Audio Opamp OPA2134.
Jan Willem Auer
The OPA 2134 is the best audio opamp from Burr-Brown, suitable for
high-end audio applications.
Does somebody know how it sounds compared to a standard NE 5532 which
is being used in my Sony DTC 59 ES datrecorder as an active filter
and output stage?
Does it sound better than the OPA 2604?
Thanks, Jan Willem Auer.
Thor...@tnt-audio.com
Yo 'dettes 'n dudes,
> The OPA 2134 is the best audio opamp from Burr-Brown,
Who says that?
> suitable for high-end audio applications.
Maybe.
> Does somebody know how it sounds compared to a standard
> NE 5532 which is being used in my Sony DTC 59 ES dat
> recorder as an active filter and output stage?
> Does it sound better than the OPA 2604?
I can only talk about NE5532 vs. OPA2604 vs. LM6172 (or LM6182 for
certain applications).
The NE5532 simply stinks. Only the 741/1458 is worse.
The OPA2604 is a notable step up, with better sound-staging and much
closer to natural timbres. But in the End, just not good enough.
I do have my own theories why, generally the Open-Loop Bandwidth of
these type of Op-Amps (conventinal style vs. current-feedback) is not
good enough (not by a few magnitudes!).
Hence my fav. Op-Amp the LM6181/82 (6181 single, 6182 dual) sounds
best to me. The 6181 is more or less comparable to the AD811, the
Dual-Chip 6182 has no other comercial equivalent I know of.
In cases where the Currentfeedback topology of the LM6181/82 makes
the usage a Problem, the LM6171/72 can be used. THis is essentially
the same Op-Amp but with the current feedback bandwidth setting done
internaly.
The LM6172 can replace the NE5532 is almost any circuit without change.
Anyway, my view on the situation.
In my CD-Player modifications (CD-67 SE) Article on TNT-Audio I go
into all of this in a bit more detail. Have a read.
http://www.tnt-audio.com/clinica/cd67.html
Kind regards Thorsten.
======================================
e-mail:
Thor...@tnt-audio.com
Visit TNT-Audio on the Web - the only advertising
free audio web-zine.
http://www.tnt-audio.com
======================================
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/ Now offering spam-free web-based newsreading
alk...@mcs.com
> The OPA 2134 is the best audio opamp from Burr-Brown, suitable for
> high-end audio applications.
[quoted text deleted -- deb]
where can I order a pair or two of these? my local supplier went oob
when op275's were the thing I yanked the 4558's and 5532's out of my
stuff and installed dip sockets for quick upgrading and A-B tests
Greg Graff
I believe that this is the same op amp that Jeff Rowland is using in
his new designs. If its the same one, it sound better than any
discrete circuits I've ever heard.
gjg :)
Tom Kite
On 26 Feb 1998 Thor...@tnt-audio.com wrote:
> The NE5532 simply stinks. Only the 741/1458 is worse.
That's just silly. There are plenty of op-amps between the 741 and
the 5532 in quality, and plenty worse than the 741. The 5532
actually measures very well in terms of distortion and noise, amongst
other things.
> I do have my own theories why, generally the Open-Loop Bandwidth of
> these type of Op-Amps (conventinal style vs. current-feedback) is not
> good enough (not by a few magnitudes!).
But your theories are absolutely false. Although you have not
presented them in full here, you did in fact do so on rec.audio.tech
on 1/18/98, and were promptly shot down. Your theory is that
enormous open-loop bandwidth is required to accurately reproduce
audio signals. Yet the AD797 and OPA132, both of which have THD
120dB below 1V, have sub-100 Hz open loop bandwidths.
Your theory of using op-amps with huge open-loop bandwidths leads to
your sacrificing all other parameters, such as distortion, noise, DC
precision, input bias current, and so on, for bandwidth that is
simply not needed.
> The LM6172 can replace the NE5532 is almost any circuit without change.
This is contrary to your assertion on r.a.t., namely:
"In addition, the widebandwidth demands extreme attention to the
powersupplies or the whole thing will oscillate."
which is an understatement.
Generally speaking, drop-in replacement should be done with great
care, especially since the compensation used for the existing op-amp
may not be known. Using an op-amp with enormous open-loop bandwidth
in a low closed-loop gain circuit is just asking for trouble.
Tom
Arny Kr|ger
Tom Kite wrote in message <6d6lpu$e...@news01.aud.alcatel.com>...
>On 26 Feb 1998 Thor...@tnt-audio.com wrote:
>
>> The NE5532 simply stinks. Only the 741/1458 is worse.
>
>That's just silly. There are plenty of op-amps between the 741 and
>the 5532 in quality, and plenty worse than the 741. The 5532
>actually measures very well in terms of distortion and noise, amongst
>other things.
While an excellent and true statement, this still understates the good
news about the 5532. 5532's are very commonly used in all kinds of pro
audio gear. It is probable that maybe a quarter or more of all
commercial recordings have passed through one or more 5532's during
their production. If all the 5532 and 5532 close clones suddenly
disappeared, the audio industry would be crippled badly.
Not only do 5532's measure quite well, but they are commonly used in
critical stages of much audio measurement gear.
Desire for low distortion and noise are why 5532's are specified, and
they often deliver. They are also dirt cheap and readily reliable and
have been around for well over a decade - almost two.
5532's downside is that they use a fair amount of power and are not
really at their best in many high impedance circuits. 5532's are not
perfect, and their performance in many key areas has been surpassed in
technical terms. Nevertheless, any claim that 5532's "stink"
establishes a high dogma level that IMO impugns the rest of the claims
that might be made in the same article.
The fact is that many modern IC's are solutions looking for
problems. Some of them deliver benefits that are surprizingly small
compared to classic audio IC's like the TL072 and the 5532. In many
other cases, substantial technical benefits are realized, but their
audible benefits remain small.
If you want to put put most audio op amp "experts" in their place,
assemble a chain of say, 10 stages based on one of the chips they
condemn and oblige them to prove they can reliably hear the difference
between that, and say 1 stage of a chip they admire. This was been
done to me about 20 years ago and it put me in my place, but it also
set the stage for learning many important things.
>> I do have my own theories why, generally the Open-Loop Bandwidth of
>> these type of Op-Amps (conventinal style vs. current-feedback) is not
>> good enough (not by a few magnitudes!).
>
>But your theories are absolutely false. Although you have not
>presented them in full here, you did in fact do so on rec.audio.tech
>on 1/18/98, and were promptly shot down. Your theory is that
>enormous open-loop bandwidth is required to accurately reproduce
>audio signals. Yet the AD797 and OPA132, both of which have THD
>120dB below 1V, have sub-100 Hz open loop bandwidths.
I'm glad that this simplistic claim was shot down because it deserves
to be. I think that many people initially become concerned when they
learn of the relatively narrow open loop bandwidth of most operational
amplifiers, be they discrete, hybrid, or monolithic. However,
operational amplifiers were never contrived to be used open
loop. Worrying about op amp open loop bandwidth because it is small is
like worrying about drowning when eating bread because bread contains
water. However, much "Audio Truth" that one sees in the popular press
and NG's is like this. The good news is that folks like you debunk it
when you can.
>Your theory of using op-amps with huge open-loop bandwidths leads to
>your sacrificing all other parameters, such as distortion, noise, DC
>precision, input bias current, and so on, for bandwidth that is simply not
needed.
Exactly. Audio is not rocket science. Whether you think that
transparent reproduction requires 20 kHz bandwidth, 15 kHz bandwidth
or 50 kHz bandwidth, its usually pretty easy with modern technology,
when that technology is correctly applied.
>
>> The LM6172 can replace the NE5532 is almost any circuit without change.
>
>This is contrary to your assertion on r.a.t., namely:
>
> "In addition, the widebandwidth demands extreme attention to the
> powersupplies or the whole thing will oscillate."
>
>which is an understatement.
Exactly. Enginering by part number leads to upgrade without technical
evaluation, which is one of the real evils in audio construction
today. The final stage is regrettably, but not infrequenly,
loudspeaker driver immolation by "upgraded" equipment, which is
peridically reported, on R.A.T. However, the real evil is fostering
claims of audible improvement without proof by any reliable means.
>
>Generally speaking, drop-in replacement should be done with great
>care, especially since the compensation used for the existing op-amp
>may not be known. Using an op-amp with enormous open-loop bandwidth
>in a low closed-loop gain circuit is just asking for trouble.
Making any kind of change without careful evaluation of the results is
a bad idea. The most common undesirable results are reduction of sound
quality when judged in an unbiassed way, destruction of correctly
operating equipment, and inflation of the ego of the person doing the
modification. I.e., the so called "constructors ear" syndrome. The
latter is a common evil that many foster and personally derive
economic returns from. IMO, the underlying drive comes from the
ego. Enhancement without reliable evaluation gives the thrill of "Look
Ma, no hands!" all over again.
Jan N Rasmussen
I have a pair of Egglestone Works "Andra" speakers. Stereophile have
rated this speaker as the speaker of the year.
After reading their test results and then listened to the speaker I
wonder why they classify this speaker to be among the best. There`s
no dynamic, the bass is far from what Stereophile describe, and the
midrange is coloured. I am very disapointed about the speaker, and
also about Stereophile.
Rest of the system is: ML 333, ML 38S, ML 36S, and ML 37 with cables
from Madrigal and Quest.
