Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Ashermusic]

Brusby wrote on Fri, 06 January 2006 03:00

Well, I promise I'm trying to get out of this discussion . . . 'just can't help myself One last point: danlavry wrote on Fri, 06 January 2006 00:08 I do not just ?dismiss those reports out of hand without first conducting some scientific inquiry to determine their validity?! I worked a whole year on developing a 192KHz gear, which I decided after much examination not to sell. I applaud your efforts.  But, I think I've been suggesting (maybe inartfully) that it's a different inquiry which needs to be made, an inquiry to identify precisely WHAT is it that people have been hearing. Here's the bottom line:  No one has yet identified whether differences some people report noticing between hi rate sampling and low are real or imagined, and if real, what the cause is.  It's entirely possible it's all just voodoo and all those people have been deceiving themselves.  But I think you must admit, we just don't have answers to those questions yet and no one seems interested in a pragmatic search to find out.   Where are the double blind tests to see if people can repeatedly identify hi sample rate conversions versus low?   Without first determining whether the effect is real we can't possibly proceed to uncovering a cause. Heck, Dan, if you could prove the claimed phenomenon is just an illusion that would end the debate and we could all get a little rest.

Brusby, people hear all kinds of things. I have had people seriously maintain that running an in the box mix through any mixer, even a cheapie, makes it sound warmer, wider, etc. Which is frankly silly.

They also see all kinds of things. Thousands saw Elvis alive last year.

What most of us want in converters (and mix monitors) is accuracy, period, and as Dan says, we turn to other things to color and individualize our sound. 192 is apparently simply not as accurate as 96.

So I really don't need a double blind study to determine why some people like 192. If it is measurably a less accurate sample rate I am not interested.

[Arf! Mastering]

Ashermusic wrote on Thu, 05 January 2006 23:17

So I really don't need a double blind study to determine why some people like 192. If it is measurably a less accurate sample rate I am not interested.

I wish I could be as sure as you are about this.  The problem is that I know so many experienced, professional, dedicated, seasoned, no-nonsense recording engineers who have listened and all say the same thing - there are aspects of a 4Fs recording that their ears tell them to be more accurate. It's easy to dismiss them all - the math says they must be wrong, that the perceived improvement in realism is an artifact.  So, how to resolve the contradiction between what is measured and what is perceived? The converse question - is there something key that is not being measured?  Dismissing the question will not make it go away.  I'm starting to gather some data in source/a/b tests with a number of converters.  If anything notable comes of it, I'll post the details.

[danlavry]

AlanS wrote on Fri, 06 January 2006 06:26

Ashermusic wrote on Thu, 05 January 2006 23:17 So I really don't need a double blind study to determine why some people like 192. If it is measurably a less accurate sample rate I am not interested. I wish I could be as sure as you are about this.  The problem is that I know so many experienced, professional, dedicated, seasoned, no-nonsense recording engineers who have listened and all say the same thing - there are aspects of a 4Fs recording that their ears tell them to be more accurate. It's easy to dismiss them all - the math says they must be wrong, that the perceived improvement in realism is an artifact.  So, how to resolve the contradiction between what is measured and what is perceived? The converse question - is there something key that is not being measured?  Dismissing the question will not make it go away.  I'm starting to gather some data in source/a/b tests with a number of converters.  If anything notable comes of it, I'll post the details.

Hi Alan,

You said:
“It's easy to dismiss them all - the math says they must be wrong, that the perceived improvement in realism is an artifact.”

I say:
Let me make a point for the sake of accuracy and clarity: the “math” does not say that people claiming to hear something are wrong! The “math” (and I would include “signal theory” and “information theory”) say that if the IMPLEMENTATION is perfect, the signal will be perfectly transparent. The “math” says that the closer you are to perfect implementation, the more accurate and transparent the transfer function – the sound going through the device. And yes, the “math” and the engineering theories tells one in great detail what kind of imperfection, and how much of it one can expect for a given deviation from perfect implementation.

You said:
“So, how to resolve the contradiction between what is measured and what is perceived? The converse question - is there something key that is not being measured?  Dismissing the question will not make it go away.”

I say:
Resolving the contradictions? In a world where golden ears have been spending dozens of years listening to the sound of re4sistors, caps, Fets vs. transistors, opamps, power supplies…  I would expect a device as complex as a AD (made out of many resistors, caps, transistors, opamps…. power supplies) to have some sonic signature.

