384kHz PCM ???
exbeatle
Pyramix DSD 16 channel Features
*2, 8, or 16 Ch DSD Record/Editing/Mixing and Mastering System.
*DSD De-noise and De-click for Re-Mastering SACD.
*SONY and PHILIPS SACD authoring compliance to Scarlet Book
specification.
*DST Encoding.
*Multi-channel Punch-in Punch-out record capability.
*Integrated Stereo Monitoring.
*Superior DSD Noise Floor Technology.
*Up/Down conversion from/to 44.1 KHz, 88.2 KHz, 96 KHz, 192 KHz, 352
KHz, 384KHz and DSD.
352KHZ, 384KHZ -
!!when did this happen, and what do you do with a 24b/384kHz PCM
recording??
.
Jay - atldigi
exbeat...@hotmail.com (exbeatle) wrote:
Pyramix downsamples DSD's 2.8224 mhz sample rate to 384kHz PCM for
signal processing. This allows for processing without using the 8 bit
"DSD wide" Sony DSD native boards. You could also upsample 96k or 192k
PCM for processing at the higher rate if you wanted to. However, there
are no 384k converters (that I know of) available at the moment, and
there's certainly some question as to whether it would be a good idea to
do so.
--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com
Mike Rivers
> !!when did this happen, and what do you do with a 24b/384kHz PCM
> recording??
Listen to it on the same machine on which it was recorded.
Down-sample to 44.1 kHz and put it on a CD
Lossy-compress it and put it on a DVD-A
Lots of possibilities there.
--
I'm really Mike Rivers - (mri...@d-and-d.com)
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo
Roger W. Norman
news:atldigi-D2CB2D...@news1.news.adelphia.net...
> Pyramix downsamples DSD's 2.8224 mhz sample rate to 384kHz PCM for
> signal processing. This allows for processing without using the 8 bit
> "DSD wide" Sony DSD native boards. You could also upsample 96k or 192k
> PCM for processing at the higher rate if you wanted to. However, there
> are no 384k converters (that I know of) available at the moment, and
> there's certainly some question as to whether it would be a good idea to
> do so.
Certainly Roger Nichols thinks so. He seems to recommend upsampling, and in
terms of gaining finer resolution on EQ and such, it might be worth it.
Then again, without the wherewithal to actually hear what might be a
difference or not, it certainly doesn't make sense, does it?
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
Romeo Rondeau
how much better it is than 192k, if they can't, their ears suck, you then
put people down that still record at 48k and consider them to be
unprofessional. Any questions? :-)
Jay - atldigi
<rno...@starpower.net> wrote:
> > are no 384k converters (that I know of) available at the moment, and
> > there's certainly some question as to whether it would be a good idea
> > to do so.
>
> Certainly Roger Nichols thinks so. He seems to recommend upsampling, and
> in terms of gaining finer resolution on EQ and such, it might be worth it.
> Then again, without the wherewithal to actually hear what might be a
> difference or not, it certainly doesn't make sense, does it?
Upsampling is another issue. The point was whether 384 kHz A to D
conversion would be a worthwhile endeavor. Upsampling for processing can
be useful, though not universally. Dynamics processing may actually
benefit more than EQ. However, performing A to D at 384 likely offers no
benefit and according to some may actually have some unintended negative
consequences. We certainly don't have the data to say definitively at
the moment. This also is not the same question as oversampling which
does offer benefits and is in common use today.
As for Roger Nichols opinion, I haven't seen exactly what he has said so
I couldn't really comment specifically. When it comes to technical
issues dealing with conversion, I may be inclined to listen more closely
to Dan Lavry (Lavry engineering, formwely db technologies, widely
regarded as the finest AD/DA conversion available) than Roger Nochols.
If it was a question of production techniques, certainly Roger's advice
would hold more weight.
Dan is soon to release a white paper on the subject of >96k conversion
that is sure to cause a stir. His website is
http://www.lavryengineering.com/ and the papers show up under the
support section. The new paper isn't there yet, but keep an eye out.
When somebody as knowledgable as Dan offers opinions on conversion, it
causes one to stop and think.
Dave Collins
> Certainly Roger Nichols thinks so. He seems to recommend upsampling, and in
> terms of gaining finer resolution on EQ and such, it might be worth it.
Except linear processes like EQ are the least likely to benefit from
upsampling.... What does "finer resolution" mean, anyway? Like you can
get 1.00001kHz?
> Then again, without the wherewithal to actually hear what might be a
> difference or not, it certainly doesn't make sense, does it?
It does to Roger. And he writes an article every month.
DC
Arny Krueger
news:40bbea52.03111...@posting.google.com
If you're coding audio, 24/384 does nothing practical, but it does waste a
lot of storage space, time and processing power. Since 24/192 has zero sonic
advantages over 24/96, and the ear is more sensitive to improvements that
are closer to its core capabilities, by logical induction it is safe to
assume that 24/384 can do nothing for sound quality.
Given that digital storage and processing power remain quite extensible at
the 24/384 point, it's hard to imagine how far this madness will go.
Arny Krueger
news:dcollins-02E935...@news04.west.earthlink.net
> In article <bp6jv0$s60$1...@bob.news.rcn.net>,
> "Roger W. Norman" <rno...@starpower.net> wrote:
>
>> Certainly Roger Nichols thinks so. He seems to recommend
>> upsampling, and in terms of gaining finer resolution on EQ and such,
>> it might be worth it.
> Except linear processes like EQ are the least likely to benefit from
> upsampling....
People need to understand that upsampling only provides advantages in
certain very specific cases. Just because it makes it so much cheaper and
easier to have good converters doesn't mean that its going to improve
*everything*.
> What does "finer resolution" mean, anyway? Like you
> can get 1.00001kHz?
Ironically, you can get 1.00001kHz-type frequency resolution with ordinary
16/44. Getting that kind of frequency resolution is just a matter of having
a long enough sample. The sample size required to get that kind of
resolution in the frequency domain is the same in both the analog and
digital domains.
Roger W. Norman
> Upsampling is another issue. The point was whether 384 kHz A to D
> conversion would be a worthwhile endeavor. Upsampling for processing can
> be useful, though not universally.
True, but we weren't necessarily discussing 384 kHz A/D conversion, at least
in terms of the original post. In that case it was 384 kHz downsampling
from 2.8 mHz of DSD, which, when one wants to edit, the conversion has to go
to PCM and 384 kHz is probably a nice figure that allows editing without
losing the "flavor" of the DSD. Certainly I wouldn't want to downsample it
to 96 kHz. However, going UP the scale in A/D conversion, then obviously
even at the 96 kHz rate one has only technical aspects that suggests this is
even necessary, such as a filter rate that is eminently more smooth than a
somewhat brickwall anti-aliasing filter at 22.5 kHz.
Just as obvious, some have argued that even 96 kHz sampling is simply a
waste of bandwidth and others suggest that it's possible to hear a
difference. More than likely it wouldn't make a difference to my war-torn
ears, nor do I think, unless the space is absolutely pristine for recording,
that one gains much benefit from 96 kHz other than maybe some spacial
placement clues that one doesn't actually hear but more likely perceives.
However, you are correct, and I did interject a somewhat different slant on
the conversation. You most definitely said 384 kHz converters, and I jumped
it to doing upsampling, so even with the original idea of downsampling I
still missed the boat! <g>
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
> In article <bp6jv0$s60$1...@bob.news.rcn.net>, "Roger W. Norman"
> <rno...@starpower.net> wrote:
>
> > > are no 384k converters (that I know of) available at the moment, and
> > > there's certainly some question as to whether it would be a good idea
> > > to do so.
> >
> > Certainly Roger Nichols thinks so. He seems to recommend upsampling,
and
> > in terms of gaining finer resolution on EQ and such, it might be worth
it.
> > Then again, without the wherewithal to actually hear what might be a
> > difference or not, it certainly doesn't make sense, does it?
>
Roger W. Norman
> Except linear processes like EQ are the least likely to benefit from
> upsampling.... What does "finer resolution" mean, anyway? Like you can
> get 1.00001kHz?
Hi, Dave. You're right, and mostly I was referring to the original post of
downsampling DSD to PCM for editing, which makes 384 kHz seem to be a
reasonable sampling rate and one that would allow some of that finer
resolution in EQ and dynamics. I interjected the upsampling statement,
disregarding Jay's 384 kHz A/D converter statement and reading it instead as
an upsampling argument, whether for or against.
Long day, no time and so I missed the boat.
As for Roger's perception, for a lot of us the difference won't be
noticeable due to rooms, equipment, condition of our hearing, etc. I don't
dispute nor argue for Roger's perceptions, since I'm obviously not
qualified. I just mentioned them because he's made the case more than once.
I still haven't made any effort whatsoever to find out on my own. I leave
that to people who have great rooms, excellent equipment and ears that
haven't been assaulted by grenade concussions! <g> Not to mention such
lumenaries as yourself who's credentials suggest a level of knowledge and
expertise most of us won't achieve.
Just chalk it up to my normal way to screw up the conversation.
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
> In article <bp6jv0$s60$1...@bob.news.rcn.net>,
> "Roger W. Norman" <rno...@starpower.net> wrote:
>
> > Certainly Roger Nichols thinks so. He seems to recommend upsampling,
and in
> > terms of gaining finer resolution on EQ and such, it might be worth it.
>
>
Erik de Castro Lopo
>
> If you're coding audio, 24/384 does nothing practical, but it does waste a
> lot of storage space, time and processing power. Since 24/192 has zero sonic
> advantages over 24/96, and the ear is more sensitive to improvements that
> are closer to its core capabilities, by logical induction it is safe to
> assume that 24/384 can do nothing for sound quality.
>
> Given that digital storage and processing power remain quite extensible at
> the 24/384 point, it's hard to imagine how far this madness will go.
There is another wrinkle in this fabric.
The digital filters used for EQ and whatever else are far harder
to design when the targeted frequency is small in comparison to
the sample rate.
For instance, attempting to design a 4th order Butterworth high
pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
difficult in 32 bit floating point but it can be done in 64
bit double precision floating point.
Once you up the sample rate to 384kHz, that same high pass filter
at 20Hz is going to be difficult the design and implement eve in
64 bit doubles.
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nos...@mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
Traditional capital was stuck in a company's bank account or investments.
It could not walk away in disgust. Human capital has free will. It can
walk out the door; traditional capital cannot.
Jay - atldigi
<rno...@starpower.net> wrote:
> However, you are correct, and I did interject a somewhat different slant
> on the conversation. You most definitely said 384 kHz converters, and I
> jumped it to doing upsampling, so even with the original idea of
> downsampling I still missed the boat! <g>
No trouble - still on topic to the thread. I just wanted to clarify
since your post had me quoted but wasn't really in response to what I
said.
One last comment: I thought I might have heard that Pyramix will edit
DSD without downsampling and only needs to downsample if you want to do
signal processing (EQ, dynamics etc). I'm not really sure. It may
downsample DSD the moment it comes in and upsample it again on the way
out. Does anybody know for sure which way it works? All downsampled, or
just for processing?
Remixer
around in pure DSD but there is no cross-fade, level change, or any other
processing possible, unless you downsample to 8Fs, render, and bump back up
to DSD.
Roger W. Norman
> One last comment: I thought I might have heard that Pyramix will edit
> DSD without downsampling and only needs to downsample if you want to do
> signal processing (EQ, dynamics etc). I'm not really sure. It may
> downsample DSD the moment it comes in and upsample it again on the way
> out. Does anybody know for sure which way it works? All downsampled, or
> just for processing?
Simply cut and paste editing, yes. Geez, I'd hate to get down into sample
accurate edits on DSD! <g> But this type of editing for DSD should be fine
considering the type of market it seems to be garnering. I don't suppose we
need worry about an AC/DC DSD recording, but in terms of bluegrass, jazz and
classical, maybe even folk or what's today called Americana, where dynamics
are key components to a performance, it seems like a natural. As I recall,
however, that inflexibility is one of the reasons that DSD never established
itself as a real contender in the DVD-A standards fight, even though there
were some very strong proponents of the technology. I remember many a
conversation on the standards fight with Stuart Robinson back in the
Compuserve days, along with Michael Gerzon just before he passed away. Too
much math to be a fun conversation sometimes, though! <g>
Arny Krueger
news:3FB7DD18...@mega-nerd.com
> Arny Krueger wrote:
>>
>> If you're coding audio, 24/384 does nothing practical, but it does
>> waste a lot of storage space, time and processing power. Since
>> 24/192 has zero sonic advantages over 24/96, and the ear is more
>> sensitive to improvements that are closer to its core capabilities,
>> by logical induction it is safe to assume that 24/384 can do nothing
>> for sound quality.
>>
>> Given that digital storage and processing power remain quite
>> extensible at the 24/384 point, it's hard to imagine how far this
>> madness will go.
>
> There is another wrinkle in this fabric.
>
> The digital filters used for EQ and whatever else are far harder
> to design when the targeted frequency is small in comparison to
> the sample rate.
Not really.
> For instance, attempting to design a 4th order Butterworth high
> pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
> difficult in 32 bit floating point but it can be done in 64
> bit double precision floating point.
Not really. There are zillions of working counter-examples.
> Once you up the sample rate to 384kHz, that same high pass filter
> at 20Hz is going to be difficult the design and implement eve in
> 64 bit doubles.
Not at all. If this were right, programs that implement your so-called
difficult filters like Adobe Audition would be far slower and more complex
than they actually are.
Kurt Albershardt
Said 8fs being 352.8k in the case of DSD, rather than 384k BTW.
Kurt Albershardt
>
> there
> are no 384k converters (that I know of) available at the moment, and
> there's certainly some question as to whether it would be a good idea to
> do so.
Claude mentioned to me awhile back that he was expecting dCS to offer a
converter with both DSD and 352.8/24 PCM outputs (using similarly gentle
filter slopes on the PCM side.)
