OK, this thread may start a firestorm, but hopefully it will be an educated and experienced firestorm.

Before getting heavily involved in making music on computers, I worked for a company that was famous for building test instruments of all varieties. Their catalog included quite a few samplers, ranging from units with extremely long record lengths but slow samples rates (audio/mechanical range) to the fastest single shot digitizer in the world. It was an interleaved quad-pipe device which sampled at 4 gigahertz, and was used by scientists who studied things like exploding atomic bombs. I was the lead software engineer for that project, and I had the pleasure of working with some of the world’s leading hardware engineers who specialized in designing devices for digital sampling.

One didn’t work for this company without acquiring a basic understanding of Nyquist, and well as a rudimentary understanding of Fouier transforms, and if one had difficulty understanding the finer points of the math one only had to ask one’s colleagues for a very clear explanation. This company invested many millions of dollars in building samplers (digitizers). The price of just one of the above mentioned 4 GHz units could outfit an entire recording studio quite nicely

When I first got involved in music on my computer, 44.1kHz @ 16 bits was still the industry standard. 44.1 seemed liked an odd number, until I learned about it’s history as a compromise with other industries/standards. But, given that the range of human hearing is 20Hz – 20kHz, 44.1kHz was an adequate sample rate. And 16 bits, or 65,536 quantization levels, was more than adequate for human ears.

Then suddenly, 48kHz/20 bit equipment became the rage. And then it was 96kHz @ 24 bits. Now it's 192kHz. I can not understand why. It all just seems like a huge industry-wide marketing scam to me, to get folks to spend their money.

I finally did run across one audio situation where a faster sampling rate was justified, but it is not one that presents itself very often. When making digital recordings from vinyl, apparently the turntable stylus can generate frequencies higher than 20kHz, and to prevent those frequencies from aliasing and consequently distorting the audible result, a sample rate twice as fast as the generated stylus frequency is required. This situation can be generalized by stating that if one is digitizing audio, and frequencies higher than 20kHz are present in the signal, those frequencies could distort the audible result through aliasing. A faster sample rate is needed to accurately capture inaudible frequencies to prevent them from influencing the audible range. Note the key word “inaudible”.

How often are you digitally recording when frequencies above 20Khz are present? Unless you’re using a spectrum analyzer, how would you even know?

I’m not an expert, nor a mathematician. But I have worked with people who were both, who not only had confidence in Nyquist and the math involved, but who committed huge quantities of dollars to building instruments based on that math. I have conducted experiments of my own, recording audio signals at 44.1kHz and 48Khz. Subtracting one result from the other produces a flat line.

Over the years, I have asked a few recording engineers who “believed” in 96kHz/24 bits (or 192kHz etc.) to explain to me why it was important. They have all responded that they can hear the difference. I have asked them to demonstrate that difference to me. Only one person actually made the attempt, and I could not hear it, although they swore they could.

Which brings me to another issue. Before getting involved in computer science, I was a professional cellist with a major symphony orchestra. When it comes to the violin family of instruments, myths abound. Musicians are always mythologizing about Amati, Stradivarius and Guarneri, and constantly comparing their own instrument to those of their colleagues. Among the classical musician crowd, buying a new fiddle is almost as big an event as having a baby.

I have a devious streak in me, and I have conducted some comparison tests. “Here is the 1697 Stradivarius ‘El Grande’”. Play and few notes, gasps, smiles, swoons. “Here is a modern machine made instrument.” Plays a few notes, frowns, boos, ughs. You get the picture. Except that neither instrument was what I said it was, and the poorer quality instrument was played first, not second. The point? People (even well trained classical musicians) often hear what they think they ought to hear. Hearing is quite subjective, and effected by many other factors than just the molecules bouncing off the eardrums.

Now that brings me to the real reason I’m writing this post. To all those people out there who “believe” in 96kHz/24 bits - what do you do with the solid mathematics which clearly and repeatedly demonstrates that it’s pointless and produces no improvement (in the audio spectrum) over 44.1/16 bits? Yes, yes, I know, you swear you can hear the difference, and I’m not trying to convince you of anything. Really! I’m just extremely curious what you do, in your own psyche, with the mathematical evidence to the contrary. Do you secretly “believe” that the math is wrong? Do you think that audio engineers have discovered new evidence which the mathematical and scientific world have yet to incorporate?

The question is interesting to me not because I’m expecting to resolve this polemic in the audio industry, but because it serves as a microcosm model for a macrocosm problem – a world in conflict, caught between religion and Scientific Rationalism. I have met extremely intelligent people who were experts in their respective scientific fields claim emphatically that the world was created in seven days and all our problems can be traced back to Adam and Eve. If I can understand how experienced audio engineers who believe in 96kHz/24 bits reconcile their belief with Nyquist, I may be able to understand some other things as well.

Thanks for reading. Flames expected.
_________________
There are three types of people -
those who can count, and those who can't

Well, I generally record at 44.1KHz/24-bit res.

The increased bit-depth allows one to play with dynamic range more. That's mathmatical as well as noticeable to the trained ear of a good mix/mastering engineer.
But I tend to agree with the sample rate thing. Mics are the ears or your recording. Maybe if i had really awesome mics with frequency responses over 20K, I'd go better on the sample rate to avoid aliasing. But my mic collection doesn't give me that, so... if the mic doesn't reproduce over 20KHz, why should my A/D converters?

And given that most non-audio professionals have trouble hearing the difference between higher-quality mp3s and pcm wave files (or don't care about the difference), I don't see the consumer level standard for digital audio changing much in the near future beyond DVD-quality stuff.

Some very good words, you write beautifully.

Personally I usually stick with 44.1/16bit particularly as eventually I have to press a CD which is of course of those figures.

I think the main argument for higher bit rates is to allow for "smarter" algorithms which have more scope to truncate the least significant bits more gently. You'll notice many DSP processing sounds hideous as it truncates the binary word in a bad way.

High sampling rates of course yields siginificantly more bandwidth for additional DSP tasks.

I have to agree with the subjective hearing opinion though, many people do hear what they wish to hear. I have friends who make violin bows and instruments in the heart of Stradivarius country and I've created similar tests, always amusing.

Another way of looking at the dynamics is that if we are making Pop music we usually end up only using a few of the least significant bits anyway.

Good thread, hope some more get stuck in.
_________________
Recording Studio Suntans