic products and intermodulation products.
When we talk about an amp with a certain kind of THD, we are really
trying to characterize the nonlinearity of the equipment.
Nonlinearity is usually characterized in terms of polynomial
expansions (a series based on various powers of X). In these
expansions, there are coefficients of various terms in the power
series. The various terms of the power series are called "orders".
This is the basis of talking about "various orders of distortion".
For a given input signal, the distortion it produces can be reliably
predicted if the spectra of the input signal and the orders of
distortion in the amplifier are known. Conversely, if the input
signal is known and reasonably simple, we can deduce the orders of
distortion it produces by looking at the spectra of the output.
>In 1953, R.O. Rowlands showed that NFB produces distortion in higher
order
>harmonics where none existed before, mathematically proven. By
bringing
>a distorted output back to the input (inverted), "harmonics of
>harmonics" are created. Overall THD will be lower, but sometimes
the
>upper harmonics are increased or generated from zero compared to no
NFB.
>(You will get a 3rd order harmonic from an amp that distorts with
only
>a 2nd order harmonic open loop by applying NFB, and then some more
in
>an infinite series).
This can be true and significant when the open-loop system is highly
nonlinear. The idea of building highly nonlinear amplifiers and
"fixing" them with massive amounts of loop feedback has (properly)
fallen into disrepute. All practical amplifers have feedback, whether
local (classic example: triode) or global (loop feedback). Most of
the usual means for making amplifiers reasonably linear prior to
application of loop feedback are based on choosing operating points
and adding nominal amounts of local feedback.
>Increasing NFB has the effect of increasing the number of harmonics
and
>IMD products but also has the effect of lowering them all as well.
This
>has been shown by experiment.
Been there, done that many times. The tie suggested here between IM
and THD is due to the issues previously discussed - they both come
from the basic nonlinearity of the amplifier.
>Subjective reports say that mild amounts of NFB can be more
detrimental
>than no NFB or using max NFB.
Most subjective reports I've seen that say this are based on
experiments that are so highly flawed that establishing the true
cause of the listener reactions is very difficult. Most such reports
come from sources with a religious belief that is against modern
technology in general and loop feedback in particular.
> We might have a clue as to why from the
>above. Scott Frankland seems to be a proponent of using max NFB for
the
>results he desires. I have found it to be wise sometimes, but
usually
>I have found that I have enjoyed better results with no NFB, or make
it
>no *intentional* (i.e. minimal possible) NFB in the circuit.
Well, that is your personal preference, and good for you. Is is good
for the world at large?
There is plenty of evidence that in general, people prefer low
distortion systems. For one thing, speech intelligibility is
generally maxmimized when all forms of distortion are minimized. I
think we all want to hear what is being said.
>Overall NFB has the additional property of destroying "soft
clipping" as
>opposed to just local NFB.
Soft clipping has questionable value for high fidelity systems. The
essence of high fidelity is minimal distortion (check the meanings of
the words in a dictionary). Therefore, the only viable clipping
alternative for high fidelity is no clipping. Soft clipping and
intentional distortion have their primary use in amplifiers that are
part of musical instruments, not systems for accurate reproduction.
>What is not known is what level of THD is inaudible. It is
virtually
>impossible to know this for sure because:
>
>Some harmonics damage more than others, and it is not a simple
progression
>as reported by Olson.
>
>All equipment used to test this distort, especially the transducers,
and their
>impact on the test is about impossible to isolate.
This would be very old-fashioned thinking. Estimating the audbility
of equipment with specific nonlinearities is not easy if those
nonlinearities are of such a level that distortion is near or above
the thresholds of audibiltiy. However, many kinds of equipment are so
precise that their distortion is well below the thresholds of
audibility.
In times past this did not include electroacoustic transducers like
loudspeakers, but many modern loudspeakers are far more linear than
many might want us to believe. Therefore, it is possibe to build
reproduction systems where nonlinear distortion is practically
inaudible, at least over a limited (but potentially useful, at least
for listening experiments) range of frequencies and SPL levels.
The form of distortion that is most difficult to control in complete
HiFi systems is linear distortion, not nonlinear distortion. This is
at least partially due to the ear's extreme sensitivity to small
variations in frequency response.
>You can hear sounds below a random noise floor - experiments show
that
>it is possible to discern signals almost 30 dB below the noise
floor.
However, the threshold of hearing roughly equates to the energy of
air molecules at standard temperature and pressure. If the ear would
respond below this, our brains would be assulted with hiss. It is
true that if you are operating well above the threshold of hearing,
discrete signals (e.g. sine waves) can be heard well below broadband
noise levels. This is true because the ear operates like a spectrum
analyzer.
>And so it is not clear as to how much of the harmonics "buried in
the
>noise floor" are not buried at all. Distortion below the noise
floor
>ought to be distinguished as "residual error" as opposed to random
noise,
>something you cannot "average out".
In times past, spectrum analysis was costly, and availible to only a
select few researchers. Therefore, broadband noise was a limit to
measurement of nonlinear distortiion. With the advent of PC-based
FFT's (some useful ones as freeware or shareware) spectrum analysis
can be the preferred means of measurement. With spectrum analysis,
measurements of signals "buried" deep in noise floors is not only
possible, it is hard to avoid! Therefore, modern test equipment can
detect and reliably measure distortion products that used to be
"buried in the noise". I would estimate the sensitivity of modern,
inexpensive test equipment for measuring nonlinear or linear
distortion to be from 10 to 100 times that of the human ear.
>NFB is not fully understood still as regards to audio
>amplifier performance to a subjective listening audience, and how it
>impacts people differently and in different situations.
Actually, the world is full of power amps that are sonically
transparent. They can be thought of as building blocks, and there is
no need to worry excessively about them. "The real" show is
elsewhere. IMO & IME most of the time spent obsessing over power amps
is time wasted.
> Generally, though,
>it is agreed that NFB makes bad amps better but good amps that run
>open loop sometimes poorer to some people. It is still loaded with
>opinion even though we know some facts about it.
There are simply so many good power amps that I can only justify my
former discussion in terms of its application to other parts of the
audio chain.
Bascially, we've got a clique of people who want to rerun the power
amplifier technological problems of the 60's and 70's ad infinitum
and milk them for all they can, in terms of money and influence. Just
about everybody who is staying modern is devoting their efforts
elsewhere. We do need a few good people to keep that supply of good
power amps coming, but apparently not nearly as many as we've got.
We've got serious techological problems elsewhere in the audio chain,
and time devoted to solving solved problems with power amps takes
away from areas that really need work.
Kurt Strain
read what the most unassuming engineer in the world has to say
on the subject. I'm sure that what Arny wrote is what the original
poster really wanted. Hey, just about everything sounds the same
and it's all perfect. That 's the big truth Arny keeps trying to say
over and over. It's time we all listened. HE is the all knowing one.
Only HIS opinion counts. That's because his is the loudest.
Kurt
--
Arny Krüger wrote in message ...
Randy Yates
> >
> >I have heard for a very long time the pooh-poohing of negative
> >feedback in audio circuits (amps, preamps, etc.). Back in the
> >70s (I believe), it was linked to excessive TIM (transient intermodulation
> >distortion).
> >
> >I have never really been able to get from "here" to "there" on this
> >issue, "here" being the level-headed, unassuming engineer that likes
> >to see everything demonstrated analytically, "there" being the
> >distortion that is purported to exist. Can anyone bridge this gap
> >with absolutely no hand-waving?
>
> generation transistor designs were based on very narrow-bandwidth
> open-loop designs that had little or no gain margin at high frequencies
> (that is, often as low as a few hundred Hz!). Look what happens to the
> summing junction of the input amp when the rest of the circuit is
> literally too slow to get out of its own way: you end up with large error
> voltages that then go through the open-loop gain. When that input amp
> saturates as a result, obviously the gain of the circuit to all but that
> saturated error signal is 0, and intermodulation results.
Are you basically saying that, due to negligence in addressing closed
loop control system issues such as stability (gain and phase margin,
etc.), the old designs had transients that saturated and thus intermodulated?
--
% Randy Yates % "My Shangri-la has gone away, fading like
%% DIGITAL SOUND LABS % the Beatles on 'Hey Jude'"
%%% Digital Audio Sig. Proc. %
%%%% <ya...@ieee.org> % 'Shangri-La', *A New World Record*, ELO
http://www.shadow.net/~yates
Arny Krüger
>
> Are you basically saying that, due to negligence in addressing
closed
> loop control system issues such as stability (gain and phase
margin,
> etc.), the old designs had transients that saturated and thus
intermodulated?
The very early days of solid state hifi power amps were tough,
particularly at the higher power levels. The major problems were that
the power devices at hand did not really have enough high frequency
response and safe operating area to be readily used to make good high
performance amplifiers.
The problems with high frequency response mostly manifiested
themselves in the form of excess nonlinear distortion at high
frequences. The SOA problems manifested themselves in the form of
amplifiers that either were easy to destroy with various loads, or
evidenced unexpected clipping-like behavior at certain frequencies
with certain loads due to protection circuits triggering, or did both
clipped unexpectedly and had short lives. Frankly, the high frequency
distortion was probably the lesser of the three evils.
Amps from the late 60's, early and mid 1970's often had these
problems, particularly if they were rated at more than about 30-40
wpc.
Numerous examples of well-engineered high-powered amplifers that
avoided both of these problem areas starting appear in the latter
part the 1970's. This was at least partially driven by the ready
availability of suitable output devices at reasonable prices.
Development of suitable output devices was largely driven by massive
demand for power devices for video sweep circuits, automotive
ignition systems, and computer disk drive access mechanisms.
However, from time to time, we see newer designs that are not
well-engineered, and the old problems come back. Given that lot of
solid state power amps were sold from the late 1960's onward, SS amps
got a bad name that many have not forgotten, even though there is no
technical reason for these problems to be present in modern
equipment.
Arny Krüger
news:7i5d6i$3rc$1...@news.ap.net...
> Well if Arny says so, it must be true.
And, if its an article about SS in rec.audio.tubes it's not suspect?
>Scratch everything I wrote, and read what the most unassuming
engineer in the world has to say
> on the subject.
OK, so I enjoy debunking bad engineering and pseudo-science. The tube
bigots have been telling lies and half-truths about SS for decades.
Hey, its a business...
> I'm sure that what Arny wrote is what the original poster really
wanted.
Regrettably for people like me who like to play with power amp
circuits, its just not a real productive thing to do any more, given
the good work that has already been done.
> Hey, just about everything sounds the same and it's all perfect.
You are lying, and are so arrogant that you think you can lie in the
face of statements from me like:
"The 'real' show is elsewhere."
That means that EVERYTHING DOES NOT SOUND THE SAME AND IT'S NOT ALL
PERFECT. One reason for we've still got some of the PRESSING PROBLEMS
we've got is that people have wasted a lot of time and money gilding
lillys - "perfecting" power amps that are already good enough. Time
spent obsessing over power amps is not devoted to problems with
recording techniques, microphones, and speakers, which remain in dire
need of more attention.
> That 's the big truth Arny keeps trying to say over and over.
What Arny's trying to say is that there is a big house fire raging
next door, but you seem to want to keep pouring water on a campfire
that is already pretty darn cold and wet.
>It's time we all listened. HE is the all knowing one.
I'm just one of many who knows a serious, unsolved problem when he
sees one, and also knows a "problem" that isn't really a problem any
more.
It ain't rocket science to see stuff that is this obvious. But it
takes discipline to act in a prioritized way.
> Only HIS opinion counts.
I'm hardly alone in this. Most of the mainstream audio business has
long since passed through its obsession with power amplifiers.
>That's because his is the loudest.
I'm just one of many, no matter how you lie and distort the truth,
and attack me personally.
Hey, fool with power amp circuits all you want. It's your life. But
stick to the truth, and stay off the personal attacks, if you can?
BTW, I notice that you made no references to the actual content of
the claims that I presented. Is that because you can find no fault
with them?
Thought so. ;-)
Randall Bradley
-chicken little
The world is FLAT.
-everyone except one guy, a long time ago now
All sufficiently high power, low distortion amps sound the same.
- you?
_-_-randy
BEAR Labs
Kurt Strain
>
>
>engineer in the world has to say
>> on the subject.
>
>bigots have been telling lies and half-truths about SS for decades.
>Hey, its a business...
>
You enjoy pretending to debunk what you think is bad engineering.
What you really do is just ridicule anyone who doesn't share
your religion. Hey, it's a business I guess. Because you think
you have all the answers, you tell everyone what they ought to do,
and how they ought to do it, all the time. You have demonstrated
the arrogance, pal. Like the time you told us all that a zobel network
is of no use, and complicates a speaker design, and you had to
add in your typical childish taunts toward me. You're still
wrong and of course you never admit to being wrong. Ever.
You made it personal, and you started in on me first.
So you can take your complaints about personal attacks to some
committe of hypocrites that share your point of view if that helps.
>
>BTW, I notice that you made no references to the actual content of
>the claims that I presented. Is that because you can find no fault
>with them?
I don't have the energy to repeat myself to a deaf person. I see
plenty of flaws, enough to write a book. I don't want to write a book
to people who have their mind set on their own religion and will take
nothing from it. I have 10 years of postings to Usenet out there
and it addresses about everything you have, but with a more
skeptical position about any side. You have one side, you have a
religion, and you have nothing to show for it except your rants. Your
religion is so strong that a balanced statement is seen by you as
another person's religious zealotry. We can't help you.
>
>Thought so. ;-)
>
Believe whatever makes you happy. You will never get another response
from me. You're just not worth my time. I would appreciate it that you
never respond to any of my posts. Throw a fit about this one if you
must. I won't care.
But since you feel so strongly about your position, why aren't you out
there working to solve the REAL PROBLEMS and instead waste time
here? My guess is you wouldn't know what to do there either.
If only you could actually let the world do what they want that doesn't
harm you. Not you. You would probably enact laws against everything
you don't see a need for. You're an angry control freak and nothing
I can do can change that. Again, you're a waste of my time.
Go get 'em, tiger! :-) (For the clueless: that was a typical Kruger-style
childish taunt, complete with
sarcastic smiley.)
Kurt
Arny Krüger
Kurt Strain wrote in message <7i78rc$j2d$1...@news.ap.net>...
>Arny Krüger wrote in message ...
>>
suspect?
>>
>>engineer in the world has to say
>>> on the subject.
>>
tube
>>bigots have been telling lies and half-truths about SS for decades.
>>Hey, its a business...
>>
>
>You enjoy pretending to debunk what you think is bad engineering.
"pretending"? ;-)
>What you really do is just ridicule anyone who doesn't share your
religion.
It takes real arrogance to call science and engineering "religion".
> Hey, it's a business I guess. Because you think
>you have all the answers, you tell everyone what they ought to do,
>and how they ought to do it, all the time.
"you know all the answers" seems to be a mantra with you... ;-)
I know a few answers, that's all.
>You have demonstrated the arrogance, pal.
What sort of arrogance does it take to go against science and
engineering? You seem to have it!
>Like the time you told us all that a zobel network
>is of no use, and complicates a speaker design, and you had to
>add in your typical childish taunts toward me.
It seems simple. I presented technical arguments. You responded with
personal attacks, now you are compounding that.
>You're still wrong and of course you never admit to being wrong.
If you want to establish in people's minds that I'm wrong, then
dealing with the technical issues I brought up would seem to be the
way to do it, wouldn't it?
>You made it personal, and you started in on me first.
Please show personal attacks in my 5/21 response to you. YOu know,
the one that you responded to with nothing but personal attacks...
>So you can take your complaints about personal attacks to some
>committe of hypocrites that share your point of view if that helps.
OK, I'll repost my 5/21 initial response to you, and everybody can
see that there are no personal attacks in it.
---------- repost of my 5/21 message that Kurt says is full of
personal attacks ------------
Arny Krüger
Randall Bradley wrote in message
<7i6bf4$1qm$1...@ultra0.rdrc.rpi.edu>...
> - you?
Hardly. Having sufficiently low noise can be audibly quite important.
Noise is usually thought of as being distinct from distortion. Ditto
for load-handing capabilities, although that is arguably about
distortion.
Trevor Wilson
Arny Krüger <ar...@flash.net> wrote in message
news:xhb13.238$nW....@news.rdc1.mi.home.com...
>
> Trevor Wilson wrote in message ...
> >**I is my belief (and that of many others), that the effects of LOOP
> NFB
> >lead to problems of interaction between the feedback path and the
> back EMF
> >from the loudspeaker.
>
> How does this relate to your former claims about zero loop NFB being
> a desirable property for a phono preamp?
**Many hours of listening tests.
>
> >This is a rather contentious issue, however. There is
> >no recogised measurement to quantify the distortion resulting from
> the use
> >of loop NFB.
>
> That's because it is imaginary, and its hard to quantify the
> imaginary.
**Because there is no presently stadardised techniques for measuring an
audible artifact, it does not automatically follow that some, yet to be
ascertained, artifacts do not exist.
>
> >It has been known as SID (Speaker Induced Distortion) and may
> >be partially responsible for the preference of many listeners for
> tube
> >amplfiers, which generally use low levels of loop NFB.
>
> When you reduce feedback, particularly loop feedback, you raise
> output impedance. When you raise output impedance, you increase
> frequency response variations due to variations in the impedance of
> the speaker with frequency and amplitude. Therefore, it is the
> low-feedback amplifiers that have the increased distortion (primarily
> linear distortion in this case).
**Quite true. Zero loop NFB amplifiers usually require more output devices,
superior power supplies and careful component matching to ameliorate the
effects, due to the absence of loop NFB. Many listeners believe it is worth
the effort.
>
>
> >Some years ago, I was involved in some DBT's, suing two identical
> >amplifiers: One utilising a small amount of loop NFB and the other,
> none.
> >All conventionally measured specs (THD, IMD, FR, Damping factor,
> etc) were
> >below audible limits, for both amps. The zero loop NFB product was
> preferred
> >by all listeners, under all conditions.
>
> Actually, Trevor has partially revealed the circumstances of these
> tests in the ppast on RAO, and they were hardly DBT's. If I can coax
> him into doing it again, the truth will be outed.
>
**I have tabled the nature of these tests, previously. They did not use the
so-called 'ABX system', yet, in their own way were double blind tests. They
were a placebo trial, where the listeners were unaware of what they were
listening to and the tester was also unaware. That constitutes double blind
and is acceptable. No independent obseravtion was involved, since the test
was for internal investigation only (ie: Was it worth the effort and expense
to use zero loop NFB amplification?).
Arny Krüger
Trevor Wilson wrote in message ...
>>
>> imaginary.
>
>**Because there is no presently stadardised techniques for measuring
an
>audible artifact, it does not automatically follow that some, yet to
be
>ascertained, artifacts do not exist.
When an allegedly audible artifact ceases to be audible when the test
is based on just listening, and levels are properly matched, then its
existence is very much in doubt. AFAIK, there simply is no evidence
in the literature of ANY "unmeasuable" distortions that are reliably
audible as inherent propreties of equipment. There have not been any
such reports that withstood scrutiny for over 30 years.
>> >It has been known as SID (Speaker Induced Distortion) and may
>> >be partially responsible for the preference of many listeners for
>> tube
>> >amplfiers, which generally use low levels of loop NFB.
>> When you reduce feedback, particularly loop feedback, you raise
>> output impedance. When you raise output impedance, you increase
>> frequency response variations due to variations in the impedance
of
>> the speaker with frequency and amplitude. Therefore, it is the
>> low-feedback amplifiers that have the increased distortion
(primarily
>> linear distortion in this case).
>**Quite true. Zero loop NFB amplifiers usually require more output
devices,
>superior power supplies and careful component matching to ameliorate
the
>effects, due to the absence of loop NFB. Many listeners believe it
is worth
>the effort.
Many listeners believe in magic stones and wooden discs. So much for
the reliability of anecdotes collected by promoters of questionable
technology.
>> >Some years ago, I was involved in some DBT's, suing two identical
>> >amplifiers: One utilising a small amount of loop NFB and the
other,
>> none.
>> >All conventionally measured specs (THD, IMD, FR, Damping factor,
>> etc) were
>> >below audible limits, for both amps. The zero loop NFB product
was
>> preferred
>> >by all listeners, under all conditions.
>>
>> Actually, Trevor has partially revealed the circumstances of these
>> tests in the ppast on RAO, and they were hardly DBT's. If I can
coax
>> him into doing it again, the truth will be outed.
>**I have tabled the nature of these tests, previously.
That has to do with the fact that your alleged evidence can't stand
scrutiny.
>They did not use the
>so-called 'ABX system', yet, in their own way were double blind
tests.
There are only two kinds of "blind listening tests":
(1) Those that are adequately blind and reduce the listening test to
the simple matter of "just listening", and
(2) Everything else, for whom the outcome is hard to attribute to
anything in particular.
>They
>were a placebo trial, where the listeners were unaware of what they
were
>listening to and the tester was also unaware. That constitutes
double blind
>and is acceptable.
Please state the means by which you ensured that was indeed, the
case.
> No independent obseravtion was involved, since the test
>was for internal investigation only (ie: Was it worth the effort and
expense
>to use zero loop NFB amplification?).
If you wanted to do a meaningful test, there are many possible
acceptable ways to do that.
You've described more, (but not all) of your procedures for this
investigation in the past. Why are you holding back now?
Trevor Wilson
**Sure and then there are promoters of technology that DOES have some merit.
You are right to be sceptical of a techology that seems to fly in the face
of accepted thinking. Assuming that THD, IMD and FR are all within
acceptable limits, then it is reasonable to compare zero loop NFB designs,
with high loop NFB ones. To paraphrase something I read several years ago:
"The soprano's voice clear and lifelike. The violins had a presence and
immediacy never heard by this listener." These words (or close to them) were
spoken about 70 years ago, after hearing a wind up gramophone, by a music
critic. It is possible that since you have (probably) never heard a well
implemented zero loop NFB amplifier, you may have no point of reference,
with which to judge the qualities of ordinary products.
>
> >> >Some years ago, I was involved in some DBT's, suing two identical
> >> >amplifiers: One utilising a small amount of loop NFB and the
> other,
> >> none.
> >> >All conventionally measured specs (THD, IMD, FR, Damping factor,
> >> etc) were
> >> >below audible limits, for both amps. The zero loop NFB product
> was
> >> preferred
> >> >by all listeners, under all conditions.
> >>
> >> Actually, Trevor has partially revealed the circumstances of these
> >> tests in the ppast on RAO, and they were hardly DBT's. If I can
> coax
> >> him into doing it again, the truth will be outed.
>
>
> >**I have tabled the nature of these tests, previously.
>
> That has to do with the fact that your alleged evidence can't stand
> scrutiny.
**Er, because an ABX box was not used, does not suggest that the test was
invalid, just different.
>
> >They did not use the
> >so-called 'ABX system', yet, in their own way were double blind
> tests.
>
> There are only two kinds of "blind listening tests":
>
> (1) Those that are adequately blind and reduce the listening test to
> the simple matter of "just listening", and
> (2) Everything else, for whom the outcome is hard to attribute to
> anything in particular.
>
>
> >They
> >were a placebo trial, where the listeners were unaware of what they
> were
> >listening to and the tester was also unaware. That constitutes
> double blind
> >and is acceptable.
>
>
> Please state the means by which you ensured that was indeed, the
> case.
**OK. Two amplifiers, marked A & B. Amplifier A was a zero loop NFB design.
Amplifier B was an IDENTICAL zero loop NFB design, with appropriate
modifications to allow the use of a small amont of global NFB, yet still
retaining IDENTICAL over all gain characteristics (+/- 0.1dB).
Specs for the zero loop NFB product were as follows:
Voltage gain 30dB
Max output Voltage 40 Volts
Max output current 40 Amps
Rise time 1uSec
THD <0.1%
IMD <0.1%
Hum & noise 100dB below rated output
Frequency resp. DC - 150kHz +/-2dB
Frequency resp. DC - 50kHz +/- 0.1dB
Damping factor 120 (DC - 50kHz) 8 Ohms.
The specs on the unit with loop NFB was superior in the areas of THD, IMD,
FR and damping factor. However the figures pertaining to the zero loop NFb
design are deemed below the limits of audibility, anyway.
Both amplifiers were shipped to 23 independent observers. The observers
were not told which amp was which. They were asked to simply supply their
feelings about what each amplifier sounded like on their systems (ranging
from small bookshelf speakers, to large, ESL's). Even if the amplfiers were
opened up, it is unlikely that anyone, outside a technically qualified
individual, could detect any differences, by simple observation. Since
neither the tester, or the testee was aware of which amplifier was being
used, at any one time, then the test qualifies as double blind. A further
advantage lay with the fact that the testee was not under any pressure, nor
were they observed by the testers at any time. Here, I invoke the
Schrodinger's Cat analogy.
All 23 listeners preferred the zero loop NFB design at all times. It is
important to note that this test was undertaken to test a theory put
foreward by the designer and manufacturer. If the high loop NFB design was
the preferred one, then that would have been the design put into
manufacture.
>
> > No independent obseravtion was involved, since the test
> >was for internal investigation only (ie: Was it worth the effort and
> expense
> >to use zero loop NFB amplification?).
>
> If you wanted to do a meaningful test, there are many possible
> acceptable ways to do that.
>
> You've described more, (but not all) of your procedures for this
> investigation in the past. Why are you holding back now?
>
**I trust that the description met with your approval. If you find fault
with the proceedure, then please let me know. Perhaps, if another test is to
be performed, then it can be modificed to incorporate such changes as you
suggest.
Jack D. Wills
>Amplifier B was an IDENTICAL zero loop NFB design, with appropriate
>modifications to allow the use of a small amont of global NFB, yet still
>retaining IDENTICAL over all gain characteristics (+/- 0.1dB).