Jay - atldigi
<ku...@nv.net> wrote:
The most exciting part of this is the gentle filter slopes, not the
sample rate (though, heck, why not offer the option if you can?).
Honestly, I'm beginning to think 96k with very gentle filters that start
a gentle rolloff not excessively far above the audible band would be the
best way to go. Perhaps they'll offer a filter choice like this for the
96k rate. That would certainly make me very interested in the converters.
Kurt Albershardt
> In article <10691844...@nnrp1.phx1.gblx.net>, Kurt Albershardt
> <ku...@nv.net> wrote:
>
>
>> Claude mentioned to me awhile back that he was expecting dCS to offer a
>> converter with both DSD and 352.8/24 PCM outputs (using similarly gentle
>> filter slopes on the PCM side.)
>
>
> The most exciting part of this is the gentle filter slopes, not the
> sample rate (though, heck, why not offer the option if you can?).
> Honestly, I'm beginning to think 96k with very gentle filters that start
> a gentle rolloff not excessively far above the audible band would be the
> best way to go.
You're far from alone in that thought. I'd spring for a dCS or
Lavry-priced converter today if I could get 88.2k - 192k conversion with
filters that started rolling off at 21-24k. I have enough faith in
the rest of their designs but this unresolved filter issue looms over my
decision process--if it's a significant as some of us think it might be,
it would mean a lot of replacement purchases.
Erik de Castro Lopo
>
> Not really.
>
> > For instance, attempting to design a 4th order Butterworth high
> > pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
> > difficult in 32 bit floating point but it can be done in 64
> > bit double precision floating point.
>
> Not really. There are zillions of working counter-examples.
So your argument is that not all filters are difficult to
design therefore they must all be "not difficult".
Sorry, that is plain and simple poor logic.
> > Once you up the sample rate to 384kHz, that same high pass filter
> > at 20Hz is going to be difficult the design and implement eve in
> > 64 bit doubles.
>
> Not at all. If this were right, programs that implement your so-called
> difficult filters like Adobe Audition would be far slower and more complex
> than they actually are.
I'm curious. Do you know anything at all about digital filter
design or are you going to stick to your faultly reasoning
above?
I've seen the arguments you get into in this group and others.
Sometimes you do know what you are talking about but in this
case I doubt it.
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nos...@mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"TLC declared bankruptcy after they received less than 2 percent of the $175
million earned by their CD sales. That was about 40 times less than the
profit that was divided among their management, production and record
companies." -- Courtney Love on the REAL piracy
Arny Krueger
news:3FBAD6E0...@mega-nerd.com
> Arny Krueger wrote:
>>
>> Not really.
>>> For instance, attempting to design a 4th order Butterworth high
>>> pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
>>> difficult in 32 bit floating point but it can be done in 64
>>> bit double precision floating point.
>> Not really. There are zillions of working counter-examples.
> So your argument is that not all filters are difficult to
> design therefore they must all be "not difficult".
Straw man argument. You named a very specific filter "4th order Butterworth
high
pass filter with a cutoff of 20Hz" and now you're trying to generalize to to
"all filters".
> Sorry, that is plain and simple poor logic.
Straw men tend to be that way.
>>> Once you up the sample rate to 384kHz, that same high pass filter
>>> at 20Hz is going to be difficult the design and implement eve in
>>> 64 bit doubles.
Why would an even higher sample rate be of benefit to the implmentation of
a 20 Hz filter, when 48,000 Hz sampling puts the Nyquist frequency more
than 10 octaves above the corner frequency of 20 Hz?
>> Not at all. If this were right, programs that implement your
>> so-called difficult filters like Adobe Audition would be far slower
>> and more complex than they actually are.
<notice that Erik doesn't respond to this very cogent point>
> I'm curious. Do you know anything at all about digital filter
> design or are you going to stick to your faultly reasoning
> above?
The faulty reasoning is something you fabricated, Erik. Please feel free to
critique it sharply, early and often.
> I've seen the arguments you get into in this group and others.
The facts are generally on my side. I've presented several and the response
is this silly straw man.
> Sometimes you do know what you are talking about but in this
> case I doubt it.
Just tell us why a sampling rate more than 10 octaves above the corner
frequency won't do the job, and why moving up an additional 3 octaves
somehow makes things better.
Erik de Castro Lopo
>
> Just tell us why a sampling rate more than 10 octaves above the corner
> frequency won't do the job, and why moving up an additional 3 octaves
> somehow makes things better.
Your have missed my point completely.
The problem is already there when the sampling rate is 10
octaves above the corner freq and it becomes very much
worse if the sampling rate moves up another 3 octaves.
The problems are due to the inherent inaccuracies of floating
point arithmetic. Computer floating point arithmetic has
accuracy problems when performing arithmetic which mixes very
large numbers and very small numbers.
Filters implementing low critical frequencies (corner for
HP/LP, center for notch/peak etc) at high sampling rates
results in coefficients with a very large spread in values.
Hence the problems.
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nos...@mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"There is no reason why anyone would want a computer in their home"
Ken Olson, DEC, 1977
Arny Krueger
news:3FBBECCD...@mega-nerd.com
> Arny Krueger wrote:
>>
>> Just tell us why a sampling rate more than 10 octaves above the
>> corner frequency won't do the job, and why moving up an additional 3
>> octaves somehow makes things better.
>
> Your have missed my point completely.
>
> The problem is already there when the sampling rate is 10
> octaves above the corner freq and it becomes very much
> worse if the sampling rate moves up another 3 octaves.
>
> The problems are due to the inherent inaccuracies of floating
> point arithmetic. Computer floating point arithmetic has
> accuracy problems when performing arithmetic which mixes very
> large numbers and very small numbers.
>
> Filters implementing low critical frequencies (corner for
> HP/LP, center for notch/peak etc) at high sampling rates
> results in coefficients with a very large spread in values.
> Hence the problems.
You said:
"For instance, attempting to design a 4th order Butterworth high
pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
difficult in 32 bit floating point but it can be done in 64
bit double precision floating point."
So let's put your claim to the test. Cool Edit Pro has a variable frequency
high pass Butterworth filter of any reasonable order including 4. It can be
run at any sample rate up to 999,999 Hz including 48 Khz and 384 KHz. AFAIK
it is implemented in 32 bit floating point arithmetic.
What bad thing to look for when operating it? Bad frequency response?
Nonlinear distortion? Bad phase response?
dan lavry
> Arny Krueger wrote:
> >
> > Just tell us why a sampling rate more than 10 octaves above the corner
> > frequency won't do the job, and why moving up an additional 3 octaves
> > somehow makes things better.
Hello,
My name is Dan Lavry, and I hope you do not mind that I barge in. I do
want to say a few words on that 192Khz. As a designer, I have long
realized that faster sampling means less accuracy. You can get 1 bit
at a GHz, 8 bits at 100MHz, 12 bits at 30Mhz…. Nearly 24 bits at a few
Hz. Well, I am not suggesting to sample at 10Hz. We need to cover the
audio bandwidth. Going too slow is bad. Going too fast is also bad.
Let's face it, you want to say charge a cap, or settle an amp, and it
does dot happen at zero time. Nothing does. Weather you get there
exponentially or you have that step with some ringing… whatever it is,
if you wait till it settled you get it closer and more accurate. These
are practical limitations. The thing that gets me is that all the EE's
I talk to make jokes about 192Khz – a common one is "my dog can not
hear it". But many in the recording and mastering side are a part of a
different crowd. There seem to be a gap between the EE designer types
and the recording types. I am not suggesting that EE's are better or
worst. There have been times when the good ear is right, and the EE
learns new things. But there should be times such as right here and
now for the recording types to be open-minded. And folks, that 192KHz
is a crock.
I have wondered about how is it possible that a whole industry can go
in the wrong direction. After all, Few musical instruments will under
special cases yield a -50dB harmonic at 50Khz, and almost nothing at
100KHz. Experiments showing that special case required special mics
and gear. Most recording mics will not pick up a thing at 60khz… The
speakers will not play it, the ear will not hear it, no matter how big
the ego is!
So the sounds are negligible at those high frequencies, the mics will
not pick them, the speaker will not play it, and you can not hear it.
So how can such a crock take place? And let me assure you the folks
that made it happen at the semiconductor houses and workstation houses
did not include the EE department... I think it was and is about
making money. I talked to many engineers that stated privately that
they are not talking because they need their jobs! Other "did not want
to make waves", "stir the pot"…
So is everyone trying to cheat? Of course not. It started with 48Khz
better than 44.1, which is true. Also, while a bit excessive, 88.2 and
96KHz are in some cases a better compromise than 44.1. That is due to
the old FIR pre shooting (pre echo) that can happen under some cases
with a lot of processing. By the time you are at 60KHz it is so far
down you will not hear it, so 70Khz is a pretty reasonable place to
be. So after we have improved some as we went higher, the expectation
was than to continue upwards. Where does it stop? 384? Why not 1Ghz?
Oh well, the will be pretty bad, will it not?
But there is more to that than that "trendy upwards worked before so
lets continue". The common sense tells you that more is better. For
example, more pixels yield more detail. More bits gives better sample
accuracy. So if we take that analog wave, made out infinite points
connecting into a line, will we not benefit from a better "tracking"
of the wave? Well here is the news, and it is pretty old stuff, though
it may not be easy to grasp with simple common sense. More pixels for
better picture- yes! More bits for better accuracy (assuming
theoretical case - no noise) – yes! More sample density (higher sample
rate) for more accuracy –NO! This is the beauty of the sampling
theory. Nyquist did not say: We take more points and we will get
closer approximation. What he said – and it is a FUNDUMENTAL THEORY,
is that once we agree to deal with a limited bandwidth (called the
Nyquist bandwidth), all we need to do is sample at a tiny amount
greater than twice that bandwidth. This will yield 100% of the
waveform information in the data stream. We may need to filter a
signal (anti alaising) to make sure we do not have energy over
Nyquist, but than we are home free. Taking 4 times as many samples
does not yield 400% of the information. You can only have 100%! How do
you retrieve the information? You use a filter and it connects the
sample points in such a way that that you get the original wave shape.
A filter does not connect the dots (sample points) with straight
lines, or parabola… It recreates the original wave! You do not need to
help things with extra point in between. It buys you nothing!
I also see a lot of confusion regarding that Nyquist, upsampling,
oversampling, gradual filters… Some folks think that a 96KHz AD will
require a sharper anti aliasing filter than say 192KHz. This is
typically very wrong. The 96KHz or 192KHz AD refers to the OUTPUT RATE
of the converter. The antialias requirement is determined by the INPUT
SAMPLING RATE which is usually way beyond 192Khz. This days, most
modern AD's are running at input sampling rates of 3-12Mhz!!! DSD is
64fs and many mulibit IC's run even faster… So even with 50Khz audio,
Nyquist is so high the a gradual 3 pole will yield 120dB at the input
rate Nyquist. The outcome of the high rate modulator (input) is than
down-sampled to whatever – 44.1, 96, 192… Of course, when the sales
guys try to stick you with it they tell you need more bandwidth, but
they also ALL I saw regarding semiconductor and gear makers alike:
specs for 192KHz device are with A weighting – which states
(indirectly?) that you do not even need to measure flat to 20Khz. So
is it a crock? It is!
Theoretically, there is "no harm" in more points, and there is "no
added good". But as I pointed out, faster is less accurate! And yes,
you double the sample rate and the processing power requirement, and
so is the file size. These are serious draw backs! Don't you say you
do not care about file size: The DVD audio has 12 to 1 compression
(Dolby AC), We do not even get near a 1 to 1, and we want to push it
higher?
I realize that with all that reasoning and science and engineering,
someone is going to tell me that they hear it and like it. In fact,
someone told me that they still hear that high frequency in the
44.1KHz CD. I will not dignify that impossibility. If you hear some
distortion you like on the 44.1K CD, you did not need to go any faster
than 44.1KHz to generate it. I am not arguing against controlled
distortions (such as tube sound and what not). If you like it is fine.
It may be artistic decision. If you feel like you need to go to 1Mhz
than down to like it, fine. I think you are letting the gear control
you instead of the other way around, but fine! Just as long as I get
you to realize that you can get those distortions with a 44Khz… And we
do not all need to double the file size and processing power, and buy
new gear that is less accurate.
192 is a crock! 382 is a super crock! 88.2/96Khz is a bit excessive,
but not too far from a good rate. I too can glue a faster IC on the
board and make more money. My 192 DA prototype is not bad, but the
96KHz bits it by a lot.
Anyone telling you that more points will give better aproximation is
lacking lacking some know how.
Thanks for your patience.
Dan Lavry
Lavry Engineering
Mike Rivers
In article <673b149b.03112...@posting.google.com> danl...@mindspring.com writes:
Hi, Dan. Welcome to rec.audio.pro.
> I have wondered about how is it possible that a whole industry can go
> in the wrong direction.
Marketing. It's easier to sell a new format than it is to sell an
improved version of an old one.
> So the sounds are negligible at those high frequencies, the mics will
> not pick them, the speaker will not play it, and you can not hear it.
> So how can such a crock take place?
There are throries (though no valid tests yet) that suggest that
what's up there that we can't hear affects what we can hear. As I said
in another posting, I related a listening demonstration where I heard
what I considered a subjective improvement when moving from a 48 kHz
to a 95 kHz A/D/A converter chain (and they were your converters, back
when you were still selling under the db name), though I have no
reason to believe (or not believe) that a similar improvement would be
realized at 192 kHz.
> Theoretically, there is "no harm" in more points, and there is "no
> added good".
The "harm" is that more samples per cycle is easier to achieve than
more accurate samples, so we are often sold the wrong technology
because it's cheaper and it looks better than the previous technology
on the ad copy.
KikeG
somewhat disagree with:
> So is everyone trying to cheat? Of course not. It started with 48Khz
> better than 44.1, which is true. Also, while a bit excessive, 88.2 and
> 96KHz are in some cases a better compromise than 44.1. That is due to
> the old FIR pre shooting (pre echo) that can happen under some cases
> with a lot of processing.