Anyone having experience with this speakers ?
Jan R
Thor...@tnt-audio.com
Hi,
> The 5532 actually measures very well in terms of distortion and noise, amongst other things.
Yes, BUT IT SOUNDS not good, amongst other things.
> > I do have my own theories why, generally the Open-Loop Bandwidth of
> > these type of Op-Amps (conventinal style vs. current-feedback) is not
> > good enough (not by a few magnitudes!).
>
> But your theories are absolutely false.
Are they. Could you please provide conclusive proof before making
these statements?
> Although you have not presented them in full here, you did in fact do so on rec.audio.tech on 1/18/98, and were promptly shot down.
Where they?
Once people shut their minds and present dogmas I simply go away. I
can do without long flame treads. This does not mean I concede
defeat. I just have better things to do than tit 4 tat arguing.
Anyway, they are thoeries only and I have had no access to suitable
mesurement equipment (if such exists) to verify them.
But at least empirically (read in practice) the use of these opamp's
works and better than either AD797 or any other conventional
monolithic Op-Amp I've tried.
> Your theory is that enormous open-loop bandwidth is required to accurately reproduce audio signals.
Yup, that's it in a nutshell. Note that high inherent linearity
features too.
> Yet the AD797 and OPA132, both of which have THD 120dB below 1V, have sub-100 Hz open loop bandwidths.
Yes, but sub 120db THD at what frequencies? And more importantly does
this (steady state) measurement have any relevance?
How about dynamic measurements with white noise bursts?
CF (Current Feedback) Topology Op-Amp's have an very interesting set
of properties.
They have an extremly wide Bandwidth Open-Lopp and unlike conventional
Op-Amp's they have no rise in THD or IMD or TIM within (and quite a
bit outside) the Audioband. The measured Distortion is usually not
worse than most "Audio Op-Amp's" at 1kHz.
Also very interestingly CF have a PSRR that is fairly constant up to
>> audio Frequencies. At above a few kHz they beat any Audio op-Amp.
So even when measured, the CF Op-Amp wins.
> Your theory of using op-amps with huge open-loop bandwidths leads to
> your sacrificing all other parameters, such as distortion, noise, DC
> precision, input bias current, and so on, for bandwidth that is
> simply not needed.
Listen to the different canidates before you say the Bandwidth is not
needed.
Distortion, noise and DC performance is not as compromised as you
think, in my applications I had never any problems due to DC Offset (I
couple directly). The only compromise is on Input Bias. In most
CD-Player or DAT Output stages this does not matter much. I have not
yet found it a problem.
Let's see some spec's:
LM6172 at +/- 15V typ. Values:
Offset: 0.4mV
Output Current: max. +90mA,-85mA
Settling Time (0.1%): 65ns
THD: 2nd Harmonic -110db (F=10kHz, Vout = 2V p-p, Rl = 100 ohm)
Slew Rate: >2000V/uS
Ein (noise Voltage): 12nV/sqrt(Hz)
Are these compromised Distortion, DC Performance and such?
Admittedly, the Noise is a bit higher than the 5532 (about 6db worse)
but similar to the OPA604 and OPA134 Op-Amps.
The LM6182's data is similar in all respects but for Noise, the LM6182
has 2.1nV/sqrt(Hz) (if memeory serves) which betters the NE5532 by
quite a bit.
> > The LM6172 can replace the NE5532 is almost any circuit without change.
>
> This is contrary to your assertion on r.a.t., namely:
>
> "In addition, the widebandwidth demands extreme attention to the
> powersupplies or the whole thing will oscillate."
I have since made a few direct substitution without changing the
PSU's, no problems. Seems I was overly cautious on RAT.
But you can hear the better PSU's.
> Generally speaking, drop-in replacement should be done with great
> care, especially since the compensation used for the existing op-amp
> may not be known.
What type of Compensation are you eluding to? The 5532 is internaly
compensated. So is the LM6172. Both are unity Gain stable.
After all this, on purely objective (measurements) and subjective
(listening tests) the Op-Amp's mentioned make excellent Op-Amp's.
On purely subjective notes, send me something better, I'll put it in
and listen and comment afterwards.
The AD797, OPA132/134/604/627 and all 5532/34 Variants need not apply
however, I tried them already. To quote the Prophet Daniel, "TE'KEL"
(weighted and found wanting - paraphrased).
As said before, this is my view and you can have different ones.
Thor...@tnt-audio.com
Hi,
> I believe that this is the same op amp that Jeff Rowland is using in
> his new designs. If its the same one, it sound better than any
> discrete circuits I've ever heard.
In his Line-Preamps Jeff Rowland uses the AD811, a current-feedback
design. As for the merits or dismerrits, there are just another line
in this tread where the 5532 is declared King of all Op-Amps, with the
AD797 Emporer.
Well You put your money down where you want to. In this case (jeff
Rowland Pre's) I agree. They sound marevelous.
The reason, Input Transformers, realy fast Op-Amps and a healty
disregard for the estabished rules how Audio is done.
Respect Jeff.
Later Thorsten
Fred Whitlock
I don't have any experience with the speakers but I have some
experience with reviewers. In the first place, a reviewer at
Stereophile liked the speakers with his system, in his listening room
with his musical preferences and his taste in sound. You are
listening to the speakers in a different environment entirely and your
own preferences and biases are different from those of the reviewer.
The fact that the speakers worked for him says absolutely nothing
about how they will work for you. It can say that the product has
merit and is worth a listen. It shouldn't say it will improve the
sound of your system.
Personally I think it's a big mistake to choose an audio component
strictly on the merits of a review. It's a popular thing to do and
yet makes no sense at all to me. A review can tell you some things
about a product but only your ears should judge a component's
suitability to your system. I've listened to many, many fine, popular
audio products with great pedigrees and reviews that were absolutely
awful in my system and to my ears. As an example, I love the sound of
electrostatic and planar speakers but I can't use them in my listening
room. They just don't work there. It doesn't matter how good they
sound elsewhere. They're not for me.
Use reviews as an input, not a guide.
--
Fred
Maplewood Photography
http://www.maplewoodphoto.com
Jan N Rasmussen <ardal...@vestdata.no> wrote in article
<6dcghc$705$1...@agate.berkeley.edu>...
gfre...@northnet.org
In article <6dcj11$812$1...@agate.berkeley.edu>,
Thor...@tnt-audio.com wrote:
>In his Line-Preamps Jeff Rowland uses the AD811, a current-feedback design.
There may be a variety of reasons he does this, as for the AD811 the
noise, CMRR, and output drive are exceptional. When specifying a part one
usually looks for a part that meets some predetermined minimum limits, if it
happens to exceed the minimum by 5 or 10 times what is required on some
parameter, so what! There as well will be other factors that influence the
choice of a specific part that have nothing to do with performance.
1) Engineering resources
Sourcing parts is one aspect of engineering that can be quite time
consuming. Unless you are a very large manufacturer and are sourcing for
production of millions of units, the time spent looking for a less expensive
part can easily eat up any savings that would be realized. For a small
manufacturer the cost of the parts is a small percentage of the total cost of
manufacturing.
2) Availability and delivery times
Once a appropriate part or parts are found there are still
considerations of availability, I have seen delivery times of 14 months, with
4 to 8 months being quite common. You don't have to be a genius to realize
that a few dollars saved on a part will be quickly lost if the product
release is delayed a substantial amount of time ( especially if it misses a
major trade show).
3) Inventory costs
Inventory costs money, it is less expensive to inventory a part of 1000
pieces than it is to inventory 500 piece lots of two different parts. Again
the actual cost of the part is further skewed. It is quite common practice to
use what is on the shelf rather than adding another part to inventory even if
the new part is cheaper!
4) Market acceptability
Then there is the consumer expectations, the marketing people know this
and they use it. Some of these expectations may be justified, many are not.
>As for the merits or dismerrits, there are just another line
>in this tread where the 5532 is declared King of all Op-Amps, with the AD797
>Emporer.
No one ever declared anything remotely close to that, you seem to
believe that the part was selected because of it's wide bandwidth and high
slew rate, I doubt it! It was probably selected for it's noise and output
drive capability and CMRR. While the bandwidth/ slew rate makes for good
marketing strategy and makes it distinct from other products.
>Well You put your money down where you want to. In this case (jeff
>Rowland Pre's) I agree. They sound marevelous.
>The reason, Input Transformers, realy fast Op-Amps and ........
Pure conjecture, you extract one parameter off the data sheet and declare
that to be the main reason for it quality of sound. Any op amp can be over
compensated to reduce its bandwidth and slew rate, I'd be real curious as to
how much of that bandwidth is preserved in that design.
>a healty disregard for the estabished rules how Audio is done.
Only the illusion of such.
Greg
gfre...@northnet.org
In article <6dcirf$802$1...@agate.berkeley.edu>,
Thor...@tnt-audio.com wrote:
>
> Hi,
> > > The 5532 actually measures very well in terms of distortion and noise,
> > >amongst other things.
> Yes, BUT IT SOUNDS not good, amongst other things.
To quote you " says who "?
> > > I do have my own theories why..........
> > But your theories are absolutely false.
> Are they. Could you please provide conclusive proof before making these
> statements?