>Specs for the zero loop NFB product were as follows:
>Voltage gain 30dB
>Max output Voltage 40 Volts
>Max output current 40 Amps
>Rise time 1uSec
>THD <0.1%
>IMD <0.1%
>Hum & noise 100dB below rated output
>Frequency resp. DC - 150kHz +/-2dB
>Frequency resp. DC - 50kHz +/- 0.1dB
>Damping factor 120 (DC - 50kHz) 8 Ohms.
>
>The specs on the unit with loop NFB was superior in the areas of THD, IMD,
>FR and damping factor. However the figures pertaining to the zero loop NFb
>design are deemed below the limits of audibility, anyway.
>
I am a little confused by the output ratings of this amplifier.
Yhe max output voltage is stated as 40 volts. Is this:
A. 40 V RMS
B. 40 V P-P (i.e. +-20 volts)
C: +-40 Volt
The reason I ask is that, at least in case A. or B., it is quite
possible that the amps were occasionally clipping. I have monitored
the output wavefrom of a variety of solid state amplifiers driving my
Spica TC-60's and found that the only amplifier that didn't clip
sometimes was a Hafler 500. This was NOT at headbanging rock and roll
levels. (I'm too old for that.)
I could monitor the clipping using a dual channel scope comparing
power amp input and output. The onset of clipping was very clear.
I used a Tektronix 7904 with 7A26 vertical amplifier and a 7B85 time
base.
In my experience, adding feedback to a zero feedback amplifier will
have a profound effect on the performace at overload.
Inn mu personal opinion, one reason people may identify differences
between amplifiers is that the amplifiers are not tested into overload
(where all sorts of differences appear) and are then listened too, and
judged, withput taking any effort to ensure that the amplifier is not
overloaded.
Dr. Jack Wills
--
Dr. Jack Wills
Teknetics
430A South Venice Blvd.
Venice, CA 90291
voice/fax: 310 821-7670
email: j...@netcom.com
Randy Yates
This sounds like a double-blind test, alright. The problem I see is that
there was no control over the test. For example, the preference for the
one amp COULD have been something as trivial as a difference in sensitivity
that caused it to sound louder at a given volume knob position.
I must admit that I am VERY skeptical about differences in amplifier sound.
Part of this comes from my experience some 11 years ago now when I attended
the 1988 AES convention in Los Angeles and participated in an amplifier
double blind test. There were three amplifiers, at least one tube and one
solid state (I believe a Crown was one of the amps). I believe they were
matched for levels. The source was a Studer 1/2 in tape with double recordings
of music passages. The two instances of each passage were played in sequence and
the object was to simply identify whether it was the same amplifier or a different
amplifier. The "controller" sat behind a curtain, while the MC directed the
test and had no knowledge of the amplifier(s) selected. Of course the audience
had no knowledge of which amplifier was being used either. My suspicions on
the inaudibility of differences were confirmed when, on one test, the MC asked
for the usual post-test comments. Several gave the typical flowery verbage about
one being "sweeter," "more detailed," etc. than the other (they even agreed on which
one sounded better). Only problem was, it was the same amp.
--
% Randy Yates % "Bird, on the wing,
%% DIGITAL SOUND LABS % goes floating by
%%% Digital Audio Sig. Proc. % but there's a teardrop in his eye..."
%%%% <ya...@ieee.org> % 'One Summer Dream', *Face The Music*, ELO
http://www.shadow.net/~yates
Trevor Wilson
Hans Beijner <Hans.B...@nrj.ericsson.se> wrote in message
news:3743933A...@nrj.ericsson.se...
> Actually at least with tube technology it IS possible to make an amplifier
without
> any form of feedback, local or global, there are even some examples
comercially
> available from Audionote for instance. But of course most of single ended
tube
> amplifiers have at least some type of global feedback.
**ALL amplifiers have some kind of NFB. The insertion of an emitter
resistor, cathode resistor, etc, constitutes NFB. Some do not use global
NFB, though.
Trevor Wilson
Jack D. Wills <j...@netcom.com> wrote in message
>
> I am a little confused by the output ratings of this amplifier.
>
> Yhe max output voltage is stated as 40 volts. Is this:
>
> A. 40 V RMS
>
> B. 40 V P-P (i.e. +-20 volts)
>
> C: +-40 Volt
>
**Sorry, my mistake. The maximum output Voltage (for both amps) is, in fact
35 Volts RMS, 50 Volts peak.
>
> The reason I ask is that, at least in case A. or B., it is quite
> possible that the amps were occasionally clipping. I have monitored
> the output wavefrom of a variety of solid state amplifiers driving my
> Spica TC-60's and found that the only amplifier that didn't clip
> sometimes was a Hafler 500. This was NOT at headbanging rock and roll
> levels. (I'm too old for that.)
**Since the experiment was conducted by listeners, not testers, then
clipping may (or may not) have occurred.
>
> I could monitor the clipping using a dual channel scope comparing
> power amp input and output. The onset of clipping was very clear.
> I used a Tektronix 7904 with 7A26 vertical amplifier and a 7B85 time
> base.
>
> In my experience, adding feedback to a zero feedback amplifier will
> have a profound effect on the performace at overload.
**It sure can.
>
> Inn mu personal opinion, one reason people may identify differences
> between amplifiers is that the amplifiers are not tested into overload
> (where all sorts of differences appear) and are then listened too, and
> judged, withput taking any effort to ensure that the amplifier is not
> overloaded.
>
**Very possibly.
Arny Krüger
Trevor Wilson wrote in message <7ibic7$bf4$1...@the-fly.zip.com.au>...
> Both amplifiers were shipped to 23 independent observers. The
observers
>were not told which amp was which. They were asked to simply supply
their
>feelings about what each amplifier sounded like on their systems
(ranging
>from small bookshelf speakers, to large, ESL's). Even if the
amplfiers were
>opened up, it is unlikely that anyone, outside a technically
qualified
>individual, could detect any differences, by simple observation.
Since
>neither the tester, or the testee was aware of which amplifier was
being
>used, at any one time, then the test qualifies as double blind. A
further
>advantage lay with the fact that the testee was not under any
pressure, nor
>were they observed by the testers at any time. Here, I invoke the
>Schrodinger's Cat analogy.
Nice story. What independent source can collaberate it?
Arny Krüger
Trevor Wilson wrote in message <7ibic7$bf4$1...@the-fly.zip.com.au>...
Trevor, this story is getting "better" every time you tell it. ;-)
Here is what you said the last time you told it...
http://x21.deja.com/[ST_rn=ap]/getdoc.xp?AN=419989929
Subject: Re: Cable comparison experiment
Date: 1998/12/08
Author: Trevor Wilson <ra...@hutch.com.au>
**Absolutely NOT TRUE. I have taken part in a double blind trial
involving two identical amplifiers, one with zero loop NFB and one
with a modest amount (10dB) of loop NFB. Of 23 listeners, none showed
a preference for the loop NFB version. BTW: Amps were perfectly level
matched and with no obvious marking to identify them. Listeners were
able to listen in their own systems, at their own liesure. I never
much believed that TIM was a real worry with amps. Perhaps back EMF
is not the problem, but I can assure you, a properly implemented zero
loop NFB amp is the solution to many listening ills.
Trevor Wilson
Arny Krüger <ar...@flash.net> wrote in message news:O_m23.2090
>
> Nice story. What independent source can collaberate it?
>
**How about the 23 listeners?
The bottom line is this, Arny: The test was done for purely investigational,
internal purposes. The designer had a theory about amplifier design and did
not trust his own listening abilities. He decided that a bunch of people who
had an interest in high end audio, would be ideal to test his theories on.
No independent observers were involved, because it was not a study that was
intended for publication, nor was is to be used for promotional purposes.
Don't forget, it was done almost 20 years ago. The design has changed little
since those tests were performed.
Trevor Wilson
Arny Krüger <ar...@flash.net> wrote in message news:uQn23.2092
>
> Trevor, this story is getting "better" every time you tell it. ;-)
>
> Here is what you said the last time you told it...
>
**I'm sorry.... your point is?
Trevor Wilson
> there was no control over the test. For example, the preference for the
> one amp COULD have been something as trivial as a difference in
sensitivity
> that caused it to sound louder at a given volume knob position.
**Yep, I see it as double blind. If you read the previous post, you will
note that the amps were level matched. Listeners were free to test the amps
in any way they wished.
**I, too, have been fooled by well set up DBT's. I was once fooled by a test
involving a tube amp and a SS one. (In any case, my arguement, in this
thread, related to a properly designed zero loop NFB amplifier vs. and amp
using large levels of loop NFB. To my ears, the difference is substantial.)
In fact, I have fooled quite a few listeners, in a similar fashion. However,
it is MUCH more difficult to fool a listener in his/her own environment. I
could argue that this is the BEST DBT that one can do, since the environment
is under much closer control.
Arny Krüger
Trevor Wilson wrote in message ...
>
>
>>
>> Trevor, this story is getting "better" every time you tell it.
;-)
>>
>> Here is what you said the last time you told it...
>>
>**I'm sorry.... your point is?
The point is that you've told this same story at least twice with a
lot of similarities, but with details changed to suit the argument at
hand:
Trevor Wilson wrote in message <7ibic7$bf4$1...@the-fly.zip.com.au> on
5/24/99
Trevor, this story is getting "better" every time you tell it. ;-)
Here is what you said the last time you told it...
Arny Krüger
Trevor Wilson wrote in message ...
>
You've forgotten the one irreducable source - the active devices
themselves. For example, the essence of things like "plate
resistance" is local negative feedback inside the tube.
Arny Krüger
Trevor Wilson wrote in message ...
>
>>
>> This sounds like a double-blind test, alright. The problem I see
is that
>> there was no control over the test. For example, the preference
for the
>> one amp COULD have been something as trivial as a difference in
>sensitivity
>> that caused it to sound louder at a given volume knob position.
>
>**Yep, I see it as double blind. If you read the previous post, you
will
>note that the amps were level matched. Listeners were free to test
the amps
>in any way they wished.
It's also true that it is a story that you've told twice with some
details changed to suit the argument at hand:
Trevor Wilson wrote in message <7ibic7$bf4$1...@the-fly.zip.com.au> on
5/24/99
Trevor, this story is getting "better" every time you tell it. ;-)
Trevor Wilson
>
> Trevor Wilson wrote in message ...
> >
> >
> >>
> >> Trevor, this story is getting "better" every time you tell it.
> ;-)
> >>
> >> Here is what you said the last time you told it...
> >>
>
> The point is that you've told this same story at least twice with a
> hand:
**Well, perhaps I should have saved the original in an easy to access file,
but I did not. The test took place many years ago and some details are not
foremost in my memory.
Mark Rehorst
test
> involving a tube amp and a SS one. (In any case, my arguement, in this
> thread, related to a properly designed zero loop NFB amplifier vs. and amp
> using large levels of loop NFB. To my ears, the difference is
substantial.)
> In fact, I have fooled quite a few listeners, in a similar fashion.
However,
> it is MUCH more difficult to fool a listener in his/her own environment. I
> could argue that this is the BEST DBT that one can do, since the
environment
> is under much closer control.
I think that "fooled" is not the right word, for it implies that there was
some trickery
involved. What you really mean is that you have been shown and you have
shown
others that there was no discernable difference between the sounds of the
amplifiers
under test.
What you are now telling us is that even after hearing the results for
yourself, you
still refuse to believe it.
Subjectivists are an interesting breed. One the one hand, they argue that
measurements don't matter. What the ears tell you is all that counts. On
the
other hand, they tell you they have been involved in listening tests and
they don't
believe the results. Then they fall back on the old "the test was flawed"
argument.
"I wasn't in MY listening room", or "the fact that I knew I was
participating in a test
induced stress that doesn't exist under non test conditions, so the results
were
skewed by the stress"...
MR
Stuart Parker
<mreh...@fmi.fujitsu.com> writes
The confusion lies in the purpose for which the two camps carry out
their tests.
One camp is looking for "explanatory" trial methodology in which the
experiment is designed to control for variables which may explain any
differences that may be heard (volume, frequency response, source
material, listening room dimensions, nature of the comparison allowed
(e.g. ABX) and standardisation of the judgement being made by the
listeners (e.g. same or different)).
The other camp is exploring the formation of preferences by individuals
under realistic domestic conditions, under which, almost by definition,
some of the various potential explanatory variables will be manipulated
by the listener until they achieve a satisfactory set of listening
conditions for each of the items under test. This is a pragmatic trial
and there is no a priori reason to suppose that the results of
explanatory and pragmatic trials will be in agreement.
In medicine for example it is known that the placebo effect differs
depending on the way in which the placebo is delivered. For example,
placebo tablets (tablets with no active ingredients) when given to
people with symptoms, may be more effective at relieving those symptoms
depending on what colour they are.
In the case of amplifier topologies it would seem to me that both points
of view are potentially correct - that is to say when the listening
conditions are controlled by the experimenter then listeners are unable
to consistently tell the difference between items under test, but when
the listeners are allowed to adjust the listening conditions according
to their preference (and within defined limits) then they are capable of
forming a consistent preference between two "black boxes".
The challenge then is to understand the way in which listeners
manipulate explanatory variables to help them to form preferences and
the influence of the contents of the "black box" on this process.
Please excuse my ignorance if experiments of this sort have already been
done. I would be grateful if the group could enlighten me (and maybe
others "lurking" nearby) if this is the case.
--
Stuart Parker
Paul Cambie
> under realistic domestic conditions, under which, almost by definition,
> some of the various potential explanatory variables will be manipulated
> by the listener until they achieve a satisfactory set of listening
> conditions for each of the items under test. This is a pragmatic trial
> and there is no a priori reason to suppose that the results of
> explanatory and pragmatic trials will be in agreement.
Wouldn't this also mean that the results of such a trial had no real
meaning outside the practical environment of each test event? The emperor
would alter the trim of his new clothes until satisfied with his appearance
and be happy with the result.
Cheers!
Paul
Stuart Parker
<cam...@ozemail.com.au> writes
Not if a large enough proportion of test subjects, faced with the same
choice (double blind) form the same preference. This result would mean
that in some generalisable sense the unit which was consistently
preferred was in some way well...preferable. The challenge then would
be to understand what it was about the test item that the test subjects
found preferable - they might all say they preferred item A over item B
because it was red!
--
Stuart Parker
Mark Rehorst
> because it was red!
Not if it were a blind test. The whole point of such a
test is to eliminate factors such as the appearance
of the equipment that have nothing to do with the
sound.
MR
Trevor Wilson
Mark Rehorst <mreh...@fmi.fujitsu.com> wrote in message
news:7iecgr$e19$1...@news.fujitsu.com...
> > **I, too, have been fooled by well set up DBT's. I was once fooled by a
> test
> > involving a tube amp and a SS one. (In any case, my arguement, in this
> > thread, related to a properly designed zero loop NFB amplifier vs. and
amp
> > using large levels of loop NFB. To my ears, the difference is
> substantial.)
> > In fact, I have fooled quite a few listeners, in a similar fashion.
> However,
> > it is MUCH more difficult to fool a listener in his/her own environment.
I
> > could argue that this is the BEST DBT that one can do, since the
> environment
> > is under much closer control.
>
> I think that "fooled" is not the right word, for it implies that there was
> some trickery
> involved. What you really mean is that you have been shown and you have
> shown
> others that there was no discernable difference between the sounds of the
> amplifiers
> under test.
**A poor choice of words, on my part. Yes, I was unable to discern the
difference between two amplifiers, that I would have automatically expected
to sound quite different.
>
> What you are now telling us is that even after hearing the results for
> yourself, you
> still refuse to believe it.
**Not at all. I accept the results for what they were. I was listening to an
extraordinarily good tube amp. I expected to hear major differences.
>
> Subjectivists are an interesting breed. One the one hand, they argue that
> measurements don't matter.
**Never have I stated such a thing. Measurements are vital. For example: the
differences between most amplifiers can be traced to frequency response
anomalies, whilst driving real world loudspeakers. However, there is far too
much emphasis placed on some parameters, widely promoted by the industry.
THD, for instance. It seems unlikely that an average (or even an
exceptional) human, could percieve THD levels of much less than 0.5%. I
believe that there are possible some aspects pertaining to mechanical sound
reproduction, which are significant to human hearing, that are yet to be
quantified by measurement techniques.
What the ears tell you is all that counts. On
> the
> other hand, they tell you they have been involved in listening tests and
> they don't
> believe the results. Then they fall back on the old "the test was flawed"
> argument.
> "I wasn't in MY listening room", or "the fact that I knew I was
> participating in a test
> induced stress that doesn't exist under non test conditions, so the
results
> were
> skewed by the stress"...
>
**A very real problem, IMHO. Listening in one's own environment is probably
the best method. I doubt that anyone would argue that point.
Trevor Wilson
Arny Krüger <ar...@flash.net> wrote in message news:4tt23.2105
>
> You've forgotten the one irreducable source - the active devices
> themselves. For example, the essence of things like "plate
> resistance" is local negative feedback inside the tube.
>
**Wrong and, right. I had not forgotten. It is much easier to demonstrate
the concept of local feedback, through variations in (for example) emitter
resistance, than the internal impedances of the devices, themeselves.
Stewart Pinkerton
>found preferable - they might all say they preferred item A over item B
>because it was red!
I think you're missing the point here. The basic tenet of a blind test
is that the subject does not *know* which unit is playing at any time,
so there is no way to 'prefer' A over B by any means other than
*audible* difference.
If you can *see* the red amplifier, then you might well prefer it,
especially if you substitute Pass (or whatever) for 'red'. That would
not however tell you if it *sounded* better, or indeed different.
--
Stewart Pinkerton | Music is art, audio is engineering
Arny Krüger
Stuart Parker wrote in message ...
>There is no conflict here.
>
>The confusion lies in the purpose for which the two camps carry out
>their tests.
>
>One camp is looking for "explanatory" trial methodology in which the
>experiment is designed to control for variables which may explain
any
>differences that may be heard (volume, frequency response, source
>material, listening room dimensions, nature of the comparison
allowed
>(e.g. ABX) and standardisation of the judgement being made by the
>listeners (e.g. same or different)).
This is a not-uncommon fallacy. For a more elegant and humorous
expression of it, please see http://www.pcavretch.cjb.net/ and click
the "ABX" icon.
What really happened can be found in an old post on Deja News:
http://x42.deja.com/[ST_rn=ps]/getdoc.xp?AN=228725465
Subject: Re: objectivist religion
Date: 1997/03/27
Author: David J. Carlstrom
Arny Krueger invented the ABX comparator hoping to prove his
amplifier sounded better than Bern Muller's. He failed and being a
scientist he was just as delighted with the result as if he'd won.
>The other camp is exploring the formation of preferences by
individuals
>under realistic domestic conditions, under which, almost by
definition,
>some of the various potential explanatory variables will be
manipulated
>by the listener until they achieve a satisfactory set of listening
>conditions for each of the items under test. This is a pragmatic
trial
>and there is no a priori reason to suppose that the results of
>explanatory and pragmatic trials will be in agreement.
This has been rebutted many times in print, but perhaps most
completely by: "Do All Amplifiers Sound the Same?" David L. Clark and
Ian G. Masters (pp. 78-84), Stereo Review Magazine, Jan 1987:
"The best solution was simply to present my plan to some audiophiles
and ask for their as assistance After all, they should see this as an
opportunity to prove the validity of their belief to the skeptics.
The cooperation I received from manufacturers, a local high-end audio
salon, and other audiophiles was more than I ever hoped for. Their
assistance and participation as listeners in this project
demonstrated that they were secure in their belief and brave enough
to risk being exposed to an uncomfortable outcome."
"Critical to hearing differences in audio equipment is the choice of
speakers and If listening room. Besides
sounding exceptionally good over a wide range of If listener posit
ions, the Magnepan speakers have fairly low
efficiency and a low impedance Both of these factors conspire to make
an amplifier work very hard. possibly
exposing a weakness. The load the Magnepans present is not highly
reactive. but it is typical of the low-
efficiency, highly damped speakers often preferred by audiophiles."
"The program matarial was highly varied, though all the music used
acoustic instruments. Both LP's and CD's were
supplied, and some If Listeners also brought their own. At each
session, the particular recordings used were
selected by the If listeners. Following are some of the more popular
ftems: Reference Recordings' "Dafos" (LP
and CD). an ethnic percuss ion album, and Capriccio espagnole (LP);
Sheffield's "Track Record" (LP and CD),
"West of OZ,. by Amanda McBroom (LP), "I've Got the Music in Me" by
Thelma Houston (CD). and "The King
James Version" by Harry James (LP and CD). Joan Baez's "Diamond s and
Rust" (Nautilus LP); Opus 3's "Tat
Record 1: Depth of Image" (LP and CD); Andrew Lloyd Webber's Requiem
(CD); Gershwin's Rhapsody in Blue on
Telarc (CD); and the Digital Music CD's by Warren Bernhardt, "Trio
'83," and Jay Leonhardt, "Salamander Pie. "
"A major feature of these If listening-test sessions was their
openness. All the equipment could be seen and
inspected. At listener request, all the amplifiers could be
auditioned with or without the ABX system prior to the
blind testing. Listeners were given as much time as they needed to
get used to the audio to system. to select
revealing program material, and to note apparent differences between
the various amplifiers. Almost all listeners
(even the skeptics) thought they could hear differences at this
point. They wrote down their sonic observations
and selected the pair of amps that seemed to differ most to compare
in the following blind test
"The listeners even decided which kind of blind testing they would
use: manual swapping of the cables feeding the
chosen amplifier s (in this case hidden behind a screen) or ABX
switching where the relay system would allow a
more raped change over. Listeners new to the ABX system had already
been trained in its operation at a separate
setup while they wafted their turn in the listening room. Most If
listeners opted for the convenience of the ABX
relay control, Feeling that the extra contacts in the system would
not degrade the signal Nine people. however,
chose to augment their ABX tests with blind cable-swap tests.
>In the case of amplifier topologies it would seem to me that both
points
>of view are potentially correct - that is to say when the listening
>conditions are controlled by the experimenter then listeners are
unable
>to consistently tell the difference between items under test, but
when
>the listeners are allowed to adjust the listening conditions
according
>to their preference (and within defined limits) then they are
capable of
>forming a consistent preference between two "black boxes".
This has been repeatedly shown to be a false claim. One particularly
egregious form of this false claim can be found at
http://www.stereophile.com./fullarchives.cgi?20
AKA "The Listeners' ManifestoBy Robert Harley, January 1992". It was
also presented at an AES Meeting. In this document Harley completely
misrepresents many of the documents he purports to quote, so that
unsuspecting audiophiles will believe the kind of false claims just
delivered by Mr. Parker.
>The challenge then is to understand the way in which listeners
>manipulate explanatory variables to help them to form preferences
and
>the influence of the contents of the "black box" on this process.
>
>Please excuse my ignorance if experiments of this sort have already
been
>done. I would be grateful if the group could enlighten me (and
maybe
>others "lurking" nearby) if this is the case.
Since you seem to have had your thinking affected by Harley's false
claims and deceptions, which John Atkinson supports to this day,
please consider the following rebuttal of them:
What Stereophile Does Not Want You to Know About Their Listening
Tests.
Introduction:
Robert Harley was Stereophile's Consulting Technical Editor at the
time his "Listener' Manifesto" appeared in Stereophile, in 1992. John
Atkinson has publicly and repeatedly offered it as a statement and
defense of current Stereophile reviewing procedures, as recently as
the first week of March, 1999.
The original text of Harley's "Manifesto" can be found at
http://www.stereophile.com./fullarchives.cgi?20 .
This is a long article, about 80% as long as Harley's original
"Manifesto". I've broken it down into 8 parts to make it more
manageable. It was originally written and posted in 8 parts on the
USENET rec.audio.opinion newsgroup in late February and early March,
1999. I've edited it and changed it some since then, and I am
continuing to do so.
In part 1 of this series I show how Harley claims that the people who
don't agree with him do so simply because they don't listen. I also
show that this is just not the case, and his critics are in fact
devoted to listening.
In part 2, I show how Harley can't even fathom how devoted the people
he criticizes are to listening, because he lacks appropriate
skepticism and follows a system of anti-scientific belief.
One probable source of Harley's logical failures is his repeated
assumption that a priori, all or almost all equipment sounds
different. A more appropriate, agnostic or skeptical view is that
perhaps some equipment sounds different, and other equipment does not
sound different. Since Harley is obviously a "true believer" in the
universal existence of audible differences, and does not practice
appropriate levels of skepticism about his own findings or beliefs,
he is hardly a reliable source of information of just about any kind
about equipment performance.
I also point out how Harley's use of the words "objective" and
"subjective" seems to be simply a rhetorical ploy to set up
"objective" to fail, and force the ascendancy of his own particular
flavor of subjective". Then I start to lay out the foundation for a
critique of Harley's flavor of "subjective".
In part 3, I show that Harley damns specs because he does not
understand them, just like he damns statistics because he does not
understand them.
Harley seems to believe that he and people like him can collect
information for reviews, which he claims are detailed and accurate,
during listening sessions. He also claims that figuring out whether X
is A or B during or after a listening session is more than ANY human
can possibly be expected to handle.
In part 4, I show that Harley has to make a ton of false distinctions
(Besides the ones about subjective and objective, as well as science
versus experience) to support his basic thesis that reliable
listening tests are a kind of work of Satan, and inherently
insensitive.
In part 5, I show that no matter what Harley says, Subjective
Critical Listening is not rejected by any of the people Harley says
reject it. In fact they do it often and grant it a ton of
credibility. They just don't buy Harley's egregiously flawed ideas
about how to do it.