This pre-echo (I'd better call it pre-ringing) is a consequence of the
brickwall FIR filter cutoff frequency, at around 20-21 KHz. Due to
this, a) this pre-ringing has a frequency of around 20-21 KHz, which
should be hardly audible, b) it happens just with signals that have
strong content at these frequencies, c), usual lengths of this
pre-ringing are in the order of a few milliseconds (less than 3 ms
with a 256-tap FIR filter), very close to the limits of audibility,
and d) this pre-ringing is important just in case of impulsive signals
where it could really make a difference.
So, it seems that this pre-ringing should have little importance in
practice. I have yet to see someone prove it is audible by means of a
reliable double-blind listening.
Even if it was proved to have some audible consequences, I think it
could be overcome by means of simply using adequate non-symmetric,
minimum pre-echo, non-linear-phase FIR filters. Our ear is little or
no sensitive to phase distortion at high frequencies, so I think the
use of this kind of filters should have little or no audible
consequences.
dan lavry
> point arithmetic. Computer floating point arithmetic has
> accuracy problems when performing arithmetic which mixes very
> large numbers and very small numbers.
>
> Filters implementing low critical frequencies (corner for
> HP/LP, center for notch/peak etc) at high sampling rates
> results in coefficients with a very large spread in values.
> Hence the problems.
>
> Erik
I agree with you and will add some additional comments:
It is not really easy to explain quickly, but there is a relationship
between filter complexity (number of coefficients) and the ratio of
sample rate to filter corner frequency. For example, making a low pass
or high pass FIR at exactly Nyquist is desirable because this is a
specific case called "half band" where half the coefficinet become
zero. The further you deviate from the Nyquist/2 point, the more
coefficients are needed for a given quality. The deviation comment is
symetrical: the 1KHz is like 21.050KHz(=22.050-1KHz) the example here
is based on say 44.1KHz sampling.
You want some intition to work for you? It is tough to explain. Maybe
you can think of it as a need to contain at least a significant part
of a cycle in the FIR pipe line. After all, 100 samples near the peak
of a 20Hz sine wave are almost all at value of 1. What can you do with
that data? On the other hand 100 samples of a higher frequency audio,
say 10KHz let you look at some significant part of the wave shape...
In any case, try to do a low pass at say 500Hz, with 100dB, and it is
not a specific filter I am talking about, very general comment here,
and you get certain number of coefficients. Now double the sample
rate, and your requirnment doubled. The point is that it is about the
RATIO between sample rate to corner frequency. A 44.1KHz 100Hz fith
certain coefficients yields say Xdb with Y transition band... Double
the sample rate to say 88.2KHz, and the SAME coefficients will give
the Same X and Y performance but at 200Hz - the ratio of sample rate
to corner frequency stays the same. You still want 100Hz? Start adding
compute power and a lot of ot!
One only needs to realize that a 100Hz FIR at 192KHz is THE SAME
FILTER as 25Hz at 48KHz... Do not anyone tell me that a 25Hz FIR is
easy. It is huge, even for real crapy DSP. That is why at very low and
very high frequency, folks do IIR's, with all the phase compromises...
And that is when coefficient acuracy may realy bite you. No one does
FIR DC removal, because the frequency ratio (sampling to corner) is
way too much at 44.1KHz. It more than 4 times worse at 192Khz.
All that talk about frequncy resolution being better with higher
sampling, that is also upside down! Say you want to do some peaking at
1Khz, and you want it to be a certain bandwidth. For any given compute
power, you have better control with the lower sampling rate.
BR
Dan Lavry
dan lavry
> This pre-echo (I'd better call it pre-ringing) is a consequence of the
> brickwall FIR filter cutoff frequency...
> So, it seems that this pre-ringing should have little importance in
> practice....
OK. I am glad you agree that preringing may not be a big deal. I was
just trying to be as carfull as posible so I brought the issue to the
discussion. You know who it is - if i didnot someone would say I
forgot some big issue...
BR
Dan Lavry
dan lavry
> In article <673b149b.03112...@posting.google.com> danl...@mindspring.com writes:
>
> Hi, Dan. Welcome to rec.audio.pro.
>
> > I have wondered about how is it possible that a whole industry can go
> > in the wrong direction.
>
> Marketing. It's easier to sell a new format than it is to sell an
> improved version of an old one.
I am gald to hear someone else say it. It feels lonely :-(
> There are throries (though no valid tests yet) that suggest that
> what's up there that we can't hear affects what we can hear.... though I have no reason to believe (or not believe) that a similar improvement would be
> realized at 192 kHz....
I read the same studies, about the alfa brain waves... but even when
you take that research into account, the issue was about extending
audio to 26KHz or so. Certainly not to 96KHz bandwidth (192KHz
sampling). 88.2KHz sampling should more than take care of that extra
bandwidth...
> The "harm" is that more samples per cycle is easier to achieve than
> more accurate samples, so we are often sold the wrong technology
> because it's cheaper and it looks better than the previous technology
> on the ad copy.
You know, right after my tutorial AES presentation on AD's, when both
Dr. Rich Cabot and I were pretty vocal about that 192KHz crock, some
industry salesmen started talking about 384KHz. I think such talk is
done delibaratly so that folks will not go against 192KHz. By making
the the conversation about 384, folks may be more willing to end up
with 192K.
The conversation should be about the fact that 192KHz is a crock
suported by major semiconductor houses, huge worksation makers and so
on. It is one thing for some industry salesman to fall for a crock. It
is another thing to get out of that serious presentation start
promoting 384KHz!
Arny Krueger
news:znr1069336652k@trad
> In article <673b149b.03112...@posting.google.com>
> danl...@mindspring.com writes:
>> I have wondered about how is it possible that a whole industry can go
>> in the wrong direction.
> Marketing. It's easier to sell a new format than it is to sell an
> improved version of an old one.
Agreed. It's the tyranny of numbers. I find it extremely ironic that some of
the people who are the first to queue up behind larger number are people who
claim they are subjectivists.
>> So the sounds are negligible at those high frequencies, the mics will
>> not pick them, the speaker will not play it, and you can not hear it.
>> So how can such a crock take place?
Agreed. People routinely publicly enthuse about mics monitor speakers that
roll off rapidly above as little as 13 KHz or as much as 23 KHz. Needless to
say, these products do right within the nominal 20-20 KHz band. AFAIK the
mics and speakers that have anything like flat response up to 44 KHz can
probably be counted on one hand. They are also anything but flat if only a
little bit off-axis. Many of the recordings that people have been praising
the SACD and DVD-A re-releases of are made from 48 KHz digital recordings or
15 ips tapes. Response plummets like a stone above 23 KHz or so.
> There are theories (though no valid tests yet) that suggest that
> what's up there that we can't hear affects what we can hear. As I said
> in another posting, I related a listening demonstration where I heard
> what I considered a subjective improvement when moving from a 48 kHz
> to a 95 kHz A/D/A converter chain (and they were your converters, back
> when you were still selling under the db name), though I have no
> reason to believe (or not believe) that a similar improvement would be
> realized at 192 kHz.
And I'm still asking about bias controls.
>> Theoretically, there is "no harm" in more points, and there is "no
>> added good".
> The "harm" is that more samples per cycle is easier to achieve than
> more accurate samples, so we are often sold the wrong technology
> because it's cheaper and it looks better than the previous technology
> on the ad copy.
Agreed. Consider the irony of so-called "24 bit" converters whose dynamic
range underperforms earlier products that were by design and application 16
or 20 bit devices.
Tommi
"dan lavry" <danl...@mindspring.com> wrote in message
news:673b149b.03112...@posting.google.com...
> I also see a lot of confusion regarding that Nyquist, upsampling,
> require a sharper anti aliasing filter than say 192KHz. This is
> typically very wrong. The 96KHz or 192KHz AD refers to the OUTPUT RATE
> of the converter. The antialias requirement is determined by the INPUT
> SAMPLING RATE which is usually way beyond 192Khz. This days, most
> modern AD's are running at input sampling rates of 3-12Mhz!!! DSD is
> Nyquist is so high the a gradual 3 pole will yield 120dB at the input
> rate Nyquist. The outcome of the high rate modulator (input) is than
Burr Brown,(a division of Texas Instruments), manufactures both one-bit and
true multibit converters. They recommend to use multi-bit converters for
"waveform synthesis applications requiring very low distortion and noise."
I'm not an expert in converter design, however, wouldn't this imply that
their finest converters are true multibit converters with an input sampling
rate of 192kHz or below? This would mean that 96kHz converters would
actually require a sharper anti-alias filter than 192kHz and so on.
xy
session...well you would need some wicked sort of raid setup to have
your hard drives keep up with the data throughput.
dan lavry
The answer to your question will be very long. First, much of what the
IC makers has to be taken with a grain of salt. Some of it is
engineering and some is marketing, Engineering is not why .why they
sell 192KHz for audio.
But one may not realize that different applications require different
focus on specific aspects. For example, audio conversion does not
require the best accuracy in terms of DC offset or gain. An +/-10V
audio converter will not cause problems if the DC offset (error) is
1mV. A 5% gain is "no much.". Just adjust the volume by a tiny portion
of a dB. Try that 1mV and 5% for an instrumentation device. You do
want your meter to give precise readings... Than there are devices
that should match (in pairs or groups) so the offsert and gain must be
right on. For audio there is a set of requirnments (including
THD+N)....
Regarding the filter, I will say it again: The input of the AD
(modulator) of the DSD and the modern mulibit noise shaping
converters, run at at least 64fs and some are at 128fs or 256fs. So
from antialiasing stand point the filter can be prteey gardual. Say a
device is a 96K. That number is NOT what you use to figure the anti
aliasing. You look at the oversampling and find it is 64fs. That means
the input is sampled at about 6Mhz and Nyquist is 3Mhz. That is the
frequncy that you look at for antialiasing. Say you want to pass
30Khz, and the INPUT Nyquist is 3MHz. Say you want 100dB rejection at
3MHz (that is a lot because you do not expect to see much signal
there). Every pole yields 20dB per decade. You have 2 decades
(30-300K, 300K-3MHz). So a pole is worth 40dB attenuation at 3Mhz. so
3 poles yield 120dB.
In the good old days, a 44.1KHz required huge number of poles. This
days, the input is sampled much higher. That modulator high speed
output is decimated down to manufacture the 192/96/44... Just like DSD
goes from 64fs (2.8M sampling or so) to 44,1KHz CD rate, so do the
multibit high rate.
For audio, a more importent distinction should be made between noise
shaping and regular type AD's. Between single bit dhaping (DSD) and
mulibit shaping.
BR Dan Lavry
Tommi
"dan lavry" <danl...@mindspring.com> wrote in message
news:673b149b.03112...@posting.google.com...
> "Tommi" <tomm...@suomi24.fi> wrote in message
news:<Vqbwb.124$dH4...@reader1.news.jippii.net>...
> > Burr Brown,(a division of Texas Instruments), manufactures both one-bit
and
> > true multibit converters. They recommend to use multi-bit converters for
> > "waveform synthesis applications requiring very low distortion and
noise."
> > I'm not an expert in converter design, however, wouldn't this imply that
> > their finest converters are true multibit converters with an input
sampling
> > rate of 192kHz or below? This would mean that 96kHz converters would
> > actually require a sharper anti-alias filter than 192kHz and so on.
>
> The answer to your question will be very long. First, much of what the
> IC makers has to be taken with a grain of salt. Some of it is
> engineering and some is marketing, Engineering is not why .why they
> sell 192KHz for audio.
> Regarding the filter, I will say it again: The input of the AD
> (modulator) of the DSD and the modern mulibit noise shaping
> converters, run at at least 64fs and some are at 128fs or 256fs.
> For audio, a more importent distinction should be made between noise
> shaping and regular type AD's. Between single bit dhaping (DSD) and
> mulibit shaping.
Dan, thanks for your informative comments. When you're talking about modern
multibit noise-shaping converters, do you actually mean 1-bit delta-sigma
converters which convert the data stream to PCM, or, say a 4-bit converter
with noise-shaping etc, or do you mean real 20-24 bit pcm converters?
I'm asking this because I don't understand why would you have to do a real
20+bit
converter with noise-shaping since the noise floor would be theoretically
over 120+dB down at all frequencies, so you wouldn't have to shift it
anywhere.
I have been under the impression that noise-shaping converters are generally
inferior
to real pcm converters without any kind of noise-shaping, one of the reasons
being that transient signals will have a poor resolution in noise-shaping
systems. If the signal doesn't endure for a long enough time, the error will
not be minimized by the noise shaper. As I said, I'm not an expert at
converters, so I'm still unsure about which type
of converter is most suitable to audio.
Scott Dorsey
>
>Burr Brown,(a division of Texas Instruments), manufactures both one-bit and
>true multibit converters. They recommend to use multi-bit converters for
>"waveform synthesis applications requiring very low distortion and noise."
>I'm not an expert in converter design, however, wouldn't this imply that
>their finest converters are true multibit converters with an input sampling
>rate of 192kHz or below? This would mean that 96kHz converters would
>actually require a sharper anti-alias filter than 192kHz and so on.
Since when do you believe marketing propaganda?
I'm still trying to figure out why none of the current Burr-Brown offerings
have as good low level performance as the long-discontinued PCM-63 multibit
converter.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Tommi
"Scott Dorsey" <klu...@panix.com> wrote in message
news:bpt7kr$71n$1...@panix2.panix.com...
My point was more on the multibit/single-bit issue. One can always say that
IC manufacturers just want people to buy the most expensive circuits they
have, but it would seem logical that multi-bit converters are generally
better for A/D for example than Delta-Sigmas which convert to pcm, since
real multibit converters can cost over
6 times as much as 1-bits. This is also related to the SACD-DVD-A issue.