I can, " Microelectronics Circuits " 3rd edition, Sedra and Smith,
Library of Congress Catalog no. 0-03-051648-X. Published by Saunders College
Publishing, a standard EE text. ( Any College EE text will provide this info )
See section 2:7, " Non Ideal Performance of Op Amps " pg 76 - 82. Now
don't go telling me about reading it in a book, I have had plenty of
opportunity to apply what is there. Your theories could be easily disproved
by any second year EE student in any reasonably equipped lab.
The burden of proof of course lies with you, if someone declared to
have discovered a planet beyond Pluto would you expect that person to provide
proof or would it be incumbent upon everyone else to prove that it didn't
exist?
> > Although you have not presented them in full here, you did in fact do so
on > >rec.audio.tech on 1/18/98, and were promptly shot down.
>
> Where they?
Yes they were.
> Anyway, they are thoeries only and I have had no access to suitable
> mesurement equipment (if such exists) to verify them.
It does exist.
> But at least empirically (read in practice) the use of these opamp's
> works and better than either AD797 or any other conventional
> monolithic Op-Amp I've tried.
This is of course a subjective judgment, putting aside possible placebo
effects, would you consider that the excessive bandwidth is causing the
circuit to ring, which you may personally find to be a desirable effect. This
is only one of a multitude of possibilities, unless you can get down and
dirty with the proper tools you will never know.
> > Yet the AD797 and OPA132, both of which have THD 120dB below 1V....
> Yes, but sub 120db THD at what frequencies?
Up to about 10 kHz, about 118 dB at 30 kHz, and 110 dB at 100 kHz. ,
the data sheets are readily available. The distortion is actually lower, 120
dB is the limits of the automated measuring equipment that Analog Devices
uses, they simply will not publish specs that cannot be confirmed by their
tests.
>And more importantly does this (steady state) measurement have any relevance?
Would you believe me if I said yes?
> The LM6182's data is similar in all respects but for Noise, the LM6182
> has 2.1nV/sqrt(Hz) (if memeory serves) which betters the NE5532 by
> quite a bit.
Quite a bit? Not in any practical audio circuit that has low gain ( were
not talking about a mic pre amp with a 200 ohm source resistance ), 2.1 nV/
Sqr ( Hz ) is the amount of thermal noise a 266 ohm resistor will produce.
With the exception of a mic pre amp the op amps input noise will be
insignificant in comparison to the noise that will be generated by the
resistors in the circuit.
> > Generally speaking, drop-in replacement should be done with great
> > care, especially since the compensation used for the existing op-amp
> > may not be known.
>
> What type of Compensation are you eluding to? The 5532 is internaly
> compensated. So is the LM6172. Both are unity Gain stable.
This is where a lot of the " substitute parts people " get stung,
they fail to realize that the PC board is part of the circuit. It has
inductance, capacitance and resistance, " Unity Gain Stable " refers to the
part only, it does not guarentee stability under all conditions. A circuit
that is stable with a 5532 may not be with a video amp due to the PC boards
inductance, capacitance and resistance, which becomes more significant as the
frequency goes up. At low frequencies current flows through the path of
lowest resistance, at high frequencies it flows through the path of lowest
inductance! I would suggest that the PC board has a lot more to do with the
way a circuit sounds than does the parts used.
Thor...@tnt-audio.com
> > Yes, BUT IT SOUNDS not good, amongst other things.
>
> To quote you " says who "?
In that case I say, about 30-40 people which I know personally and many more
whose comments I read.
> > Are they. Could you please provide conclusive proof before making these
> > statements?
>
> I can, " Microelectronics Circuits " 3rd edition, Sedra and Smith,
> Library of Congress Catalog no. 0-03-051648-X. Published by Saunders
> College Publishing, a standard EE text. ( Any College EE text will provide this > info ) See section 2:7, " Non Ideal Performance of Op Amps " pg 76 - 82.
Now agreed, my own EE is a few years old (about 15) and in recent years I
have had little practice, but still, I don't think what this book says in
much different from what I was thought.
Even 15 Years ago there was a wide (aknowledged) gap between common theory,
measurements and percieved audible quality.
> Your theories could be easily disproved by any second year EE student
> in any reasonably equipped lab.
Do the white noise burst test in a reasonably equipped lab and report back
please. The test is so old that I was thought about it in my course. It
should pose no problems to your abilities. I did it back then on the
uA709,uA741 and TL081 to find what ended up in the EQ Circuitry for my
Mixing-Desk, I sayed discrete for all the hassle it did.
I was back then working in E-Germany for a small Company making specialised
recording Consoles for Radio and TV.
> The burden of proof of course lies with you, if someone declared to
> have discovered a planet beyond Pluto would you expect that person to
> provide proof or would it be incumbent upon everyone else to prove that
> it didn't exist?
I have said before, that I have a few theories and am currently not in a
position to evaluate them in a Lab that is well equipped enough to allow me
measurements that would proof or disproof these.
Hence theories and not facts.
> > Anyway, they are thoeries only and I have had no access to suitable
> > mesurement equipment (if such exists) to verify them.
>
> It does exist.
Actually for about half of the stuff does yes. The rest is a bit more
problematic.
> > But at least empirically (read in practice) the use of these opamp's
> > works and better than either AD797 or any other conventional
> > monolithic Op-Amp I've tried.
>
> This is of course a subjective judgment, putting aside possible placebo
> effects, would you consider that the excessive bandwidth is causing the
> circuit to ring, which you may personally find to be a desirable effect.
It MEASURABLY DOES NOT RING. Hey I do have an EE. I can make stable Circuits
at least as well as you do and I do measure my stuff.
But I have neither an AP-1 nor a realy good FFT Analyser. The stuff I've got
at hand is good enough to tell me if something works to a point, but not to
quantify the differences I hear. But with squarewaves and on a good 'scope,
no they don't ring, oscillate or misbehave otherwise.
My stuff can't measure THD below 0.1%, though luck. Would you like to do the
measurements for me, howzdat?
> This is only one of a multitude of possibilities, unless you can get down and
> dirty with the proper tools you will never know.
Probably, but then, I do know how a ringing circuit sounds.
> >And more importantly does this (steady state) measurement have any relevance?
>
> Would you believe me if I said yes?
I believe that you believe that.
I do not say that the AD797 is a bad Op-Amp (indeed there are lot's of places
where it has applications, some even in Audio), but still steady state THD is
a comparably irelevant measurement. How about settling time?
It tells you when the actual loop has corrected the errors inherent to the
Op-Amp.
Now the AD797 looks real good here. The 5532 and a few others look bad. Most
Video-OpAmps excel, due to higher inherent linearity combined with lower OLG.
But that all is 1st year EE.
> > The LM6182's data is similar in all respects but for Noise, the LM6182
> > has 2.1nV/sqrt(Hz) (if memeory serves) which betters the NE5532 by
> > quite a bit.
>
> Quite a bit?
6db. My neigbours would say it's quite a bit when I turn my stereo up by that
amount.
> Not in any practical audio circuit that has low gain ( were not talking
> about a mic pre amp with a 200 ohm source resistance )
Who says we where not - yes the context was a DAT output stage.
I raised this specifically as the other poster implied that the Input Noise
of my suggested Replacement Op-Amps was inferior to a 5532.
I agree in almost any Circuit for Audio, any of these is good enough for
noise - considerations. Other problems are much more relevant.
> > What type of Compensation are you eluding to? The 5532 is internaly
> > compensated. So is the LM6172. Both are unity Gain stable.
>
> This is where a lot of the " substitute parts people " get stung,
> they fail to realize that the PC board is part of the circuit. It has
> inductance, capacitance and resistance, " Unity Gain Stable " refers to the
> part only, it does not guarentee stability under all conditions.
Again, I do know that.
> A circuit that is stable with a 5532 may not be with a video amp due to the
> PC boards inductance, capacitance and resistance, which becomes more
> significant as the frequency goes up.
Agreed to a point, I had one case where I ever had problems.
And there I expected them. The PSU was not even following the recommendations
for the 5532 with regards to layout and decoupling.
> I would suggest that the PC board has a lot more to do with the
> way a circuit sounds than does the parts used.
I would not dispute that in a slightly more moderated form. In the end the
whole design needs to be evaluated (and often amended) to achieve the best
results. At the same time, in practice I have not found many stability
problems when using the LM6172.
The LM6182 is a bit more difficult, specifically due to the fact the any
parasitic Capacitance or Inductance in the NFB loop can make it go crazy. But
when implemented correctly it (and the AD811) sound best of all Op-Amps to my
Ears (and a few other peoples too).
As for PCB's well I mostly hardwire the stuff these days with plenty of SMD
components when I mess about with Solid State.
Kind regards Thorsten.
======================================
e-mail:
Thor...@tnt-audio.com
Visit TNT-Audio on the Web - the only advertising
free audio web-zine.
http://www.tnt-audio.com
======================================
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
Arny Kr|ger
Thor...@tnt-audio.com wrote in message <6dcirf$802$1...@agate.berkeley.edu>...
>Hi,
>> The 5532 actually measures very well in terms of distortion and noise,
>> amongst other things.
>Yes, BUT IT SOUNDS not good, amongst other things.
Actually, it sounds fine to me and literally 10,000's of others.
However, I can imagine and may have built circuits where the 5532 is
not a good choice (High impedance), and this might even be audible
sometimes.
I would speculate that you base your opinion on an open evaluation
where your demonstrated prejudices about this device were able to
contaminate the evidence you developed and lead to a pretty unique
conclusion.