I also show that no matter what Harley says, his (egregiously flawed)
subjective tests can't be hoped to be sensitive enough. I mostly
focus on the fact that we often use components in varied cascades
where distortion adds, but Stereophile's "single presentation method"
has a big blind spot.
In part 6 I show how Harley should criticize his own listening
technique, his "single presentation method".
In part 7 I point out how Harley's' basic critical style is one-sided
and highly biased towards serving his own viewpoint.
In part 8, I show how Harley's claims about published blind tests
that he references are the exact opposite of what their authors
actually wrote.
---------------------------------------------------------------
Part 1: A "Listener's Manifesto" or just more posturing from the
religious high end?
The first problem with this piece is its title and tone. Harley
clearly seems to be living in a fictional world where there are
"listeners" and "not listeners".
The fact is that no matter how Harley postures, pretty much everybody
is a listener of some sort. Harley pursues this tack on and off
through the whole piece, despite the fact that he also flatly denies
it is true.
Harley seems to think that people in the big mainstream companies and
research areas who disagree with him don't really listen. I'm sitting
here thinking about all the industrial-strength and industrially
situated listening rooms I've been in, used, and/or or seen pictures
of.
Harley's posturing about "not listeners" seems like a conceit of his
given all the money the mainstream audio business spends on doing
listening tests.
Probably one of the most impressive accounts I've heard of is the
AT&T labs listening room that RAO's own "not listener" JJ has spoken
to me of. First off, the thing was built way to close to a freeway.
So, they blew a few 10's or 100's of thousands of dollars on making
it quiet. It goes uphill from there. ;-)
This kind of attention to detail in corporate listening rooms is not
unusual.
Another corporate giant I'm familiar with actually maintains several
more-or-less independent development teams that concurrently develop
for the same general requirement. Each team follows different general
philosophies as to what constitutes good sound. Each team has its own
engineers and listening rooms, implemented in accordance with their
various philosophies, ranging from "Meter Reader", to "Golden Ear".
When the competing teams finish their candidate solutions to the same
problem, there is a double blind "Bake Off".
Judging by their "edifice complexes", people who don't agree with
Harley are willing to pay at least lip service to the idea of serious
listening in dedicated, engineered places. In many cases listening is
paid a lot more than lip service - it is the most important thing and
it gets a top salary.
Harley writes: "The theme of the 91st Audio Engineering Society
Convention, "Audio Fact and Fantasy: Reckoning with the Realities,"
reflects the increasing polarization of the audio community over
so-called "subjective" and "objective" audio---also known as "The
Great Debate." At one extreme is the "if you can't measure it you
can't hear it" school of thought. This camp rejects the listening
experience, believing that nothing more can be known about audio
equipment quality beyond the numbers generated by "objective"
testing. At the other extreme are those who reject any role of
science in audio engineering and make absurd pseudoscientific claims
about the audibility of certain phenomena. "Harley actually does not
seem to understand the true extremes. While the idea that "if you
can't measure it you can't hear it" exists, it is hardly extreme. The
extreme version is more like ""if you can't measure it with a Simpson
260 it just does not matter to the customers".
The mainstream view of audio measurements and hearing is more like
"If you hear it, there is some corresponding physical effect that
leaves measurable tracks, or it's not a reliable, inherent property
of the equipment, or it's a dramatic new discovery like we have not
had in over 30 years."
Everybody is open to proof of some new thing, if the evidence is
worthwhile. As this paper evolves, I'll show how Harley's listening
evaluations are egregiously flawed. I think he confuses people being
skeptical about his egregiously flawed listening evaluations with
people not trusting listening evaluations at all.
Nobody with a brain rejects the listening experience. They just
understand that it is not always all that it seems, taken at face
value. Therefore, open listening evaluations, such as those practices
by Stereophile to this day, are not a be-all and end-all all by
themselves. Harley says: "At the other extreme are those who reject
any role of science in audio engineering and make absurd
pseudoscientific claims about the audibility of certain phenomena."
Well, that is one way to say it, another is "There are people who
routinely claim they can hear a cricket farting at 300 feet in a
thunderstorm". Many such people seem to write for Stereophile. ;-)
Harley says: "I propose there is a third approach that incorporates
the methods of rational, scientific inquiry without rejecting the
very real and important role critical listening can play in advancing
audio engineering. "Again, nobody with a brain rejects that.
However, Harley's specific "third approach" is so flawed and so
filled with religious beliefs, poor science and poor critical
thinking that it leads to absurd pseudoscientific claims about the
audibility of certain phenomena.
Harley continues: "This approach doesn't reject measurement, yet
recognizes its limitations. This approach repudiates claims that
established physical laws are suspended, yet believes in the direct
reality of the listening experience. This approach considers
subjective critical listening as an expansion of thought, not as a
rejection of rationality."
This approach existed for at least a decade before Harley wrote his
"Manifesto".
I call this approach "reliable listening". Reliable listening takes
place when people recognize the basic flaws of open listening
evaluations and take some relatively simple steps to deal with them.
One interesting consequence of reliable listening tests is that they
pretty well agree with what is scientifically known about hearing,
the properties of the human ears, and technical evaluations of audio
equipment.
Harley says: "The objectivists' claim that no sonic differences exist
between competently designed and manufactured audio components (or
those having similarly good measured performance) is an absurd
premise that is anathema to the experience of hundreds of thousands
of critical listeners." If course, this would be true if Harley could
show it was true in even one case. But in his rush to judgement and
flawed view of commercial reality, he throws out the baby with the
bath water and condemns himself to a life of claims and acts that are
just not that relevant to most mainstream audio engineers.
Interestingly enough, Harley contradicts himself by saying: "No one
doubts the necessity or utility of subjective listening." OK, so he
just threw out the first how many pages of his rant? ;-)
Harley says: "My experience overwhelmingly indicates that many
aspects of audio equipment quality are revealed in the listening room
and not in the laboratory." There is a simple explanation for this,
being that Harley does not approach the listening room with the rigor
that is pretty much an inherent part of evaluation of audio gear
using test equipment. I'm not saying that test equipment evaluation
is perfect, but its most common abuse involves chasing numbers for
the sake of numbers, as opposed to the fantastic and unreliable
findings that are fairly universal when Harley's ideas about
listening evaluations are applied.
Regrettably, subjective evaluations have a rotten reputation in much
of the mainstream scientific world. To many, the very word
"subjective" conjures up an antonym for "objective". Harley seems to
do everything he can to perpetuate this. The reason is one of simple
rhetorical technique: Harley seems to define "subjective" and
"objective" the ways he does so he can destroy "objective" and leave
"subjective" standing as the only option.
In contrast, many old-timers like me, tired of many years of watching
a battle that need not be fought, define objective and subjective in
ways that correspond to human experience: they can coexist and agree,
or not: depending on the circumstance. This pretty well parallels the
obvious truth about reliable, inherent, audible differences in audio
equipment: they exist or not, depending on the circumstance. Probably
the three most important events in the process of making the study of
subjective responses profitable were:
(1) The Hawthorne effect was discovered by Western Electric, ca.
1935.
(2) The Placebo Effect and means to control it were publicized by the
AMA ca. 1953-55.
(3) Common knowledge of basic statistical tests for reliability by
engineers which really picked up serious steam in the 50's and 60's.
---------------------------------------------------------------
Part 2: The Hawthorne and Placebo Effects. Why didn't Harley mention
them?
Hawthorne is the name of a Western Electric plant near Chicago that
made telephone equipment. At one time it was one of the largest
manufacturing plants in the world. Obviously, this plant was a
critical resource of AT&T and they wanted it to run as efficiently as
possible. I once got lost coming back from a Midwest Acoustics
conference and drove by it, in the 70's.
Being a "modern" company, AT&T tried to apply some new technology,
the relatively new art of "Operations Research" (OR) to the
production procedures at this plant. After making some striking gains
in production efficiency, the OR staff began to suspect their own
work, particularly when trying to get additional gains out of a
production task that had already been optimized once.
AT&T's OR team (which included Demming) found is that whenever they
started to study a group of workers in order to improve their
efficiency, the efficiency of the workers increased. The team began
to see the potential for even nonsensical "changes" to "improve"
worker efficiency. Furthermore, once the study was over, worker
efficiency would tend to slide. So, the gains made were not reliable.
A formal term for this is "Experimental demand characteristics that
may change the motivation". The basic idea is that the simple fact of
making a change may cause increased motivation on the part of the
subject to provide a favorable outcome. As applied to changes in your
stereo, simply making a change may cause you to listen more
carefully, and therefore hear more detail and a better soundstage, as
we have many anecdotal reports of. See
http://www.psy.gla.ac.uk/~steve/confidence.html for more information
on this. Another way to look at this is shown in
http://www.sytsma.com/phad530/expdesig.html:
"The Hawthorne Effect refers to the behavior of interest being caused
by subject being in the center of the experimental stage, e.g.,
having a great deal of attention focused on them. This usually
manifests itself as a spurt or elevation in performance or physical
phenomenon measured. Although the Hawthorne Effect is much more
frequently seen in behavioral research, it is also present in medical
research when human subjects are present. Dealing with this problem
is handled by having a control group that is subject to the same
conditions as the treatment groups, then administering a placebo to
the control group. The study is termed a blind experiment when the
subject does not know whether he or she is receiving the treatment or
a placebo. The study is termed double blind when neither the subject
nor the person administering the treatment/placebo knows what is
being administered knows either."
There is an obvious application of the Hawthorne effect in audio: If
you make any change in your system, there is a strong possibility
that you are going to perceive it is an improvement, even if there is
no reliable and inherent benefit. In many cases, people characterize
a system enhancement as "improving the resolution of details" or "
better soundstage." Whether those perceived improvements are due to
the listener simply becoming more attentive is not known without good
experimental controls.
Now, you may wonder why Robert Harley and the rest of the Stereophile
writers never talk about the Hawthorne effect. There are several
reasons. One is that Harley has no credentials that would make anyone
suspect that he's competent in Experimental Design and Operations
Research. He's basically a recording and CD mastering engineer. He
might not have any formal education after high school, except maybe a
number of weeks of OJT or in some technical school. Nothing wrong
with that if what you want to do is be a recording or mastering
engineer.
But, as a designer of experiments involving complex factors, it
really helps to have some formal education in statistics,
experimental design, etc. For me, appreciation of statistics started
when I read a little tract called "How to Lie with Statistics" in 9th
grade. I've taken a fair number of engineering statistics and signal
analysis type courses. I think that getting a BS in engineering
included quite a bit of structured OJT in experimental design for me.
Another probable reason Harley's Stereophile "Manifesto" did not
mention the Hawthorne effect is that the Hawthorne Effect leads one
to a state of mind that is not present in your typical Stereophile
Review: skepticism. A key factor in Stereophile's credibility with
its audience is their suspension of disbelief, not their skepticism.
It's hard to read Harley's paper for any amount of time without
concluding that he lives in a state of suspended disbelief, and that
he is encouraging others to do the same. One of his major supporting
references (5 cites) is a deconstructionalist, post-critical tract by
Polanyi.
Deconstructionalism is a philosphy that in recent days has attacked
all of science, as we know it on the grounds that science is just a
fabrication of white, European, males; which they use to keep
themselves in power. While I won't deny that some use science to get
where they are and stay there, it takes major suspension of disbelief
to really believe that science is a fabrication. It is highly
probable that Harley promotes the thinking of Polanyi because
deconstructionalists are generally death on objectivism. Of course,
to do that, deconstructionalists define objectivism and subjectivism
as antonyms, which a review of the standard US English dictionary
definitions of the words won't support as an absolute truth.
It's really pretty amazing that to me it took as long as 20 years
from the discovery of the Hawthorne effect to develop the early
knowledge of the Placebo Effect, because they are really pretty
similar. The Placebo Effect is generally recognized as being one of
the major discoveries of the 20th century, and understanding and
controlling the Placebo effect is one reason why medicine has
advanced as it has in the second half of this century.
The fundamental observation behind the Placebo effect is that giving
people as sugar pill can have an observable and even lasting health
benefit. A corollary is that when presented with two very similar or
identical alternatives, people will tend to say they are different,
even when they are identical in every way.
A fundamental issue in the "Great Debate" that Harley refers to is
the idea that some audio equipment performs so well that it does not
have any audible effect on the signals it amplifies or stores and
plays back. In the 60's, this led several audio writers including
Julian Hirsch to conclude that all really good equipment
(particularly amplifiers) should sound the same. This led some of
them to observe that in fact, this was at least occasionally the
case.
It is axiomatic that a reviewer does his best work when he can see
the thing he is reviewing in the clearest possible light. If a
reviewer's biases prevent him from noticing an obvious flaw, then his
review has a serious defect. If a reviewer's biases cause him to
observe a flaw that is not actually part of the equipment, but yet
the reviewer attributes it to that equipment, then his review has a
serious defect.
If you review equipment, it helps to be even more than a little
skeptical about not only the equipment, but also your own
observations of it. Ever hear of double-checking your results by
using an alternative means of analysis?
Harley clearly shows his bias to claim the existence of audible
differences, and his comments show that he is hardly skeptical of his
own beliefs. He rants: "When blind listening tests, despite their
effect of obscuring audible differences, indicate that an audible
phenomenon does exist (a phenomenon denied by the engineering
community), the results are either incorrectly reported as a null, or
judged "not statistically significant". The "disinterested"
experimenter often chooses to believe that certain subjects enjoyed
an amazing run of luck rather than that they could discriminate a
difference the experimenter had previously concluded in his own mind
to be inaudible.
For example, during the power-amplifier listening tests conducted at
the 85th AES Convention in Los Angeles, a prominent reviewer of
high-end equipment---a trained, skilled listener---identified a
particular power amplifier in five out of five double-blind trials.
His performance was dismissed by the experimenter as that of a "lucky
coin." The experimenter explained the use of this term to the
subject: If one flips a coin enough times, five heads in a row will
appear on occasion."
The most obvious flaw in this paragraph is that it is a tacit
admission by Harley that he is very naive about experiments and
statistics. In an undergraduate class in statistics or experimental
design, a common, early lab exercise is to flip a coin and observe
what happens. What happens is that you get 5 heads or 5 tails in a
row, from time to time. If you flip a coin 100 times, it's about
certain that you will observe at least 3 runs of "5 in a row".
The second flaw is far more insidious. Harley clearly presumes that
the "right answer" is identification of the power amp. A correct,
skeptical view is that the power amp may or may not be identifiable,
no matter how sensitive and accurate the listener's ears are. Harley
shows this insidious flaw in his arguments later on:
"Perhaps the strongest indictment of blind listening tests is,
ironically, the very test cited by the objectivists that all power
amplifiers sound alike. This test "revealed" that power amplifiers of
widely varying designs and price were sonically identical(31). "
"Amplifiers as diverse as an output-transformerless tube design, an
expensive solid-state unit, and a $220 Japanese receiver, were all
judged---under blind conditions---to be sonically identical. These
amplifiers were as different from each other---on an objective
basis---as one could assemble. Despite the large measurable
differences between these amplifiers, the listeners could not
distinguish among them."
This is a classic case of presuming an answer, and then judging a
test based on the presumed answer. Here is an important question:
Does the existence of measurable differences mean that there is
necessarily an audible difference?
If you study audio technology, the properties of the ear, and
psychoacoustics, the answer is very resounding "no". Just because a
difference is measurable, does not mean it is audible. I routinely
measure frequency response difference of 0.001 dB in my work that I
post at http://www.pcavtech.com . I pinch myself and remember that
the fundamental idea of the decibel is that it is a difference in
level that is in some sense, "just detectable". Well, if 1 dB is in
some sense detectable, then what is 0.001 dB? Well, rather obviously
it is 1/1,000th of what is in some sense, "just detectable". I must
add that the original definition of the dB seems gross in the context
of modern listening tests. I actually think that 1 dB is an amount of
level difference that can be reasonably expected to be heard in the
outdated sorts of listening tests that Harley advocates: his
so-called "single presentation method".
If you want to reliably hear tenths of a dB, the easiest possible
way, you use rapid presentation of alternatives. If you want to hear
hundreths of a dB, you pray! And if you want to hear thousandths of a
dB, you find an alternative universe! ;-)
Harley rants on:
"The objectivists can't have it both ways. If "blind testing shows up
differences very sensitively," yet the same methodology led to the
absurd conclusion that an output-transformerless tubed amplifier, a
high-end solid state design, and a $220 Japanese receiver, all having
very different objective performances (including different linear
performances), were sonically identical, then the inescapable
conclusion is that blind listening tests are fundamentally flawed. If
blind testing is truly sensitive to revealing differences, why were
such gross differences between amplifiers in the cited test not
detected?"
What is lacking from Harley's rant is any proof that the differences
were, in fact "gross" from the perspective of audibility. Actually,
some have pointed out that the amplifiers shared one obvious
similarity: They lacked output transformers. It is well known that
output transformers are a major but sometimes manageable detriment to
flat frequency response, and reasonably flat frequency response is an
absolute requirement for equipment to be sonically transparent.
The audible differences between these amplifiers are "gross" because
Harley says they are "gross", I guess. It's proof by authority and
the authority is Harley. He says he is right, so obviously he is
right! ;-)
Again, Harley's apparently weak and outdated education leads him
astray. Harley's presumption seems to be that an audible difference
ALWAYS or almost always exists. He is arguing that ANY listening test
that fails to show a difference is either highly suspect or a priori,
wrong.
This is a rather gross failure of the basic idea of skepticism.
Admittedly, perfect objectivism is like all perfect things:
impossible in this world. However, does one have to be as
closed-minded and biased as Harley? What we've got in Harley is a
"True Believer", not an analytical reviewer, sophisticated
technologist or apt critic.
Harley's disbelief is suspended from the heights of non-scientific
and even anti-scientific tradition. Hey, he writes for Stereophile
and John Atkinson cites him as an authority. Need I say more? Well, I
will say more! ;-)In the next section I'll deal with the inherent
contradictions of Harley's self-serving rush-to-judgement of blind
tests.
---------------------------------------------------------------
Part 3: Harley damns engineering specifications. Does he know what
they mean?
In this part I'm going to show that Harley damns specs because he
does not understand them. And that he seems to believe that he and
people like him can collect information for reviews, which he claims
are detailed and accurate, during listening sessions; but he also
claims that figuring out whether X is A or B during a listening
session is more than ANY human can possibly be expected to handle.
Harley writes:
"But what is "quality" in an audio component? Is it merely the
ability to meet certain "objective" criteria? I think not. I propose
that audio equipment quality is irreducible to an arbitrary set of
numbers."
This is a rather classic straw man argument. Harley claims that
"audio equipment quality is irreducible to an arbitrary set of
numbers." and of course, everybody with a brain agrees with him. The
reason is that when REAL engineers attempt to reduce the performance
of ANYTHING to a set of numbers, the numbers are hardly "arbitrary".
The American Heritage Dictionary says:
Arbitrary adj. 1. Determined by chance, whim, or impulse, and not by
necessity, reason, or principle. 2. Based on or subject to individual
judgment or preference. 3. Established by a court or judge rather
than by a specific law or statute. 4. Not limited by law; despotic.
If Harley is using the American English language, (and American
English seems to be the appropriate standard for a publication
published in the US) he's made a straw man by claiming that ALL
equipment specs are "Determined by chance, whim, or impulse, and not
by necessity, reason, or principle".
In fact, necessity, reason, and principle determine reasonable
equipment specs. The necessity for equipment specs is the need for
an abstract way to describe equipment performance. The reason is that
suitable abstractions exist, are very applicable, and can be reliably
applied. The principle is reliability. Obviously, Harley is not
talking about reasonable equipment specs in "A Listener's Manifesto".
Since I want to limit my discussion to only items that seem clearly
stated in Harley's paper, I therefore dismiss further discussion of
Harley's clearly irrelevant and erroneous ideas about equipment
specs.
Harley goes on:
"Audio component quality is defined in the listening room---by its
ability to convey the music's essence and meaning without imposing
itself on the musical experience. Some components produce an intimacy
with the music that makes the listener forget the playback system;
others seem to do their best to prevent such a reaction."
I'm going to presume that Harley is trying to create a poetic view of
sonic accuracy, which seems appropriate: he's a much better poet than
a technologist. In more technical terms, the "...ability to convey
the music's essence and meaning without imposing itself on the
musical experience" would be like sonic accuracy.
"My experience suggests that this fundamental
characteristic---perhaps related to the listener's holistic reaction
to the reproduced sound---is a far more meaningful indication of
audio component quality than a set of numbers produced in the test
lab"
There is a real problem here which everybody who has experience AND
understanding of measurements knows, but Harley specifically denies:
"More specifically, there are myriad audible differences between
components whose causes we haven't begun to understand, much less
measure and quantify."
The rather obvious flaw with this claim is the presumption that you
have to fully understand something in order to measure and quantify
it. Do I fully understand all the chemicals and chemical reactions
that are involved in cooking? I took two semesters of inorganic
chemistry in college, and I've read some on the topic, but sorry, no
full understanding of food chemistry is in me. Can I measure
ingredients and follow a recipe and make reasonable-tasting food? The
reaction of dozens and hundreds of people who eat dinners I've been
known to single-handedly prepare for charitable and volunteer groups
seems to indicate: "Yup, Arny can cook up a storm".
One of the key aspects of the business of measurement and
quantification is that you don't need anything like full
understanding to do it in a useful fashion. Harley seems to be
denying that this is true, and in a paragraph shows that he really
does not understand what measuring, quantifying and characterizing is
all about.
Harley seems to think that you need to fully understand something to
measure it in a meaningful way. This means that only skilled senior
automotive engineers can meaningfully weigh a car. Obviously, this is
just not the case.
"Such aspects of musical presentation as soundstage depth, sharpness
of instrumental image outlines, sense of space between individual
instruments, how well soundstage width is maintained toward the rear
of the presentation, and natural reproduction of timbral shadings,
are far beyond the abilities of existing technology to measure are
just a few examples of the currently unmeasurable differences between
components."
Since Harley does not seem to understand that measurements are
abstractions, his inability to resolve the little problem he presents
here is easy to understand. But just because we understand that
Harley's poor abstract reasoning causes him to err significantly,
does not mean that we have to agree with him.
It's true, we can't directly measure "imaging" with extreme accuracy.
However, we can measure a laundry list of things that are well-known
to ruin imaging. In the case of simple, well-understood components
like amps, preamps, and most digital audio equipment, we know that if
we reduce three things below well-known limits, then "imaging" passes
unmolested. Those three things are: noise, linear distortion and
nonlinear distortion.
This isn't to say that the qualities of imaging are somehow mystical
because they defy measurement, only that the resolution of today's
instruments is far greater that of the human auditory system. That
the ear has higher resolution than test equipment was a false claim
10 years ago, and with recent advances, it's a false-claim squared.
The problem with modern measurements is not lack of resolution it's
in the analysis of the reams of data that their extremely high
resolution drops in our lap. Hence, the need for abstract indicators
of performance, AKA specifications.
"Indeed, most of the measurements in use today were developed decades
ago as design tools, not as representations of musical reality."
Again, Harley stumbles over the concept of abstraction. While
claiming to support the sonic accuracy model, Harley seems to believe
that musical reality would somehow not be preserved by a system that
has sonic accuracy.
"The advances made in digital audio data-compression techniques
underscore the role of subjective critical listening in evaluating
audio equipment quality. Data-compression schemes produce huge
objective errors in the signal, errors reportedly masked by the
correctly coded wanted signal. No measurements exist that reveal the
relative quality of data-compression systems: all evaluations are
made by critical listeners."
This is a "wonderful" mixture of correct information and
misinterpretation. Yes, it is correct that as Harley says:
"Data-compression schemes produce huge objective errors in the
signal, errors reportedly masked by the correctly coded wanted
signal." Far be it from a person who believes in the universal
existence of audible differences to admit that despite the huge
objective errors, hearing differences among perceptual coders can be
difficult or impossible.
The proof of how easily high end listeners can be fooled by
perceptual coders can be seen right here on RAO. In my "Perceptual
Challenge", none of the RAO listeners reliably heard audible
differences in program material that was intentionally coded with
poorer perceptual coders, based on my intent that they actually hear
a difference. Only when I chose a perceptual coding technique that
was so egregious in its effects that all signals over 11 KHz were
eliminated did the RAO listeners hear differences. OTOH, two
Detroit-area listeners with years of experience with blind listening
tests did far better than the RAO high-end listeners. One produced
statistically significant results for even the "best" coder I had on
hand, and all the poorer ones. This is not an exceptional result -
the MPEG group tests showed that the same coders produce audible
artifacts.
Harley falsely claims: "No measurements exist that reveal the
relative quality of data-compression systems: all evaluations are
made by critical listeners." The fact is that measurements do exist
that show real differences between data-compression schemes. For
example, compare
http://www.pcavtech.com/play-rec/Fraunhofer-l3enc271-winamp209/index.
htm and
http://www.pcavtech.com/play-rec/Psytel%201068-PB109-winamp209/index.
htm.
In 1992, the same year Harley wrote "Listener's Manifesto", we also
find this little document: "ABXing DCC", David L. Clark, Audio, April
1992, pp.32-34. I believe that Clark's preliminary work was done in
1991, and talked about in industry circles well before publication.
AFAIK Clark found both audible and measurable differences using a
measurement scheme and by doing listening tests based on multi-tones.
I've described on RAO and other USENET newsgroups the fact that
measurements of some modern perceptual coders show signficant
differences at levels that are clearly in the audible range. See
http://www.pcavtech.com for examples. For example, some perceptual
encoders summarly remove program material above some seemingly low
frequency like 15 kHz (for example, see
http://www.pcavtech.com/play-rec/Psytel%201068-PB109-winamp209/index.
htm.