Scott Dorsey
>
>My point was more on the multibit/single-bit issue. One can always say that
>IC manufacturers just want people to buy the most expensive circuits they
>have, but it would seem logical that multi-bit converters are generally
>better for A/D for example than Delta-Sigmas which convert to pcm, since
>real multibit converters can cost over
>6 times as much as 1-bits. This is also related to the SACD-DVD-A issue.
Why does expense have anything to do with quality?
Multibit converters are difficult to make linear, because of the trimming
issues. Lots of resistors that have to be very precise. So any good multibit
converter will be hand-trimmed and that is expensive.
Sigma-delta converters are easy to make linear, but it's hard to make them
quiet. Idle tones become a bit issue.
You pays your money and you takes your chance.
dan lavry
Hello Scott,
I agree with completely. I hope you do not mind if I use your comment
as a starting point to take it in a somewhat different direction.
Based on your comments, I suspect you will agree with me. So here I
go, with a long one:
One needs to make a distinction between a multi bit and a noise
shaping multi bit.
The classical multi bit is, as Scott points out, a more difficult
design, in terms of matching resistors or capacitors in some designs.
Most of these converters are beyond a simple R-2R ladder, with
improved architectures (such as segmented design) providing faster
settling and less resistor dependence. But still, they are precision
made devices with a lot of trimming. They provide a theoretical flat
noise floor to Nyquist. Given the high number of bits, the circuits
tend to operate slower than their noise shaping counter parts. That is
not to say that running the front end is not best to run at some
faster rate than X2 the required bandwidth, to help the anti aliasing
filter requirements.
Now, think of a 1MHz 16 bit AD. Say we decide to take the average of
every 2 adjacent samples. That will yield 500KHz, but we gain 3dB of
dynamic range. In theory we get 3dB improvement each time we divide
the clock by 2. So a 64fs which is 2*2*2*2*2*2 yield 6*3=18dB (3 bits
improvment). Of course one does not want to go 64fs for 18dB, but it
does demonstrate that speed and accuracy are a tradeoff, which is one
of the reasons why 192KHz - too fast for audio, is a crock!
The noise shaping multi bits are a different animal. After some
evolution, from a 2nd order 1 bit, to 3rd order, to 5th order 1 bit at
64fs (such as much of DSD), we stated getting into the area of
diminishing returns, in terms of increased filter order. The only way
to improve 1 bit significantly was to go faster, which is fine for an
AD maker, but too much data for DSD, which already takes a lot of
space.
But DSD offered that "perfect differential linearity". True it has a
large amount of noise rising fast at about 22KHz all the way to
1.4MHz, but it was the "linear way" to go. A 1 bit has only one
comparator and the decision is simple: Is the signal over zero or
under zero. If the threshold is not exactly zero, it is still ok (it
is just a DC offset). With more than 1 bit you have steps and they
must match extremely well (on the DA in the feedback loop of the noise
shaper). Say you have the following comparator levels -2,-1,0,1V. It
is no longer just a question of say "greater than or less than" -1V or
0V. The difference between thresholds must be the same, very
accurately so.
True, if you go for, say 3 bits noise shaper, the trimming and
precision is about 8 levels, much less than the classical (no noise
shaping) AD. But 8 levels at very high speed is no walk in the park.
Again, the faster you go, the less accurate! Keep that in mind - There
is alwayse a speed vs. accuracy tradeoff.
But some smart folks figured way to deal with it. The main one methode
is called DEM (dynamic elements matching). You basically take a lot of
elements (say resistors) and keep rotating which one to use. Given
that you do it very fast, you end up with some nice "average value".
Say one resistor is really off, but it is only 1 out of 16. It gets
used only 1/16 of the time, thus it does not "pull" the error "that
bad"…
The multi bit noise shaping AD's has less noise amplitude rise than
DSD, and with DEM, better dynamic range than the 1 bit DSD. But they
can not be used as a format. A say 3 bit noise shaping AD with 64fs is
makes 3 times more data than DSD. Now, some of the converter makers
choose 128fs or even 256fs. That is not going to go well for the
consumer 64fs 1 bit DSD format. But the amount of data, modulator
sampling rate, high frequency noise, are all just an part of an
intermediate step. That 3 bit 128fs modulator (for example) will be
made into an audio PCM by means of decimation.
Now here is an important point. Please think of a balloon, partially
filled with air – you squeeze at one place, the air goes to another
place… Given an oversampling rate and a given an N bit shaping, one
can design the modulator and decimator so that the noise will start
rising at say 22KHz (for a 44.1KHz sampling AD outcome), or 48KHz
audio (for a 96KHz outcome), or for a 96KHz audio bandwidth (for
192KHz sampling rate outcome). Guess what: The 22KHz case will yield
better audio band performance than the 48KHz and the 96Khz audio (for
192K sampling) provides much worst than the 0-22KHz performance. It is
like that balloon. You want more audio bandwidth and something has to
give. It is important to "park" the design at a good point. If you
just keep rasing the sampling rate, you may end up at 3 dB at a GHz
:-)
I can not argue with those that say we hear 22KHz. I doubt it that we
need more than 44 or 48KHz audio (88 or 96KHz sampling). But 192KHz is
a CROCK!!! 384 is a SUPER CROCK!!!
We love to talk about how to get from point a to b, by way of topics
such as bipolar vs FET's, comparing transient response or jitter of
sigma delta to the classical one… and so on. But here we are dealing
with FUNDUMENTALS. Humans can not hear 96KHz audio, most microphones
do not pick it, speakers don' play is, and there is hardly any sound…
We are at the place where the fundamental theory (Nyquist) states that
we do not get ANY benefit from adding samples, because the properly
filtered (anti aliasing) sampled signal, contains 100% of the
information – read it as "100% of the wave shape". More samples is not
more money, more pixels, more bits. More samples is more NOTHING. And
that is in THE FUNDUMENTAL THEORY OF SAMPLING. In practice, more
samples means everything is speeded up so the overall accuracy is
LESS! The old speed accuracy compromise.
So 192KHz is a CROCK!!! Just the mention of 384KHz or 192KHz is an
outrages attempt to have folks think "faster is better". Folks that
push that either lack some serious background, or are trying to figure
a way to sell more stuff. It is eiteher disrespectfull to reality or
to the customers. I wish we concentrated on improving what needs
improving, instead of screwing things up. I am doing my best to try
and stir that big ship back in the right direction. I can not do it
alone, but I am not alone. I am finding out many real good ears and
serious talent on the recording and mastering side that know what I am
saying "by ear". I do believe history will prove that science, math
and engineering will prevail over ignorance and short term marketing
plans.
BR
Dan Lavry
Rob Adelman
dan lavry wrote:
> With more than 1 bit you have steps and they
> must match extremely well (on the DA in the feedback loop of the noise
> shaper). Say you have the following comparator levels -2,-1,0,1V. It
> is no longer just a question of say "greater than or less than" -1V or
> 0V. The difference between thresholds must be the same, very
> accurately so.
Dan, thanks for all of the information, though I have to admit most of
it is over my head. But somehow I am guessing that the above section is
a pretty critical piece of the puzzle.
-Rob
dan lavry
> dan lavry wrote:
>
> Dan, thanks for all of the information, though I have to admit most of
> it is over my head. But somehow I am guessing that the above section is
> a pretty critical piece of the puzzle.
>
> -Rob
I am sorry. I have a massage and felt compelled to explain it with
some solid arguments. That I just got too technical. Let me rewind and
start again, though I may take some liberties to be less accurate and
more intuitive.
I someone told you that in order to draw a straight line, all you need
to know are 2 point, you will believe it. It talks to your common
sense. If some one tried to tell you that you need more points, you
will probably dismiss them.
Lets try drawing a circle through 3 points. Are 3 points enough to
draw that circle line? Slightly less intuitive is it not? But
manageable.
If I tell you that I have a curve that can go up or down or sideways
in a totally unpredictable way, you will realize that the more points
and the closer they are, the better the representation. So some folks
are saying: audio is like that complex curve, so give me more points –
increase the sample rate.
It may not be easy to grasp, but while audio is very complex, there
are some restrictions there, and it is not true that "anything goes".
The fact that we are dealing with some limited bandwidth (frequency
range) will, for example restrict that curved line representing the
sound from moving too fast (think of putting a restriction on the
maximum allowed slop). This is true for any wave, video, audio,
medical, instrumentation… The lower the bandwidth, the lower the
slope. I am not being completely accurate with slope, but higher
frequencies move faster.
The point is that the restrictions define the signal well enough so
that you do not need too many points to draw that line. Too few points
will not cut it, but you get to a certain level that allows you to
draw the line correctly. Just like 2 point for a straight line.
Some folks are trying to sell you on doubling the points, when you do
not need to. They call it 192KHz sampling. The extra points (samples)
take space, require you to double the processing power of your
machine, and in fact lower the quality of the outcome.
The argument is based on "more is better" which is often true, but not
always. Those that study EE and math know that it is not. Those that
do not have the background are just as likely to buy the BS, as they
are to buy the truth.
There are a lot of forces out there, from huge semiconductor houses to
huge workstation makers and their whole support network that have been
promoting that crock. With so much combined clout, few want to stand
up to it. And of course, such a myth gets propagated to the sales guys
that mostly lack the know how, and latch into that "more is merrier"
wrong explanation. We have a whole industry going in the wrong
direction.
But there are a lot of things in audio that are in the "gray area".
The above is not. We are dealing with fundamentals. The only argument
I can not deal with is: but I like it. Or It sounds great to me. Fine
if it does, but my point is: You do not need to go and double the data
and also double the processing to get that thing you like, If you have
a certain characteristic (distortion) you like, I can make it for you
with a 96KHz AD or lower.
I too want to improve quality, and there are things to do. But going
above 96KHz is screwing things up. Math engineering and science is on
my side of the argument. History will prove it, and hopefully very
soon. Meanwhile I am sorry to see folks pay good money to be taken to
a ride in the wrong direction.
Of course, those that got influenced to belive they are getting better
sound, are in a bind. It takes a "hack of a man" or a woman to go back
on it, to admitt you were wrong. Certainly such is the case in this
industry. And it is always "acceptble" to just say "but I like it". In
audio, you call it "an artistic decision" and no one will argue, well
almost no one...
I do not want to make anyone uncomfortable. I just want that 192 and
to have 96 accepted as morer than enough. I hoe it does and soon.
While observing some recording and mastering guys "go with the flow"
of faster is better, I am very pleased to see some top notch ears
that figured it out ""by ear". That is encouraging.
I hope this is clear and direct enough.
Dan Lavry
Tommi
> I do not want to make anyone uncomfortable. I just want that 192 and
> to have 96 accepted as morer than enough. I hoe it does and soon.
> While observing some recording and mastering guys "go with the flow"
> of faster is better, I am very pleased to see some top notch ears
> that figured it out ""by ear". That is encouraging.
>
> I hope this is clear and direct enough.
>
>
> Dan Lavry
Dan, you have explained your point about 192kHz problems very well!
It's sad to see some of these so-called 24-bit/192kHz converters which
really only have dynamics of about 100dB with bad distortion etc., marketed
with a 24/192 tag.
It appeals to some people because it's new, even though I must admit that I
myself didn't know about the 192kHz problems until you presented some
serious information.
Even though my head is still trying to figure out the various differences
between the variations of single- and multibit, shaping and non-shaping
converters,and their unique flaws/merits, your point is well made.
Mike Rivers
> If I tell you that I have a curve that can go up or down or sideways
> in a totally unpredictable way, you will realize that the more points
> and the closer they are, the better the representation. So some folks
> are saying: audio is like that complex curve, so give me more points –
> increase the sample rate.
>
> It may not be easy to grasp, but while audio is very complex, there
> are some restrictions there, and it is not true that "anything goes".
> The fact that we are dealing with some limited bandwidth (frequency
> range) will, for example restrict that curved line representing the
> sound from moving too fast (think of putting a restriction on the
> maximum allowed slop).
The way I like to explain this is to imagine that you're driving a
race car on a waveform-shaped track. If that track is a perfect sine
wave, you'll have to go at a certain speed in order to complete a
cycle in a given time. If you now put some more bends in the track,
you'll be going a greater distance from end to end. If you want
to match your time for the simple curved track, you'll have to drive
faster. Put more kinks in it and you'll have to drive still faster.
When the track gets sufficiently contorted, ignoring things like
centrifical force and coefficient of friction that cause you to slow
down for turns, eventually the path will be long enough so that you
simply can't get up enough speed to get to the end of the track within
the proscribed time.
If you can somehow increase your maximum speed (like by putting a
bigger engine in the car) you can then again meet your mark. The speed
of the car represents the sample rate, the turns in the track
represent the frequency because they increase the distance traveled in
a fixed amount of time.
Hp Widmer
some additional comments:
I did some measurements using an analog burst 10khz signal who is
complete not in sync with the sample clock.
To get a fine none-overshouting (pre/post) sampled signal a high
sample rate is required to get ride of the ringing (about 10 times
related to burst sine freq.).
I read some time ago in a paper that 1/100% ringing is audible....
Cheers
Hp
Len Moskowitz
Tommi <tomm...@suomi24.fi> wrote:
>It's sad to see some of these so-called 24-bit/192kHz converters which
>really only have dynamics of about 100dB with bad distortion etc., marketed
>with a 24/192 tag.
Well, there are other converters that are spec'd at 24/192 that offer
120+ dB of dynamic range, very fine distortion specs and very low
noise. And they sound fine.
What surprises me most these days are the folks who complain about the
sound of high res digital audio but extol LPs (vinyl) with their 30 dB
max of separation and surface noise.
--
Len Moskowitz PDAudio, Binaural Mics, Cables, DPA, M-Audio
Core Sound http://www.stealthmicrophones.com
Teaneck, New Jersey USA http://www.core-sound.com
mosk...@core-sound.com Tel: 201-801-0812, FAX: 201-801-0912
Mike Rivers
> Dan, you have explained your point about 192kHz problems very well!