Furhtermore, if you did your tests with commerical program material,
it is very possible that program material passed through any number
of 5532's because they are so commonly used in studio and production
equipment.
>Once people shut their minds and present dogmas I simply go away.
Your posts present what many of us feel to be dogma. Does that mean
we should go away?
>I can do without long flame treads. This does not mean I concede defeat. I
>just have better things to do than tit 4 tat arguing.
I think that you ought to set up a reliable listening test, one in
which your eyes and prejudices are not controlling the results, and
prove your point; first to yourself, and then to the rest of us.
>Anyway, they are thoeries only and I have had no access to suitable
>mesurement equipment (if such exists) to verify them.
Actually, just about anybody with a Pentium PC, and a few hundred
dollars for a good sound card and some software can make audio
measurements that are at or well beyond the limits of what can be
heard reliably.
>But at least empirically (read in practice) the use of these opamp's works
>and better than either AD797 or any other conventional monolithic Op-Amp
>I've tried.
By what criteria that is reliably audible?
>> Your theory is that enormous open-loop bandwidth is required to
>> accurately reproduce audio signals.
>Yup, that's it in a nutshell. Note that high inherent linearity features
>too.
Any modern audio op amp profoundly beats the human ear's ability to
detect nonlinear and linear distortion when applied with even
mediocre levels of care, in almost every audio application.
>> Yet the AD797 and OPA132, both of which have THD 120dB below 1V, have
>> sub-100 Hz open loop bandwidths.
>Yes, but sub 120db THD at what frequencies? And more importantly does this
>(steady state) measurement have any relevance?
Ah, the old "I am ignorant of the meaning of the Fourier and Laplace
transforms and I think that steady state and transient performance
exist in total isolation" approach. ;-(
>How about dynamic measurements with white noise bursts?
A very hard row to hoe. However, if you like to test with things that
are not pure steady state tones, try working with complex multitones.
I do and find them very helpful.
>Listen to the different canidates before you say the Bandwidth is not
>needed.
I very seriously suspect that you look at the alternatives and know
which equipment is on, while you are listening and therefore we can't
tell if your opinions are based on your prejidice or actual reliable
listening.
Sam
gfre...@northnet.org wrote in message
<6deefb$n...@jamesv.Warren.MENTORG.COM>...
>>In his Line-Preamps Jeff Rowland uses the AD811, a current-feedback design.
About 8 years ago I added an 8 pin dip socket to the simple final
inverting analog output buffer of a CD player to determine if I could
hear a difference between op-amps. I would play the same song then
swap op-amps then play the same song again. I used several brands of
then current op-amps: the 5532 (National, JRC?), the LM833? NE5535,
TI TL0XX series.
It was very easy to hear each op-amp's affect upon the sound quality.
All were lacking in there own ways. The sound was certainly mid-fi
from all. They compressed the timing and slam and inserted their own
'personality' into the music. Looking back at how bad these old
designs were, perhaps they should be re-classified as signal
processors! Overall, I liked the LM833? for its smoothness.
I currently use the AD826. Maybe its time to upgrade again to the
lastest generation.
The above is my opinion of a simple experiment from which I learned a
lot.
Enjoy the music.
gfre...@northnet.org
In article <6df10p$q...@jamesv.Warren.MENTORG.COM>,
Thor...@tnt-audio.com wrote:
Thorsten said on 1998/03/01, rec.audio.high-end
" I do have my own theories why,..... "
Then it is time to revisit them lest they get lost in the noise.
On rec.audio.tech 1998/01/18, Thorsten stated:
"..... the Open-Loop bandwidth of the Amplifier MUST be at least one
octave more extended than the Bandwidth of the signal to be amplified. "
No, by definition the " unity gain bandwidth ", " small signal gain
bandwidth " ect. ect., is the point where the amplifier gain decreases to 1,
at any frequency below that the amplifier has a gain greater than 1. If you
look at any data sheet that specifies the " unity gain bandwidth " and also
provides a graph of the open loop gain vs.. frequency, you will see that the
" unity gain bandwidth " is the point on the graph where the gain becomes 1
volt / volt, or 0 dB. In practice the " unity gain bandwidth " should be much
larger than bandwidth of the signal of interest, one octave is not adequate.
Thorsten then states in the same post:
" Most Audio-Opamps are by my rules only good for signals of up to 100-200Hz
"
By Thorsten's rules this means the " unity gain bandwidth " of " Most
Audio-Opamps " is 200 to 400 Hz.
Thorsten states on 1998/02/26, rec.audio.high-end
" The NE5532 simply stinks. Only the 741/1458 is worse."
Examining the data sheets of the NE5532 reveals that it has a unity gain
bandwidth of 10 MHz, that's 21 octaves above what Thorsten says is required.
Even the 741 with a unity gain bandwidth of 1.5 MHz, is more than 11 octaves
higher than his requirement.
So what is going on here? Thorsten appears to think the 3-dB or " break "
frequency is the unity gain bandwidth, it is not! For further proof of the
definition of " unity gain bandwidth and " break " frequency "see "
Microelectronic Circuits " Sedra and Smith, Copyright 1991 by Saunders
College Publishing, pg. 78; all further references to this text will be
denoted by " (1) ".
So how much bandwidth is required? The transfer function of an amplifier
using feedback is: ( ref (1) pg.572 )
A(f) = A/ (1+AB)
Where:
A(f) is the amplifier gain with feedback
A is the open loop gain in volts / volt
AB is called the loop gain ( the B is really the Greek letter beta )
and for a simple amplifier with: ( ref (1) pg 81 )
R1 connected from the negative input to ground
R2 connected from the output to the negative input
B= R1 / ( R1 + R2 )
We can now analyze the maligned 5532 and determine the amount of error
that is introduced due to it bandwidth limitations. To achieve a gain of 1,
we allow R1 to approach infinity and R2 approach zero, we therefore see that
B approaches 1. The gain ( A ) of the 5532 at 100 kHz is about 45 dB or about
180 v / v, plugging into our first equation we arrive at a gain of .994475 or
-.048 dB. Now for the LM6172 at 100 kHz the gain is about 55 dB or 600 v / v,
again plugging into the first equation we arrive at .998336 or -.014 dB and a
difference of .034 dB. As we go down in frequency the open loop gain of the
amplifier goes up and the gain error is reduced. At 20 kHz for the 5532 the
gain error is .9996668 v / v or -.0029 dB. The 6172 on the other hand has a
open loop gain of 82 dB before the " break " frequency resulting in a gain
error of .00087 dB only .002 dB difference between the two amplifiers. If we
look at the AD 797 at 20 kHz we find that it too has 82 dB of open loop gain
at 20 kHz. The AD 797's maximum gain is about 130 dB while the LM 6172 is
only 82, the 6172 will actually have more gain error over the audio
bandwidth! Making claims that the audio op amps do not have enough open loop
gain is a bit absurd. Feel free to run the numbers on different gains, I'm
just too tired to do it.
It should also be noted that at some point the signal has to pass
through a power amplifier which is just a big op amp and will simply act as a
low pass filter anyway.
Some other statements of interest were:
>...........that I have a few theories and am currently not in a
>position to evaluate them in a Lab that is well equipped enough to allow me
>measurements that would proof or disproof these.
I would suggest you learn what is and has been accepted theory first.
Referring to the text book I used as a reference Thorsten said:
>I don't think what this book says in much different from what I was thought.
I believe you mean taught, I fail to see any evidence you ever
learned any of it. If you wish to propose a theory it would be helpful to
specify some of the mechanisms that you believe are at play, subjective
test's are worthless in any field and should be avoided.
And concerning the THD from the semiconductors data sheet:
>And more importantly does this (steady state) measurement have any relevance?
Knowing at the time it was a loaded question I decided answer this one with a
question.
" Would you believe me if I said yes?"
Followed by Thorsten's response
>I believe that you believe that.......but still steady state THD is
>a comparably irelevant measurement.
This answer leads me to believe that he thinks this is the same bogus THD
measurement that the audio equipment manufacturers use, It is not! So
Thorsten could you tell me how they are different?
He further touched on:
>How about settling time? It tells you when the actual loop has corrected the
errors >inherent to the Op-Amp.
The data sheets for the 5532 don't even specify it, at a quick glance
the test circuits are not even the same if they exist at all. Start adding
capacitance to the output and watch the settling time go up.
>Video-OpAmps excel, due to higher inherent linearity combined with lower OLG.
>But that all is 1st year EE.
And I thought first year was Thevenin's and Norton's theorems, forcing
functions, complex frequency, state variable, ect. ect. ect.
some more from Thorsten:
>>>the LM6182 has 2.1nV/sqrt(Hz) (if memeory serves) which betters the NE5532
by
>> > quite a bit..............6db.
If you would please provide the circuit configuration and values of
said circuit, include the source resistance of the input source. As I stated
before the noise from the resistors in the circuit will reduce this
difference considerably, here is your chance to prove me wrong.
And more:
> Hey I do have an EE.
Excuse me if I find that hard to accept.
> I can make stable Circuits at least as well as you do ......
Excuse me if I find that hard to believe.
>My stuff can't measure THD below 0.1%, though luck. Would you like to do the
>measurements for me, howzdat?
Sure, $150 / hour, 4 hour minimum.