One question is: does the industry have sufficient background to conclude that there is such a thing as a “particular 192KHz sonic signature”. A few dozen AD makers use an AKM IC, the others use a Cirrus IC. So with the same 2 IC’s, do you think you can conclude that what you hear is due to faster sampling? Or is it due to the use of multibit conversion while utilizing new semiconductor processing…. (I can list dozens of different things that make these IC’s different than previous technology).

You see, the industry is full of terribly wrong conclusions regarding components, architecture approaches…

My favorite example is a person that after a lot of listening concluded that some opamp sounds great. The conclusion is “opam XXX is great”. But the test was conducted with “SETUP A”: an opamp in an inverting configuration, which does not required good common mode. The power supply was exceptionally clean, which did not require good supply rejection. The signal level was high, which did not demand low noise characteristics, and the load was 100KOhms which did not require strong drive….

Now, the same OPamp in “SETUP B” (another circuit) may be real bad news. Say you place it in a non inverting configuration, somewhat noisy supply, for processing a small signal level and driving say 600Ohms. This will require 4 major characteristics that were not required by the “SETUP A”.

So is opampXXX great? Is it a “dog”? There are thousands of opamps out there, each one with tones of characteristics and various sets of tradeoffs. Selecting a proper one is a job for a designer equipped with a lot of analog expertise, to “integrate” the proper parts into a given circuit design.

There are tons of “old wives tales” floating around in audio, regarding caps, resistors, transistors, opamps… I do not question the listening reports and the sonic preferences, be it objective or subjective. I often do take issue with the CONNECTIONS that are being made between the sonic reports and their cause. The ear is the best tool to tell you how things sound, but it is not the proper tool to tell you why.

You can certainly listen and decide for yourself what sound you like and prefer. But can you really stand by a statement CONNECTING what you hear to a higher sample rate? Say for example, that the coloration you like is due to “pushing a sample hold circuit” a bit too hard (too fast). If so, do you need faster sampling? Why not screw up the sample hold circuit performance at low frequencies, to give you the sound you like? My point is –  there are too many variables to reach a conclusion.

My issue is: Industry forces including large manufacturers and some visible but unqualified individuals decided to steer audio to higher rate. Therefore, the new generation IC’s are being made to cover 192KHz. The proper thing to do is to optimize the new IC’s for operation around 88-96KHz  (or lower if there were a 60-70KHz standard). Audio is being compromised by marketing.

You said: is it possible that we do not know how to test things? I have been in that pro audio environment for a long time, and it is very often that we hear about the conflict between measurements and sound. Yes, proper measurements and interpretation of the measurements is a whole subject by itself. It is not a simple and easy matter, and there is much room for evolution and improvements. There are many that do not know how to measure, what to measure and what is really important. But there are people that DO KNOW A LOT, and the ongoing popular strong dismissal of the CONNECTION between measuring and what the ear hears is a dangerous one. It allows marketing and sales people to dismiss measurements that DO COUNT.

Back to the beginning of my post – the math and theory GURANTEES US that perfect implementation will yield perfect results. There is NO ROOM for any doubts here. We will NOT find some new earth shaking discovery about some “part of the signal” that has been “left out” or “mis-represented”. A deviation from perfect implementation is the only cause for deviation from perfect transparency. Knowledge of the imperfection is knowledge about the sonic deviation, not always easy to measure, but true. Going to faster rate then needed is forcing more deviation from perfect implementation. Listening to a huge number of variables will not tell you the specific cause of sonics.

All of what I said is not analogous to listening and liking the tube sound. In the case of tubes, one can do the math, one can measure signals with real hardware… we know a lot about what it is that causes the sonic alterations. We can CONNECT what we measure to what we hear. With tubes, we are not saying “we like it but we do not know why”. We may not know it all but we do know a lot about tubes. I can plot a non linear transfer function of a tube. I can account for other factors such as inter electrode capacitance and so on.

With 192KHz, we know nothing. We have no credible theory or engineering fact other then larger files, more processing and more errors, all down sides, no up side.