> It's sad to see some of these so-called 24-bit/192kHz converters which
> really only have dynamics of about 100dB with bad distortion etc., marketed
> with a 24/192 tag.
They get away with it because people see "24/192" and don't see the
actual dynamic range specifications.
In the early days of 96 kHz products, a lot converters (in both
directions) that sounded pretty good at 48 kHz didn't sound quite as
good at 96 kHz. I think those problems have been fixed, but the
numbers for S/N ratio and dynamic range are limited (at least with
good quality chips) by analog circuitry, short term clock stability,
board layout (which affects all of the above) and physics.
dan lavry
A burst is not a band limited signal. In fact, making a sharp corner
takes infinite bandwidth. So if you take a zero signal and all of a
sudden you "shoot up" into the first cycle of the burst, and also the
ending of it- sudenly to zero, it takes huge bandwidth.
That is why we window FFT's, and why we also the best FIR filters are
done with the window method. Loosly speaking, windowing anounts to a
very gradual taper at the start and end of the wave, sort of like fade
in and fade out that mastering and music editors do.
I am not sugesting you need to window andything, but I sugest that you
need to make sure that the burst is FILTERED with proper anti alaising
filter. If you did not filter the high frequency content of the burst,
you will have alaising.
If you filter it properly and the rest is done correctly, your
conclusion will change. If you get ANY more detail when sampling
faster there are 2 possibilities:
1. Something is wrong with the test or the setup
2. Nyquist was wrong,. Shannon was wrong, math does not work and
science are wrong.
I will not get into much back and forth regarding issues that are as
solid as the law of garvity.
I do not know why after my long post, I have to deal with such a
response. You are saying "I tested it and sampling theorm is wrong". I
would be inclined to figure out what is wrong with the test or the
setup.
Dan Lavry
Glenn Zelniker
> You said:
>
> "For instance, attempting to design a 4th order Butterworth high
> pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
> difficult in 32 bit floating point but it can be done in 64
> bit double precision floating point."
>
> So let's put your claim to the test. Cool Edit Pro has a variable frequency
> high pass Butterworth filter of any reasonable order including 4. It can be
> run at any sample rate up to 999,999 Hz including 48 Khz and 384 KHz. AFAIK
> it is implemented in 32 bit floating point arithmetic.
Aaahhh. A discussion that's near and dear to my day's work.
You *did* misinterpret what Erik said. His point was that moving to
sample rates *complicates* matters considerably. In other words, it's
something that a DSP guy dreads.
I can assure you from extensive personal experience that if you design,
say, a highpass filter of fixed order and a cutoff frequency that's very
low relative to Fs, implementing a filter with the same cutoff frequency
(in absolute Hz, *not* in rad/sample) at 2Fs is much more difficult. 8Fs
is even worse. And maintaining the same *transition* band makes it even
uglier, as the required order grows with Fs.
What do I mean by difficult? Specifically, I'm talking about the
roundoff noise power and its attendant spectral distribution. For an IIR
filter, this metric gives an indication of the filter's output's
dependence on and vulnerability to re-circulated round-off noise errors.
And it also gives some indication of the relative amplitudes of the
filter's internal states. Here's a simple experiment you can do yourself
if you have a good analysis tool like Matlab. Design a HP filter with a
cutoff of 28 Hz and a transition band of 10 Hz and a sampling rate of
44.1 kHz. Now use those same specs and do it for 88.2 kHz. Now do it for
176.4 kHz. On each of those filters, use Matlab's qfilt objects to
generate finite-wordlength equivalents and set the quantization format
to {float,32,8}, architect the thing as a direct-form filter and run a
noise-load analysis on it. You'll see that direct-form is horrible at
44.1 kHz, hopeless at 88.2 kHz, and lethal at 176.4 kHz. We're talking
about so much noise that the filter won't work.
One's first thought is to re-implement the thing as cascaded
second-order sections. The results are stil bad, terrible, useless,
respectively. Doing it as a lattice/ladder only helps a little. In fact,
the only way to make it "work" is to resort to cascaded exotic
section-optimal minimum-roundoff structures, a la Roberts and Mullis.
The point here is not that it can't be done. But one must take heroic
measures to make it work properly at higher sample rates. (By
*properly*, I mean in the worst case, stable; in the best case, as good
as the "easy" filters). The theory is very, very solid on this. Have a
look at Dick Roberts and Cliff Mullis's book -- they give the most
elegant geometric explanation of the phenomenon to date. But try the
expermient yourself. It's maddening!
(In case anybody is wondering, I have to live with these rates because
of DSD. It wasn't my choice!)
> What bad thing to look for when operating it? Bad frequency response?
> Nonlinear distortion? Bad phase response?
Overwhelming noise and graininess on the output, clipping, limit cycles,
grunge, etc. Just make sure you force the transition band to be tight.
Or, if you like, try a peaking filter centered at 28 Hz with a Q of 10
or greater and a boost of 8 dB. And just because you don't hear any
"nasties" doesn't mean the filter is really doing what you asked of it!
I've seen many a designer cheat and implement a fatter filter in order
to avoid the problem.
Or, it may indeed be the case that your program *does* work properly, in
which case the designer applied some real TLC to the DSP programming.
But it's still waaaaaaay harder to make it work at higher Fs.
enjoy!
Glenn @ Z-Systems
dan lavry
> In article <Tt_wb.350$Wn1...@reader1.news.jippii.net> tomm...@suomi24.fi writes:
>
> > Dan, you have explained your point about 192kHz problems very well!
> > It's sad to see some of these so-called 24-bit/192kHz converters which
> > really only have dynamics of about 100dB with bad distortion etc., marketed
> > with a 24/192 tag.
>
> They get away with it because people see "24/192" and don't see the
> actual dynamic range specifications.
You are correct. First, as I stated before, all the 192 gear and IC's
I saw, and I look all the time (!!!) is specfied with A weighting.
Most of the other gear 96KHz gear is is specified A weighted. Next,
many of the AD converter companies just copy the IC specifications as
if it "exsists in mid air". You look at the finished design, including
the front end circuitry, power supply, clocks... you will be surprised
at the discrapancy there.
I am not saying the "ordinary specs" are the only thing that matter.
There are a lot of things not on the spec sheet, and should be there,
that make it or break it.
Here is an example: I am often amazed by how folks insist on ,1dB
flatness response. Not that it is difficult to do. I glad to comply
with that. But did you look at a speaker resonse latley? It is up and
down by dB's. It is a mess. Somewhere smack in the middle, there is a
crossover network that breaks that tone made out of say 500Hz into
(Assume 1.2KHz cross over for the example):
A. 500Hz fundumental and 1KHz first harmonic go to to the large cone.
B. 1.5Khz, 2KHz,.... 3rd forth and so on harmonics to the smaller
speaker cone
And guess what, there is a huge phase shift in there around the cross
over range...
And guess what else, that single say 500Hz piano note is played so
that some of the sound comes from the lower cone and the rest from the
upper cone...
If you play 100Hz, now you have the fundumntal and 10 harmonics or so
come out of the bottom large speaker...
When folks demand flatness I give it, but I do think they are barking
up the wrong tree. When the talk 192 I know they are full of it. When
they like a well preserved vinal, I at least understand some of what
they like. .
But transfering vinal to a 192 cracks me up.
BR
Dan Lavry
Mike Rivers
> You are correct. First, as I stated before, all the 192 gear and IC's
> I saw, and I look all the time (!!!) is specfied with A weighting.
I've noticed that too. It's amazing how much power supply hum goes
away with that filter, and stray clock hash too. The first computer
audio interface that I reviewed, the Echo Layla, on quick measurement
had an outrageous amount of noise, though it didn't sound noisy. A
look at the output with a scope showed lots of stuff above 22 kHz.
Applying a 20 kHz filter ahead of the meter let me get closer to the
manufacturer's published noise spec.
> many of the AD converter companies just copy the IC specifications as
> if it "exsists in mid air".
I've noticed that, and also just using the theoretical number of
96 or 144 dB (depending on the word length) as well.
> I am not saying the "ordinary specs" are the only thing that matter.
> There are a lot of things not on the spec sheet, and should be there,
> that make it or break it.
The sad thing is that a lot of people who buy this stuff have nothing
to go on but specifications. When you're dealing with people with so
little experience and knowledge that they think the only difference
between a mic and a line input is the size of the connector, it's easy
to absorb a lot of irreleveant information and then be confused when
trying to sort it out.
> Here is an example: I am often amazed by how folks insist on ,1dB
> flatness response. Not that it is difficult to do. I glad to comply
> with that. But did you look at a speaker resonse latley?
There must be something to frequency response. I hear people talk all
the time about fixing a "problem" in mastering by boosting something a
couple of tenths of a dB. I can't say as I hear it myself, but then
I'm just average.
Scott Dorsey
> complete not in sync with the sample clock.
>
> To get a fine none-overshouting (pre/post) sampled signal a high
> sample rate is required to get ride of the ringing (about 10 times
> related to burst sine freq.).
This is what bandlimiting does. That ringing is the result of the
bandlimiting, and the stuff you are seeing is all well above 20 KHz.
Dave Collins
wrote:
> There must be something to frequency response. I hear people talk all
> the time about fixing a "problem" in mastering by boosting something a
> couple of tenths of a dB. I can't say as I hear it myself, but then
> I'm just average.
People say a lot of things. What the those engineers should realize, is
that a tenth dB "correction" is absoultely meaningless in the real
world! No speaker, ear, room, etc. is nearly that accurate...
It's just a desire on the part of the engineer to do _something_ rather
that cut it flat.
DC
Arny Krueger
news:bq2mcs$5an$1...@panix1.panix.com
> Tommi <tomm...@suomi24.fi> wrote:
>
>> It's sad to see some of these so-called 24-bit/192kHz converters
>> which really only have dynamics of about 100dB with bad distortion
>> etc., marketed with a 24/192 tag.
>
> Well, there are other converters that are spec'd at 24/192 that offer
> 120+ dB of dynamic range, very fine distortion specs and very low
> noise. And they sound fine.
>
> What surprises me most these days are the folks who complain about the
> sound of high res digital audio but extol LPs (vinyl) with their 30 dB
> max of separation and surface noise.
It's got to be the sentimentality factor. It ain't the technical
performance, that's for sure!
Mike Rivers
> > There must be something to frequency response. I hear people talk all
> > the time about fixing a "problem" in mastering by boosting something a
> > couple of tenths of a dB.
> It's just a desire on the part of the engineer to do _something_ rather
> that cut it flat.
Wait a minute. Aren't you a mastering engineer? But then most of us
have "fixed" something by turning a control on something that we
didn't realize was bypassed.
--
I'm really Mike Rivers (mri...@d-and-d.com)
Bob Cain
Mike Rivers wrote:
>
> Wait a minute. Aren't you a mastering engineer? But then most of us
> have "fixed" something by turning a control on something that we
> didn't realize was bypassed.
Whew! I'm glad to hear you say that. I feared I was the
only one who had fallen into that trap. :-)
Bob
--
"Things should be described as simply as possible, but no
simpler."
A. Einstein
Dave Collins
wrote:
> Wait a minute. Aren't you a mastering engineer? But then most of us
> have "fixed" something by turning a control on something that we
> didn't realize was bypassed.
Yes, I am. One that is realistic enough to know that you can't "fix"
anything by boosting 0.1dB... I think it's more of a canine-equine show
for the client "Ooh, listen to the air when I add .2 @20k Q1" That's
bound to increase sales.
Back when everything sounded great, the eq's were in 2dB steps.....
DC
Roger W. Norman
news:znr1070019730k@trad...
> Wait a minute. Aren't you a mastering engineer? But then most of us
> have "fixed" something by turning a control on something that we
> didn't realize was bypassed.
All too often.
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
Roger W. Norman
the drag coefficient, weight and balance, and g force so that the bigger
engine doesn't turn in any better time. Like in drag racing, after 190 mph
it costs an addition 14 horse power to get even one mph more. But if you
don't spend the money to redesign the drive chain, etc., the extra horse
power won't get you going any faster.
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
"Mike Rivers" <mri...@d-and-d.com> wrote in message
news:znr1069853646k@trad...
Mike Rivers
In article <dcollins-A02768...@news06.west.earthlink.net> dcol...@earthlink.net writes:
> Yes, I am. [a mastering engineer] One that is realistic enough to know
> that you can't "fix" anything by boosting 0.1dB...
> Back when everything sounded great, the eq's were in 2dB steps.....
Back then, in mastering, what you "fixed" wasn't what's wrong that's
keeping the record from selling millions, you fixed what would keep
you from cutting it properly. When you got the pressings, you expected
them to be missing a little of what you heard in the recording studio.
You didn't expect them to sound a whole lot better.
Roger W. Norman
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
"Mike Rivers" <mri...@d-and-d.com> wrote in message
news:znr1070112071k@trad...
dan lavry
> Hp Widmer <h...@no.junk.com> wrote in message news:<a1g9svci8v2d06k3u...@4ax.com>...
> > I did some measurements using an analog burst 10khz signal who is
> > complete not in sync with the sample clock.
> >
> > To get a fine none-overshouting (pre/post) sampled signal a high
> > sample rate is required to get ride of the ringing (about 10 times
> > related to burst sine freq.).
>
> This is what bandlimiting does. That ringing is the result of the
> bandlimiting, and the stuff you are seeing is all well above 20 KHz.
> --scott
It is easier to relate to the ringing with low frequency squate waves.
That 10KHz burst is "mixing up" a lot of things.
Say you take a 1KHz square wave. Say we talk about 44.1KHz. Than there
are 11 harmonics (decaying in amplitude). For square wave only odds
exist (no even harmonics) and as you add them, you find the decaying
ringing, and yes it is at high frequency. If you go to 100Hz, you have
110 harmonics.