Greg
cha...@hotmail.com
Jan N Rasmussen <ardal...@vestdata.no> wrote:
>I have a pair of Egglestone Works "Andra" speakers. Stereophile have
>rated this speaker as the speaker of the year.
Can you tell us about the one component you left out - your listening room
(size, walls, floor, speaker position)?
Richard Chaife
gfre...@northnet.org
In article <6djsqi$3...@news01.aud.alcatel.com>,
"Sam" <pau...@flash.net> wrote:
>
> gfre...@northnet.org wrote in message
> <6deefb$n...@jamesv.Warren.MENTORG.COM>...
> >In article <6dcj11$812$1...@agate.berkeley.edu>,
> > Thor...@tnt-audio.com wrote:
>
> About 8 years ago I added an 8 pin dip socket to the simple final
> inverting analog output buffer of a CD player to determine if I could
> hear a difference between op-amps.
As far back as I can remember the use of sockets for audio
circuits has been a no no!
>I used several brands of then current op-amps: the 5532 (National, JRC?),
>the LM833? NE5535,TI TL0XX series.
I am familiar with all but the NE5535, I do have a reference to
it in a old Signetics data book but no data sheet, perhaps it is
discontinued.
> It was very easy to hear each op-amp's affect upon the sound quality.
> All were lacking in there own ways.
There are several problems with your comparison, a circuit
designed for one op amp will not necessarily work for another. In the
case of the 5532 vs. the LM833 or the TL0XX it is more than likely.
The 5532 input impedance can be as low as 30 k ohm, to avoid
having the inputs load the circuit excessively neither input should
see an impedance greater than about 3 k ohms. ( Although the input
impedance is typically 300 k ohms it is poor engineering practice to
use that number as a design parameter unless you are willing to test
for it ). All the resistors will usually be smaller when using the
5532 than would be found in the same circuit using the other op amps
you sighted. This will in many cases, but not all, require the output
current drive capability to be larger, which is in fact the case with
the 5532 ( it will drive 10 volts RMS. into 600 ohms ). Some of the
effects of substituting a 5532 in a circuit designed for a TL0XX or
LM833 are significant change of the gain, excessive offset that causes
the signal to clip prematurely at one of the supply rails or, high
frequency instability and oscillation.
Consider the reverse case of substituting a TL0XX into a circuit
designed for a NE5532. The input impedance of the TL0XX is on the
order of 10^ 12 ohms, a fair bit higher than the 5532, so as far as
the input impedance there should be little concern. The problem lies
in the output drive capability, the TL0XX series are designed to drive
loads greater than 2 k ohms and preferably they should be greater than
10 k ohm to avoid any loss of headroom. A circuit designed for a 5532
is likely to have impedance levels considerably lower than 2 k ohm.
The LM833 is the middle ground here, the data sheet does not
specify the input impedance, I would guess that it is in the Mega ohm
range which is typical of bipolar inputs. The output drive capability
LM833 is plotted with a 2k load and is slightly better than the TL0XX.
The DAC output, being a current source, determines the allowable
values for the resistors to a fairly narrow range. This in turn places
restrictions on the choice of op amps for the circuit and appears to
be a question of which op amp works the best in this particular
circuit.
Valvano
Those speakers are extremely room sensitive. If you don't have them set
up properly you will not get their optimum sound. It is also foolish to
buy a component based solely on a Stereophile rating. You can use a
rating as a guide but if you don't listen and audition a component
carefully you will more than likely be disappointed by basing your
purchase on someone elses ears.
John V.
Arny Kr|ger
Sam wrote in message <6djsqi$3...@news01.aud.alcatel.com>...
>gfre...@northnet.org wrote in message
><6deefb$n...@jamesv.Warren.MENTORG.COM>...
>>In article <6dcj11$812$1...@agate.berkeley.edu>,
>> Thor...@tnt-audio.com wrote:
>>>In his Line-Preamps Jeff Rowland uses the AD811, a current-feedback design.
>About 8 years ago I added an 8 pin dip socket to the simple final
>inverting analog output buffer of a CD player to determine if I could
>hear a difference between op-amps. I would play the same song then
>swap op-amps then play the same song again. I used several brands of
>then current op-amps: the 5532 (National, JRC?), the LM833? NE5535,
>TI TL0XX series.
>It was very easy to hear each op-amp's affect upon the sound quality.
>All were lacking in there own ways. The sound was certainly mid-fi
>from all. They compressed the timing and slam and inserted their own
>'personality' into the music. Looking back at how bad these old
>designs were, perhaps they should be re-classified as signal
>processors! Overall, I liked the LM833? for its smoothness.
In other words what you did is plug in an op amp, do some listening,
plug in another, do some listening, and so on. Please consider the
possibility that your reactions to what you heard were contaminated
by your preconceived notions about what each chip "should" sound
like. This is a well known effect called the "Placebo Effect", and if
you did the same kind of "experiment" in other contexts, and
presented your results to a group of peers, you would be laughed out
of the room.
I've been party to a different experiment. The purpose of the
experiment was to determine and demonstrate the audibility of phase
shift. Two sets (i.e., stereo) of 20 1 KHz all-pass filters, each
composed of a TL07x type op amp, were built, connected in series. A
switch was used to select taps on the series of all-pass filters.
We performed measurements on this device and it performed as
expected, including adding noise and distortion and small frequency
response variations. However, frequency response remained within 0.1
dB, SNR was at least 80 dB, and THD was less than 0.1 %. Under blind
conditions, "hearing" up to 10 stages of this device with musical
program material was very difficult or impossible. When the device
was "heard" the impression that led to reliable detection was clearly
of the phase shift - it made things sound kinda "sproingy".
This device still exists and every once in a while we drag it out to
see if modern recordings and other associated components have made a
difference. Not yet. This experiment has been peer reviewed, and
there has been no laughter! ;-)
Let us contrast this peer-reviewed test where multiple stages of
really not-the-best op amp were not "heard", and your evaluation
where one stage of a really-pretty-good op amp was "heard". I notice
one significant difference - your test involved both sight and sound,
and ours involved only sound.
Thor...@tnt-audio.com
Hi,
> >> The 5532 actually measures very well in terms of distortion and noise,
> >> amongst other things.
> >Yes, BUT IT SOUNDS not good, amongst other things.
> Actually, it sounds fine to me and literally 10,000's of others.
Yes. So you and all these people say.
Only because the Pro-Audio Industry uses them does not make them
good. A few people whose recordings I respect as some of the very
best sounding classical recordings currently made make a point not to
use anything of the current generation "pro-audio" gear with 5532's
and such all over the place.
Many prefer either various "extreme" discrete solid-state units while
others had the Tim de Paravicini Valve chip fitted. In either case
the results speake for themselfes.
I can reliably demonstrate the audible difference. Care to come
around and listen? The diference from NE5532 (or in my case the Super
NE5532 - NJN 2114) to a LM6172 is not small. And it is a very
positive one.
> However, I can imagine and may have built circuits where the 5532 is
> not a good choice (High impedance), and this might even be audible
> sometimes.
I can imagine any number of circuits where it's not good. Most Audio
Circuits. I won't go into much detail, but a bit food for thought.
Negative Feedback will attempt to correct an error once it has
happened.
The speed in which the circuit can reliably reduce the error to a
acceptable minimum (say 0.1% - is 0.1% Harmonic Distortion okay with
you?) will determine how much junk these "Audio-Opamp's" produce.
> I would speculate that you base your opinion on an open evaluation
> where your demonstrated prejudices about this device were able to
> contaminate the evidence you developed and lead to a pretty unique
> conclusion.
Doubtfull. I usually crosscheck what I hear by testing it on other
people who do not know what has been done.
> Furhtermore, if you did your tests with commerical program material,
> it is very possible that program material passed through any number
> of 5532's because they are so commonly used in studio and production
> equipment.
Most of my programme Material which I care to listen to has not. One
the reasons for the very poor sound of many current recordings is all
that junk in the signal chain. But then in as large, the Music
Industry is non too worried either about soundquality or indded
artistical quality. All they care is if it sells.
That's why we have got the Spice-Girls.
> >Once people shut their minds and present dogmas I simply go away.
>
> Your posts present what many of us feel to be dogma. Does that mean
> we should go away?
Maybe.
> >I can do without long flame treads. This does not mean I concede defeat. I
> >just have better things to do than tit 4 tat arguing.
>
> I think that you ought to set up a reliable listening test, one in
> which your eyes and prejudices are not controlling the results, and
> prove your point; first to yourself, and then to the rest of us.
I have offered that before. Just come around. I have also just this
weekend done such a test (stock CD-Player against modified - blind)
with a reliably detected difference which most listeners felt was a
very notable improvement.
> Actually, just about anybody with a Pentium PC, and a few hundred
> dollars for a good sound card and some software can make audio
> measurements that are at or well beyond the limits of what can be
> heard reliably.
Sorry, I care to dispute that. But you won't believe that either of
course.
My hearing has a dynamic range of at least 120 to 130db. Where do I
get a Soundcard (or indeed a AD-Chip) which will go anywhere close?
> By what criteria that is reliably audible?
Various distortion artifacts specifically on human voices and cymbals
on excelent minimally miked recordings.
> Any modern audio op amp profoundly beats the human ear's ability to
> detect nonlinear and linear distortion when applied with even
> mediocre levels of care, in almost every audio application.
Does it?