). Other MP3 coders have very limited channel separation in the 12
kHz range. Others seem to add quite a bit of nonlinear distortion to
low bass. This list of measurable flaws is even more remarkable
because high end listeners can't reliably hear them. However, since
people who ignore Harley's claims about listening tests can reliably
hear the artifacts in the best perceptual coders, and people who
ascribe to Harley's claims don't seem to be able to, we've got a
little conundrum here... ;-)
I've also posted an equipment test of a device incorporating a
perceptual coder at
http://www.pcavtech.com/play-rec/mds-j920/index.htm , mores
specifically
http://www.pcavtech.com/play-rec/mds-j920/index.htm#FR_DD that's
shows a number of artifacts that should be audible (see
http://www.pcavtech.com/soundcards/techtalk/FR/index.htm ).
Measurable problems, known to be in the range that is audible, were
known to exist in perceptual coders in 1992, and 1998 and 1999.
Harley was wrong 7 years ago and is still wrong.
Harley seems to believe that he and people like him can collect
information for reviews, which he claims are detailed and accurate,
during listening sessions. However he also claims that figuring out
whether X is A or B during a listening session, a simple task is more
than ANY human can possibly be expected to handle.
Harley writes: "During blind testing, the subject's tendency is to
focus on a specific aspect of the presentation to aid him in
identifying a particular component." This is just a presumption of
his. While analytical listening can be practiced during blind tests,
the exact format of listening used is up to the listener.
Gestalt listening is probably the form of listening that is probably
the farthest removed from analytical or focussed listening. In
gestalt listening, the listener tries to sense a configuration or
pattern of sonic elements that are so unified as a whole that the
properties of the entire sound cannot be derived from a simple
analysis of its parts. There is nothing about a blind listening test
that prohibits this.
It has been the experience of many participants in blind listening
tests, that near the limits of audibility, gestalt listening is
simply how one hears differences. I first personally experienced this
in the early 1980's, shortly after I built the first ABX comparator.
IME, gestalt listening works some of the time. It's a good listening
technique for some situations. It is not a be-all or end-all, no
matter what Harley says.
---------------------------------------------------------------
Part 4: Objectivism versus Subjectivism. Is there a conflict?
In this part I'm going to show that Harley has to make a ton of false
distinctions (besides the ones about subjective and objective, as
well as science versus experience) to support his basic thesis that
reliable listening tests are a "work of Satan", and inherently
insensitive.
"Going beyond the nuts and bolts of blind testing, the procedure is
suspect in that the entire reason we listen to music is subverted:
They turn an emotional experience into an intellectual exercise.
Music isn't merely an arbitrary collection of pitches of varying
amplitude; it is filled with meaning, expression, and feeling.
Indeed, there would be no rational reason for listening to music if
it were merely an incomprehensible and meaningless assortment of
sound. It is the expression of the artist or composer that compels us
to listen. The expression inherent in music is what drives the entire
audio software and hardware industries. Why else would people spend
millions of dollars per year on audio hardware and software?"
Harley's first artificial distinction is the apparent claim that
logic and emotion can't exist concurrently in the same person. He
rants: "They turn an emotional experience into an intellectual
exercise." Problem is that he also wants people to believe that
subjective reviews have the trappings of intellect and science:
"Careful controls are also maintained during subjective critical
listening. Levels between components under audition are matched to
0.1dB or less. Linear differences, such as whether the unit is
polarity-inverting or not, are accounted for. Listening sessions are
conducted virtually daily for weeks or even months before the review
is written. A wide range of familiar source material is used over
long periods of time and over a variety of equipment, precluding the
possibility of ascribing a particular characteristic to a component
that is actually a characteristic of the recording."
However, these are trappings of intellect and science, probably not
the real thing. Some of his claims could be true, but many cannot.
For example, he makes an unqualified claim that "Levels between
components under audition are matched to 0.1dB or less". But
inspection of the literature shows that he and his associates and
cohorts review speakers. It is simply not possible or even useful to
try to match listening levels of speakers within 0.1 dB at all the
possible listening locations in a room.
Harley claims: "Linear differences, such as whether the unit is
polarity-inverting or not, are accounted for". Well if he really
matches levels within 0.1 dB, then AFAIK they ARE accounted for. But
this is not always possible because much equipment, even equipment
among which audible differences are actually difficult or impossible
to reliably hear, does not inherent have levels matched within 0.1 dB
20-20kHz. Furthermore, some of the types of equipment that he
reviews, particularly low-feedback vacuum tube equipment, is subject
to trivial amounts of gain drift over time, but enough
to violate the 0.1 dB rule he claims.
Furthermore, within the context of Harley's "single presentation
method", matching levels within 0.1 dB can be tough. Within the "dual
presentation method" used in ABX tests, level matching is facilitated
by the fact that both pieces of equipment are concurrently present,
warmed-up, and ready to go.
Absolute level matching is more difficult than relative level
matching. For one thing, it requires meters that are within 0.1 dB
from 20-20kHz. Good examples of such equipment are HP 400E, 3400 or
Fluke 8060 meters.
Harley claims: "Listening sessions are conducted virtually daily for
weeks or even months before the review is written." Does each
Stereophile reviewer have this sort of precision equipment at his
disposal and use it daily or hourly (if the equipment is prone to
drift) over the weeks and months that their tests run? Are they
checking for matched levels every time they listen? I kinda doubt it!
It seems incredible to me that Harley's listeners can manage all the
technology he claims they manage during their listening tests, and
not be able to figure out whether "X" is more like "A" or "B". ;-)
---------------------------------------------------------------
Part 5: Subjective Critical listening: All in favor, say "I"!
In this part I'm going to show that no matter what Harley says,
Subjective Critical Listening is not rejected by any of the people
Harley says reject it. In fact they do it often and grant it a ton of
credibility. They just don't buy
Harley's egregiously flawed ideas about how to do it.
Harley rants:
"If the methods of critical listening are invaluable in assessing
audio equipment quality, why would rational scientific minds
summarily reject these techniques and brand their practitioners
frauds and charlatans? The question cuts to the core of a much deeper
conflict between science and values."
While stated as a question, we all know Harley's answer to the
question.
Critical listening is invaluable for assessing audio equipment. But,
it's not the only way, and it's not always the best way.
As shown previously, there is a whole class of audio equipment that
is (often professionally) sonically accurate. If used in relatively
long chains. Distortion and noise can build up in the chain. If the
chain is to be sonically accurate, then each component must have
distortion that is way below the threshold of audibility. However,
Harley's one and only way, his much-vaunted "single presentation
method" does not address this kind of requirement because, after all,
he only uses a single presentation.
Long chains are becoming more common in high-end systems as people
cascade players, crossovers, amps, signal processors. A minimal
system is the cascade of at least 3 components. There is always a
long chain if the recording equipment is considered. Because we can't
control the entire chain, we don't know how close any recording is to
the thresholds of audibility or irritation.
Therefore, a system that is composed of as much equipment with
inaudible
distortion as reasonably possible makes sense. However, Harley's
"single presentation method" is very weak for doing that. After all,
how does one characterize inaudible levels of distortion with a test
based solely on audibility?
"The rejection of critical listening stems from the mistaken belief
that any acceptance of listening impressions is tantamount to a
rejection of rationality itself.
Given that nobody who practices reason rejects critical listening as
a rule, nobody is approaching the rejection of rationality that
Harley claims.
"The listening experience contains no matter, has no energy, cannot
be quantified by instrumentation, and is therefore considered merely
a creation of the mind."
We now know that the state of the mind is defined by states of matter
and energy. We can't resolve what our crude brain-state measuring
equipment tells us into specific thoughts, but we can surely tell the
difference between a mind that is clearly capable of thought and
reason from one that clearly can't. Regrettably I have personal
experience with such determinations, due to
the recent death of my 16 year old son caused by a spontaneous brain
hemorrhage. When sophisticated tests showed that his brain was truly
dead, his body was as pink and healthy as ever. Little pieces of his
brain were probably operating his body as they always had. But
thought and reason was no longer possible, and it all was collapsing
rapidly. He was dead.
The listening experience should never be a creation solely of the
mind, but a response of the mind to stimulus. If the listening
experience is solely a creation of the mind, why have speakers,
recordings and amplifiers? If we accept that the listening experience
is a response to stimulus, then we have to face the facts that some
potential stimuli are either so small that even our basic sensory
organs don't respond to them, or they are so similar that there is no
difference in how our basic sensory organs respond to them.
The obvious false claim in Stereophile's "Recommended Equipment List"
is that
so many things that lack any known physical effect, or have physical
effects that are so identical to other things are rated or rated
differently. Harley seems to claim that is because somehow they do
something that we don't understand. However, many of these components
are simple amplifiers and preamplifiers that perform a simple
two-dimensional function. We can measure that two dimensional
function almost to the point of absurdity, and certainly several
orders of magnitude below what the ear can reliably hear. While we
can't resolve the complexity of the music in that two dimensional
function, we can tell if it has significantly changed.
In many cases Stereophile claims substantial audible differences when
we know for sure that the simple, two-dimensional function is not
significantly changed. Sometimes the changes are less than the
ridiculously small changes that we can measure at all, yet the
reviewers at Stereophile and comparable magazines seem to claim they
"hear" substantial and meaningful differences.
This would be a mystery if Stereophile's listening tests were
rigorous and performed in accordance with the best available
technology. Regrettably, Stereophile has repudiated the best
available technology and claims that all existing approaches to
reliable testing are not applicable or worthy of development by
anyone, particularly Stereophile, themselves.
Since the listening tests by Stereophile are not done in accordance
with the best available technology, and it is easy to show that their
purported arguments against the best available technology are rooted
in anti-scientific post-modernist mysticism, outright error, and
out-of-date knowledge. There is no reason to grant them much
credibility, particularly when they contradict common sense and what
is known about the properties of the human ear and psychoacoustics.
---------------------------------------------------------------
Part 6: The Single Presentation Method: Poetry or Pain?
Harley writes:
"The "single presentation method" is the preferred technique of
assessing a component's quality. In this method, the component under
review replaces a component in a known reference playback system, and
the reviewer spends weeks or months listening to music through it.
The same level-matching controls and awareness of relative response
errors are used as in direct comparison listening. Although some A/B
comparisons with other known or comparably priced components are
made, the single presentation method is the best way to determine the
long-term quality of the component in question (8)."
Now, this paragraph contains a basic contradiction and/or a fallacy:
"The same level-matching controls and awareness of relative response
errors are used as in direct comparison listening"
As a rule, one of the major hassle parts of any listening test is the
area of level matching. As I pointed out two days ago, it is always
harder to match levels in a single presentation situation than in a
direct comparison. In a single presentation, long-term stability of
the test signal, the test instruments, and the equipment under tests
is paramount. In direct comparisons, only short-term stability
matters.
Since Harley says: "Listening sessions are conducted virtually daily
for weeks or even months before the review is written." (a claim I
provided contradictory evidence for in part 5), we are talking long
term stability which just might not exist at all in units under test,
and is not trivial to provide in test equipment and test tones
sources. This would be particularly true in an analog-oriented
environment.
The second problem is that the single presentation method requires
the listeners to perform superhuman acts of audible memory that even
Harley criticizes,
Harley says:
"Research indicates that the limits of consciousness are far lower
than previously assumed---limits that are routinely exceeded during
blind testing. Csikszentmihalyi writes: "At this point in our
scientific knowledge we are on the verge of being able to estimate
how much information the central nervous system is capable of
processing. It seems we can manage at most seven bits of
information---such as differentiated sounds, or visual stimuli, or
recognizable nuances of emotion or thought---at any one time, and
that the shortest time it takes to discriminate between one set of
bits and another is about 1/18 of a second. By using these figures
one concludes that it is possible to process at most 126 bits of
information per second, or 7560 per minute, or almost half a million
per hour. It is out of this total that everything in our life must
come---every thought, memory, feeling, or action. It seems like a
huge amount, but in reality it does not go that far."
At several points I've dealt with Harley's ludicrous claim that these
limitations don't apply to "single presentation method" listening
sessions. While Harley may suffer hysteria during blind tests due to
his mind set and fear of being shown to be a charlatan (IMO, a
valid, rational fear for him!),not everybody does.
Hysteria during listening tests is not even likely unless you've made
a lot of wild claims you don't want to "eat" either personally or
privately. After all, blind listening is just a thing that you do to
find some stuff out. Unless you have a big emotional stake in a
particular outcome, what matters about the outcome during the
listening?
Here is another example of how adopting a agonistic, skeptical mind
set leads to the most sensitive and reliable outcome. And, since
Harley has tacitly admitted that he thinks everything sounds
different and any test that does not show this is fatally flawed,
he's just not gonna be a sensitive reliable listener until he settles
down into a more scientific, agonistic, and skeptical viewpoint.
Furthermore, critical listening can be pretty much be critical
listening as long as the listener can get into the right mind set.
Similarly, gestalt listening can be pretty much be gestalt listening
as long as the listener can get into the right mind set. Exactly what
that mind set is, obviously varies with the listener and the nature
of the things being listened to and for. At some point, it all comes
down to listening and remembering enough about what you heard to make
a judgement.
However, Harley has made an estimate of quantities of data, per time
unit, associated with listening.
The problem is that this data builds up across the listening session,
and is essentially what the next listening experience is compared to.
Since Harley claims that he relies on single presentations that
obviously last for minutes or hours or even days or months, he's
requiring that people remember and compare massive amounts of data in
order to come up with a judgement.
In contrast, the listener-controlled direct comparison technique (a
keystone of ABX testing, for example), allows the user to adjust his
listening presentations to suit him and the listening task at hand.
This minimizes the requirement that the listener remember a lot about
the last piece of equipment that was presented.
Another way to look at this is that mental comparisons of sonic
alternatives are based on what the listener remembers. Reducing the
time over which information needs be remembered can minimize the
degree to which the information becomes flawed by human memory.
Therefore, the remembrance of the older alternative is likely to have
more flaws. Flaws in human memory can lead to the perception of an
audible difference that based on memory flaws, not sound quality
differences that are inherent in the alternatives being compared.
Reducing the time between listening to sonic alternatives can lead to
more sensitive and reliable results.
---------------------------------------------------------------
Part 7: How Harley contradicts Harley
Today, I'll show another example of Harley's one-sided critical
technique.
For background, lets consider how reasonable people do a balanced
critique of two alternatives. What many (if not most) skilled
analysts do is come up with a list of reasonable criteria for
judgement and explain them. Then the analyst applies the criteria to
both alternatives evaluates them and explains that. Then the analyst
does a summary and a recommendation. Access to the criteria and the
application lets the reader determine how much credibility he wants
to give the summary and the recommendation. A good list of reasonable
criteria for judgement shows the good and the bad in both
alternatives. This is how life is, right? Some good, some bad, and
nothing is perfect.
In Harley's "Manifesto", this is clearly not done. Not even a vague
approximation. Aside from the absence of any kind of rigorous
structure, Harley fails on the most important part of all, and that
is the good-and-bad evaluation of both alternatives. Harley seems to
say only good things about how Stereophile does things, and he seems
to say only says bad things about how Stereophile's critics does
things.
In its simplest form, the story about open evaluations and reliable
listening tests is that open evaluations are easy to set up and do,
and reliable listening tests usually take more work to set up and do.
OTOH, open evaluations are always at risk for being contaminated by
the placebo effect, while good reliable listening tests are not. Some
experts think that discovering the placebo effect in subjective
tests, and learning how to control it in subjective tests was the
most important discovery of the second half of this century.
Of course, as I pointed out before, Harley's "Manifesto" NEVER
mentions the placebo effect. Too much like presenting a balanced
view?
---------------------------------------------------------------
Part 8: Harley's intellectual dishonesty and lying about the
documents he references.
(Introduction by Arny Krueger) Paragraphs preceded by a letter are
from http://www.stereophile.com./fullarchives.cgi?20
"The Listeners' Manifesto" By Robert Harley and John Atkinson,
January 1992.
Paragraphs preceded by quotation marks are from "Do All Amplifiers
Sound the Same?" David L. Clark and Ian G. Masters (pp. 78-84),
Stereo
Review Magazine, January 1987, which was cited and criticized
specifically and repeatedly by Atkinson's and Harley's "Manifesto".
Notice that "Do All Amplifiers Sound the Same?" was published 5 years
before "The Listeners' Manifesto". Harley and Atkinson
had plenty of time to research their work, and there should be no
reason why Harley's and Atkinson's claims about "Do All Amplifiers
Sound the Same?
would be so flawed and largely irrelevant.
WHAT STEREOPHILE'S "LISTENER'S MANIFESTO" CLAIMS:
a) The experimenter's agenda is often to prove that no audible
differences exist rather than to discover if differences do exist.
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"Clark's task was to set up a series of tests that would not only
satisfy his own technical standards and those of Stereo Review
magazine but that would be conducted in such a way as to meet must
of the potential criticisms of the Believers (that amplifiers sound
different). As far as possible the aim was to forestall claims that
the procedures were not adequate to reveal amplifier differences."
"The best solution was simply to present my plan to some audiophile
be- in significant sonic differences between amplifiers and ask for
their assistance. After all, they should see this as an opportunity
to prove the validity of their belief to the skeptics. The
cooperation I received from manufacturers, a local high-end audio
salon, and other audiophiles was more than I ever hoped for. Their
assistance and participation as listeners in this project
demonstrated that they were secure in their belief and brave enough
to risk being exposed to an uncomfortable outcome."
"Skeptics and Believers were never combined in the same test, and the
Believers' tests were conducted by an audiophile Believer."
"The design of the present test, one that would make it as easy as
possible to hear differences between amplifiers, therefore presented
some special challenges. First, I had to keep my old skepticism about
the audibility of amplifier differences from influencing the test.
Second, a highly pedigreed sound system, acceptable to the most
critical listeners, had to be assembled.
"Third was the matter of finding those listeners. The SMWTMS group
provides my usual pool of experienced listeners. But most had
previously participated in such tests and had become as skeptical as
I.
"In addition, I wanted to end up with more than statistics from the
tests: I wanted to record the emotional experiences of the listeners
as they discovered how small (if not inaudible) the differences are
between gain-matched amplifier s operated below clipping."
WHAT STEREOPHILE'S "LISTENER'S MANIFESTO" CLAIMS:
b) There is an adversarial relationship between subject and
experimenter, and
the subject is aware that he will
be exposed to ridicule if he "fails."
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"Practically all listeners, including Skeptics, felt at this point
that there were audible differences - some with satisfaction, some
with amazement. Even so, it was immediately apparent that whatever
differences there were tiny, although many of the Skeptics began to
feel that they could now understand what the Believers had been
talking about."
"Skeptics and Believers were never combined in the same test, and the
Believers' tests were conducted by an audiophile Believer."
"The listeners even decided which kind of blind testing they would
use:
Manual swapping of the cables feeding the chosen amplifiers (in this
case hidden behind a screen) or ABX switching where the relay system
would allow a more rapid change over. Listeners new to the ABX system
had already been trained in its operation at a separate setup while
they waited their turn in the listening room. Most listeners opted
for the convenience of the ABX relay control, Feeling that the extra
contacts in the system would not degrade the signal Nine people.
however, chose to augment their ABX tests with blind cable-swap
tests."
"The best solution was simply to present my plan to some audiophile
be- in significant sonic difference s between amplifiers and ask for
their as assistance. After all, they should see this as an
opportunity to prove the validity of their belief to the skeptics.
The cooperation I received from manufacturers, a local high-end audio
salon, and other audiophiles was more than I ever hoped for. Their
assistance and participation as listeners in this project
demonstrated that they were secure in their belief and brave enough
to risk being exposed to an uncomfortable outcome."
(Listener comment about the tests) "I prefer using the ABX switcher
because the plug-unplug time limits the accuracy of my comments."
(Listener comment about the tests) "The source material, different
types of amplifiers and ancillary equipment would appear to be
capable of showing differences. However, I was hard pressed to tell
the differences."
WHAT ATKINSON AND HARLEY CLAIM:
c) The playback system, music, room, and other conditions are all
foreign to
the subject.
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"A major feature of these listening-test sessions was their openness.
All the equipment could be seen and inspected. At listener request,
all the amplifiers could be auditioned with or without the ABX system
prior to the blind testing. Listeners were given as much time as they
needed to get used to the audio to system. to select revealing
program material, and to note apparent differences between the
various amplifiers. Almost all listeners (even the skeptics)thought
they could hear differences at this point."
"Two options were offered for the tests themselves. The listeners
could choose to have the operator manually swap cables between the
two units being listened to in any given session if they felt that
would contribute to the accuracy of their responses. It was expected
that some of the Believers would prefer this method, as many high-end
audiophiles are leery of instantaneous A/B testing, and this proved
to be the case."
"At the beginning of each session, every listener was given a form
that asked whether or not the test conditions were adequate to prove
whether differences between the amplifiers were audible.
"After a preliminary round of listening, which lasted about an hour,
all but three subjects signed these sheets. One of those who did not
said conditions were not adequate, one had reservations and one gave
no answer."
"Nevertheless, a majority of listeners including some of the
Believers, approved of the test methods both going in and coming out,
the amplifiers chosen varied widely in design and price, and the
sample of listeners was diverse and large, as these things go. And
the results indicated no audible differences."
"After the formal tests, each listener was again asked whether or not
the tests were adequate to reveal audible differences and also
whether or not the tests could be considered relevant to consumers.
In pan this inquiry was simply for information, but in part it was
also to gauge how attitudes toward the tests changed when the results
were known. In a number of cases, Believers' feelings about the
adequacy of the tests were modified or reversed."
"The listeners even decided which kind of blind testing they would
use: manual swapping of the cables feeding the chosen amplifier s (in
this case hidden behind a screen) or ABX switching where the relay
system would allow a more rapid change over. Listeners new to the ABX
system had already been trained in its operation at a separate setup
while they waited their turn in the listening room. Most listeners
opted for the convenience of the ABX relay control, Feeling that the
extra contacts in the system would not degrade the signal Nine
people, however, chose to augment their ABX tests with blind
cable-swap tests."
(Listener comment about the tests) "I prefer using the ABX switcher
because the plug-unplug time limits the accuracy of my comments."
(Listener comment about the tests) "The source material, different
types of amplifiers and ancillary equipment would appear to
be capable of showing differences. However, I was hard erased to tell
the differences."
WHAT STEREOPHILE'S "LISTENER'S MANIFESTO" CLAIMS:
d) The experimenter controls all aspects of the test, including the
music used, playback level, how long the subject can hear each
presentation, how many times the subject can hear each presentation,
the rapidity of switching between presentations, and in which musical
passage the switching occurs.
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"Skeptics and Believers were never combined in the same test, and the
Believers' tests were conducted by an audiophile Believer."
"The best solution was simply to present my plan to some audiophile
be- in significant sonic difference s between amplifiers and ask for
their as assistance. After all, they should see this as an
opportunity to prove the validity of their belief to the skeptics.
The cooperation I received from manufacturers, a local high-end audio
salon, and other audiophiles was more than I ever hoped for. Their
assistance and participation as listeners in this project
demonstrated that they were secure in their belief and brave enough
to risk being exposed to an uncomfortable outcome.
"ABX switching and selection among the available LP's and CD's was
performed by the listeners. Throughout all the tests the equipment
could be seen at the front of the room, although there were no hints
as to which amplifier was playing at any one time."
"The listeners even decided which kind of blind testing they would
use: manual swapping of the cables feeding the chosen amplifiers (in
this case hidden behind a screen) or ABX switching where the relay
system would allow a more rapid change over. Listeners new to the ABX
system had already been trained in its operation at a separate setup
while they waited their turn in the listening room. Most listeners
opted for the convenience of the ABX relay control, Feeling that the
extra contacts in the system would not degrade the signal Nine
people, however, chose to augment their ABX tests with blind
cable-swap tests."
(Listener comment about the tests) "I prefer using the ABX switcher
because the plug-unplug time limits the accuracy of my comments."
(Listener comment about the tests) "The source material, different
types of amplifiers and ancillary equipment would appear to be
capable of showing differences. However, I was hard erased to tell
the differences".
WHAT STEREOPHILE'S "LISTENER'S MANIFESTO" CLAIMS:
e) The experimenter controls the number of successive trials without
regard for the subject's fatigue factor, increasing the number if a
trend indicating reliable identification appears
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"In all, some fifty-four tests were run, most of them requiring
sixteen choices by each listener (to save time, the cable-swap tests
required I only five choices). A total of 772 choices were made."
"Out of all those decisions, one could expect 386 correct choices
through chance alone. In fact, the overall score was 388. So for this
panel of listeners, overall, and this group of amplifiers, no
statistically significant audible differences were detected."
WHAT ATKINSON AND HARLEY CLAIM:
f) The number of successive trials is far too high in an attempt to
get a greater statistical sample size.
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"In all, some fifty-four tests were run, most of them requiring
sixteen choices by each listener (to save time, the cable-swap tests
required I only five choices).
(Comment by Arny Krueger) In recent RAO correspondence, John Atkinson
claimed that the measurable differences between the equipment in
this test was such that it was reasonable to expect that there were
audible differences, and seemed to reiterate his "Manifesto" claim
that since audible differences were not found, the Clark and Masters
test was invalid.
