My white paper "Sampling Theory" is at
www.lavryengineering.comunder "suport". I posted it a few weeks ago, and have since been involved in conversations on the proaudio NG, comp.dsp, mastering web board. I am pleased with the response. A number of highly respected engineers know the push to 192KH is about manufacturers making money, not about scientific and engineering issues.
The few that argue on behalf of 192KHz talk about "things we do not know" and about listening tests. Such listening tests, taking a device designed to do 192 and switching in a final X2 decimation to 96KHz is NOT VALID, because it tests the last stage decimator and nothing else. The real test should be based on the comparing a 192KHz device against a 96KHz device. By the time you do 192KHz, way too far from what is optimal for audio, the damage is already done.
I am not saying that all distortions are bad. There are those that like tube distortions and more. I am saying that whatever you hear is NOT outside the audio hearing range, thus can be contained within 20KHz or so, certainly within 44-48KHz (88-96KHz sampling). Moreover, in audio, the lowest bandwidth device is the bottleneck (week link in the chain). Lets see:
1. Mics? Who is using 96KHz mics? Most Mics drop at about 20KHz...
2. Speakers? Who is using 96KHz mics? Most Mics drop at about 20KHz...
3. How much energy do musical instrument put out above 40KHz?
4. Ear - my dog does not hear 48KHz
So all of a sudden we are presented with that gizmo- 192KHz?
It is true that there is a lot we can not explain. I have been catering to the ear for many years, and do not know all the answers. But that does not mean that everything is fuzzy. There are some things we know, and know well. We know that sampling must exceed twice the bandwidth of interest. Not by much, just a tiny bit.
Only in audio there is such a disconnect between content bandwidth and sampling. No one else (medical, instrumentation, telecom, video) goes nuts with faster sampling because it yields no positives, only negatives:
1. Less accuracy (there is always a tradeoff between speed and accuracy)
2. More data (storage, data transfer...)
3. More processing required (often traded with lower quality processing)
Again, for those that like a certain distortion associated with 192KH sampling, it is all contained under 20KHz or so. One should not take a whole industry into having to lower transparency, twice the storage requirement and so on. Why not instead, manufacture that distortion (if you like it) with a 96KH device? (96KHz is already an overkill, 60KHz would have been an optimal rate, taking care of ALL the issues including filters, pre ringing and there is nothing else to worry about!).
My paper may be of interest to some, while other "ear types" may find it "too much". I kept the math to a minimum, so it is about graphs, plots and text. I put some energy to explaining what Nyquist finding is all about (he was a major contributor to modern technology).
Common sense may be misleading folks to think that "the more the better". What is true for say pixels and video or computer screen, is not true for sampling limited bandwidth signal. Making an analogy here is wrong! You need 2 points to draw a straight line. No need for more. You need 3 point for a circle. Well, the bandwidth restriction ends up with: you need only to exceed twice the highest frequency you deal with, thus 88.2KHz accommodates 44.1KH of audio.
I have heard numerous folks that understand math, engineering and science talk about the 192KHz being all about selling new gear to make money. In fact, there are some well respected engineers making such statements on the PRO Audio NG right now. I talked to a lot of folks that admitted privately that they are afraid, uncomfortable, consider it unwise... to raise objections to their employer. I see a strong correlation between the who promotes 192 and who sells such gear. I get ZERO scientific arguments suggesting that my comments are not on solid grounds. I get some folks that want to promote 192 to IGNORE all the science and engineering and math. All I hear is "maybe it is something we do not understand", and “golden folks hear this” (a 192KHz design, against a 192KHz design plus a less than perfect decimator to 96KHz, which is how you listen to a final stage decimator). Folks, if that is what you like, lets do a real good X2 up-sampler than use the imperfect X2 down sampler (decimator), and there is your 192KHz sound. This is the ONLY VARIABLE in most of the listening test. Many folks reported that they like the sound of that X2 last stage decimator. They were confused into thinking it is a test comparing 192KHz to 96KHz.
I am sad to see audio going in the wrong direction. I am not going to be a part of it, though it does impact my economics. The upside: I can look at myself in the morning. I did not sell out.
I few preemptive comments:
1. Nyquist is immuned to real life imperfections. All the noise, be it analog or quantization or any linearity causing distortions follows Nyquist. It is solid and does not need verification for every (or any) piece of gear.
2. Again, I do not tell folks what they hear and what they do not. The argument is not about what is good or bad. It is about signals distortions and what not. The standard for “people with ears” is to go for a listening test. My points are valid independently of listening test, because I stay away from tests and good vs bad.
It is possible to figure some things with math. I stop where math and science and real world engineering do, and it still lets me tell a story full of facts.
BR
Dan Lavry
Lavry Engineering