> Ah, the old "I am ignorant of the meaning of the Fourier and Laplace
> transforms and I think that steady state and transient performance
> exist in total isolation" approach. ;-(
No, not realy. But a single sinewave does not remotly approach any
relevance with regards to the behaviour of the circuit when faced
with signals having complex harmonic structures. Try a Cymbal. It's
almost a white noise burst.
> I very seriously suspect that you look at the alternatives and know
> which equipment is on, while you are listening and therefore we can't
> tell if your opinions are based on your prejidice or actual reliable
> listening.
Indeed, so do I. But my test listeners?
Later Thorsten
Thor...@tnt-audio.com
Hi,
> Thorsten said on 1998/03/01, rec.audio.high-end
> " I do have my own theories why,..... "
>
> Then it is time to revisit them lest they get lost in the noise.
>
> On rec.audio.tech 1998/01/18, Thorsten stated:
> "..... the Open-Loop bandwidth of the Amplifier MUST be at least one
> octave more extended than the Bandwidth of the signal to be amplified. "
>
> No, by definition the " unity gain bandwidth ", " small signal gain
> bandwidth " ect. ect., is the point where the amplifier gain decreases to 1,
Yes. But I do not care for a penny about the UNITY GAIN Bandwidth. I
said Bandwidth, meaning as in Audio. That is the -3db Bandwidth my
honored collegue.
> at any frequency below that the amplifier has a gain greater than 1. If you
> look at any data sheet that specifies the " unity gain bandwidth " and also
> provides a graph of the open loop gain vs.. frequency, you will see that the
> " unity gain bandwidth " is the point on the graph where the gain becomes 1
> volt / volt, or 0 dB. In practice the " unity gain bandwidth " should be much
> larger than bandwidth of the signal of interest, one octave is not adequate.
Of course.
> Thorsten then states in the same post:
> " Most Audio-Opamps are by my rules only good for signals of up to 100-200Hz
> "
>
> By Thorsten's rules this means the " unity gain bandwidth " of " Most
> Audio-Opamps " is 200 to 400 Hz.
So my dear collegue. I must consider that you made your assumptions as
to what bandwidth means and decided to disragard what I actually
wrote. Unity Gain Bandwith? Yes, far shot.
Most Audio Op-Amp's have a -3db Gain-Bandwidth of a few 100Hz. And
only this of any interest to me. Whatever happens above that - I don't
much care.
Do you really think I cannot read either a spec-sheet or a bode-plot?
> So what is going on here? Thorsten appears to think the 3-dB or " break "
> frequency is the unity gain bandwidth, it is not! For further proof of the
> definition of " unity gain bandwidth and " break " frequency "see "
> Microelectronic Circuits " Sedra and Smith, Copyright 1991 by Saunders
> College Publishing, pg. 78; all further references to this text will be
> denoted by " (1) ".
Well, to further this, plaese show me where I use the term Unity Gain
Bandwidth. Indeed, I should have specified that I am talking about the
-3 db point in the Gain or the -3db Bandwidth. In general I infer this
from the basic point of refference being AUDIO.
> the 6172 will actually have more gain error over the audio bandwidth!
Indeed. The LM6172 has much less excess gain available than the AD797
or NE5532 or indeed any Audio-Op-Amp.
At the same time the actual measured THD is similar (-110db). So how
is this possible? The Designer of the LM6172 actually traded inherent
linearity and speed for gain. Interstingly it also sounds better.
> Making claims that the audio op amps do not have enough open loop
> gain is a bit absurd. Feel free to run the numbers on different gains, I'm
> just too tired to do it.
It is not the numbers. Textbooks. I say again, do a test with a
suitable signal. I may even go along with a mulitone (but not a
2-tone) signal. Bandwidth limited noise-bursts are better. Observe
how much each device adds.
I only myself did this over 12 years ago with the uA741, uA709 and
TL080 and 081 as well as a few discrete circuits.
The results where both interesting and enlightning. I did not consider
any of the integrated Devices fit for Audio-Duties after that. But
then I had access to a propper lab back then.
> It should also be noted that at some point the signal has to pass
> through a power amplifier which is just a big op amp and will simply act as a
> low pass filter anyway.
It will indeed. The same Bandwidth Axinome is true there too. It
applies even to Zero-Nfb Circuits of any type. It alplies anywhere in
Audio.
> I would suggest you learn what is and has been accepted theory first.
Learned that ages ago. Found it insufficent back then. If existing
theory is insufficent to explain reality, the theories are either
- incomplete
- based on wrong basic assumptions
- or simply false
> I believe you mean taught,
Indeed. Excuse my rather insufficent knowledge of English. I have not
been taught much English in school behind the Iron Curtain. Care to
switch to russian?
> This answer leads me to believe that he thinks this is the same bogus THD
> measurement that the audio equipment manufacturers use, It is not! So
> Thorsten could you tell me how they are different?
I have not specifically looked at recent standards and specifications,
but I note that:
a) the concrete conditions are specified (load, frequency, voltage)
b) measurements are usually undertaken near the limits of the
performance envelope of the Circuits.
I make the assumption, that any all spuriae may it be noise or
harmonic distortion is taken in account.
At the same time, I do not find notes that tests with bandwidt limited
noise-bursts where made. If so, tell me.
> He further touched on:
> >How about settling time? It tells you when the actual loop has corrected the
> errors inherent to the Op-Amp.
>
> The data sheets for the 5532 don't even specify it, at a quick glance
> the test circuits are not even the same if they exist at all. Start adding
> capacitance to the output and watch the settling time go up.
Correct. Read the datasheets. Put a 100 Ohm Resistor in series. Then
you isolate this.
And indeed the 5532 does not specify Setteling Time. I wonder why?
Measure it, just for fun.
> some more from Thorsten:
> >>>the LM6182 has 2.1nV/sqrt(Hz) (if memeory serves) which betters the NE5532
> by
> >> > quite a bit..............6db.
>
> If you would please provide the circuit configuration and values of
> said circuit, include the source resistance of the input source.
Can you read a datasheet? You seem to have it handy anyway.
> As I stated
> before the noise from the resistors in the circuit will reduce this
> difference considerably, here is your chance to prove me wrong.
I never disputed that. But who say that I'm not using 100 Ohm Source
Resistances?
> > Hey I do have an EE.
>
> Excuse me if I find that hard to accept.
Fine.
> > I can make stable Circuits at least as well as you do ......
>
> Excuse me if I find that hard to believe.
Fine.
> >My stuff can't measure THD below 0.1%, though luck. Would you like to do > >the measurements for me, howzdat?
>
> Sure, $150 / hour, 4 hour minimum.
Ahh, but surely you would gladly give your time freely to advance the
existing knowledge.
Later Thorsten
Tom Kite
On 5 Mar 1998 Thor...@tnt-audio.com wrote:
> I can imagine any number of circuits where it's not good. Most Audio
> Circuits. I won't go into much detail, but a bit food for thought.
> Negative Feedback will attempt to correct an error once it has
> happened.
Nonsense, Thorsten.
> The speed in which the circuit can reliably reduce the error to a
> acceptable minimum (say 0.1% - is 0.1% Harmonic Distortion okay with
> you?) will determine how much junk these "Audio-Opamp's" produce.
This is absolute poppycock. As if the distortion reduces in time as
the feedback network catches up! Where do you get this from?
> My hearing has a dynamic range of at least 120 to 130db. Where do I
> get a Soundcard (or indeed a AD-Chip) which will go anywhere close?
Absurd. As has been pointed out several times recently, the dynamic
range of a normal room/system is somewhere around 80 dB at best, which
is easily achieved by a decent soundcard.
Tom
peter kalinkov
gfre...@northnet.org wrote:
Greg,
Have you ever seen a real DAC circuit? All your comments on the
differences between the op-amps are really nitpicking. The current o/p
of the DAC will happily drive the i/p of any reasonable op-amp (or a
fixed value resistor as some have showed to be better) without
imposing any drastic values fot the feedback resistor. Moreover
commercial circuits don't get as adventurous as trying to drive 600ohm
loads at max o/p voltage. From the original post I was under the
impression that we are not even dealing directly with the DAC but only
with the harmless o/p buffer where impedance mismatches are even more
ridiculous. So lets face it: to some of us op-amps do sound very
different and surprise: the more expensive ones rutinely sound better.
Then again it is well known that we shouldn't be doing this outside of
a double blind listening test (as opposed to double deaf ) , it's just
more tricky to get the op-amp right in the socket when swapping,
ha,ha.
peter
Ben Gilsdorf
In article <6d2072$g...@news01.aud.alcatel.com>, Jan Willem Auer <janw...@condor-post.com> writes:
> The OPA 2134 is the best audio opamp from Burr-Brown, suitable for
> high-end audio applications.
> and output stage?
> Does it sound better than the OPA 2604?
> Thanks, Jan Willem Auer.
I have stayed out of this discussion till now but during a long
evaluation of current opamps in standard aplications (CD playback) we
found that the much maligned 5534 outperformed most of the currently
referenced opamps. The only parts which reliably performed better
were the LT1028 for singles and the LT1124 for duals. They are more
expensive that the generic parts, but they were consistantly more
defined with better imaging.
Ben
gfre...@northnet.org
In article <6dot1n$4...@jamesv.Warren.MENTORG.COM>,
pe...@iafrica.com (peter kalinkov) wrote:
>The current o/p of the DAC will happily drive the i/p of any reasonable op-
>amp (or a fixed value resistor as some have showed to be better) without
>imposing any drastic values fot the feedback resistor.