As scott pointed out, the added harmonics are never enough to express
a square wave. With limited bandwith, the missing high frequncy energy
shows up as deviation from perfect square, by means of high frequency
overshoot and ringing. the DIFFERENCE between the perfect and
bandlimited square wave IS HIGH FREQUENCY CONTENT.
BR
Dan Lavry
Hp Widmer
Hi Dan,
>It is easier to relate to the ringing with low frequency squate waves.
>That 10KHz burst is "mixing up" a lot of things.
The real question will be : Bandlimit vs. transients given from some
natural instruments and the required sample rate. In other words how
do we really hear or we sime mask above 1xkHz.
When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
California) or Neil Young (Harvest) there is so much more MUSIC coming
out...
Hp
Roger W. Norman
> When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
> California) or Neil Young (Harvest) there is so much more MUSIC coming
> out...
>
>
> Hp
>
Than what? How do you make a comparative when you give nothing by which the
comparison is made?
Scott Dorsey
>When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
>California) or Neil Young (Harvest) there is so much more MUSIC coming
>out...
How much of this is because the DVA-A is well done, and how much of it
is because the CD release was just butchered?
The CD release of Hotel California is just nasty-sounding. Harsh and
screechy, but with no real detail. It sounds nothing at all like the
original LP. That's no slam against the CD format, that is a slam against
the people who horribly bungled the CD release.
On the other hand, does it really matter? If the DVD-A sounds good, that
is a reason to go to DVD-A, whether it's a technical improvement or just
a social one. I mean, I'm the first one to acknowledge that there are serious
problems with LPs, but I'm not getting rid of my turntable because of the
number of recordings out there that sound better on LP because the CD release
was so poorly handled.
Rob Adelman
Roger W. Norman wrote:
>>When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
>>California) or Neil Young (Harvest) there is so much more MUSIC coming
>>out...
> Than what? How do you make a comparative when you give nothing by which the
> comparison is made?
Then before he hits play? ;)
Rob Adelman
Scott Dorsey wrote:
> The CD release of Hotel California is just nasty-sounding. Harsh and
> screechy, but with no real detail. It sounds nothing at all like the
> original LP. That's no slam against the CD format, that is a slam against
> the people who horribly bungled the CD release.
I have this LP. While the songs are classics, and some really good stuff
I was always disappointed with the overall sound. It seems a little 2d
to me. I never had the CD but I can imagine.
-Rob
Justin Ulysses Morse
> How much of this is because the DVA-A is well done, and how much of it
> is because the CD release was just butchered?
>
> The CD release of Hotel California is just nasty-sounding.
Yeah, but I'm pretty sure the Eagles played on the DVD-A release also.
ulysses
Remixer
that 44.1k sampled reproduction does not retain the detail, air, and open
sounding quality of an analog source. This is not the case with 2Fs or 4Fs
sampling, at least not in practice. You get to recognize the closed in,
blurry quality of CDs in general vs the more natural sound of DVD-A or SACD.
Before someone rises up to smack me down with the sampling theorem, and how
1Fs is all we'll ever need because of the bandwidth the human auditory blah
blah blah, the fact is that in *practical implementation* the promise of the
sampling theorem is not fulfilled, even with the best available 1Fs dacs.
I've recently been doing some tests with 44.1k 16-bit recordings upsampled
to DSD with a new "trellis" algorithm from Philips. The playback of the
upsampled-to-DSD files through a DSD dac has better detail and transient
response than what can be heard from the original 44k file played through a
variety of the best 1Fs dacs in the industry. This may be due to the fact
that there is actually more information encoded in a good 44.1k recording (a
la the sampling theorem) than we can playback with real world
top-of-the-line PCM dacs, but more of that information does come through the
upsampled DSD playback. Not trying to prove anything, only reporting what I
can hear.
"Roger W. Norman" <rno...@starpower.net> wrote in message
Rob Adelman
Remixer wrote:
> Not trying to prove anything, only reporting what I
> can hear.
But, but, but..... ;)
dan lavry
You seem to be insisting that in suport of 192KHz despite the fact
that it is against all aspects of physics and math. There is a
research artical done with the greatest care, that after searching
around for tons of cases, talks about some special case showing a
muted trumpet yielding -55dB ay around 45KHz and almost zero at 90KHz.
(96KHz AD will be enough for that).
The mics used are special measuring types. Not the kind folks use for
recording. "Ordinary" mics drop off way below the 120KHz test mic. The
speakers do also. The human ear does too.
So here is the old BS about the mysterious transients. Transients are
subject to bandwidth limitations. It may be true that a Fourier series
is based on periodic waves so it is not the way to approach
transients. So what? So you use different math. Nyquist was still
right. You bandlimit the signal to a frequency called Nyquist, and
sample over twice that rate, and you got ALL the information in the
data. Not just sine waves that last for ever. The transients you hear
are made out of energy in the same bandwidth called audio.
I know I did not say it that way earlier. I though it would be enough
to say that if the mics don't pick it, the speakers do not sound it...
it places the same limitation on the bandwidth of all music - which
includes transients.
Anything outside constant preiodic waves that started at the beginng
of time and will last for ever, includes some transient energy! I
guess you are trying to talk about "fast transients". There are huge
misconceptions around that. For example, many folks thing that a
bandlimited (say to 22KHz) 1KHz square wave rises faster (more slew
rate - volts per second) than a sine wave at 20KHz. I can go on but
will not.
I have been designing for a long time and know about transient
problems such as capacitor dialectric absorption in sample hold. That
is not a 192KHz related. It is at the oposite side - low frequencies
way within the ausio band. If it fixing it took changing from 96 to
192K sampling, you would not hear it! That is the whole point. 192KHz
will not fix a thing for audio.
Folks keep throwing misleading garbage based on lack of understanding,
and I can not answer it for ever. I need to work for a living, but I
am right, so the billion dollar conglamerates will chock me and the
truth but constatly throwing garbage at such volume that it will be
unanswered.
And whan it all gets too much for them to handel, there will always be
that last resort. That last argument that no one can stand to: It
sounded better. This argument will yield best results if you get some
real big names. Guess what- I have some real big names agreeing with
me that 96KHz is better than 192. I did not ask permision to say who,
but I put my inegrity on the line.
In either case, I came accross a big name that refused to audition 96
on the grounds that 192 contains "something" that staed with the music
at 44.1 (after decimation. Now, I will not ever calim that I know 1%
about recording when compared to that gentelman. I just wish that I
was recipicated with the same respect regarding Math and Engineering.
I would like to belive that there is a differance between the
stereophile and the pro community. Not all stereophile stuff is bad,
but much of it is very ridiculess. You do not hear a differance
between a yellow and orange cable. Arrowa on speakers is insanly
stupid. The argument is always about "It sounds better to me"...
Followed by someone getting ripped off.
Folks this is about getting our industry back on track. Double the
data. double the processing and LOWERING THE QUALITY OF CONVERSION is
a high price to pay. Science, and egineering and math on one hand, led
by greats such as Shanon and Nyquist, and the billion dollars
industial companies using every trick in the book (including avoiding
the physics and engineering books) to sell thiere BS. I belive it is
about marketing and money.
I realize that if one fell for it, it is difficult to admit. But
192KHz for audio is a crock! The only argumet left is "but I like it".
No one can stand to that one. If I tell you you added 10% distortions,
and you say "I like it", you win. So lets compromise: I say 192KHz s a
crock. You agree it is but you get to like it.
Sorry for my tone. I put out some energy to do good. I lost a big sale
because I refuse to play that 192KHz game. I try to educte folks about
it. First it was "more dots are better". Than it was "better
antialiasing". Than "transients". All very wrong and based on far from
sufficient understanding of the basics. What's next?
I am still here, but getting tired of it.
Dan Lavry
Bobby Owsinski
danl...@mindspring.com (dan lavry) wrote:
While I don't even dream to have even a portion of the technical
knowledge and experience that you have, Dan, I have done emperical tests
of the same material at 48, 96 and 192k.
We used acoustic sources; piano, light percussion (bells, shakers,
etc.), acoustic guitar, cymbals, room tone and a ping-pong match. Every
person involved (5 of us) agreed that there was no comparison; all 192k
material from every source sounded far better than the others.
After listening to the 192k material, we were able to pick out
undesirable characteristics at the lower sampling rates that were not
that evident before. 96k had a slight hardness in the upper mids and
48k was really closed, small and hard (not scientific descriptions, I
admit) with even more of the upper mid edgyness. 192k had none of the
hardness that the lower sample rates exhibited. After listening to the
192k stuff, 48k was practically unlistenable and 96k only slightly
moreso. In blind tests, the musicians especially picked the 192k
material every time.
So why is this? I always thought it was because of the higher sampling
rate but if you can provide an alternative explanation I'm certainly
open to it.
Until then, I've got to say that 192k sounds A LOT better to me than
anything else.
--
Bobby Owsinski
Surround Associates
http://www.surroundassociates.com
dan lavry
> Before someone rises up to smack me down with the sampling theorem, and how
> 1Fs is all we'll ever need because of the bandwidth the human auditory blah
> blah blah...
And the laws of garvity? and basic math? and the whole of science just
because you hear something good enough to have every thing wrong. This
is pretty disrespectfull to engineering and math.
Why do not figure what is it about your setup blha blah blah blah...
> the fact is that in *practical implementation* the promise of the
> sampling theorem is not fulfilled, even with the best available 1Fs dacs.
We are talking 192KHz fs, not a 1fs DA. It has been years simce 1fs
DA. Stay on the subject. You are mixing it up blah blah blah blah
> I've recently been doing some tests with 44.1k 16-bit recordings upsampled
> to DSD with a new "trellis" algorithm from Philips. The playback of the
> upsampled-to-DSD files through a DSD dac has better detail and transient
> response than what can be heard from the original 44k file played through a
> variety of the best 1Fs dacs in the industry.
The best 1fs Dac in the in the industry is long gone. Wake up to
upsampling. The problem with 1 bit DA was the need for exreamly high
order anti imaging filter. So we went to upsdampling. Also, mixing 16
bits into the argument is out of place. Clearly 20 bits is better than
16. We are talking sample rates.
> This may be due to the fact
> that there is actually more information encoded in a good 44.1k recording (a
> la the sampling theorem) than we can playback with real world
> top-of-the-line PCM dacs, but more of that information does come through the
> upsampled DSD playback. Not trying to prove anything, only reporting what I
> can hear.
I really do not think you know how significent the sampling theorm is.
It is not about a couple of jerks drinking beer a spouting garbage. It
is about the fundumentals of modern signal theory, by great minds. Not
much of the electronics around you would work if the therory was
wrong. Which camp do I respect? Reality science math and engineering.
And a few top notch audio engineers that know their stuff.
Dan Lavry
Remixer
"dan lavry" <danl...@mindspring.com> wrote in message
news:673b149b.03113...@posting.google.com...
> And the laws of garvity? and basic math? and the whole of science just
> because you hear something good enough to have every thing wrong. This
> is pretty disrespectfull to engineering and math.
>
> Why do not figure what is it about your setup blha blah blah blah...
>
I realize my use of "blah blah blah" was sufficiently ambiguous to offend
just about anybody, but it was not intended for Mr. Lavry and was not meant
to disparage any of the well-reasoned arguments he has put forth in this
thread.
> We are talking 192KHz fs, not a 1fs DA. It has been years simce 1fs
> DA. Stay on the subject. You are mixing it up blah blah blah blah
>
This thread has already split off into tangents. My post was in response to
Mr. Norman's questioning of HP's ability to recognize the perceived
superiority of a DVD-A without the benefit of a direct comparison under
scientifically controlled circumstances. (That's why I back quoted Mr.
Norman's post.) Sorry if it was taken any other way.
> The best 1fs Dac in the in the industry is long gone. Wake up to
> upsampling. The problem with 1 bit DA was the need for exreamly high
> order anti imaging filter.
The pcm dacs in my test were all oversampling dacs. I guess I should have
been more explicit in comparing the DSD upsampling of a 44.1k source against
the same source played through an upsampling PCM dac. Yet another tangent
and another target for a flame war.
> I really do not think you know how significent the sampling theorm is.
> It is not about a couple of jerks drinking beer a spouting garbage. It
> is about the fundumentals of modern signal theory, by great minds. Not
> much of the electronics around you would work if the therory was
> wrong. Which camp do I respect? Reality science math and engineering.
> And a few top notch audio engineers that know their stuff.
>
Again, no attempt to question the validity of a proven mathematical theorem,
but it is all too often used as a blunt instrument to bludgeon those who
find fault with 1Fs recording and reproduction, upsampled or not.
Mr Lavry, please don't get caught by the remarkable power of newsgroups to
get one's dander up. With so many cross-currents going on at once, and
without the benefit of nuances, it is easy to take offense when none was
intended, or even directed.. Your recent posts have already caused me to
rethink a lot of what I have taken for granted about 4Fs. Take it all in
stride and keep contributing. Every little bit of good information helps
somebody somewhere in spite of the punting, trolling, and sniping that does
go on.
Rob Adelman
dan lavry wrote:
> And the laws of garvity? and basic math? and the whole of science just
> because you hear something good enough to have every thing wrong. This
> is pretty disrespectfull to engineering and math.
The problem is that on paper, you could have 2 sets of criteria, one
seeming to be far more significant than the other. But in reality, the
lesser one could have far more significance in the way the brain
interprets it for reasons we just don't understand.
-Rob
Bob Cain
dan lavry wrote:
>
> "Remixer" <rem...@gaol.com> wrote in message news:<c3qyb.173334$Gq.21...@twister.nyc.rr.com>...
>
> > Before someone rises up to smack me down with the sampling theorem, and how
> > 1Fs is all we'll ever need because of the bandwidth the human auditory blah
> > blah blah...
>
> And the laws of garvity?