WHAT CLARK AND MASTERS ACTUALLY WROTE:
"Equally important was the choice of amplifiers to be tested. They
had to run the gamut from truly exotic to mass-market cheap, with
some interesting things in between. Since the number of amplifiers
that could be included in the test was limited, the probability that
there would be differences within the group had to be reasonably
high. For this reason, the low end was represented by a modest
Pioneer receiver, the Model SX-1500, priced at $220, while the upper
end of the scale was represented by New York Audio Labs' Julius
Futterman OTL-I tube amplifier, which weighs in at a hefty $6,000 per
channel. (The pair of Futterman's, with their separate power
supplies, made a stack roughly the size of a small stove and gave off
about as much heat.) In between these two extremes were one
audiophile favorite, the Mark Levinson ML-I I ($2,000), and two well
respected mid-price units, the Hafler DH-1 120 (S320) and the AND
2200 ($548). The Counterpoint SA-12 (S995) represented
tube-transistor hybrids, but its untimely demise during the early
listening tests prevented its full participation. The gains of all
the amplifiers were equalized by attenuators in the tape monitor loop
of the Audio Research preamplifier."
"Inserted in the tape loop of the preamp was a precision attenuator
that I trimmed gain to match the outputs of amplifier s in the test
within +/-0.05 dB. This attenuator as well as the ABX Comparator
relay module (when used) was connected by short lengths of the highly
acclaimed Hitachi LC-OFC cable. All connections were treated with a
small amount of Tweek or Cramolin contact enhancer"
(Comment by Arny Krueger) In fact, I know of no sufficiently
comprehensive, comparable technical tests of ALL this equipment. I
have invited Mr. Atkinson to share what he has in this regard, if he
has anything to present at all, but he seems unable to respond in any
meaningful way even after several weeks. I simply don't think he has
or ever had definitive technical information in this matter.
I recently interviewed a technician who assisted Clark, Mark Ziemba.
He said that the NYAL amps were surprisingly like the solid state
equipment also tested, in terms of low distortion and flat frequency
response.
Jan Didden
that personal) into an argument about DBT etc etc. May I come back to
feedback?
The late Peter Baxandall did some testing on the relative levels of
harmonic distortion in an amp while moving from 0dB to 40dB feedback.
Interesting result: 2nd harmonic started to decrease with increasing
feedback (which what we would expect), but all the other (higher order)
harmonics started to INCREASE! With further increase in feedback, all
harmonics (in order of ... eh ... order) eventually also started to
decrease, but the higher the harmonic number, the more feedback was
needed to start to decrease it. Looking at the curves, there are regions
where many higher order components were higher than with either more or
less feedback.
Now, whether this is audible, I don't know, but it certainly was clearly
measurable. Also, this was measured on a particular single amplifier
stage, and I'm not sure how this applies to other configurations.
However, some conclusions stands out:
1 - More feedback does not automatically mean less overall distortion;
2 - The distortion level has a great influence on the distortion spectrum
as well.
Any comment, anyone?
Jan Didden
(Engage brain before operating mouth)
"Arny Krüger" wrote:
> Trevor Wilson wrote in message ...
> >>
> >> >Some years ago, I was involved in some DBT's, suing two identical
> >> >amplifiers: One utilising a small amount of loop NFB and the
> other,
> >> none.
> >> >All conventionally measured specs (THD, IMD, FR, Damping factor,
> >> etc) were
> >> >below audible limits, for both amps. The zero loop NFB product
> was
> >> preferred
> >> >by all listeners, under all conditions.
> >>
> >> Actually, Trevor has partially revealed the circumstances of these
> >> tests in the ppast on RAO, and they were hardly DBT's. If I can
> coax
> >> him into doing it again, the truth will be outed.
>
> >**I have tabled the nature of these tests, previously.
>
> That has to do with the fact that your alleged evidence can't stand
> scrutiny.
>
> >They did not use the
> >so-called 'ABX system', yet, in their own way were double blind
> tests.
>
> There are only two kinds of "blind listening tests":
>
> (1) Those that are adequately blind and reduce the listening test to
> the simple matter of "just listening", and
> (2) Everything else, for whom the outcome is hard to attribute to
> anything in particular.
>
> >They
> >were a placebo trial, where the listeners were unaware of what they
> were
> >listening to and the tester was also unaware. That constitutes
> double blind
> >and is acceptable.
>
> Please state the means by which you ensured that was indeed, the
> case.
>
robert_...@my-deja.com
nite...@voicenet.com (Barry Mann) wrote:
> In <3741FAFA...@shadow.net>, on 05/18/99
> at 10:33 PM, Randy Yates <ya...@shadow.net> said:
>
> >Hello Group,
>
> >I have heard for a very long time the pooh-poohing of negative
feedback
> >in audio circuits (amps, preamps, etc.). Back in the 70s (I believe),
it
> >was linked to excessive TIM (transient intermodulation distortion).
>
> >I have never really been able to get from "here" to "there" on this
> >issue, "here" being the level-headed, unassuming engineer that likes
to
> >see everything demonstrated analytically, "there" being the
distortion
> >that is purported to exist. Can anyone bridge this gap with
absolutely
> >no hand-waving?
>
> Go to the Audio Engineering Site and look through the JAES (Journal of
the
> Audio Engineering Society) archives. In the early 80's there were a
bunch
> of papers on the subject. In many cases the citations will be more
useful
> than the papers.
>
> Basically the unwary assumed that, given enough feedback, any trash
open
> loop amplifier could be turned into something wonderful. Perhaps this
was
> true for a 1000 Hz steady signal, but at higher frequencies that sad
> little open loop amplifier would cause grief.
>
> Even though the problem was officially "discovered" by the audio
community
> at about 1980 (I don't recall the exact date of Otala's paper), I can
> remember a few voices in the wilderness warning of the problem in the
mid
> 1960's. I can't imagine anyone who graduated from a good undergraduate
> school after the mid 60's (and who paid attention) was surprised by
the
> "discovery". I suspect that many of the senior audio designers in the
70's
> had never attended a formal course on feedback. In the late 40's and
50's,
> the feedback calculations that we can perform on a napkin, were
doctorial
> stuff and beyond.
>
> -----------------------------------------------------------
> nite...@voicenet.com (Barry Mann)
> -----------------------------------------------------------
>
>
When cheap op-amps first came on the market in the early 1970s, I made a
flat amplifer stage from a 741. This op-amp has poor open loop
distortion performance. As I recall the loop had 70 db of negative
feedback. I expected the high negative feedback would eliminate all
distortion. I was very disapointed when I (and several others) listened
to.
Bob S.
--== Sent via Deja.com http://www.deja.com/ ==--
---Share what you know. Learn what you don't.---
Richard D Pierce
>When cheap op-amps first came on the market in the early 1970s, I made a
>flat amplifer stage from a 741. This op-amp has poor open loop
>distortion performance. As I recall the loop had 70 db of negative
>feedback.
Not at anything higher than about 10 Hz, it didn't.
>I expected the high negative feedback would eliminate all
>distortion. I was very disapointed when I (and several others) listened
>to.
The open loop gain of a 741 was on the order of about 100 dB
or so, AT DC! However, the first pole in the open loop response
was on the order of 1-2 Hz, and THAT'S one of the many obvious
factors that was ignored at the time.
So, by the time you're getting up in the realm of high
frequencies, your 70 dB of feedback has degenerated to well less
than that. For example, how high in frequency do you think you
have to go before your feedback dropped from 70 dB to, say, 30
dB? Would you be shocked to learn that it was well below 20 kHz?
How about 1 kHz.
And this is but ONE aspect about the simplifying assumptions
made about feedback which is simply wrong.
--
| Dick Pierce |
| Professional Audio Development |
| 1-781/826-4953 Voice and FAX |
| DPi...@world.std.com |
Stewart Pinkerton
>When cheap op-amps first came on the market in the early 1970s, I made a
> flat amplifer stage from a 741. This op-amp has poor open loop
>distortion performance. As I recall the loop had 70 db of negative
>feedback. I expected the high negative feedback would eliminate all
>distortion. I was very disapointed when I (and several others) listened
>to.
Check out the open loop gain, it's flat to about 1Hz! How much
negative feedback (if any) you think you had at 10kHz? :-)
Stewart Pinkerton
>I realize that the forward part of this thread degenerates (don't take
>that personal) into an argument about DBT etc etc. May I come back to
>feedback?
Yuk Yuk, very good!
>The late Peter Baxandall did some testing on the relative levels of
>harmonic distortion in an amp while moving from 0dB to 40dB feedback.
>Interesting result: 2nd harmonic started to decrease with increasing
>feedback (which what we would expect), but all the other (higher order)
>harmonics started to INCREASE! With further increase in feedback, all
>harmonics (in order of ... eh ... order) eventually also started to
>decrease, but the higher the harmonic number, the more feedback was
>needed to start to decrease it. Looking at the curves, there are regions
>where many higher order components were higher than with either more or
>less feedback.
>Now, whether this is audible, I don't know, but it certainly was clearly
>measurable. Also, this was measured on a particular single amplifier
>stage, and I'm not sure how this applies to other configurations.
>However, some conclusions stands out:
>1 - More feedback does not automatically mean less overall distortion;
>2 - The distortion level has a great influence on the distortion spectrum
>as well.
>
>Any comment, anyone?
>
>Jan Didden
>(Engage brain before operating mouth)
OK, I'll try, just this once!
Baxendall's results were geared to the technology available at the
time (more than forty years ago!), with limited gain-bandwidth
product. Basically, negative feedback will reduce harmonic distortion
so long as the open loop gain exceeds the closed loop gain in vector
terms (i.e. including phase angle). Above this frequency (which could
be just a few hundred hertz in many commercial designs!) the feedback
is no longer negative and distortion will therefore increase. In 1999,
An *open loop* bandwidth of 50kHz is easily achievable at very high
gain, allowing vanishingly low THD and IMD across the whole audio
band.
It's a very simple concept, often conveniently forgotten by those who
are promoting 'zero feedback' amplifiers because that's what they are
selling..............................
Stuart Parker
<ar...@flash.net> writes
>
>Stuart Parker wrote in message ...
>>There is no conflict here.
>>
>>The confusion lies in the purpose for which the two camps carry out
>>their tests.
<27 pages of reply snipped>
Sorry for the delay in replying but I've had a bit of reading to do ;-).
Thanks for an informative reply. I can find nothing to disagree with in
any of it.
I think the poster to whom I had replied said that 23 different
individuals made the same choice of preferring amp A over amp B and they
were blind to the contents of the amp "black box", but the listening
test conditions were not otherwise controlled. These listeners were
being asked a different question than if they were being asked to tell
if X is A or B under particular listening conditions and the purpose of
the experiment appears to have been to identify a consumer preference.
It may be that the preference was formed according to some measurable
characteristic of the amps, but the experiment was simply designed to
discern if a consistent preference would be formed by potential
purchasers. Perhaps they all preferred the heavier one or something - I
don't know - and, for the experimenter who is trying to decide which box
to try and sell, it doesn't make any difference to the decision which
the trial result will help to make. The results can't be used to draw
generalised conclusions about the mechanism(s) underlying the formation
of the preference and what proportion (if any) of those mechanisms were
purely acoustic in nature. To do that you need a quite different
experimental approach as you describe so well in your 27 page post.
These concepts are discussed rather more eloquently than I can manage,
in relation to clinical (i.e. medical) trial design in :
Schwartz, Lellouch J. Explanatory and pragmatic attitudes in
therapeutic trials. J Chron Dis 1967;20:637-48.
Schwartz D, Flamant R, Lellouch J. Clinical Trials. London. Academic
Press 1980.
You may be interested to know (of course you may not) that I come from a
medical background but have come to start thinking about this particular
question only recently through an interest in hobby electronics. I
built myself a stereo power amp out of interest to learn more about
electronics and when I'd built it I was eager to know if it sounded any
good (it measured fine by my amateur testing). I soon realised that I
was unable to tell if it was better or worse than the commercial
amplifiers I compared it with - possibly because it sounded the same or
possibly because of the powerful placebo effect that "constructors
pride" was likely to have induced. Blind testing is the next obvious
step, but double blind might be quite difficult to manage securely
without a randomising switching device.... (been there? done that?)
Regards, and thanks once again for an informative post
--
Stuart Parker
Stuart Parker
<pat...@popmail.dircon.co.uk> writes
>Stuart Parker <Stu...@parker-towers.demon.co.uk> writes:
>
>>In article <01bea6ec$39859120$86536ccb@currentu>, Paul Cambie
>><cam...@ozemail.com.au> writes
>>>> under realistic domestic conditions, under which, almost by definition,
>>>> some of the various potential explanatory variables will be manipulated
>>>> by the listener until they achieve a satisfactory set of listening
>>>> conditions for each of the items under test. This is a pragmatic trial
>>>> and there is no a priori reason to suppose that the results of
>>>> explanatory and pragmatic trials will be in agreement.
>>>
>>>meaning outside the practical environment of each test event? The emperor
>>>would alter the trim of his new clothes until satisfied with his appearance
>>>and be happy with the result.
>>
>>Not if a large enough proportion of test subjects, faced with the same
>>choice (double blind) form the same preference. This result would mean
>>that in some generalisable sense the unit which was consistently
>>preferred was in some way well...preferable. The challenge then would
>>be to understand what it was about the test item that the test subjects
>>found preferable - they might all say they preferred item A over item B
>>because it was red!
>
>I think you're missing the point here. The basic tenet of a blind test
>is that the subject does not *know* which unit is playing at any time,
>so there is no way to 'prefer' A over B by any means other than
>*audible* difference.
>
>If you can *see* the red amplifier, then you might well prefer it,
>especially if you substitute Pass (or whatever) for 'red'. That would
>not however tell you if it *sounded* better, or indeed different.
>
identical black boxes. Spray half of amplifiers A blue and half red.
Do the same for amplifier B. Now mix up the A's and B's so that you
have 200 red and blue boxes but can't tell which is which (randomised).
Now make 100 pairs of amplifiers by taking one red and one blue box for
each pair. You don't know whether the pairs you've assembled contain 2
A's 2 B's or an A and a B (single blind). Send a pair of amplifiers to
100 listeners and ask the question "which amplifier do you prefer, the
red one or the blue one?". The listeners can not tell which box
contains amplifier A and which box contains amplifier B (double blind).
After the listener has expressed their preference, open the amps to find
out which was which. It may be that in this randomised, double blind
trial, a consistent preference might be expressed for the red
amplifiers. It's still a double blind trial with respect to the
amplifier topology under test.
--
Stuart Parker
Arny Krüger
Stuart Parker wrote in message ...
>identical black boxes. Spray half of amplifiers A blue and half
red.
>Do the same for amplifier B. Now mix up the A's and B's so that you
>have 200 red and blue boxes but can't tell which is which
(randomised).
>Now make 100 pairs of amplifiers by taking one red and one blue box
for
>each pair. You don't know whether the pairs you've assembled
contain 2
>A's 2 B's or an A and a B (single blind). Send a pair of amplifiers
to
>100 listeners and ask the question "which amplifier do you prefer,
the
>red one or the blue one?". The listeners can not tell which box
>contains amplifier A and which box contains amplifier B (double
blind).
>After the listener has expressed their preference, open the amps to
find
>out which was which. It may be that in this randomised, double
blind
>trial, a consistent preference might be expressed for the red
>amplifiers. It's still a double blind trial with respect to the
>amplifier topology under test.
You buying the 200 amps? ;-)
Arny Krüger
Stuart Parker wrote in message ...
>Blind testing is the next obvious
>step, but double blind might be quite difficult to manage securely
>without a randomising switching device.... (been there? done that?)
It turns out that I invented such a switching device over 20 years
ago. You can read about it at
http://www.oakland.edu/~djcarlst/abx.htm .
Lately I devised a software version of it which I intend to
distribute widely once I get it where I want it.
Trevor Wilson
Stewart Pinkerton <pat...@popmail.dircon.co.uk> wrote in message >
> OK, I'll try, just this once!
>
> Baxendall's results were geared to the technology available at the
> time (more than forty years ago!), with limited gain-bandwidth
> product. Basically, negative feedback will reduce harmonic distortion
> so long as the open loop gain exceeds the closed loop gain in vector
> terms (i.e. including phase angle). Above this frequency (which could
> be just a few hundred hertz in many commercial designs!) the feedback
> is no longer negative and distortion will therefore increase. In 1999,
> An *open loop* bandwidth of 50kHz is easily achievable at very high
> gain, allowing vanishingly low THD and IMD across the whole audio
> band.
**Actually, this kind of open loop bandwidth has been available for several
decades.
>
> It's a very simple concept, often conveniently forgotten by those who
> are promoting 'zero feedback' amplifiers because that's what they are
> selling..............................
>
**There is a point which has not been mentioned, in this discussion. It
relates to the way global feedback loops distribute unwanted information to
the front end of an amplifier. A serious case, may involve the pickup of
spurious RF transmissions through speaker cables. Of course, shielded
speaker cables, may alleviate this problem, but zero loop NFB amplifiers may
eliminate it. Ditto, any spurious back EMF effects from dynamic
loudpseakers. Since a loudpspeaker can act as a generator, those signals may
traverse the NFB line, thus causing rather unpleasant effects.
Todd Krieger
ya...@ieee.org wrote:
> Hello Group,
>
> I have heard for a very long time the pooh-poohing of negative
> feedback in audio circuits (amps, preamps, etc.). Back in the
> 70s (I believe), it was linked to excessive TIM (transient
intermodulation
> distortion).
>
> I have never really been able to get from "here" to "there" on this
> issue, "here" being the level-headed, unassuming engineer that likes
> to see everything demonstrated analytically, "there" being the
> distortion that is purported to exist. Can anyone bridge this gap
> with absolutely no hand-waving?
> % Randy Yates
Transient intermodulation distortion (TIM) is a phenomenon which is
basically excessive "spikes" which occur at the start and end of a
transient signal. (It has no relation to classical intermodulation
distortion, which is how well an amp reproduces two sine wave signals of
closely spaced frequencies without adding extra similarly-spaced
"signals" above and below those two signals.) The most common
occurrence of TIM in a real-world system is when a signal through an amp
stops playing, the speaker cone will not stop on its own, but will try
to keep moving in a certain direction. The output or global feedback
loop will see the back EMF from the cone movement as an error signal,
relative to the input, and the differential will then make the amp send
an inverted ("negative" to the back EMF) output signal to oppose the
cone movement. (The faster the amp can stop the movement of the speaker
cone, the higher the amp's "damping factor.") If the cone movement is
excessive, the amp could overcompensate, and send an audible "spike" in
the effort to stop the movement of the cone. (It's not audible as a
spike per se, but the effect is transient noise artifacts added to the
music signal.) TIM can be shown in oscilloscope traces (time domain)
tied to an amp's output to a loudspeaker, but to the best of my
knowledge (someone correct me if I'm wrong), there is no known method of
actually measuring or quantifying TIM.
TIM is also dependent on the interface between a specific amp and
specific loudspeaker. One amplifier may have TIM problems driving
speaker A, and may be almost free of TIM driving speaker B. Yet a
different amp may have a TIM problem on speaker B, but not speaker A.
(Other amps may have a TIM problem with both speakers, or with neither
speaker.) It is not easy to predict how an amp will interface with a
specific complex impedance of a loudspeaker. Generally, TIM is most
likely to occur with speakers with large or massive diaphragms, since it
often requires more energy to stop the movement. And with amps with a
lot of negative feedback.
If an amp has less negative feedback, it will put less inverted signal
on the speaker to stop cone movment. With less feedback, the chances of
excessive correction signal go down, hence TIM goes down. Amps with no
feedback at all are often close to free of TIM, driving any speaker of
competent design. (TIM can exist for other reasons, such as momentary
oscillation if the amp isn't stable driving complex loads.)
Since line-level signals almost never encounter electro-mechanical
devices (other than maybe headphones) or complex impedances, TIM is
usually not a factor with signals at the line level.
Todd Krieger
pfjel...@my-deja.com
into clipping it is very difficult to tell one amplifier from another.
However one piece of information was missing as far as I could find and
that is the actual listening levels used during the test. Since low
efficiency speakers were used (82db/1m) and some medium power, <= 100W,
amplifiers were used. The average sound level must have been fairly low.
So I would like to know:
1/ Does the listening level used during the test equal or exceed
preferred listening level for concentrated listing?
2/ If not, would it not be better to make the test at a
realistic domestic sound level?
3/ Shouldn't testreports include some information on the sound levels
achievable with a given amplifier/speaker combination before it starts
to sound really bad.
//PF
Todd Krieger
> >no recogised measurement to quantify the distortion resulting from
> the use
> >of loop NFB.
>
> That's because it is imaginary, and its hard to quantify the
> imaginary.
>
imaginary?
TIM can be illustrated rather clearly, with the "right" amp-speaker
combination, monitoring the amp's output with a scope. Preferably a
digital storage scope, where you can compare traces taken at different
times. The best way to illustrate TIM is to look at the same transient
passages using an amp susceptible to TIM versus one not so susceptible,
driving the same loudspeaker. (The hard part is matching the output
levels.) You will see extra "spikes" (or "fuzz") on the output signal
of the high-TIM amp.
If I recall, an issue of Audio Magazine from the late '70's or early
'80's shows such comparative illustrations. In fact, I also recall
Onkyo (? Could have been Harman/Kardon) having such illustrations in
their sales brochures.
Todd Krieger
robert_...@my-deja.com
a...@borealis.com wrote:
> robert_...@my-deja.com writes:
>
> >When cheap op-amps first came on the market in the early 1970s, I
made a
> > flat amplifer stage from a 741. This op-amp has poor open loop
> >distortion performance. As I recall the loop had 70 db of negative
> >feedback. I expected the high negative feedback would eliminate all
> >distortion. I was very disapointed when I (and several others)
listened
> >to.
>
> Check out the open loop gain, it's flat to about 1Hz! How much
> negative feedback (if any) you think you had at 10kHz? :-)
If you have a circuit analysis program, click on a 741. The open loop
gain will be: 80 dB at 100 Hz, 60 dB at 1kHz, and 40 dB at 10kHz.This
will give feedback factors of: 50dB, 30dB and 10 dB. Clearly not good
enough.
A more modern op-amp such as the SE5533 will have 96 dB gain at 100Hz,
80 dB at 1kHz, and 60 dB at 10kHz.
In my opinion 40 dB of feedback is necessary to linearize most op-amps.
So, the SE5533 is adaquate at all frequenies. Just to be safe, I only
design stages for 20 dB gain max. The ear test says: it sounds good.
Thank you for your input. In your opinion, how much feedback is
necessary to linearize a low cost op-amp?
Bob S.
robert_...@my-deja.com
DPi...@world.std.com (Richard D Pierce) wrote:
> In article <7ihjd7$9iq$1...@nnrp1.deja.com>,
<robert_...@my-deja.com> wrote:
made a
> >flat amplifer stage from a 741. This op-amp has poor open loop
> >distortion performance. As I recall the loop had 70 db of negative
> >feedback.
>
>
> >distortion. I was very disapointed when I (and several others)
listened
> >to.
>
> or so, AT DC! However, the first pole in the open loop response
> was on the order of 1-2 Hz, and THAT'S one of the many obvious
> factors that was ignored at the time.
>
> So, by the time you're getting up in the realm of high
> frequencies, your 70 dB of feedback has degenerated to well less
> than that. For example, how high in frequency do you think you
> have to go before your feedback dropped from 70 dB to, say, 30
> dB? Would you be shocked to learn that it was well below 20 kHz?
> How about 1 kHz.
>
> And this is but ONE aspect about the simplifying assumptions
> made about feedback which is simply wrong.
>
> --
> | Dick Pierce |
> | Professional Audio Development |
> | 1-781/826-4953 Voice and FAX |
> | DPi...@world.std.com |
>
op-amps with "heavy" feedback don't perform well.
To reduce power consumption, the push-pull output stage is biased to
near cutoff. This causes a type of distortion called crossover
distortion. When the signal goes through the zero axis, one transistor
cuts off before the other transistor turns on! There is a moment when
both output transistors are turned off. During this moment feedback
can't do much to correct distortion.
You may have experienced that if the op-amp is not properly laid out on
the PC board, a ground loop can cause hum in spite of the heavy
feedback. This is hum that feedback can not correct. If the
power supply has no hum, you won't be aware of the ground loop problem.
If the supply voltage pulls down (under load), the change will
modulate the audio output, thru the ground loop.
Bob S.
"Things should be as simple as possible, but never simpler."
Stewart Pinkerton
Hmmmmmmmmm. Actually, it's *way* simpler than all that. If you design
the input stage such that it can handle twice the input voltage
required to clip the output stage, over the full operating frequency
range, then you will not suffer from slewing-induced distortion, aka
SID or TIM.
Control and Instrumentation engineers knew this in the fifties......
Mark Rehorst
> on the speaker to stop cone movment. With less feedback, the chances of
> excessive correction signal go down, hence TIM goes down.
What about the still moving speaker cone which is producing sound?
What do you call that distortion?
MR
Todd Krieger
Todd Krieger
Sent via Deja.com http://www.deja.com/
Randy Yates
>
> > If an amp has less negative feedback, it will put less inverted signal
> > on the speaker to stop cone movment. With less feedback, the chances of
> > excessive correction signal go down, hence TIM goes down.
>
> What about the still moving speaker cone which is producing sound?
> What do you call that distortion?
Isn't TIM a phenomenom that is seen in preamps as well as power amps?
Hence the output can be a simple load resistor and we can do
away with the electromechanical complications. They seem to be
a whole 'nuther question.
--
% Randy Yates % "How's life on earth?
%% DIGITAL SOUND LABS % ... What is it worth?"
%%% Digital Audio Sig. Proc. % 'Mission (A World Record)',
%%%% <ya...@ieee.org> % *A New World Record*, ELO
http://www.shadow.net/~yates
Francis Vaughan
Umm, there seems to be a lot of really rather misguided
stuff being bandied about about the nature of feedback
and TIM. Much of it one would recognise as almost being
part of the background noise of audio conversations of the
last decade odd. Much of it is simply wrong.
First up, it is important to realise where TIM comes from.