You really missed the point here, I never said anything about
the ability of the DAC to drive the the input. Since Sam did not
provide any details of the circuit it is very difficult to make a
reasonable assessment of the effects. Through one example I attempted
to point out that there are significant differences between the parts
which must be accounted for in the design. A reasonable analogy might
be what spark plug, results in the best performance for a particular
car, let's call it spark plug X. If we replace spark plug X with
spark plug Y and notice that Y does not perform as well as X are we
to conclude that X is therefore better than Y? NO, given a different
car you may find that spark plug Y is the better choice. The only
logical conclusion we can arrive at is that under these conditions,
with this particular car, that X is better.
>.... I was under the impression that we are not even dealing directly with
>the DAC but only with the harmless o/p buffer....
Again I will reiterate, to make a assessment of any part all
other variables must be accounted for.
>So lets face it: to some of us op-amps do sound very different and surprise
>the more expensive ones rutinely sound better.
And any conclusions drawn that routinely ignore the fact that the
op amps operate as part of a system is flawed.
>..... it's just more tricky to get the op-amp right in the socket when
>swapping,
It's even trickier to account for the parasitic inductive and
capacitive effects a socket adds to the circuit.
Greg
gfre...@northnet.org
>> No, by definition the " unity gain bandwidth ", " small signal gain
>> bandwidth " ect. ect., is the point where the amplifier gain decreases to
1,
Thorsten said :
>Yes. But I do not care for a penny about the UNITY GAIN Bandwidth. I said
>Bandwidth, meaning as in Audio.
No, you said:
On rec.audio.tech 1998/01/18, Thorsten stated:
"..... the Open-Loop bandwidth of...."
Perhaps I wasn't clear enough, they are by definition one in the
same! It follows the same convention as " open loop gain ", which is
the largest gain a amplifier has under open loop conditions. So it
goes that the " open loop bandwidth " is the largest bandwidth that
the amplifier is capable of amplifying under open loop conditions.
>............I must consider that you made your assumptions as to
>what bandwidth means .......
No, these are established conventions, see Sedra and Smith,
pg. 77, under " Non Ideal Performance of OP Amps ", subsection "
Finite Open-Loop Gain and Frequency ".
>.......Indeed, I should have specified that I am talking about the -3 dB
>point in the Gain or the -3 db Bandwidth. In general I infer this from the
>basic point of reference being AUDIO.
Which always applies to " closed loop amplifiers ", we are
talking about open loop.
>The LM6172 has much less excess gain available than the AD797 or NE5532 or
>indeed any Audio-Op-Amp. At the same time the actual measured THD is similar
B(110db).
It depends what you mean by similar, the AD797 is at least 10 dB
better. I say at least because that is as low as their automated test
equipment goes, it is actually lower, they simply will not print a
spec that cannot be tested! What makes this interesting is your post
on this NG (1998/03/05 ), you state:
" My hearing has a dynamic range of at least 120 to
130db. Where do I get a Soundcard (or indeed a AD-Chip) which will go
anywhere close? "
On To THD measurements made by the semiconductor manufacturers.
>Thorsten's response
>........but still steady state THD is a comparably irelevant measurement.
To which I responded:
>> This answer leads me to believe that he thinks this is the same bogus THD
>> measurement that the audio equipment manufacturers use, It is not! So
>> Thorsten could you tell me how they are different?
To which he responds:
>I have not specifically looked at recent standards and specifications,
but.....
So you don't know, do you?
To my suggestion:
>> I would suggest you learn what is and has been accepted theory
first.
>Thorsten responded
>Learned that ages ago. Found it insufficent back then. If existing theory is
insufficent >to explain reality, the theories are either.....
So far you have failed to provide proof that the theories are
insufficient, the largest problem for the IC manufacturer is
accounting for all parasitic elements in their SPICE models, not with
the theory. First you criticize THD measurements and then admit you
don't know how they are done. Now you tell us that " existing theory
is insufficient to explain reality ", which begs the question, do you
know what existing theory is?
Now we get on to settling time:
Two points to be made here, the 5532 does not specify settling
time, Thorsten makes a statement about how bad the 5532 is on settling
time and then admits that is not specified.
Thorsten said concerning settling time:
" Now the AD797 looks real good here. The 5532 and a few others look
bad. Most Video-OpAmps excel. " ( 1998/03/02, rec.audio.hi-end )
And Now:
>And indeed the 5532 does not specify Setteling Time. I wonder why?
At one point he asks:
> Do you really think I cannot read either a spec-sheet or a bode-plot?
Anybody can read a spec sheet, this does not mean that everybody
will understand what they read, and in this case it appears you did
neither.
Why isn't settling time specified, for the same reason video
amps don't specify 1/f noise, because it is not all that important
when you consider what function the amp was designed for!
Concerning the 5532 Thorsten said:
> Measure it, just for fun.
So guess what I did Friday? Just for fun I measured it, I used
the same test circuit as the Philips data sheet with one exception, I
did not load the output with a 100 pf cap ( None of the video amps
were loaded with a cap ). The schematic may be obtained from:
http://www.geocities.com/CapeCanaveral/Hangar/6027/schematic.gif
The input frequency is about 1 kHz and 3 v pk to pk.
See:
http://www.geocities.com/CapeCanaveral/Hangar/6027/tek0.gif
The view was expanded in the time domain to:
http://www.geocities.com/CapeCanaveral/Hangar/6027/tek1.gif
Next the scope zoom mode is used to expand the view of the waveform
overshoot, while the bottom half continues to display the whole
waveform.
http://www.geocities.com/CapeCanaveral/Hangar/6027/tek2.gif
The overshoot is about 80 MV and it takes about 215 ns from the
beginning of the overshoot to the point where it has settled.
The question remains, how significant is this overshoot? The
next step was to insert a first order low pass filter with a cutoff
frequency of about 1.6 MHz.
http://www.geocities.com/CapeCanaveral/Hangar/6027/tek3.gif
Obviously the harmonic content of the overshoot is far removed
from the from the audio band even with this somewhat crude setup!
Some more from Thorsten:
" the LM6182 has 2.1nV/sqrt(Hz) (if memeory serves) which
betters the NE5532 by quite a bit..............6db.".........and
"6db. My neigbours would say it's quite a bit when I turn my stereo up
by that amount."
As a side note it should be noted that the LM6182 has a noise
voltage of 4nV / root Hz and that Thorsten ignores totally the input
current noise. In the case of the LM6182, it is 3 and 16 pico amp /
root Hz for the + and - inputs respectively. We then calculate how
much noise is added on the negative input with the recommended 820 ohm
resistor, and arrive at a value of 10.7 nV / root Hz. In the NE5532
the current noise is .7 pA / root Hz, so for the feedback resistor to
generate a equivalent amount of noise would require a 15 k ohm
resistor ( a unnecessarily large value ). If the NE5532 is used in a
voltage follower configuration the resistor could be replaced with a
short making the " input current noise " contribution equal to zero.
I also made this request:
" If you would please provide the circuit configuration and
values of said circuit, include the source resistance of the input
source. As I stated before the noise from the resistors in the circuit
will reduce this difference considerably, here is your chance to prove
me wrong.
Now he says:
>I never disputed that. But who say that I'm not using 100 Ohm Source
Resistances?
I asked for the source resistance because it is one of the
contributors of noise, but not the only one! Judging from your failure
to understand that the " equivalent input noise voltage " is not the
only factor involved in determining the noise performance, I felt it
was necessary to make sure it was included. You made a technical
claim, now back it up!
Greg
Kazushi Endoh
> The OPA 2134 is the best audio opamp from Burr-Brown, suitable for
> high-end audio applications.
It seems all-mighty OPamp not yet made. I read OPA 134 has specially
low noise CURRENT density, so OPA 134 is suitable for mic-(pre)amp,
MC head amp, feed back type RIAA equalizer, and such high gain stage.
> Does it sound better than the OPA 2604?
It depends on application. As a line buffer, I guess OPA 2134 will
make little difference as compared to 2604.
> Does somebody know how it sounds compared to a standard NE 5532 which
> is being used in my Sony DTC 59 ES datrecorder as an active filter
> and output stage?
Desireble feature of devices are different among each stages.
For IV stage, first choice is high slew and quick settle, because
the signal contains HF image noise here.
LPF stage, I prefer FET input stage OPamp, since output impedance
of IV stage was not main concern.
Buffer stage, main concern of this stage is output impedance and
current. BUF 634, OPA 633, etc. are designed for buffer use. 5534
or OPA 604 also work fine, in addition these are't nervous.
I have three different choice here,
"I souldn't use high slew buffer such as OPA 633 to avoid HF noise"
"I'd like to use famous OPA 633 even with band limitation"
"I need high slew buffer for some reason, even if the band width
of interest is upto 100kHz, not 100MHz".
Mr. Nagae's (of DENON) choice was
IV: band limited AD 797 (the other candidate was AD 811)
LPF: OPA 627 (something for more transparent timbre than AD 817)
BUF: BUF 634 (he added high slew buffer for some reason)
gils...@adtaz.sps.mot.com (Ben Gilsdorf) wrote:
> The only parts which reliably performed better
>were the LT1028 for singles and the LT1124 for duals. They are more
>expensive that the generic parts, but they were consistantly more
>defined with better imaging.