Hey, dood, I dropped a feather and a BB that weighed the
same from my window yesterday and they definitely did _not_
fall at the same rate. Those Norton and Bernstein guys got
something wrong.
Seriously, it is a pleasure having someone with your
recognized authority debunking all the nonsense that plugs
the gullet here. I wish you luck in convincing people that
they should be looking for the factors that are really
causing the perceived differences.
Dave Collins
> I've recently been doing some tests with 44.1k 16-bit recordings upsampled
> to DSD with a new "trellis" algorithm from Philips. The playback of the
> upsampled-to-DSD files through a DSD dac has better detail and transient
> response than what can be heard from the original 44k file played through a
> variety of the best 1Fs dacs in the industry.
First of all, you don't have a 1fs dac. 64fs is entry level, yours is
probaby more. Second, that "better detail and transient response" could
just be distortion that you like better. Nothng wrong with that. A
little 3rd harmonic added sounds like detail.... How much new music
would you say the upsampling creates?
I guess the question is: If you run it through the trellis 20 times,
does it just get better and better?
DC
Remixer
> probaby more. Second, that "better detail and transient response" could
> just be distortion that you like better. Nothng wrong with that. A
> little 3rd harmonic added sounds like detail.... How much new music
> would you say the upsampling creates?
>
> I guess the question is: If you run it through the trellis 20 times,
> does it just get better and better?
>
> DC
I don't think it's 3rd harmonic distortion, it doesn't really sound like
that well known effect, (I've been doing this since before the Aphex Aural
Exciter and HEDD) and doubtful that Philips would let enough distortion
creep into their rather pricey trellis algorithm to make such an audible
difference.
Tommi
"Bobby Owsinski" <poly...@earthlink.net> wrote in message
news:polymedia-D565A...@news01.west.earthlink.net...
> While I don't even dream to have even a portion of the technical
> knowledge and experience that you have, Dan, I have done emperical tests
> of the same material at 48, 96 and 192k.
>
> We used acoustic sources; piano, light percussion (bells, shakers,
> etc.), acoustic guitar, cymbals, room tone and a ping-pong match. Every
> person involved (5 of us) agreed that there was no comparison; all 192k
> material from every source sounded far better than the others.
>
> After listening to the 192k material, we were able to pick out
> undesirable characteristics at the lower sampling rates that were not
> that evident before. 96k had a slight hardness in the upper mids and
> 48k was really closed, small and hard (not scientific descriptions, I
> admit) with even more of the upper mid edgyness. 192k had none of the
> hardness that the lower sample rates exhibited. After listening to the
> 192k stuff, 48k was practically unlistenable and 96k only slightly
> moreso. In blind tests, the musicians especially picked the 192k
> material every time.
>
> So why is this? I always thought it was because of the higher sampling
> rate but if you can provide an alternative explanation I'm certainly
> open to it.
>
> Until then, I've got to say that 192k sounds A LOT better to me than
> anything else.
Where do I begin?
What converters did you use? What speakers? Mics? Preamps? How can you say
anything about the characteristics of different sample rates without any
valid data?
It's like saying your kitchen television has a better picture than the
monitors at Lucasfilm, no one can argue about your opinion, but there is no
science backing it up.
Remixer
the converter variable, and compared with the feed from the studio,
eliminating the mic and pre-amp variables. BTW, when you go to the
supermarket and ask for change for your twenty do you demand proof of the
validity of integer arithmetic? At least some of the people who post here
are professional listeners who don't need to prove the earth is a planet to
take a walk. Dismissing the reasoned observations of working engineers
because they're not EEs under a double-blind test is such an old saw and not
particularly productive.
"Tommi" <tomm...@suomi24.fi> wrote in message
news:DxCyb.20$xj...@reader1.news.jippii.net...
KikeG
> In blind tests, the musicians especially picked the 192k
> material every time.
So you are saying you were able to pick apart the 48/96/192 KHz
material, under blind conditions, with no problems at all, being the
differences evident, aren't you? This is an extraordinary claim, that
I'd like to know more about. Were the test samples properly level
matched (< 0.1 dB difference) and time aligned? Was the test
double-blind? How many trials and correct identifications did you get?
Could you give us more information about the equipment used at the
test? Could you provide us with same of the samples used at the test?
Dave Collins
"Remixer" <rem...@gaol.com> wrote:
> Dismissing the reasoned observations of working engineers
> because they're not EEs under a double-blind test is such an old saw and not
> particularly productive.
I'm a professional listener, not an EE, yet a working engineer. Somehow
it makes me more skeptical......
DC
Roger W. Norman
> This thread has already split off into tangents. My post was in response
to
> Mr. Norman's questioning of HP's ability to recognize the perceived
> superiority of a DVD-A without the benefit of a direct comparison under
> scientifically controlled circumstances. (That's why I back quoted Mr.
> Norman's post.) Sorry if it was taken any other way.
What, was my Dad here? <g>
The basic premise of language is to convey ideas and concepts, create common
ground and to be able to make verbal comparisons that represent intangible
items. An example of the latter would be something as "An apple tastes
better THAN shit." If one simply says "An apple tastes better" particularly
with the appended "..." then we don't have a comparison. Of course, the
other aspect is that when one makes a comparison they have the experience of
both the apple's taste and that of shit's taste.
Now one could make the digital to digital assumption that Hp was saying his
comparison was against CDDA, but it might have been against Sony's PCM-F1
playback, or MP3, or any number of current or previous digital formats.
So, as you may see, I wasn't making a statement about his injection of
comment on the subject, nor questioning his ability to even be able to make
a comparison, I was simply saying that the language of the comment did not
make a comparison. He allowed you or I to infer one and make our own
decision about what the comparison meant.
Roger W. Norman
Unfortunately it had to float on the sea, of which it obviously did a pretty
piss poor job.
--
Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.
"Rob Adelman" <SPAMLESS...@mn.rr.com> wrote in message
news:bqegfh$21jemr$1...@ID-75267.news.uni-berlin.de...
Remixer
experience been in comparing 1Fs, 2Fs, 4Fs, and DSD formats? What gear and
conditions have led to your opinions?
"Dave Collins" <dcol...@earthlink.net> wrote in message news:dcollins-
Remixer
cylinder either for that matter.
"Roger W. Norman" <rno...@starpower.net> wrote in message
news:bqfc30$g8f$1...@bob.news.rcn.net...
Tommi
> Presumably they used the same converters at each sample rate, eliminating
> the converter variable, and compared with the feed from the studio,
> eliminating the mic and pre-amp variables. BTW, when you go to the
> supermarket and ask for change for your twenty do you demand proof of the
> validity of integer arithmetic? At least some of the people who post here
> are professional listeners who don't need to prove the earth is a planet
to
> take a walk. Dismissing the reasoned observations of working engineers
> because they're not EEs under a double-blind test is such an old saw and
not
> particularly productive.
Yes, but the point was that no-one knows _why_ the 192kHz sounded better.
If there is nothing to back it up except your ears, I don't think there's
any serious reason to think 192 kHz must be better. All sorts of tests are
done all over the world all the time, each with different results. Of
course, your ears are the most valuable tool at your disposal in the music
world, but ears can deceive you.
Take psychoacoustics for example; there actually were some "serious" tests
in the beginning of the mp3 era which claimed that downgrading a 16/44.1 wav
to an 128bps results in no audible change in most cases..
Likewise, if there's no science to back up 192, it all just comes down to
the "it sounds good" argument. What "sounds good" can be anything from
harmonic distortion to a gentle background hiss, it can be ultrasonic noise
downmixing to audible band in the air, etc. The point is that what "sounds
good" to the human ear doesn't mean that it is the truth about the sound
source. Heck, phaser can sound good when we use it in the mix to "widen" a
sound, but phase shift certainly isn't a thing we want in our amplifier
characteristics.
If a man who designs converters for a living says that 192kHz converters
_currently_ have much more jitter than 96kHz ones, that they're not good
enough for audio just yet, why couldn't you take his word for it? Science
usually beats our perception, even though we don't want always to believe
that.
The big leaps in our technology and knowledge always have always been due to
theoretical calculations, empirical measurements, and to a certain extent,
our own perception. There are always folks who base their opinion on one or
two of the above things, but the real pioneers always check that their
"thing" (whatever it is) is correct with all three aspects. Einstein was
one, but Aristotle for example wasn't since he relied mostly on perception
and logic..
Bobby Owsinski
"Tommi" <tomm...@suomi24.fi> wrote:
There is no science, just listening under normal studio conditions on a
PT HD rig. All sources were recorded with the same signal chain and
reproduced with the same signal chain. Signal path was a very nice C12
through a Hardy mic amp directly into PTHD. Playback was out of the PT
HD into an SSl 9k out to Genelec 1031's, HD-1's and large soffit mounted
monitors (dual 15's with an SLS ribbon tweeter). No EQ or dynamics in
the signal path.
Before anyone else gets upset about this, please just go and listen for
yourself. I clearly heard a difference. Maybe your will or maybe you
won't. But for now, I can hear it.
Bobby Owsinski
ki...@arrakis.es (KikeG) wrote:
These were unscientific tests of a PT HD system, just to see if we could
hear any real differences between the sample rates. We could with no
problem, but there was no attempt to calibrate, level match, etc. Just
listening under somewhat normal studio conditions.
Agreed, you could drive a truck through the scientific holes in the
testing, but we all thought that the differences were not subtle.
Scott Dorsey
>
>Before anyone else gets upset about this, please just go and listen for
>yourself. I clearly heard a difference. Maybe your will or maybe you
>won't. But for now, I can hear it.
I did, and I heard a difference, but it was no greater than the difference
between different A/D units at 44.1. Admittedly this is a fairly big
difference, but if the converters all sound different then the sample
rate differences are the least of our worries.
dan lavry
> After listening to the 192k material, we were able to pick out
> undesirable characteristics at the lower sampling rates that were not
> that evident before....
>
> So why is this? I always thought it was because of the higher sampling
> rate but if you can provide an alternative explanation I'm certainly
> open to it.
>
> Until then, I've got to say that 192k sounds A LOT better to me than
> anything else.
Great! We are finaly getting somewhere! Alternative explanation! That
is what I have been saying all along.
It all started really bugging me when I heard folks say that when
recording at 192KHz, than decimating down to 44.1KHz for CD, they
still had that special 192KHz "quality" in the music. I heard it from
more than one person I rerspect.
I did not stop respecting them as great recording and mastering
engineers, but I did start on that path, claiming that there is no way
on earth that a 44.1KHz CD can ever contain anything with higher than
22.050KHz content.
So I thought it would be easy to state that it can not possibly be
about that "extra bandwidth". "Easy my foot"! I am now hearing
arguments similar to: I have the best supply and the best meter and it
measured 5.1V. It should be 5V. Can anyone there do something about
Ohms Law?
One of the last posting was calling someone to do something about that
sampling theorm. There are probably a quater million EE's in the IEEE
alone. Probably over a million EE's out there, in medical, telecom,
instrumentation, video, and yes audio and more. Are we talking about
50 million engineer years? And no one wants to step up to the plate
and chalange Nyquist?
It is like siding with the primitive tribal medicine man, against
what? Penicilin?
A couple of weeks ago, in Times science special publication, it stated
(based on a pole) that 1/2 of Americans belive in ghosts. So what am I
complaining about?
Back to the issue at hand - alternative explanations:
We need to agree on the bandwidth needed (for a given application,
video, audio, instrumentation...). There is no escape, we need to do
that! Going with too little, you loose importent data. Going with too
much you end up with too much data (large files) and difficult
processing. If your sensor can not go above 1MHz, sampling at much
higher than 2MHz is a waste. If your sound source and microphone are
limited to say 40KHz, than 96KHz is fine.
Once we agree on bandwidth, let us not get confuse how to get there. A
much higher front end sampling rate with less bits for further
decimation is a common method, and it helps make the anti aliasing
filter order into a non argument. The sampling rate will be the data
output rate.
Within the context of the given bandwidth, and that includes ALL music
including ALL the transient and horns and bells... there are 2
fundumental approaches:
1. Get the output waveform to track the input waveform as precisly as
possible.
2. Have some distortions.
To have 1. you need linear phase, low noise, quick recovery from
overdrive, lowest jitter possible, great components and material and
so on and on....
To have 2. opens a huge a vast field to talk about. Is there a type of
distortion that you like? You all know what I am saying - tube sound,
2nd and 3rd harmoics, wide main lobe of a SRC, jitter sidebands...
Some of such distortions are better understood than others in terms of
hearing and little is documented, thus a designer with more experience
can have some ideas for what combinations or tradeoffs to make. That
is where analog becomes importent...
I like to have that "waveform in" is the same as "waveform out", but I
do face tradoffs all the time. I do have some ideas of where the
tradoffs should be, and I am not going to spell it out and let the
competion catch up.
The 192KHz does create more distortions than 96KHz. Again, more speed
results in less accuracy, regardless to what voodo one throws at it.
There is a certain nature to those distortions, something to do with
"pushing" things a certain way. For example, If you are trying to
charge a cap via a reasistor, the longer you wait, the closer you will
be (EE call it "how many time constants"). This example is
logarithmic! If you have an OPamp with a certain Bode plot
(characteristics), how will it settle? The longer time (slower), the
further say 2nd order rolloff. It is a mess to figure out everything,
but I see some light, in terms of explaining it.
Clearly one way is to measure the outcome of 192KHz. 96, 44.1 and see
what happens. That is why I am dicusted with the A weighing specs of
192. It hides things.
So I will not answer your question as to what you hear in 192. I will
not answer what you hear in 44.1 or with a single transistor. But
since it is not about added conent that slower system can not
accomodate, than it is about DISTORTIONS.
I learned a long time ago that distortions is not always a bad word. I
can give you the same distortions you like at 192 with a 96K system,
without the penelty of twice file size, double processing requirnment.
The "Forces to be" are trying to say otherwise so they can sell you
that 192 gear. I think it is a bad thing to force a whole industry to
192KHz to get that distortion.