Note that no other industry using digital technology has gone to sampling at way too fast a rate then the required optimum. It is an audio industry related crock, and if we do not resist it, we will end up at 384KHz and then 768KHz…

Best Regards
Dan Lavry
www.lavryengineering.com

[Arf! Mastering]

Thanks for the reasoned reply, Dan.  I understand that the survey I am running cannot possibly isolate the sample rate as the single cause of any of the results - too many variables.  The listening tests will employ  four converter sets that run at 4Fs and two that run at 2Fs.  Perhaps some element will appear that is common to one group or another.

[squeegybug]

kraster wrote

Is there anywhere in your work as a designer that you must make an aesthetic decision after weighing up all the technical aspects? I know the ultimate goal is to reproduce the source signal without non-linearities but have you come across a point where you are up against the limits of components and had to make choices based on your aesthetic judgment?

How about here -- from the AD122-96 manual:

Quote:

Dithered noise shaping technology has been incorporated into a handful of hardware devices. While all are based on the same concepts, some perform better than others. After simulating and listening to all available public domain algorithms, Lavry Engineering came to some conclusions in forming a basis for Acoustic Bit Correction™. The  principal conclusions are: a. The practice of greatly amplifying low level signals to determine triangular flat PDF (probability density function) dither reveals the effectiveness of distortion and noise modulation elimination. This practice yields misleading results when testing unflattened dithers and/or noise shapers. It conflicts directly with L. Fielder’s findings showing completely different threshold delectability curves for quiet and loud levels. Noise shaping listening tests must be done at "reasonable" volume levels. b. Given the above requirement, our listening tests concluded a strong preference for "triangle high pass" dither (this dither is produced by simultaneously adding a new random number and subtracting the previous value). Such dither is frequency-shaped to carry more high frequency energy (the energy content at low frequencies is minimal). c. Listening tests revealed a preference for smoothly varying noise-shaping curves. Peaks and notches seem to irritate the listener (admittedly while turning the volume up). In addition, despite the temptation to optimize the noise shaping curve to the average listener’s hearing threshold, given a significant variation from listener to listener requires reasonable compromises in tailoring such a curve. In other words, smooth the curve. The improvements offered by dither and noise shaping vary with source material and final word length. An A/B/X test at 16-bit level, requires a quiet environment and low level (loudness) audio. The listener must resist the temptation to turn the volume up to unreasonable levels. The practice of truncating to short word length (8-12 bits) should be avoided. The ideal noise-shaping curve may be irritating at loud levels. Lavry Engineering’s listening tests were based on test tones and repeating loops of quiet passages of various material (mostly classical music) with flat amplifier response. Listening to test tones was straightforward: we used the Model AD122-96 MKIII test tone generator mode switching the Acoustic Bit Correction™ on and off. The frequency and amplitude programmability was very useful.

Steve

[danlavry]

squeegybug wrote on Mon, 09 January 2006 04:45

kraster wrote Is there anywhere in your work as a designer that you must make an aesthetic decision after weighing up all the technical aspects? How about here -- from the AD122-96 manual: Quote: Dithered noise shaping technology has been incorporated into a handful of hardware devices. While all are based on the same concepts, some perform better than others..... Steve

It is difficult for me to draw a line between "aesthetic decisions" and "audibility of art effect". In the noise shaping case, the idea is to "push the noise" from the range that the ear is very sensitive to (say 1-4KHz) to where the ear is less sensitive (very low or very high frequencies). There is a rather well know curve showing hearing sensitivity vs. frequency. But I was not sure I would be ready to "submit" the that curve, for 2 reasons:

1. The curve shape does change when you change loudness
2. The curve may be statistically correct, but what about people way outside the average? (Can you dress a 7 foot guy with a 6 foot pants?).

Therefore, I decided to provide 4 curves (very mild, mild, a bit aggressive and aggressive). After many years of feedback, I have some statistics about what curves most people like to use.

I do not make statements about people liking the sound of dither 1 more then dither 2. The design is driven by a concept, to shape the noise. It is very difficult to listen to that sort of low level truncation noise, unless one amplifies it a lot. It is understandable that such noise would alter the perception, and my goal is to make it disappear as much as possible.

I believe that music without truncation noise sounds better. Do you call it artistic decision? Or is it engineering decision.

True, a designer is going to face choices and possible compromises at every step of the way. If one faces say a cost driven a trad off between say a supper clean 4KHz and 20KHz, I would recomand to choose the clean 4KHz. Why? Because I know some fundamentals about the ear. Is that artistic or engineering driven? Where do you draw the line?

Regards
Dan Lavry
www.lavryengineering.com