It is caused by the output stage of a feedback amp reaching
its slew rate limit. When this happens the amp can no
longer operate as a feedback amplifier, since the output
stage is (for the period it is in rate limit) not able to
follow the voltage gain stage; the feeback loop is essentially
open. Typically the amp will appear to lock-up in some sense,
producing arbitarily bad distortion.
This has essentially nothing whatsoever to do with the back
EMF of the speaker load. A speaker is not able to create the
"spikes" that are so often described. A conventional speaker
is a bandwidth limited electro-mechanical device. We so often
hear descriptions whereby a heavy bass driver is somehow suddenly
overshooting and creating high slew rate spikes. It cannot be.
The back EMF is defined by the same laws that govern its acoustic
properties in the radiative direction. It is a nice big spring/mass
system with defined and rather mundane charateristics.
If the input signal is bandwidth limited (it is) and the output
transducer is bandwidth limited (similarly) we will find it very hard
to make a case for any mechanism for back EMF _spikes_, no matter
what topology amplifier is in use.
However, this does not disclaim TIM. Cherry published 25 years
ago on this subject. His nested differential feedback loop
design regime shows how to construct an amplifier with essntially
arbitary feedback that is guarenteed never to suffer from TIM.
Better still, he shows how multiple loops allow you to provide a
desired feedback factor out to whatever frequency you desire.
With guarenteed stability.
The issue of feedback ratio vs frequency is really a crucial
issue is all these discussions about the realtive merit of
feedback. As Douglas Self is often at pains to point out, there is
no such thing as a single quotable number for feedback ratio. Indeed
I would go as far as suggesting that any designer who quotes you such
a number for his amp is simply showing off his ignorance, and by
implication tarnishing any regard I would have for his designs.
Maybe he got it right anyway, but is if so, by accident,
not design.
We have heard of a double blind trial of two amps, one with, one
without feedback. This is really a worry, simply because we know
so little about the underlying designs. It is not possible to
draw any conclusions the merits of the designs, in particular it
is entierly possible that the feedback design was incompetantly
done. Minimally we would want to know what the feedback ratios
were at various frequencies. If the designer is unable to furnish
these (perhaps having never calculated them) we would probably
be justifed in having grave doubts of the amplifier's credentials.
I also remain quite worred about the claims of essentially
identical overall design, apart from the application of global
feedback. Topologically global feedback looks a trivial change.
It isn't. Just about every parameter of the design requires
rework, and may require very substantial changes in realisation.
Comparisons assuming that it is a one parameter alteration are
flawed from the outset.
Feedback is a complicating factor in amplifier design, it is
not enough to simply regard it as a topological change. It is
(sadly and manifestly) much easier to create an amplifier with
benign (if undistinguished) distortion charatersitics with little or
no global feedback.
It is easy to produce a feedback design that has terrific
but trivial (steady state, single tone etc) measured performance that has
quite awful real world charateristics.
It is entierly reasonable with a little more care and understanding
to create an amplifier with absolutly immaculate real world distortion
charateristics through the application of appropriate global feedback.
The methodologies are not new, and not rocket science. They are however
often ignored.
Francis Vaughan
Trevor Wilson
Francis Vaughan <fra...@cs.adelaide.edu.au> wrote in message
<SNIP>
>
> We have heard of a double blind trial of two amps, one with, one
> without feedback. This is really a worry, simply because we know
> so little about the underlying designs. It is not possible to
> draw any conclusions the merits of the designs, in particular it
> is entierly possible that the feedback design was incompetantly
> done. Minimally we would want to know what the feedback ratios
> were at various frequencies. If the designer is unable to furnish
> these (perhaps having never calculated them) we would probably
> be justifed in having grave doubts of the amplifier's credentials.
> I also remain quite worred about the claims of essentially
> identical overall design, apart from the application of global
> feedback. Topologically global feedback looks a trivial change.
> It isn't. Just about every parameter of the design requires
> rework, and may require very substantial changes in realisation.
> Comparisons assuming that it is a one parameter alteration are
> flawed from the outset.
**Interesting and valid points. I am unable to answer these points, at this
time. I will attempt to furnish more information, as it becomes available. I
can say, that the level of global NFB was not large.
>
> Feedback is a complicating factor in amplifier design, it is
> not enough to simply regard it as a topological change. It is
> (sadly and manifestly) much easier to create an amplifier with
> benign (if undistinguished) distortion charatersitics with little or
> no global feedback.
**I disagree. Building a low distortion, low output impedance, zero loop NFB
amplifier, is not a trivial exercise. Generally more attention must be paid
to various parameters, such as:
Power supply, device matching, number of output devices (pretty much equates
to output Z) and a number of other factors. These items can be *almost*
ignored with high levels of loop NFB.
>
> It is easy to produce a feedback design that has terrific
> but trivial (steady state, single tone etc) measured performance that has
> quite awful real world charateristics.
**Absolutely.
>
> It is entierly reasonable with a little more care and understanding
> to create an amplifier with absolutly immaculate real world distortion
> charateristics through the application of appropriate global feedback.
> The methodologies are not new, and not rocket science. They are however
> often ignored.
>
**Agreed. There are many very fine amplifiers utilising high levels of loop
NFB. To my ears, however, the best designs use no loop NFB.
--
Cheers,
Trevor Wilson
http://www.hutch.com.au/~rage
> Francis Vaughan
>
>
>
>
>
Johnny
(Stewart Pinkerton) wrote:
>Jan Didden <did...@wxs.nl> writes:
>
>>The late Peter Baxandall did some testing on the relative levels of
>>harmonic distortion in an amp while moving from 0dB to 40dB feedback.
>>Interesting result: 2nd harmonic started to decrease with increasing
>>feedback (which what we would expect), but all the other (higher order)
>>harmonics started to INCREASE! With further increase in feedback, all
>>harmonics (in order of ... eh ... order) eventually also started to
>>decrease, but the higher the harmonic number, the more feedback was
>>needed to start to decrease it. Looking at the curves, there are regions
>>where many higher order components were higher than with either more or
>>less feedback.
>>Now, whether this is audible, I don't know, but it certainly was clearly
>>measurable. Also, this was measured on a particular single amplifier
>>stage, and I'm not sure how this applies to other configurations.
>>However, some conclusions stands out:
>>1 - More feedback does not automatically mean less overall distortion;
>>2 - The distortion level has a great influence on the distortion spectrum
>>as well.
>>
>>Any comment, anyone?
>>
>>Jan Didden
>>(Engage brain before operating mouth)
>
>OK, I'll try, just this once!
>
>Baxendall's results were geared to the technology available at the
>time (more than forty years ago!), with limited gain-bandwidth
>product. Basically, negative feedback will reduce harmonic distortion
>so long as the open loop gain exceeds the closed loop gain in vector
>terms (i.e. including phase angle). Above this frequency (which could
>be just a few hundred hertz in many commercial designs!) the feedback
>is no longer negative and distortion will therefore increase. In 1999,
>An *open loop* bandwidth of 50kHz is easily achievable at very high
>gain, allowing vanishingly low THD and IMD across the whole audio
>band.
An open loop amplifier with 50Khz open loop bandwidth can't have very
much feedback. Modern audio amplifiers are usually designed with a
unity-gain point of around 600kHz. This equates to 30dB of negative
feedback at 20kHz and 56dB at 1kHz, and 90dB at 20Hz.
Assuming a unity loop gain bandwidth of 600kHz, your amplifier with
50kHz open loop bandwidth can have no more than 22dB of negative
feedback at any frequency in the audio bandwidth.
The effectiveness of feedback has nothing to do with the open loop
bandwidth. The open loop transfer function does determine the
stability of the amp, but it only matters at the frequency where the
open loop gain equals the closed loop gain. At all other points
within the closed loop bandwidth, the 'vector-difference' as you you
describe it, is unimportant. As long as the amp is stable (by Nyquist
theorem as above) only the amount of NFB matters.
Could it be that you were thinking of the vector sum of the open loop
gain and the feedback transfer function (usually a constant). This
vector sum is what many engineers refer to as the 'loop-gain'.
Johnny.
Jan Didden
Stewart Pinkerton wrote:
> Jan Didden <did...@wxs.nl> writes:
>
> >I realize that the forward part of this thread degenerates (don't take
> >that personal) into an argument about DBT etc etc. May I come back to
> >feedback?
>
> Yuk Yuk, very good!
>
> It's a very simple concept, often conveniently forgotten by those who
> are promoting 'zero feedback' amplifiers because that's what they are
> selling..............................
>
> --
>
> Stewart Pinkerton | Music is art, audio is engineering
Stewart, I beg to disagree.
As I said, Baxandall analysed a single amp stage (he used a FET stage). I
would venture that a common drain source 20 yrs ago did not differ too much
(technology wise) from one today. But the point was another one. As an
axample, if you went from 0dB feedback to 20 dB feedback, the 2nd harmonic
dcreased from -40dB to -56dB (good), but the 3rd harmonic went UP from say
-65dB to -50dB (don't quote me on the numbers, but you get the idea). That
means that by increasing feedback you INCREASE some harmonics, not just
relative but also in absulute terms! This stuff can readily be measured; as I
get the impression on these forums that audibily is often more sensitive than
measurements, this could be one scientific basis of audible differences. Your
premise that QUOTE basically, negative feedback will reduce harmonic
distortion so long as the open loop gain exceeds the closed loop gain in
vector terms (i.e. including phase angle) UNQUOTE need to be heavily caveated.
Jan Didden
[Minds are like parachutes - they work best when open]
Jan Didden
Todd Krieger wrote:
Sorry, not true. TIM is created when an amplifier's internal circuits are
slewing so fast (hopelessly trying to follow the signal) that the causal
relationship between input and output is lost (the amp is in slew rate
limiting condition). That means that the feedback can supply all the error
signals it wants, but it will have no effect on the output signal. THAT
means that the amp is running completely open loop, and running open loop
(usually) generates large amounts on IM distortion. The slew limiting occurs
because of fast signal risetimes (transients) hence TransientIM.
Jan Didden
>
>
> TIM is also dependent on the interface between a specific amp and
> specific loudspeaker. One amplifier may have TIM problems driving
> speaker A, and may be almost free of TIM driving speaker B. Yet a
> different amp may have a TIM problem on speaker B, but not speaker A.
> (Other amps may have a TIM problem with both speakers, or with neither
> speaker.) It is not easy to predict how an amp will interface with a
> specific complex impedance of a loudspeaker. Generally, TIM is most
> likely to occur with speakers with large or massive diaphragms, since it
> often requires more energy to stop the movement. And with amps with a
> lot of negative feedback.
>
> If an amp has less negative feedback, it will put less inverted signal
> on the speaker to stop cone movment. With less feedback, the chances of
> excessive correction signal go down, hence TIM goes down. Amps with no
> feedback at all are often close to free of TIM, driving any speaker of
> competent design. (TIM can exist for other reasons, such as momentary
> oscillation if the amp isn't stable driving complex loads.)
>
> Since line-level signals almost never encounter electro-mechanical
> devices (other than maybe headphones) or complex impedances, TIM is
> usually not a factor with signals at the line level.
>
> Todd Krieger
>
> --== Sent via Deja.com http://www.deja.com/ ==--
> ---Share what you know. Learn what you don't.---
Jan Didden
Todd Krieger wrote:
Yeah, I remember those pictures. They were 2 scope shots, one with, the
other without TIM. Actually, if you looked carefully enough, you could see
that they actually were the SAME pictures, but with the scope sensitivity
settings differing. That was already an old scam at those times.
Todd Krieger
Jan Didden <did...@wxs.nl> wrote:
>
>
> Todd Krieger wrote:
> > If I recall, an issue of Audio Magazine from the late '70's or early
> > '80's shows such comparative illustrations. In fact, I also recall
> > Onkyo (? Could have been Harman/Kardon) having such illustrations
in
> > their sales brochures.
>
> Yeah, I remember those pictures. They were 2 scope shots, one with,
the
> other without TIM. Actually, if you looked carefully enough, you could
see
> that they actually were the SAME pictures, but with the scope
sensitivity
> settings differing. That was already an old scam at those times.
>
How would changing the scope sensitivity cause extra spikes to show up
in a waveform?? The waveform would be taller or shorter vertically, but
thats it.
Todd Krieger
Sent via Deja.com http://www.deja.com/
Karl_Uppiano
engineering show up (well, except for the tubes vs. transistors and maybe
the analog vs. digital discussions).
Quite simply, you can't build a stable analog system without negative
feedback. Period. You can fiddle around with how much feedback to use, and
how big to make the loop, but if your amplifier has a resistor in the
emitter (or cathode) circuit, then it has negative feedback! And that's just
fine. It gets bad when you take a bad design and try to fix it with
feedback.
A lot of the feedback discussion derives from some of the early solid state
designs, that were poorly implemented (we were still learning back then),
and newer designs that deliberately attempt to use no feedback. Most of the
no-feedback designs sound different because of their instability. Some
people even prefer the distortion they get without feedback. Field-effect
devices (tubes and FETs) produce a lot of "nice" even harmonic distortion
when they're not controlled by negative feedback. Whether this is good or
bad depends on your religion, but most engineers would opt for stability and
accuracy, and add "effects" as a separate function.
Arny Krüger <ar...@flash.net> wrote in message
news:9db13.237$nW....@news.rdc1.mi.home.com...
>
> Randy Yates wrote in message <3741FAFA...@shadow.net>...
> >Hello Group,
> >
> >I have heard for a very long time the pooh-poohing of negative
> >feedback in audio circuits (amps, preamps, etc.). Back in the
> >70s (I believe), it was linked to excessive TIM (transient
> intermodulation
> >distortion).
> >
> >I have never really been able to get from "here" to "there" on this
> >issue, "here" being the level-headed, unassuming engineer that likes
> >to see everything demonstrated analytically, "there" being the
> >distortion that is purported to exist. Can anyone bridge this gap
> >with absolutely no hand-waving?
>
> It can't be done. Fear of negative feedback is a religion, not
> science or technology.
>
>
Karl_Uppiano
one. I think it's far too easy to develop a religion about one aspect of
electronics design, whether it's feedback, tubes or whatever. The engineer
has to juggle many conflicting parameters to come up with a good design:
Cost, size, power, distortion, frequency response, feedback, component
suitability and availability, layout, heat removal and so on. It just
doesn't make sense to pick any one item and say it's "bad". Feedback is just
one tool in the engineer's toolkit, and used appropriately and correctly, it
is extremely useful. Used inappropriately and carelessly, it can be
extremely nasty. But I wouldn't throw the baby out with the bathwater.
Trevor Wilson <ra...@hutch.com.au> wrote in message
news:o_233.1057$PN5....@newsfeeds.bigpond.com...
>
> Stewart Pinkerton <pat...@popmail.dircon.co.uk> wrote in message >
> > OK, I'll try, just this once!
> >
> > Baxendall's results were geared to the technology available at the
> > time (more than forty years ago!), with limited gain-bandwidth
> > product. Basically, negative feedback will reduce harmonic distortion
> > so long as the open loop gain exceeds the closed loop gain in vector
> > terms (i.e. including phase angle). Above this frequency (which could
> > be just a few hundred hertz in many commercial designs!) the feedback
> > is no longer negative and distortion will therefore increase. In 1999,
> > An *open loop* bandwidth of 50kHz is easily achievable at very high
> > gain, allowing vanishingly low THD and IMD across the whole audio
> > band.
>
> **Actually, this kind of open loop bandwidth has been available for
several
> decades.
>
> >
> > It's a very simple concept, often conveniently forgotten by those who
> > are promoting 'zero feedback' amplifiers because that's what they are
> > selling..............................
> >
>
> **There is a point which has not been mentioned, in this discussion. It
> relates to the way global feedback loops distribute unwanted information
to
> the front end of an amplifier. A serious case, may involve the pickup of
> spurious RF transmissions through speaker cables. Of course, shielded
> speaker cables, may alleviate this problem, but zero loop NFB amplifiers
may
> eliminate it. Ditto, any spurious back EMF effects from dynamic
> loudpseakers. Since a loudpspeaker can act as a generator, those signals
may
> traverse the NFB line, thus causing rather unpleasant effects.
>
Jan Didden
Todd Krieger wrote:
> In article <374F0155...@wxs.nl>,
> Jan Didden <did...@wxs.nl> wrote:
> >
> >
> > Todd Krieger wrote:
>
> > > If I recall, an issue of Audio Magazine from the late '70's or early
> > > '80's shows such comparative illustrations. In fact, I also recall
> > > Onkyo (? Could have been Harman/Kardon) having such illustrations
> in
> > > their sales brochures.
> >
> > Yeah, I remember those pictures. They were 2 scope shots, one with,
> the
> > other without TIM. Actually, if you looked carefully enough, you could
> see
> > that they actually were the SAME pictures, but with the scope
> sensitivity
> > settings differing. That was already an old scam at those times.
> >
>
> How would changing the scope sensitivity cause extra spikes to show up
> in a waveform?? The waveform would be taller or shorter vertically, but
> thats it.
Yes, and that was exactly what was shown.
Jan Didden
Jan Didden
Karl_Uppiano wrote:
> As with anything, a well designed amp will work better than a badly designed
> one. I think it's far too easy to develop a religion about one aspect of
> electronics design, whether it's feedback, tubes or whatever. The engineer
> has to juggle many conflicting parameters to come up with a good design:
> Cost, size, power, distortion, frequency response, feedback, component
> suitability and availability, layout, heat removal and so on. It just
> doesn't make sense to pick any one item and say it's "bad". Feedback is just
> one tool in the engineer's toolkit, and used appropriately and correctly, it
> is extremely useful. Used inappropriately and carelessly, it can be
> extremely nasty. But I wouldn't throw the baby out with the bathwater.
No disagreement here.
Jan Didden
Karl_Uppiano wrote:
> Ooh, I love the negative feedback threads! Nowhere else does so much voodoo
> engineering show up (well, except for the tubes vs. transistors and maybe
> the analog vs. digital discussions).
>
> Quite simply, you can't build a stable analog system without negative
> feedback. Period. You can fiddle around with how much feedback to use, and
> how big to make the loop, but if your amplifier has a resistor in the
> emitter (or cathode) circuit, then it has negative feedback! And that's just
> fine. It gets bad when you take a bad design and try to fix it with
> feedback.
>
> A lot of the feedback discussion derives from some of the early solid state
> designs, that were poorly implemented (we were still learning back then),
> and newer designs that deliberately attempt to use no feedback. Most of the
> no-feedback designs sound different because of their instability.
I think you got your arguments 180 degrees out of phase. It's the negative
feedback that introduces instability, not the absence of feedback. In the light
of such an elementary error, I'm not sure what to think of your other arguments
(or is it just opinion?)
Jan Didden
Karl_Uppiano
tone control design in use on practically every system that have tone
controls. The tone is controlled by adjusting the amount of negative
feedback from a pair of frequency selective networks. The two networks
cancel each other out at the control's center position, resulting in flat
response. These circuits specifically require negative feedback to operate.
Of course, some people object to perceived coloration due to tone controls,
and bypass them, or buy equipment without them. But the design dates back
well into the vacuum tube era.
Karl_Uppiano
opposite! Negative feedback will always stabilize an amplifier. The problem
is, in a poor design, the designer neglects worst case signals and worst
case loads, and under those conditions, the feedback is insufficient, or
even of the wrong phase (in which case it's positive feedback) and you've
got an uncontrolled, unstable situation, perhaps even an oscillator! In the
oscillator scenario, the amp (or your speakers) will self-destruct almost
immediately.
Jan Didden <did...@wxs.nl> wrote in message
news:37513CDD...@wxs.nl...
Joe Berry
|Typically zero loop NFB designs eschew the use of
|output Zobels, simply because that are so stable that
|any combination of RLC on the output will cause no
|serious difficulties for the amplifier.
A good point, and I would also note that not just Zobels,
but also internal compensation components and even
whole gain stages can be omitted with a ZNF approach.
The absence of these components can produce sonic
benefits which are only secondarily due to the absence
of global NFB.
<jnb>
Trevor Wilson
Karl_Uppiano <Karl_U...@email.msn.com> wrote in message
news:OtyYXNoq#GA.113@cpmsnbbsa03...
> As with anything, a well designed amp will work better than a badly
designed
> one. I think it's far too easy to develop a religion about one aspect of
> electronics design, whether it's feedback, tubes or whatever. The engineer
> has to juggle many conflicting parameters to come up with a good design:
> Cost, size, power, distortion, frequency response, feedback, component
> suitability and availability, layout, heat removal and so on. It just
> doesn't make sense to pick any one item and say it's "bad". Feedback is
just
> one tool in the engineer's toolkit, and used appropriately and correctly,
it
> is extremely useful. Used inappropriately and carelessly, it can be
> extremely nasty. But I wouldn't throw the baby out with the bathwater.
>
**All good words. At the outset, I stated (to paraphrase), that I had heard
some excellent high loop NFB amplifiers, however, the FINEST amps, that I
have heard, have zero loop NFB. Building amps with zero loop NFB, will
always be more expensive than those using loop NFB to remove distortion (all
things being equal).
Trevor Wilson
> Ooh, I love the negative feedback threads! Nowhere else does so much
voodoo
> engineering show up (well, except for the tubes vs. transistors and maybe
> the analog vs. digital discussions).
>
> Quite simply, you can't build a stable analog system without negative
> feedback. Period. You can fiddle around with how much feedback to use, and
> how big to make the loop, but if your amplifier has a resistor in the
> emitter (or cathode) circuit, then it has negative feedback! And that's
just
> fine. It gets bad when you take a bad design and try to fix it with
> feedback.
**In this thread, a distinction was drawn early on that references were to
either LOOP NFB or the absence of loop NFB. All amplifiers contain NFB of
some kind. Some amplifier designers choose to use no loop NFB.
>
> A lot of the feedback discussion derives from some of the early solid
state
> designs, that were poorly implemented (we were still learning back then),
> and newer designs that deliberately attempt to use no feedback. Most of
the
> no-feedback designs sound different because of their instability. Some
> people even prefer the distortion they get without feedback. Field-effect
> devices (tubes and FETs) produce a lot of "nice" even harmonic distortion
> when they're not controlled by negative feedback. Whether this is good or
> bad depends on your religion, but most engineers would opt for stability
and
> accuracy, and add "effects" as a separate function.
>
**If we are talking zero loop NFB designs, vs. high loop NFB ones, then zero
loop NFB designs are inherently FAR more stable than those using large
amounts of loop NFB. Typically zero loop NFB designs eschew the use of
output Zobels, simply because that are so stable that any combination of RLC
on the output will cause no serious difficulties for the amplifier.
robert_...@my-deja.com
Jan Didden <did...@wxs.nl> wrote:
>
> Sorry, not true. TIM is created when an amplifier's internal circuits
are
> slewing so fast (hopelessly trying to follow the signal) that the
causal
> relationship between input and output is lost (the amp is in slew rate
> limiting condition). That means that the feedback can supply all the
error
> signals it wants, but it will have no effect on the output signal.
THAT
> means that the amp is running completely open loop, and running open
loop
> (usually) generates large amounts on IM distortion. The slew limiting
occurs
> because of fast signal risetimes (transients) hence TransientIM.
>
> Jan Didden
>
>
>
point out that during the time interval when the amplifier is slewing,
it is not acting as an amplifier. Signals applied to the input are not
passed through to the output (until after the amplifier finishes its
slew). It is incorrect to think of the amplifier as "distorting" during
this time period. It is only putting out an unpleasent impulse.
Intermodulation distortion means two or more signals are creating beats.
During the slew time there are no signals passing throught the
amplifier. It is locked up, until the slew finishes. While slewing,
amplifier is operating in the time domain, not the frequeny domain.
Thinking of this condition as a form of intermodulation distortion is
not correct.
Bob
Jan Didden
Karl_Uppiano wrote:
> Negative feedback, properly applied does not create instability! Quite the
> opposite! Negative feedback will always stabilize an amplifier.
Could you explain that? First time I ever heard this.
> The problem
> is, in a poor design, the designer neglects worst case signals and worst
> case loads, and under those conditions, the feedback is insufficient, or
> even of the wrong phase (in which case it's positive feedback)
In that case, the feedback is not 'insufficient', but too much. If you have an
amp that is instable because of feedback, cut the feedback loop and presto!
stability.
In fact, negative feedback ALWAYS becomes positive feedback if you go high
enough in frequency. The trick is to be sure that the loop gain becomes less
than one before that point, so that the instability cannot sustain oscillations.
What you call 'worst vase loads and worstr case signal' modify the loop gain
and/or phase shifts in ways not foreseen by the designer, and lead to
instability in an otherwise perfectly stable amp. Of course, instability does
not always lead to oscillations, but can cause 'only' peaking and oscillatory
busts as well. But, in any case, cut out the neg feedback and there is no more
mechanism to sustain instability. Not the other way around.
Jan Didden
> and you've
> got an uncontrolled, unstable situation, perhaps even an oscillator! In the
> oscillator scenario, the amp (or your speakers) will self-destruct almost
> immediately.
>
>
> Jan Didden <did...@wxs.nl> wrote in message
> news:37513CDD...@wxs.nl...
> >
> >
> > Karl_Uppiano wrote:
> >
> > > Ooh, I love the negative feedback threads! Nowhere else does so much
> voodoo
> > > engineering show up (well, except for the tubes vs. transistors and
> maybe
> > > the analog vs. digital discussions).
> > >
> > > Quite simply, you can't build a stable analog system without negative
> > > feedback. Period. You can fiddle around with how much feedback to use,
> and
> > > how big to make the loop, but if your amplifier has a resistor in the
> > > emitter (or cathode) circuit, then it has negative feedback! And that's
> just
> > > fine. It gets bad when you take a bad design and try to fix it with
> > > feedback.