I read LT1028 is more power hungry than others, and tend to be angry
with poor supply. Plug and play upgrade likely to fail with LT1028.
--
Kazushi Endoh Med.Univ.Yamanashi
ken...@res.yamanashi-med.ac.jp
Arny Kr|ger
Thor...@tnt-audio.com wrote in message
<6dmmfe$p...@news01.aud.alcatel.com>...
>I can imagine any number of circuits where it's not good. Most Audio
>Circuits. I won't go into much detail, but a bit food for thought.
>Negative Feedback will attempt to correct an error once it has
>happened.
Nope, the errors are corrected continuously as they happen.
>The speed in which the circuit can reliably reduce the error to a
>acceptable minimum (say 0.1% - is 0.1% Harmonic Distortion okay with
>you?) will determine how much junk these "Audio-Opamp's" produce.
That is not how feedback works. The errors are detected when they
happen. OK, since these op amps have gain 1-10 bandwidth in the
megahertz range, there is a lag in the nanosecond range between the
current input and the feedback output. However, we are talking Audio,
where 20 kHz takes 50 microseconds per wave. Because of the
nanosecond-level delay what might be perfect cancellation is
imperfect, but still good enough to get the distortion way below
audiblity. Furthermore, this nanosecond-level delay is for *ALL*
signals, both transient and steady-state, so any error that is not
corrected shows up in steady-state tests. Steady state and transient
testing shows that the distortion after correction by feedback is
way, way below audiblity for audio signals.
>> I would speculate that you base your opinion on an open evaluation
>> where your demonstrated prejudices about this device were able to
>> contaminate the evidence you developed and lead to a pretty unique
>> conclusion.
>Doubtfull. I usually crosscheck what I hear by testing it on other
>people who do not know what has been done.
So, you have shown yet another example of the Placebo effect as a
universal problem with listeners who are not protected from "The lust
of the eyes". The pervasiveness of the Placebo efffect has been
demonstrated over and over again, starting with landmark papers
published in 1953. It took from 1953 to 1969-70 for Audio engineers
to become fully informed about this. However, any serious subjective
test that is done today is done under controlled conditions. The only
hold-outs on this 45 or 21 year old technology are found in High End
Audio, AFAIK.
>> Furhtermore, if you did your tests with commerical program material,
>> it is very possible that program material passed through any number
>> of 5532's because they are so commonly used in studio and production
>> equipment.
>Most of my programme Material which I care to listen to has not. One
>the reasons for the very poor sound of many current recordings is all
>that junk in the signal chain.
I think that over-production is the problem, not good op amps being
not good enough.
>> I think that you ought to set up a reliable listening test, one in
>> which your eyes and prejudices are not controlling the results, and
>> prove your point; first to yourself, and then to the rest of us.
>I have offered that before. Just come around.
Well, if you are anywhere near the home of your domain, that would be
Italy, and I am in Michigan, USA.
>I have also just this
>weekend done such a test (stock CD-Player against modified - blind)
>with a reliably detected difference which most listeners felt was a
>very notable improvement.
Via private email, we talked about the characteristics of this test.
The test failed virtually every criteria for reliable listening
tests.
>> Actually, just about anybody with a Pentium PC, and a few hundred
>> dollars for a good sound card and some software can make audio
>> measurements that are at or well beyond the limits of what can be
>> heard reliably.
>Sorry, I care to dispute that.
Dispute as you will.
>But you won't believe that either of course.
Depends on how good your "story" is, whether I'll believe it.
>My hearing has a dynamic range of at least 120 to 130db. Where do I
>get a Soundcard (or indeed a AD-Chip) which will go anywhere close?
See www.pcavtech.com/soundcards/gina. Here is a card that tests out
with 120 dB+ dynamic range.
>> Any modern audio op amp profoundly beats the human ear's ability to
>> detect nonlinear and linear distortion when applied with even
>> mediocre levels of care, in almost every audio application.
>Does it?
You would like us to believe otherwise. Show some good evidence and I
will salute you!
>> Ah, the old "I am ignorant of the meaning of the Fourier and Laplace
>> transforms and I think that steady state and transient performance
>> exist in total isolation" approach. ;-(
>No, not realy. But a single sinewave does not remotly approach any
>relevance with regards to the behaviour of the circuit when faced
>with signals having complex harmonic structures. Try a Cymbal. It's
>almost a white noise burst.
If you want a seriously taxing signal, try jangling some keys on a
chain near a microphone. Makes Cymbals look like a cakewalk. Cymbals
were designed to produce as much sound as possible at the high end of
the audible range. It has been known for decades that jangling keys
are easily as tough - acutally far tougher - because they were not
designed to produce sound in the audible range - they just do what
they do and it rises at about 6 dB per octave way, way past 20 kHz.
The point of all this is that you were really savvy about taxing real
world sounds, you would have cited this. It's well known among pros.
But, and here is the rub, the only modern components seriously taxed
by jangling keys are microphones. Yup, even with some fairly heavy
stuff, you have to stand back when you jangle those keys because the
mic may very well clip internally. With cheap mics, you may have to
stand back several yards. With mics with no internal circuitry, or
those that roll off, the clipping action next moves to the mic amp,
and 20 dB pads are usually in order! Once you get the jangling keys
as a clean audio signal, most better equipment, including that
containing the "dreaded" 5532, do just fine.
>> I very seriously suspect that you look at the alternatives and know
>> which equipment is on, while you are listening and therefore we can't
>> tell if your opinions are based on your prejidice or actual reliable
>> listening.
>Indeed, so do I. But my test listeners?
So far, it has been clear that your listeners are working in a rich
field of non-audible (i.e., visual) clues about the identity of the
UUT's. Are you planning to do reliable tests that avoid this
well-known pothole in the road to audio truth?
gfre...@northnet.org
In article <6e6clr$t...@news01.aud.alcatel.com>,
ken...@res.yamanashi-med.ac.jp (Kazushi Endoh) wrote:
> In article <6d2072$g...@news01.aud.alcatel.com>,
> Jan Willem Auer <janw...@condor-post.com> writes:
> I read OPA 134 has specially low noise CURRENT density, so OPA 134 is
>suitable for mic-(pre)amp, MC head amp, feed back type RIAA equalizer, and
>such high gain stage.
It is more dependent on the source resistance, on a low
impedance mic, the noise with any reasonable design will be more
dependent on the " input noise voltage ". With most amplifiers
intended for this type of application the " current noise " can be
ignored without any significant effect on the total equivalent input
noise. With hi impedance sources such as a RIAA amplifier the
dominant noise source will be the source itself. Typical values of a
phono cartridge is 47 K ohms, so your starting with 28 nV / root Hz.
National Semiconductor has a application note or two on noise and how
to calculate it, I think it is called " Noise Specs Confusing " or
something like that. You can download it from their web site if you
can find it.
> For IV stage, first choice is high slew and quick settle, because
> the signal contains HF image noise here.
You want a fast amp but you also want a wide bandwidth, these
two parameters although related are not necessarily the same. The
LM833 has a slew rate of 7V/us and a open loop bandwidth of 10 MHz,
while the TL07X has a slew rate of 12V/us and a open loop bandwidth
of only 3MHz. The slew rate is to keep up with the DAC and the wide
bandwidth is to maintain gain accuracy.
> Buffer stage, main concern of this stage is output impedance and
> current. BUF 634, OPA 633, etc. are designed for buffer use.
They were designed for VIDEO, the BUF634 only specifies the
noise voltage at 10KHz. The 1/f noise (low frequency noise ) is not
usually specified for video amps. In audio amps the 1/f corner
frequency is only a few Hz, in video amps it can vary from hundreds
to thousands of Hz. The OPA633 specifies the noise voltage for 10 Hz
to 1 MHz, they do not provide the spectral distribution, so again
there is no way to know how much of it lies in the audio bandwidth.
> 5534 or OPA 604 also work fine, in addition these are't nervous.
The NE5534 makes me nervous, they are internally compensated
for gains of 3 or more. The word " buffer " usually implies a gain of
1 and if you want to operate the NE5534 at unity gain then you have
to add external compensation. The OPA 604 was designed for buffering
audio signals, probably the best choice of the lot.
Hartmut
gfre...@northnet.org wrote:
> noise. With hi impedance sources such as a RIAA amplifier the
> dominant noise source will be the source itself. Typical values of a
> phono cartridge is 47 K ohms, so your starting with 28 nV / root Hz.
This is definitely wrong, 47k is the load of the cart implemented in
the phono stage. But 47k is the noise source only when circuit is
open, that means without cartridge.
A typical moving magnet phono cart is app. 300 Ohms DC resistance, and
app. 1k impedance over the listening band. The Linn K5/9/18 and Audio
Technica moving magnet cartridges are examples for 1k impedance.
The input of the phono stage sees 47k in parallel with the source
impedance of 1k, which gives app. 1k again, so you start with 4nV/root
Hz. If you use a hi-output moving coil, you have 100 Ohms source
impedance and thus only app. 1.5nV/root Hz.
Moving coil cartridges have source impedances of 2 Ohms (like the
Ortofon and Lyra cartridges) over 30 Ohms (like a Denon103) up to 100
Ohms (Benz Glider hi-output, which should be connected to a moving
magnet stage with 47 input impedance).
Hartmut from Munich