I do not tell anyone what to do. You want to use 100MHz and down
sample to 44.1KHz, fine! You like 1% distortion? Who am I to object?
If it sounds good, I'll buy the CD. I just don't like to have folks
twisted into beliving that the 192 gives better representaion of the
signal, when in fact it is the oposite.
I am of the opinion that folks that alrerady comitted to 192 will be
resisting my arguments. But with 192KHz it is distortions you hear,
and it does not take high sampling rate to make the sort of such
distortions. The 96KHz provides much better "waveform in" = "waveform
out", and if the job is about picking up air vibrations (sound) and
making it as identical as possible in playback, 192KHz is barking up
the wrong tree.
Again, I know some distortions are fine and fall under the category of
artistic decision. In fact, much of mastering is about just that -
musical taste. I apppreciate a well recorded and master CD. It is one
thing to have less distortions than make the adjustments you want.
Once stuck with distortionms, you can not remove them. I would hope
that there are other toys out there to play with, without the need to
fall for what I think is a marketing scheme attempting to move a whole
industry into a horrible mess - double the data and double the
processing (buy all new gear) and get more distortions for it.
So what is the next argument going to be? I already answered the
issues:
more bandwidth
more points is better
transients
Let's someone adjust that gravity stuff so we can fly
I do not want to say too much about the ego aspect (I can hear it so
no one tells me nothing).
BR
Dan Lavry
Glenn Booth
In message <bqfrfs$4ti$1...@panix2.panix.com>, Scott Dorsey
<klu...@panix.com> writes
>Bobby Owsinski <poly...@earthlink.net> wrote:
>>
>>Before anyone else gets upset about this, please just go and listen for
>>yourself. I clearly heard a difference. Maybe your will or maybe you
>>won't. But for now, I can hear it.
>
>I did, and I heard a difference, but it was no greater than the difference
>between different A/D units at 44.1. Admittedly this is a fairly big
>difference, but if the converters all sound different then the sample
>rate differences are the least of our worries.
Taking it a step further, in my experience it's quite possible for the
same converter to sound different at different sample rates. Who's to
know how much testing a given 192kHz capable converter gets at 96kHz?
--
Regards,
Glenn Booth
Chris Hornbeck
wrote:
>We need to agree on the bandwidth needed (for a given application,
>video, audio, instrumentation...).
I'm having a very hard time following this discussion. The numbers
(44.1, 96, 192, etc.) are thrown around as if everyone but me
understands the useage.
With modern ADC's these are not the input sampling rate. No one
has discussed processing. If they're related to the output
resampling rate, that's been left unclear, and is also unlikely
with modern DAC's.
So, what ARE we talking about?
Thanks,
Chris Hornbeck
"That is my Theory, and what it is too."
Anne Elk
dan lavry
> Yes, but the point was that no-one knows _why_ the 192kHz sounded better.
> If there is nothing to back it up except your ears, I don't think there's
> any serious reason to think 192 kHz must be better. All sorts of tests are
> done all over the world all the time, each with different results. Of
> course, your ears are the most valuable tool at your disposal in the music
> world, but ears can deceive you.
YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION, THERE IS NO
DIFFERANCE BETWEEN A 48KHZ BANDLIMITED SIGNAL SAMPLED AT 96KHZ AND A
48KHZ SAMPLED AT 192KHZ. YOU CAN USE 2 POINT TO DESCRIBE A LINE, OR
DOUBLE TYHE FILE SIZE TO 4 POINTS. IT IS A STRIGHT LINE. IN THEORY
THERE IS NO DIFFERANCE.
IT IS NOT THAT WE CAN NOT EXPLAIN IT. ASUMMING PERFECT CONVERSION.
THEY WILL YIELD THE SAME SIGNAL OUT OF THE DA. NOT A FIMPTO VOLT OF
DIFFERANCE!!!
SO ANY DIFFERANCE IS ABOUT PRACTICE, NOT THEORY. AND WE CAN EXPALIN IT
JUST FINE:
1. 192 YIELDS MORE NOISE, THOUGH YOU CAN NOT HEAR MUCH OF IT
2. 192 IS LESS ACCURATE AND THAT IS WHAT YOU HEAR.
IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
EE". YES THE EARS ARE VERY VALUBLE FOR MUSIC. AND ALSO, DO NOT ASSUME
THAT A GOOD CIRCUIT DESIGN DOES NOT TAKE ENGINEERING AND TECHNICAL
KNOWHOW. THEY ARE BOTH IMPORTENT.
AGAIN: IN THEORY WE KNOW THAT THE OUTCOME WAVE OUT OF THE DA WILL BE
THE SAME IDENICAL OUTCOME. IT IS NOT THAT WE DO NOT KNOW HOW TO
EXPLAIN THINGS. THIS IS CAST IN STONE. IF I GIVE YOU A SCOPE PROBE AND
YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
"I CAN NOT EXPLAIN IT BUT THERE" IS NO STRANGER TO ALL OF US IN AUDIO.
BUT IS NOT ALWAYS THE CASE. WE DO KNOW A COUPLE OF THINGS! SUCH AS
NYQUIST, AND OHMS LAW AND 1+1=2. I AM NOT STATING THE THEORY OF
RELATIVITY HERE. JUST FUNDUMENTAL STUFF.
BR
DAN LAVRY
LeBaron & Alrich
> Seriously, it is a pleasure having someone with your
> recognized authority debunking all the nonsense that plugs
> the gullet here. I wish you luck in convincing people that
> they should be looking for the factors that are really
> causing the perceived differences.
Amen. Now what's the best ADC for less than ten bucks? <g>
--
ha
LeBaron & Alrich
> Before anyone else gets upset about this, please just go and listen for
> yourself. I clearly heard a difference. Maybe your will or maybe you
> won't. But for now, I can hear it.
I'm not upset about this, nor do I have access to a Digi 192 rig. But it
does occur to me that a manufacturer offering a "more is more so it's
also better" product might well, in businesslike self-interest, might
try to make sure the "more" also sounded "better".
So one might wonder if PT 192 sounds "better" than Paris at 48, for
example.
--
ha
dan lavry
Sorry yiou are confused about it. I am pretty sure I made a clear
distinction between the input sampling and output sampling. I did it
at least twice, pointing out that the input sample is the one
determinimg the requirnment for anti aliasing filter. I also stated
that modern AD's mostly sample at 64fs to 256fs input rate thus the
antialiasing filter does not need be high order.
You sound like you understand it, so I will assume you just did not
read my comments. I do not balme you - they are long.
Regarding the procesing, input rate output rate and the rest, I have
been trying to explain here that most conversion starts with at most
few bits at high rate, and than we go through a process of "trading
off" output speed for
performance. The old 1 bit at 64fs can be made to say 105dB at 44.1KHz
(noise start rising at 22Khz). Of course one could have opted for say
a 96KHz output rate with 48KHz, but at lower accurcy. Or a 192KHz with
96KHz at yet lower accuracy. Folks tend to relate to the AD at the
output rate, so a 192KHz is the AD with 192KHz output (what you called
resampled, I called decimated, and we both call output rate).
Your comments give me one more chance to say the same thing but from a
different point of view: The whole concept of the modern AD, be it 1
bit or a few bits at high oversampling is based on "moving unwanted
energy" of low bits perfrmance from a low frequncy band (audio) to
high frequency (above audio). The amount of tradeoff (thus the final
outcome) is greatly a matter of oversampling ratio. It also depends on
the noise shaping order and other factors. But all things being equal,
when the ratio of sampling rate to what we decide to call audio
bandwidth is high, we get better results. If you started at 64fs 3
bits and wanted to keep 64fs your audio bandwidth, you get 3 bits! Go
to 32fs, and you get a few more bits and so on. 192KHz is twice the
rate of 96KHz so it will not yield as good performance to say 40KHz...
So the same converter at 96 is better than 192KHz. And it should sound
so unless someone did something to skew the test. Who would do such a
thing? Even those that stand to gain a lot of money selling that
192KHz crock would not go that low. I did not test for that, so I will
assume they did not.
BR
Dan Lavry
Tommi
> YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION, THERE IS > NO
> DIFFERANCE BETWEEN A 48KHZ BANDLIMITED SIGNAL SAMPLED > AT 96KHZ AND A
> 48KHZ SAMPLED AT 192KHZ. YOU CAN USE 2 POINT TO DESCRIBE A LINE, OR
> DOUBLE TYHE FILE SIZE TO 4 POINTS. IT IS A STRIGHT LINE. IN THEORY
> THERE IS NO DIFFERANCE.
> IT IS NOT THAT WE CAN NOT EXPLAIN IT. ASUMMING PERFECT CONVERSION.
> THEY WILL YIELD THE SAME SIGNAL OUT OF THE DA. NOT A FIMPTO VOLT OF
> DIFFERANCE!!!
Erm..You're either replying to the wrong post, or missed the tone of my
text.
I did never say that 192kHz sounds better to me, OR that faster sampling
results in better accuracy! I was just replying to the post where someone
said that.
> SO ANY DIFFERANCE IS ABOUT PRACTICE, NOT THEORY. AND WE CAN EXPALIN IT
> JUST FINE:
> 1. 192 YIELDS MORE NOISE, THOUGH YOU CAN NOT HEAR MUCH OF IT
> 2. 192 IS LESS ACCURATE AND THAT IS WHAT YOU HEAR.
> IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
> EE". YES THE EARS ARE VERY VALUBLE FOR MUSIC. AND ALSO, DO NOT ASSUME
> THAT A GOOD CIRCUIT DESIGN DOES NOT TAKE ENGINEERING AND TECHNICAL
> KNOWHOW. THEY ARE BOTH IMPORTENT.
Did I assume that? Did I EVER say that "the ears are better than EE"??
My point was EXACTLY the opposite, my point was that I understand people who
say that their ears hear a difference at 192, but that if there's no science
to back it up, then the "better" results they hear are due to
psychoacoustics! Harmonic distortion can have a pleasing effect, but
converters should be as close to the truth as possible without adding any
elements.
> AGAIN: IN THEORY WE KNOW THAT THE OUTCOME WAVE OUT OF THE DA WILL BE
> THE SAME IDENICAL OUTCOME. IT IS NOT THAT WE DO NOT KNOW HOW TO
> EXPLAIN THINGS. THIS IS CAST IN STONE. IF I GIVE YOU A SCOPE PROBE AND
> YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
> ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
> HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
> "I CAN NOT EXPLAIN IT BUT THERE" IS NO STRANGER TO ALL OF US IN AUDIO.
> BUT IS NOT ALWAYS THE CASE. WE DO KNOW A COUPLE OF THINGS! SUCH AS
> NYQUIST, AND OHMS LAW AND 1+1=2.
I really don't understand this reply. Either I didn't express my opinion
correctly, or it was severely misunderstood. Either way, I don't disagree
with any of the above since my point was more or less the same.
Luke Kaven
> the converter variable, and compared with the feed from the studio,
> eliminating the mic and pre-amp variables. BTW, when you go to the
> supermarket and ask for change for your twenty do you demand proof of the
> validity of integer arithmetic? At least some of the people who post here
> are professional listeners who don't need to prove the earth is a planet to
> take a walk. Dismissing the reasoned observations of working engineers
> because they're not EEs under a double-blind test is such an old saw and not
> particularly productive.
There are a lot of fallacies in this:
1) Using the same converters at different sample rates does not
eliminate the "converter variable" if there were such a thing. There
may be a quantitative shift in performance of the same device that can
not be explained by the mere doubling/halving of sample rate. Those
quantitive shifts are what the experimenter needs to somehow control.
The secondary problem is how one knows that one has controlled all
relevent variables, and that may not even be possible to know.
2) Asking for proof of integer arithmetic when getting change at the
supermarket is not the same as verifying the validity of one's
auditory judgments. First of all, integer arithmetic is not an
empirical phenomenon. It is true in all possible worlds. What is at
issue here are *empirical* questions, which are contingent, contingent
upon all kinds of things.
3) Being a professional listener says nothing about the validity of
one's auditory judgments. Auditory judgments are among the most
unreliable there are, even among experts, because they are subject to
many kinds of short term to long-term context effects.
4) Nobody here claimed that only electrical engineers were qualified
to make such judgments. Anyone schooled in experimental method can
suggest a set of controls. But because the phenomena in question is
empirical, we can only control for things *so far as we know*, unless
there are *analytical* facts (true in all possible worlds) supporting
it.
I've got to conclude that your "argument from common sense" does not
work.
Luke
Luke Kaven
[...]
> There is no science, just listening under normal studio conditions on a
> PT HD rig. All sources were recorded with the same signal chain and
> reproduced with the same signal chain. Signal path was a very nice C12
> through a Hardy mic amp directly into PTHD. Playback was out of the PT
> HD into an SSl 9k out to Genelec 1031's, HD-1's and large soffit mounted
> monitors (dual 15's with an SLS ribbon tweeter). No EQ or dynamics in
> the signal path.
>
> Before anyone else gets upset about this, please just go and listen for
> yourself. I clearly heard a difference. Maybe your will or maybe you
> won't. But for now, I can hear it.
One needn't deny that you *heard* a qualitative difference. The claim
is that you haven't successfully *explained* what you heard, and you
have given no grounds to support any claim of what the *causes* are.
This is not to deny your *practical knowledge*, but as we know,
practical knowledge is partly the act of reasoning under uncertainty,
with all the error that entails.
Luke
Luke Kaven
> Agreed, you could drive a truck through the scientific holes in the
> testing, but we all thought that the differences were not subtle.
You're committing the same fallacy again. You have no grounds to
attribute those difference to any particular causes, whether sampling
frequency, or elsewise. Your implied reference to sampling rate in
citing "the differences" is vacuous. Undoubtedly, you are an
impressively skilled audio engineer. But that is not the same thing
as science, and science is what you are trying to argue.
Luke