> > >
> > > A lot of the feedback discussion derives from some of the early solid
> state
> > > designs, that were poorly implemented (we were still learning back
> then),
> > > and newer designs that deliberately attempt to use no feedback. Most of
> the
> > > no-feedback designs sound different because of their instability.
> >
> > I think you got your arguments 180 degrees out of phase. It's the negative
> > feedback that introduces instability, not the absence of feedback. In the
> light
> > of such an elementary error, I'm not sure what to think of your other
> arguments
> > (or is it just opinion?)
> >
> > Jan Didden
> >
> > > Some
> > > people even prefer the distortion they get without feedback.
> Field-effect
> > > devices (tubes and FETs) produce a lot of "nice" even harmonic
> distortion
> > > when they're not controlled by negative feedback. Whether this is good
> or
> > > bad depends on your religion, but most engineers would opt for stability
> and
> > > accuracy, and add "effects" as a separate function.
> > >
Jan Didden
robert_...@my-deja.com wrote:
> In article <374F008E...@wxs.nl>,
> Jan Didden <did...@wxs.nl> wrote:
> >
> > Sorry, not true. TIM is created when an amplifier's internal circuits
> are
> > slewing so fast (hopelessly trying to follow the signal) that the
> causal
> > relationship between input and output is lost (the amp is in slew rate
> > limiting condition). That means that the feedback can supply all the
> error
> > signals it wants, but it will have no effect on the output signal.
> THAT
> > means that the amp is running completely open loop, and running open
> loop
> > (usually) generates large amounts on IM distortion. The slew limiting
> occurs
> > because of fast signal risetimes (transients) hence TransientIM.
> >
> > Jan Didden
> >
> >
> >
> First, I agree that what you said was essentially correct. I want to
> point out that during the time interval when the amplifier is slewing,
> it is not acting as an amplifier. Signals applied to the input are not
> passed through to the output (until after the amplifier finishes its
> slew). It is incorrect to think of the amplifier as "distorting" during
> this time period. It is only putting out an unpleasent impulse
Maybe I should have said 'changes to the input signal are not passed to the
output'. There obviously is a signal at the input, and the output IS
reacting on it.
> .
>
> Intermodulation distortion means two or more signals are creating beats.
> During the slew time there are no signals passing throught the
> amplifier. It is locked up, until the slew finishes. While slewing,
> amplifier is operating in the time domain, not the frequeny domain.
Agreed.
Jan Didden
Trevor Wilson wrote:
> Karl_Uppiano <Karl_U...@email.msn.com> wrote in message
I wouldn't say it is more expensive per se, but it does require a very detailed
knowledge of analog design and much more carefull design work.
Jan Didden
Todd Krieger
fra...@cs.adelaide.edu.au (Francis Vaughan) wrote:
>
>
> Umm, there seems to be a lot of really rather misguided
> stuff being bandied about about the nature of feedback
> and TIM. Much of it one would recognise as almost being
> part of the background noise of audio conversations of the
> last decade odd. Much of it is simply wrong.
>
> First up, it is important to realise where TIM comes from.
> It is caused by the output stage of a feedback amp reaching
> its slew rate limit. When this happens the amp can no
> longer operate as a feedback amplifier, since the output
> stage is (for the period it is in rate limit) not able to
> follow the voltage gain stage; the feeback loop is essentially
> open. Typically the amp will appear to lock-up in some sense,
> producing arbitarily bad distortion.
>
> This has essentially nothing whatsoever to do with the back
> EMF of the speaker load. A speaker is not able to create the
> "spikes" that are so often described.
The speaker itself does not "create" the spikes. The amplifier creates
the spikes because it's too slow to correct the speaker's back EMF. The
spike is a result of the correction signal from the amp "countering" the
EMF of the speaker *after* the EMF subsides. The slower the amp's slew
rate, the more time it takes for the amp to generate the correction
signal, and the worse and more delayed the spike.
> A conventional speaker
> is a bandwidth limited electro-mechanical device. We so often
> hear descriptions whereby a heavy bass driver is somehow suddenly
> overshooting and creating high slew rate spikes.
It's not the speaker, once again. The amp (due to insufficient slew
rate) does not react quickly enough to the error signal at the output.
(It's impossible to design a speaker which doesn't generate EMF back to
the amplifier.) Think of the spike as the signal which should have
countered the back EMF, but came too late.
This problem exists with any electromechanical system (such as a servo
control, for example) which involves feedback and a finite maximum
reaction time. Not just amps with speakers.
> It cannot be.
> The back EMF is defined by the same laws that govern its acoustic
> properties in the radiative direction. It is a nice big spring/mass
> system with defined and rather mundane charateristics.
>
I wish this was the case- it would be much easier to design amplifiers.
Back EMF has very little to do with a speaker's acoustic properties,
other than the fact the generator (voice coil in a magnetic field) is
coupled to the diaphragm.
> If the input signal is bandwidth limited (it is) and the output
> transducer is bandwidth limited (similarly) we will find it very hard
> to make a case for any mechanism for back EMF _spikes_, no matter
> what topology amplifier is in use.
>
Now you do have a point here, but if the slew rate is limited, so is the
bandwidth of the *output*. The feedback loop would have no problem
tracking with a slower output if there were no external signals acting
upon the feedback loop. The problem is when an external signal (the
speaker's back EMF) introduces error signals at the output, and hence
the feedback loop. It becomes a moving target. Since an EMF cannot be
countered instantaneously, the limited slew rate of an amp becomes a
liability.
> However, this does not disclaim TIM. Cherry published 25 years
> ago on this subject. His nested differential feedback loop
> design regime shows how to construct an amplifier with essntially
> arbitary feedback that is guarenteed never to suffer from TIM.
> Better still, he shows how multiple loops allow you to provide a
> desired feedback factor out to whatever frequency you desire.
> With guarenteed stability.
>
I remember there used to be an amp manufacturer which used to advertise
a "nested feedback" design. Can someone recall who it was?
> The issue of feedback ratio vs frequency is really a crucial
> issue is all these discussions about the realtive merit of
> feedback. As Douglas Self is often at pains to point out, there is
> no such thing as a single quotable number for feedback ratio. Indeed
> I would go as far as suggesting that any designer who quotes you such
> a number for his amp is simply showing off his ignorance, and by
> implication tarnishing any regard I would have for his designs.
> Maybe he got it right anyway, but is if so, by accident,
> not design.
>
> We have heard of a double blind trial of two amps, one with, one
> without feedback. This is really a worry, simply because we know
> so little about the underlying designs. It is not possible to
> draw any conclusions the merits of the designs, in particular it
> is entierly possible that the feedback design was incompetantly
> done.
I personally think the feedback loop should be as short and well coupled
as possible. I've heard some great amps, some designed with more
feedback, others with less feedback. It's just one design parameter.
> Minimally we would want to know what the feedback ratios
> were at various frequencies. If the designer is unable to furnish
> these (perhaps having never calculated them) we would probably
> be justifed in having grave doubts of the amplifier's credentials.
> I also remain quite worred about the claims of essentially
> identical overall design, apart from the application of global
> feedback. Topologically global feedback looks a trivial change.
> It isn't. Just about every parameter of the design requires
> rework, and may require very substantial changes in realisation.
> Comparisons assuming that it is a one parameter alteration are
> flawed from the outset.
>
> Feedback is a complicating factor in amplifier design, it is
> not enough to simply regard it as a topological change. It is
> (sadly and manifestly) much easier to create an amplifier with
> benign (if undistinguished) distortion charatersitics with little or
> no global feedback.
>
> It is easy to produce a feedback design that has terrific
> but trivial (steady state, single tone etc) measured performance that
has
> quite awful real world charateristics.
>
> It is entierly reasonable with a little more care and understanding
> to create an amplifier with absolutly immaculate real world distortion
> charateristics through the application of appropriate global feedback.
> The methodologies are not new, and not rocket science. They are
however
> often ignored.
>
> Francis Vaughan
>
I'm afraid we still don't know enough to get a definitive opinion, we
can share ideas, but something tells me a designer needs to come up with
a breakthrough design (whether it involves feedback or not) to advance
the amp art. Amps designed today are not that much better than those
designed 20 years ago.
Todd Krieger
Ed Berger
> > First up, it is important to realise where TIM comes from.
> > It is caused by the output stage of a feedback amp reaching
> > its slew rate limit.
Call me confused, I thought stages before the output stage slew
rate limit, or lack the drive current for their sucessive stages,
such as a driver stage could slew rate limit into its output stage,
or a secondary voltage gain stage might slew rate limit into the driver
stage,
or the input voltage amplifier might slew rate limit into a secondary
voltage gain stage. If not a "slew limit" perhaps we could say a
discontinuity in the open loop transfer function.
> The speaker itself does not "create" the spikes. The amplifier creates
> the spikes because it's too slow to correct the speaker's back EMF.
Isn't "the back EMF" usually damped by internal losses in the loudspeaker?
Isn't "the back EMF" usually damped by the natural output impedence of
the amplifier output stage before global feedback? Is there really enough
"error" signal coming back to the global feedback port from "back EMF"
to be a real world problem? Which speakers have low "back EMF" and which
high "back EMF"? Or does the amplifier just have a problem with itself,
and would exhibit the same "spikes" without "back EMF" being present
such as driving a passive network with similar frequency dependent
impedences?
> (It's impossible to design a speaker which doesn't generate EMF back to
> the amplifier.)
What kind of back EMF is generated by full range electrostatic panels with
direct
drive high voltage amplifiers (instead of step up transformers) compared
to traditional dynamic loudspeakers and "regular" amplifiers?
What changes when audio step up transformers are used with electrostatic
panels?
There is no "coil/magnet generator" here, yet electrostatic speakers were
said
to benefit from "low-TIM" or "feedbackless" amplifiers.
If you were designing a new two-way dynamic loudspeaker, and the drivers
you were interested in were available in either 4ohm or 8ohm versions, all
else
supposedly being equal, which would you choose for reducing the "back EMF
related amplifier problems" your customers might have?
> Amps designed today are not that much better than those
> designed 20 years ago.
Yes, there were some very good amplifiers made 20 years ago,
both with and without "global feedback loops".
Richard D Pierce
Todd Krieger <tkri...@concentric.net> wrote:
>
>The speaker itself does not "create" the spikes. The amplifier creates
>the spikes because it's too slow to correct the speaker's back EMF. The
>spike is a result of the correction signal from the amp "countering" the
>EMF of the speaker *after* the EMF subsides. The slower the amp's slew
>rate, the more time it takes for the amp to generate the correction
>signal, and the worse and more delayed the spike.
Sorry, this is pure poppycock.
Precisely where do you think the largest amount of this
so-called "back-EMF" originates? AT what frequencies?
In fact, the single largest contributor to the "back-EMF" is the
fundamental mechanical resonance of the woofer, and
this is at frequencies WELL below 100 Hz. With some very rare
exception, that is the dominant source.
Now, you are trying to tell us that the "signals" generated by
this exceed the slew rate capacity of modern amplifiers?
Absolute, pure, nonsensical bunk.
First, unless you have discivered a new operating mode of
physics governing energy storage in electrical or mechanical
resonant systems, there is NOW way on earth the a loudspeaker
can generate more energy than it receives from the amplifier.
And for it to returneven a sizable portion of the energy to the
amplifier, it's dissipative losses, be they electrical or
mechnical, MUST be essentially non-existant.
Further, AT resonance, where the back-EMF is at its greates, the
so-called back EMF is exactly in-phase with the driving signal.
That's why the impedance is so high at that point. And the
pohase of the impedance is such that it is a purely resistive
load. So the load seen by the amplifier at resonance is
IDENTICAL to that of a pure resistor whose resistance equals
that of the woofer at that frequency.
>It's not the speaker, once again. The amp (due to insufficient slew
>rate) does not react quickly enough to the error signal at the output.
>(It's impossible to design a speaker which doesn't generate EMF back to
>the amplifier.)
Absolute nonsense. All one has to do is design a speaker whose
impedance is essentially resistive> There is NO back emf as you
claim. The KEF 104-2 is but one example of such.
And it precisely looking at the problem using the extremely
limited and flawed model of "back-EMF" that leads to this sort
of nonsensical analysis. A speaker is NO more than a resonant
system. Period. Instead of a voice coil moving in a magnetic
field, you can just as readily consider it a parallel LRC tank
cirvuit: the behavior as far as the amplifier is concerned is
IDENTICAL.
>Think of the spike as the signal
>which should have countered the back EMF, but came too late.
Give us a bloody break, please.
The slew rate of these signals is determined by the bandwidth of
the signals themselves. Since it is the woofer that is
contributing most to these issues. Now, at 30 Hz, you are
claiming that the so-called "back-emf" generated exceeds the
slew rate of the amplifier? If so, why are you using amplifiers
with bandwidths less than 50 Hz?
And please do not tell me about the "back-emf" of tweeters. Show
it to me.
>This problem exists with any electromechanical system (such as a servo
>control, for example) which involves feedback and a finite maximum
>reaction time. Not just amps with speakers.
And this is preciesly because in most servo-system, the
resonances being controlled are at the HIGH end of the systems
bandwidth, NOT the low end, as is the case with loudspeakers.
>> It cannot be.
>> The back EMF is defined by the same laws that govern its acoustic
>> properties in the radiative direction. It is a nice big spring/mass
>> system with defined and rather mundane charateristics.
>>
>I wish this was the case- it would be much easier to design amplifiers.
> Back EMF has very little to do with a speaker's acoustic properties,
>other than the fact the generator (voice coil in a magnetic field) is
>coupled to the diaphragm.
And as such, it behaves IDENTICALLY as a simple RLC parallel
resonant circuit whose resonance is at the BOTTOM of the
bandwidth, not the top. Slew rate IS NOT A FACTOR, despite your
claims to the contrary.
>> what topology amplifier is in use.
>>
>Now you do have a point here, but if the slew rate is limited, so is the
>bandwidth of the *output*. The feedback loop would have no problem
>tracking with a slower output if there were no external signals acting
>upon the feedback loop. The problem is when an external signal (the
>speaker's back EMF) introduces error signals at the output, and hence
>the feedback loop.
Come on, this is nonsense, and does not get any less so by
repetition.
In order for your claim to hold, you must show that the
resonance that reflect themselves as the speaker's impedance
MUST be high-Q, higyh-amplitude AND high-frequency. In fact, the
vast majority of speakers exhibit exactly the opposite. The
speaker's impedance is closer to resistive at high-frequencies
than at low.
There are NO "external signals" acting upon the feedback loop.
ALL signals originate in the amplifier and as long as that
amplifier is acting as a reasonable approximation of a voltage
source (which they do VERY well at the frequencies where your
onbjections hold, e.e. low frequencies), then ANY external
signals are going to see that amplifier as a dead short anyway.
There are no high-frequency signals as you claim, otherwise they
would show up as anomolously high and overly reactive
impedance at high frequency. And such simply is not the case.
Sorry, but the whole slew rate thing is a complete crock, and is
one of the things that sent Otala completely off the deep end.
--
| Dick Pierce |
| Professional Audio Development |
| 1-781/826-4953 Voice and FAX |
| DPi...@world.std.com |
DUNCLIF
>a "nested feedback" design. Can someone recall who it was?
Was that Spectrascan? The Accustat TNT 200 also used multiple feedback paths.
>I'm afraid we still don't know enough to get a definitive opinion, we
>can share ideas, but something tells me a designer needs to come up with
>a breakthrough design (whether it involves feedback or not) to advance
>the amp art. Amps designed today are not that much better than those
>designed 20 years ago.
>
>Todd Krieger
This is the reason for some of the hype. It is hard to interest the public
with claims of "Hey, we havent changed anything in 20 years". The easiest part
of getting an amp to market is the design. The hardest part is marketing a
design that hasen't changed in 20 years as something fersh and innovative.
With so many companies doing that, the few truely fresh and innovative ideas
may never make it to market.
Jan Didden
DUNCLIF wrote:
> >I remember there used to be an amp manufacturer which used to advertise
> >a "nested feedback" design. Can someone recall who it was?
Nested (differential) feedback was invented (and patented) by D. Edward Cherry from
Australia. I have never found any reference to a marketed implementation (which
doesn't mean there isn't one, of course).
Jan Didden
>
>
> Was that Spectrascan? The Accustat TNT 200 also used multiple feedback paths.
>
> >I'm afraid we still don't know enough to get a definitive opinion, we
> >can share ideas, but something tells me a designer needs to come up with
> >a breakthrough design (whether it involves feedback or not) to advance
> >the amp art. Amps designed today are not that much better than those
> >designed 20 years ago.
> >
> >Todd Krieger
>
Jan Didden
Ed Berger wrote:
> Todd Krieger wrote:
>
> > > First up, it is important to realise where TIM comes from.
> > > It is caused by the output stage of a feedback amp reaching
> > > its slew rate limit.
>
> Call me confused, I thought stages before the output stage slew
> rate limit, or lack the drive current for their sucessive stages,
> such as a driver stage could slew rate limit into its output stage,
> or a secondary voltage gain stage might slew rate limit into the driver
> stage,
> or the input voltage amplifier might slew rate limit into a secondary
> voltage gain stage. If not a "slew limit" perhaps we could say a
> discontinuity in the open loop transfer function.
>
> > The speaker itself does not "create" the spikes. The amplifier creates
> > the spikes because it's too slow to correct the speaker's back EMF.
The speaker's back EMF is caused by the movement of the coil in the magnetic
field. Are you suggesting that somehow some speakers (a mechanical system) move
faster than the electronics of the amp can react? Can you explain this a bit
better please?
Jan Didden
>
>
> Isn't "the back EMF" usually damped by internal losses in the loudspeaker?
> Isn't "the back EMF" usually damped by the natural output impedence of
> the amplifier output stage before global feedback? Is there really enough
> "error" signal coming back to the global feedback port from "back EMF"
> to be a real world problem? Which speakers have low "back EMF" and which
> high "back EMF"? Or does the amplifier just have a problem with itself,
> and would exhibit the same "spikes" without "back EMF" being present
> such as driving a passive network with similar frequency dependent
> impedences?
>
> > (It's impossible to design a speaker which doesn't generate EMF back to
> > the amplifier.)
>
> What kind of back EMF is generated by full range electrostatic panels with
> direct
> drive high voltage amplifiers (instead of step up transformers) compared
> to traditional dynamic loudspeakers and "regular" amplifiers?
>
> What changes when audio step up transformers are used with electrostatic
> panels?
>
> There is no "coil/magnet generator" here, yet electrostatic speakers were
> said
> to benefit from "low-TIM" or "feedbackless" amplifiers.
>
> If you were designing a new two-way dynamic loudspeaker, and the drivers
> you were interested in were available in either 4ohm or 8ohm versions, all
> else
> supposedly being equal, which would you choose for reducing the "back EMF
> related amplifier problems" your customers might have?
>
> > Amps designed today are not that much better than those
> > designed 20 years ago.
>
send me no spam
Spectral!) and
they advertised the nested feedback as a feature key to the product's very low
distortion across the audio bandwidth
In article <3756C4B7...@wxs.nl>, did...@wxs.nl says...
Francis Vaughan
Jan Didden <did...@wxs.nl> writes:
|> Nested (differential) feedback was invented (and patented)
|> by D. Edward Cherry from Australia. I have never found any reference
|> to a marketed implementation (which doesn't mean there isn't one, of course).
Cherry tried for a while to interest Australian manufacturers in the NDFL
technology, before finally selling the patent rights to one of the big
Japanese conglomerates (Matushita I think.) Since he was (and still is)
a lecturer at Monash University in Melbourne Australia the money came
to the University. I have heard that it acted as a very welcome cash
cow for some years. However for unknown reasons the technology was
never marketed. There were a few sporadic implementations and licensed
products but very very few. The patents have now all lapsed.
Cherry's contribution was not the "invention" of NDFL, but rather a fully
worked design regime for the generation of the design. From a simple
basis of knowing the charateristics of the output stage he provides
a methodology that allows a designer to very simply calculate the
location of all the poles in the NDLF circuit. A dominant pole on the
output ensures the output stage adheres to the design, and the poles
located on each feedback look ensure unconditional stability and
provide a mechanism by which the full feedback ratio can be preserved
to significantly higher frequencies, thus providing for an ampilfier which
does not suffer from the conventional problem of fast dropping feedback
with increasing frequency.
There does seem to be some belief that with the modern output drivers
feedback can be preserved enough without resorting to NDFL.
Certainly Self seems to put forward this view. However it is a
close run thing. Self is, I think, too quick to dismiss the NDFL
regime as "too complex"; an opinion I find hard to understand. However
the NDFL regime does require non-inverting gain stages, which is unusual
in conventional amplifer design. Thus it is not possible to take
an existing amp design and retro-fit an NDFL topology over it. Input,
driver and output stages are preserved, but the voltage gain stage
needs total redesign. Of course this is an over-simplification, these
stages are not orthogonal, a new gain stage may need changes to
input and driver stages. Also the design really does rather like to
have some real gain in the driver/output stage, so MOSFETs are hard to
use.
Overall however the NDFL regime must represent one of the most undersold
contributions to amplifier design.
I would venture to suggest that the combination of the design regimes
proposed by Self, plus the addition of NDFL would create an amplifier
that truely met Self's "blameless" tag under all circumstances.
Francis Vaughan
Jan Didden
Francis Vaughan wrote:
> Jan Didden <did...@wxs.nl> writes:
>
> |> Nested (differential) feedback was invented (and patented)
> |> by D. Edward Cherry from Australia. I have never found any reference
> |> to a marketed implementation (which doesn't mean there isn't one, of course).
>
> Cherry tried for a while to interest Australian manufacturers in the NDFL
> technology, before finally selling the patent rights to one of the big
> Japanese conglomerates (Matushita I think.) Since he was (and still is)
> a lecturer at Monash University in Melbourne Australia the money came
> to the University. I have heard that it acted as a very welcome cash
> cow for some years. However for unknown reasons the technology was
> never marketed. There were a few sporadic implementations and licensed
> products but very very few. The patents have now all lapsed.
Just. 1 feb 98. And as far as the cash cow is concerned, it was a worldwide patent,
(there even is a Netherlands version) and that cannot be done for less then a 100k $
US at the minimum, and there was very limited marketing as you said.
It is a very elegant topology indeed, but you listing of differences with a
'run-of-the-mill' amp and specific requirements may be one reason why there was no
extensive marketing. Something similar happened to Sandman's 'class S' topology,
very elegant as well, and the switching output stage, I forgot from whom. What Self
seems to have accomplished with his series in EW/WW is to suggest that there is only
one way to design a power amp, which I cannot for the life of me see as contributing
to the body of knowledge and pointing out the way forward.
BTW, you seem to be quite knowledgeable. What do you think of my
'on-demand-feedback' topology (AES preprint 4597 ;Sep 97 convention)?
Jan Didden
Todd Krieger
DPi...@world.std.com (Richard D Pierce) wrote:
> In article <7iuu40$mia$1...@nnrp1.deja.com>,
> Todd Krieger <tkri...@concentric.net> wrote:
> >
> >The speaker itself does not "create" the spikes. The amplifier
creates
> >the spikes because it's too slow to correct the speaker's back EMF.
The
> >spike is a result of the correction signal from the amp "countering"
the
> >EMF of the speaker *after* the EMF subsides. The slower the amp's
slew
> >rate, the more time it takes for the amp to generate the correction
> >signal, and the worse and more delayed the spike.
>
> Sorry, this is pure poppycock.
>
> Precisely where do you think the largest amount of this
> so-called "back-EMF" originates? AT what frequencies?
>
It does not occur as a specific frequency, but is **instantaneous.**.
Why do they call it **transient** intermodulation distortion??
> In fact, the single largest contributor to the "back-EMF" is the
> fundamental mechanical resonance of the woofer, and
> this is at frequencies WELL below 100 Hz. With some very rare
> exception, that is the dominant source.
>
It may be the largest **steady state** contributor, which is indeed at a
frequency. We're talking about a totally different phenomenon here.
> Now, you are trying to tell us that the "signals" generated by
> this exceed the slew rate capacity of modern amplifiers?
> Absolute, pure, nonsensical bunk.
>
Sounds like you have a chip on your shoulder...
> First, unless you have discivered a new operating mode of
> physics governing energy storage in electrical or mechanical
> resonant systems, there is NOW way on earth the a loudspeaker
> can generate more energy than it receives from the amplifier.
Oops. Are you saying when a loudspeaker driver is moving, and the
signal driving the driver is suddenly removed, the driver diaphragm
would then stop **instantaneously??**
> And for it to returneven a sizable portion of the energy to the
> amplifier, it's dissipative losses, be they electrical or
> mechnical, MUST be essentially non-existant.
>
How?? (Don't tell me dissipated as heat.)
> Further, AT resonance, where the back-EMF is at its greates, the
> so-called back EMF is exactly in-phase with the driving signal.
You're stating something irrelevant. We're talking about a driver
moving at an instantanous time frame. What you just stated is true, but
it is also irrelevant.
> That's why the impedance is so high at that point.
Obviously. In this case, the back EMF is moving in sympathetic motion
with the amp's output, so the amp basically "sees" a higher impedance.
> And the
> pohase of the impedance is such that it is a purely resistive
> load.
***Right exactly at that frequency.*** It becomes capacitive right
above that frequency, and inductive right below it.
> So the load seen by the amplifier at resonance is
> IDENTICAL to that of a pure resistor whose resistance equals
> that of the woofer at that frequency.
>
Once again, obvious but irrelevant. You're talking about steady state
signals, which if they don't have HF components, will cause the amp to
generate **no** TIM whatsoever, regardless of slew rate or feedback
used.
> >It's not the speaker, once again. The amp (due to insufficient slew
> >rate) does not react quickly enough to the error signal at the
output.
> >(It's impossible to design a speaker which doesn't generate EMF back
to
> >the amplifier.)
>
> Absolute nonsense. All one has to do is design a speaker whose
> impedance is essentially resistive> There is NO back emf as you
> claim. The KEF 104-2 is but one example of such.
>
If there exists a voice coil and magnet, the system will **always**
generate an EMF. Unless one can design a "massless" diaphragm. (Hill
Plasmatronics??) Whether the impedance of an electromagnetic system is
resistive or not has nothing to do with it.
In order for a speaker to have a "resistive" impedance, it must be flat
in regard to frequency. Dunlavy speakers have a flat impedance curve.
Does that mean Dunlavy speakers generate no back EMF?? Absolutely not.
It is still an electromagnetic system.
> And it precisely looking at the problem using the extremely
> limited and flawed model of "back-EMF" that leads to this sort
> of nonsensical analysis.
It may seem nonsensical if the basis is irrelevant.
> A speaker is NO more than a resonant
> system. Period.
I guess you're referring to a moving mass system...
> Instead of a voice coil moving in a magnetic
> field, you can just as readily consider it a parallel LRC tank
> cirvuit: the behavior as far as the amplifier is concerned is
> IDENTICAL.
>
An LRC can approximate an impedance curve in the steady state, but in
instantaneous conditions, an LRC does not have electromotive forces
coming from it. So the above statement is not correct for instantaneous
analysis, which is what we're dealing with here.
> >Think of the spike as the signal
> >which should have countered the back EMF, but came too late.
>
> Give us a bloody break, please.
>
OK. You got a break!! [-;
> The slew rate of these signals is determined by the bandwidth of
> the signals themselves. Since it is the woofer that is
> contributing most to these issues. Now, at 30 Hz, you are
> claiming that the so-called "back-emf" generated exceeds the
> slew rate of the amplifier?
The component of instantaneous back EMF is essentially a DC signal (in
the time frame we're dealing with) and how fast the amp can counter it.
So it does indeed involve slew rate. Think of one's hand suddenly
pushing on your hand, and you then counter it, to keep the movement to a
minimum. If the reaction takes longer, the farther the hands would move
(more "error signal"), and the harder the person reacting would have to
push to move the "system" back to its original location.
> If so, why are you using amplifiers
> with bandwidths less than 50 Hz?
>
50 Hz?? I think you mean 50 kHz. Don't understand the question.
> And please do not tell me about the "back-emf" of tweeters. Show
> it to me.
>
They exist, because anything with a voice coil, magnet, and moving mass
is capable of generating back EMF. The tweeter generates so little,
however, to make it an issue with TIM from an amplifier. (This is
because the moving mass and motor size of a tweeter is a tiny fraction
of that of lower-frequency diaphragms.)
> >This problem exists with any electromechanical system (such as a
servo
> >control, for example) which involves feedback and a finite maximum
> >reaction time. Not just amps with speakers.
>
> And this is preciesly because in most servo-system, the
> resonances being controlled are at the HIGH end of the systems
> bandwidth, NOT the low end, as is the case with loudspeakers.
>
There do exist servo-controlled speakers. (I think in the form of
powered subwoofers.) Don't sound very good because they often generate
a relatively high-frequency resonance components. (Such systems often
have a relatively high frequency resonance at a **fixed** frequency.
Audible as a variable-amplitude hum at that frequency.)
> >> It cannot be.
> >> The back EMF is defined by the same laws that govern its acoustic
> >> properties in the radiative direction. It is a nice big
spring/mass
> >> system with defined and rather mundane charateristics.
> >>
> >I wish this was the case- it would be much easier to design
amplifiers.
> > Back EMF has very little to do with a speaker's acoustic properties,
> >other than the fact the generator (voice coil in a magnetic field) is
> >coupled to the diaphragm.
>
> And as such, it behaves IDENTICALLY as a simple RLC parallel
> resonant circuit whose resonance is at the BOTTOM of the
> bandwidth, not the top. Slew rate IS NOT A FACTOR, despite your
> claims to the contrary.
>
Who says an RLC parallel resonant circuit has to have its resonance at
the *bottom* of the frequency band? Once can just as easily make an RLC
with a high-frequency resonance.
> >> what topology amplifier is in use.
> >>
> >Now you do have a point here, but if the slew rate is limited, so is
the
> >bandwidth of the *output*. The feedback loop would have no problem
> >tracking with a slower output if there were no external signals
acting
> >upon the feedback loop. The problem is when an external signal (the
> >speaker's back EMF) introduces error signals at the output, and hence
> >the feedback loop.
>
> Come on, this is nonsense, and does not get any less so by
> repetition.
>
> In order for your claim to hold, you must show that the
> resonance that reflect themselves as the speaker's impedance
> MUST be high-Q, higyh-amplitude AND high-frequency. In fact, the
> vast majority of speakers exhibit exactly the opposite. The
> speaker's impedance is closer to resistive at high-frequencies
> than at low.
>
Once again, if you introduce a sudden **DC** error signal on the amp's
output, how does the amp react to it? With instantaneous signals, we're
dealing with momentary potentials, not what frequency the system is
operating at.
> There are NO "external signals" acting upon the feedback loop.
I don't like saying this, but this is totally wrong. Maybe you need to
take a course in electromechanics. You have to realize that anything
with moving mass and that mass is attached to a coil in the presence of
a magnetic field, such as a loudspeaker or even a motor, will have the
capability of generating EMF signals. Once the diaphragm moves, it will
continue to move (momentum) and generate back EMF when the signal is
removed.
If you're skeptical, hook up a storage oscilloscope to the output of a
typical feedback amplifier, and then set off a firecracker in front of
the woofer of the speaker. I can guarantee you that you'll see a spike
at the time the firecracker goes off. (If the global feedback is high
enough, you may even pick it up at the amplifier's input.)
> ALL signals originate in the amplifier and as long as that
> amplifier is acting as a reasonable approximation of a voltage
> source (which they do VERY well at the frequencies where your
> onbjections hold, e.e. low frequencies)
If a woofer generates EMF, and an amp takes too long to counter it, the
TIM component will often be audible through the tweeter.
> then ANY external
> signals are going to see that amplifier as a dead short anyway.
> There are no high-frequency signals as you claim, otherwise they
> would show up as anomolously high and overly reactive
> impedance at high frequency. And such simply is not the case.
>
TIM, like the EMF it reacts to, is not steady state of a continuous
frequency. It is, once again, an amp's overreaction to an
**instantaneous** error signal.
Todd Krieger
> Sorry, but the whole slew rate thing is a complete crock, and is
> one of the things that sent Otala completely off the deep end.
>
> --
> | Dick Pierce |
> | Professional Audio Development |
> | 1-781/826-4953 Voice and FAX |
> | DPi...@world.std.com |
>
Todd Krieger
dun...@aol.com (DUNCLIF) wrote:
>
> This is the reason for some of the hype. It is hard to interest the
public
> with claims of "Hey, we havent changed anything in 20 years". The
easiest part
> of getting an amp to market is the design. The hardest part is
marketing a
> design that hasen't changed in 20 years as something fersh and
innovative.
> With so many companies doing that, the few truely fresh and innovative
ideas
> may never make it to market.
>
Actually, I'm auditioning the Australian Symfonia Opus 10 amplifier, and
it might be a breakthrough design in regard to sonics. I have not seen
details regarding the design of this solid-state amp, other than the
fact the internal circuitry uses exceptionally high-quality parts.
(Then again one can say that about several companies' products.) The
amp, still "burning in," seems to have very little of the "nasties"
which have plagued even the finest in solid-state design. (And also
has a lot of the "magic" one hears in the finest tube designs.) But you
are correct about the market. Very few people have heard of the
company, and there's no guarantee this company will be around five or
ten years from now. The only thing I can say is to my ears, it is the
best amp, tube or solid state, I've ever tried in my own system.
But I remember companies which made exceptional equipment, from Hegemann
to Holman to BRB to GAS/Sumo, which went belly-up. And there's a lot of
so-called "mid-fi" going on stronger than ever. Not everyone buys audio
for absolute recreation of the recording. People buy what they like,
whether it's high-end audiophiles, musicians, or people who love zippy
highs and boomy, overwhelming bass. The companies that prosper make
what's popular, which is not necessarily the best, or even close to the
best. (I think Polk Audio realized this, and eventually transformed
from "budget high end" company to "consumer" company.)
André Huisman
>Oops. Are you saying when a loudspeaker driver is moving, and the
>signal driving the driver is suddenly removed, the driver diaphragm
>would then stop **instantaneously??**
And how would this sudden removal happen??? Not in a passive system where
the filter prevents this from happening. NOT in an active system too where
the active filter prevents this from happening. So it only happens during
clipping. During clipping, all kinds of nasty things go on (like OT's
needing to get out of saturation and such).
SO: How would this "suddenly removed" signal happen in a bandwidth limited
system to begin with
>> And for it to returneven a sizable portion of the energy to the
>> amplifier, it's dissipative losses, be they electrical or
>> mechnical, MUST be essentially non-existant.
>How?? (Don't tell me dissipated as heat.)
OK, let's look into this back EMF thingie from a numbers point of view. We
apply a signal to a speaker. This wonderfull speaker then turns 1% of that
into movement. At the end of a movement, your magical "suddenly removed"
thingie happens. So now we have a speaker that is going to perform as a
generator (with an efficiency of 1%). That means that in essence (if the amp
driving the speaker had infinite output impedance) we get a back EMF of 1%
of 1% of the original signal. In dB's that would be a signal at -80dB. BUT
wait. Let's take a more realistic amp. This amp will probably have an open
loop output impedance of less than 0.5 Ohms (most amps are far better than
that). This means that there's a nice voltage dividing going on on top of
the already terribly big losses. 0.5/6.5 for an 8Ohms speaker. So now we
have 1% of 1% of 7% which translates to a "back EMF" signal that has a level
of -102dB compared to the stimulus (best case scenario). BUT wait. Most amps
with high feedback have a filter at their output meaning that the signal
gets even more dampened. EVEN MORE. And this signal will give a so called
"spike"??? One would think that the RFI noise on the speaker cords would
exceed this -102dB signal AND they would be far greater in bandwidth than
your "back EMF" signal.
But wait, let's assume this nasty -102dB signal error IS finding it's way
into the input again (unaltered). Let's also assume a nice big 100W amp with
a sens of 775mV for full power output and let's even assume we want to
torment our ears to that kind of power. SO now we have a signal at the
output of the amp of around 28V. We get a "back EMF" signal (for whatever
strange "stopping on a dime" reason) of around 100dB below that. That would
be a signal of 0.001% of the signal -> 0.28mV. The A of the amp would have
to be around 36 times, so the error signal would represent itself as an
error signal of 0.00775mV at the input of the amp. Hey, that's also 100dB
below the input signal.
YEAH, I can really see how this -100dB signal would shop up on a scope
picture, really I can ;-)
I've seen strange spikes and other things on some amps. BUT I also saw them
on a resistive load (a dummy load) and I just know that most resistors don't
have any "back EMF behavour".
>The component of instantaneous back EMF is essentially a DC signal (in
>the time frame we're dealing with) and how fast the amp can counter it.
You really picture a speaker with a very limited bandwidth to stop on a
"dime"??? Hmm, can't really picture that big bulky heavy cone stopping like
that. IF the speaker can't "stop" like that, it can NOT generate this
strange spike you claim.
>50 Hz?? I think you mean 50 kHz. Don't understand the question.
Open loop bandwidth. If you can't even understand this joke then....
>Once the diaphragm moves, it will
>continue to move (momentum) and generate back EMF when the signal is
>removed.
And any form of filtering (active/passive) prevents this from happening.
>If you're skeptical, hook up a storage oscilloscope to the output of a
>typical feedback amplifier, and then set off a firecracker in front of
>the woofer of the speaker. I can guarantee you that you'll see a spike
>at the time the firecracker goes off. (If the global feedback is high
>enough, you may even pick it up at the amplifier's input.)
And this exactly shows how increddibly lossy your "back EMF" thingie is to
begin with. You need a fire cracker (at an SPL far greater than whatever the
speaker could put out to begin with) to even be able to see the error
signal. Do this while the speaker is being driven by a nice big signal and
the "fire cracker induced back EMF" probably drowns in the signal already
present at the speaker terminals.
André Huisman
New Line licht & geluid
new...@xs4all.nl
http://www.xs4all.nl/~newline/
--- pardon my French, I'm Dutch ---
Richard D Pierce
anything else. I'll just point out some of them.Frankly, I am
groung weary of trying to propvide reasonable information for
people not interested in facts.
In article
<7jd6ai$oe2$1...@nnrp1.deja.com>, Todd Krieger
<tkri...@concentric.net> wrote:
>IYO...
>
>> Precisely where do you think the largest amount of this
>> so-called "back-EMF" originates? AT what frequencies?
>>
>It does not occur as a specific frequency, but is **instantaneous.**.
>Why do they call it **transient** intermodulation distortion??
Sorry, this is so nonsencially, utterly wrong. NOTHING in a
loudspeaker, nothing fed to a loudspeaker, nothing out of a
croissover or an amplifier behaves instantaneously or even
remotely like it. You may choose to ignore this fact all you
want, Mr. Krieger, but it does not becaome any truer because you
do so. Instantaneous changes in signal, as you claim, require
infinite bandwidth, it's as simple as that. The fact that ALL of
these drivers are connected via crossover further, by
definition, limits the rate at which ALL signals can change.
>
>> In fact, the single largest contributor to the "back-EMF" is the
>> fundamental mechanical resonance of the woofer, and
>> this is at frequencies WELL below 100 Hz. With some very rare
>> exception, that is the dominant source.
>>
>It may be the largest **steady state** contributor, which is indeed at a
>frequency. We're talking about a totally different phenomenon here.
No, we are not, and your attempts at proof by vigorous
assertion notwithstanding.
>> First, unless you have discivered a new operating mode of
>> physics governing energy storage in electrical or mechanical
>> resonant systems, there is NOW way on earth the a loudspeaker
>> can generate more energy than it receives from the amplifier.
>
>Oops. Are you saying when a loudspeaker driver is moving, and the
>signal driving the driver is suddenly removed, the driver diaphragm
>would then stop **instantaneously??**
Oops, are YOU saying the signal is remobed "instantaneously?"
>> And for it to returneven a sizable portion of the energy to the
>> amplifier, it's dissipative losses, be they electrical or
>> mechnical, MUST be essentially non-existant.
>>
>How?? (Don't tell me dissipated as heat.)
Why, because it's another physical fact that is a fatal
to your argument?
Sorry, but unless you can show that under NAY conditions, steady
state or transient, that the reac6tive portion of the system
exceeds the resistive portion over a sufficiently wide
bandwidth, then your argument fails miserably.
And yes, the VAST majority of energy delivered to a loudspeaker
(in the case of most direct radiator system, exceeding 96%) IS
DISSIPATED AS HEAT. That is a simple fact of physics.
>> Further, AT resonance, where the back-EMF is at its greates, the
>> so-called back EMF is exactly in-phase with the driving signal.
>
>You're stating something irrelevant. We're talking about a driver
>moving at an instantanous time frame. What you just stated is true, but
>it is also irrelevant.
Your continued, erroneous and mistaked use of "instantaneous is
but one of your irrelevancies.
>> That's why the impedance is so high at that point.
>
>Obviously. In this case, the back EMF is moving in sympathetic motion
>with the amp's output, so the amp basically "sees" a higher impedance.
>
>> And the
>> pohase of the impedance is such that it is a purely resistive
>> load.
>
>***Right exactly at that frequency.*** It becomes capacitive right
>above that frequency, and inductive right below it.
HOW capacitve? HOW inductive? SO you know? Are you going to be
shocked to learn that, under ALL conditions, the resistive
portion dominates?
>> So the load seen by the amplifier at
resonance is
>> IDENTICAL to that of a pure resistor whose resistance equals
>> that of the woofer at that frequency.
>>
>Once again, obvious but irrelevant. You're talking about steady state
>signals, which if they don't have HF components, will cause the amp to
>generate **no** TIM whatsoever, regardless of slew rate or feedback
>used.
No, I am talking about ALL signals. One of the predicitions of
your "hypothesis" is that the impedance mreasured under
transient conditions should be radically different than that
measured under steady state conditions. And, having in fact done
just that literally thousadns of rimes, I and assure you that it
is not.
>> A speaker is NO more than a resonant
>> system. Period.
>
>I guess you're referring to a moving mass system...
You guess quite incorrectly, not the first of your guesses to go
awry.
>> Instead of a voice coil moving in a magnetic
>> field, you can just as readily consider it a parallel LRC tank
>> cirvuit: the behavior as far as the amplifier is concerned is
>> IDENTICAL.
>>
>An LRC can approximate an impedance curve in the steady state, but in
>instantaneous conditions, an LRC does not have electromotive forces
>coming from it. So the above statement is not correct for instantaneous
>analysis, which is what we're dealing with here.
This is pure nonsense. The elctrical impedanceos a loudspeaker
approximates that of a resonant circuit BECAUSE IT IS A RESONANT
CIRCUIT. Whether it is mechanically resonant circuit or an
electrically resonant circuit is irrelvant.
And, once again, you assertion fails miserably in that the
transient behavior of the iimpedance matches that of the steady
stae impedance quite accurately, thank you.
>> The slew rate of these signals is determined by the bandwidth of
>> the signals themselves. Since it is the woofer that is
>> contributing most to these issues. Now, at 30 Hz, you are
>> claiming that the so-called "back-emf" generated exceeds the
>> slew rate of the amplifier?
>
>The component of instantaneous back EMF is essentially a DC signal (in
>the time frame we're dealing with) and how fast the amp can counter it.
You have a real semantic problem here. The notion of
"instantaneous" and "DC" are contradictory. It's either DC and
therefore steady state or it is not.
>> If so, why are you using amplifiers
>> with bandwidths less than 50 Hz?
>>
>50 Hz?? I think you mean 50 kHz. Don't understand the question.
No, given yoyur absurd assertion, I meant 50 Hz. Given the
DYNAMIC behavior of loudspeaker, YOUR claims about slew rate
suggest that YOU are using amplifiers of VERY limited bandwidth.
>> And please do not tell me about the "back-emf" of tweeters. Show
>> it to me.
>>
>They exist, because anything with a voice coil, magnet, and moving mass
>is capable of generating back EMF. The tweeter generates so little,
>however, to make it an issue with TIM from an amplifier. (This is
>because the moving mass and motor size of a tweeter is a tiny fraction
>of that of lower-frequency diaphragms.)
You loose. Look at your EMF. WHat is the magnitude of the EMF?
It is, in fact, the product of the field density, the length of
wire in that field, and the velocity normal to that field. Where
do the highest possible velocities in the voice coil take place?
Care to guess, or might you be interested in facts instead?
>> >This problem exists with any electromechanical system (such as a
>servo
>> >control, for example) which involves feedback and a finite maximum
>> >reaction time. Not just amps with speakers.
>>
>> And this is preciesly because in most servo-system, the
>> resonances being controlled are at the HIGH end of the systems
>> bandwidth, NOT the low end, as is the case with loudspeakers.
>>
>There do exist servo-controlled speakers. (I think in the form of
>powered subwoofers.) Don't sound very good because they often generate
>a relatively high-frequency resonance components. (Such systems often
>have a relatively high frequency resonance at a **fixed** frequency.
>Audible as a variable-amplitude hum at that frequency.)
What on earth are you smokin'?
>> And as such, it behaves IDENTICALLY as a simple RLC parallel
>> resonant circuit whose resonance is at the BOTTOM of the
>> bandwidth, not the top. Slew rate IS NOT A FACTOR, despite your
>> claims to the contrary.
>>
>Who says an RLC parallel resonant circuit has to have its resonance at
>the *bottom* of the frequency band? Once can just as easily make an RLC
>with a high-frequency resonance.
Look, you want to play non-sequitur, go somewhere else.
The electrical impedance of a driver is a reflection of the
motional impedance, pure and simple. All electrodynamiuc
speakers in use, with VERY few exceptions, are operated ABOVE
fundamental resonance, so that resonance is at the BOTTOM of
their range.
>> In order for your claim to hold, you must show that the
>> resonance that reflect themselves as the speaker's impedance
>> MUST be high-Q, higyh-amplitude AND high-frequency. In fact, the
>> vast majority of speakers exhibit exactly the opposite. The
>> speaker's impedance is closer to resistive at high-frequencies
>> than at low.
>>
>Once again, if you introduce a sudden **DC** error signal on the amp's
>output, how does the amp react to it? With instantaneous signals, we're
>dealing with momentary potentials, not what frequency the system is
>operating at.
What is the magnetidue of that signal?
It CANNOT exceed the input signal. Where is the extra coming
from. It was the amplifer that set the cone into motion, and
that cone CANNOT return more energy to the system than was put
in it. And MOST of that energy is dissipated by losses in the
system (whether you're smart enough to understand or admit it is
beside the point).
Further,WHATEVER mechnical energy is left to generate your EMF,
NOW must see the Thevenin attenuator formed by the voice coil
resistance and the amplifier output resistance.
NOW what has happend to this mythical signal of yours. It's
gone.
>Maybe you need to take a course in electromechanics.
No, maybe after 30 years of doing this professional, of
measuring thousands upon thousands of speakers and doing
precisely the study and measurement, I should give up trying to
educate people.
>If you're skeptical, hook up a storage oscilloscope to the output of a
>typical feedback amplifier, and then set off a firecracker in front of
>the woofer of the speaker. I can guarantee you that you'll see a spike
>at the time the firecracker goes off.
Great. He's invented the low-frequency microphone. Give the man
a patent.
> (If the global feedback is high
>enough, you may even pick it up at the amplifier's input.)
Prove it. Show us.
Sir, it is obvious that you have NEVER performed a single one of
these experiment. I do them DAILY.
>> ALL signals originate in the amplifier and as long as that
>> amplifier is acting as a reasonable approximation of a voltage
>> source (which they do VERY well at the frequencies where your
>> onbjections hold, e.e. low frequencies)
>
>If a woofer generates EMF, and an amp takes too long to counter it, the
>TIM component will often be audible through the tweeter.
And you have totally, ignoprantly, ignored the fact that your
assertion requires the bandwidth of the speaker to be wider than
that of the amplifier, something that you have utterly failed
and WILL utterly fail to demonstrate.
>TIM, like the EMF it reacts to, is not steady state of a continuous
>frequency. It is, once again, an amp's overreaction to an
>**instantaneous** error signal.
There ARE no such "instantaneous" signals as you claim, pure and
simple.
INstead of more handwaving amd vigourous assertion, SHOW us the
data. Not some wild specualtion as you have show here, but DO
those measurements you claim and SHOW us the data.
DUNCLIF
> The only thing I can say is to my ears, it is the
>best amp, tube or solid state, I've ever tried in my own system.
Well that's only because you haven't heard one of my amps.
>But I remember companies which made exceptional equipment, from Hegemann
>to Holman to BRB to GAS/Sumo, which went belly-up.
I got to meet all those guys. I guess I feel better about my bussiness failure
with company like this. Marketing ruels!
Todd Krieger
> anything else. I'll just point out some of them.Frankly, I am
> groung weary of trying to propvide reasonable information for
> people not interested in facts.
>
I'm not arguing anything. The original intent was to explain TIM to one
who was curious about it, and details if one questions such details.
And what I'm saying is not-at-all the last word in why amplifiers do
what they do. You go around with the attitude that you know it all, and
it seems like you get bent out of shape when **someone else** tries to
explain something. The dead giveaway is that you've been countering
these explanations with utterly irrelevant (and sometimes false)
information, intertwined with ridicule. (There are others here who use
the very same "technique" in countering such explanations.) And often
confusing the readers in the process.
Why would one **want** to post "myths" regarding the pheonomenon of TIM?
(Or any other phenomenon, for that matter?) As if I had an agenda
promoting those who design low-TIM amplifiers or something.
I'm **not** implying that TIM is definitely occurring (nor that it
unequivocally sounds better), but explaining a real world possiblility
on how and why it **could** be occurring. Nothing more than a theory.
For if one has the attitude about what is occurring is *definitely* one
way or the other ("poppycock", "nonsense", "myth"), maybe that's why we
have been in this rut regarding the recent lack of progress in advancing
the state-of-the-art in amplifiers. The best designs in anything
involve investigation of **all** possible phenomena, whether established
in textbooks or new conditions suspected in recent studies.
I'm not saying Matti Otala was correct in his papers regarding TIM, just
saying it should be one of many findings to be investigated. And
discounting any phenomenon as "poppycock" is closed-mindedness at best,
arrogance at worst.
Todd Krieger
Todd Krieger
dun...@aol.com (DUNCLIF) wrote:
> >Todd Krieger wrote:
>
> > The only thing I can say is to my ears, it is the
> >best amp, tube or solid state, I've ever tried in my own system.
>
> Well that's only because you haven't heard one of my amps.
>
equipment!!
Todd Krieger
DPi...@world.std.com (Richard D Pierce) wrote:
> And, once again, you assertion fails miserably in that the
> transient behavior of the iimpedance matches that of the steady
> stae impedance quite accurately, thank you.
>
What do you say when an "engineer" cites **everything** in a post as
wrong, and then posts a statement like this?? Wow.
> No, maybe after 30 years of doing this professional, of
> measuring thousands upon thousands of speakers and doing
> precisely the study and measurement, I should give up trying to
> educate people.
>
Maybe **this** is why audio hasn't advanced much recently. With
educators like this, who needs BS'ers??
That was the most accurate statement I've ever seen you post.
Todd Krieger