Elektron Musical Instruments

Here's some thoughts on the audio sample rate influence on the... Audio. Are you one of those who thinks 96 khz (DVD-A) is more precise than 44.1 khz (CD)? Well, it's not, in fact, it's as precise as analog... Find out why.
Before going into the issue, let me let you know that I work in 96 khz, or 48 khz (enough) or above (192 khz for fun as I write), but not because 44.1 khz is less precise - the higher sample rate tends to compensate for worst A/D conversion stage, ergo, it's a cheaper way of making good converters, at the expense of more storage and CPU.
More ahead... In English, as much as possible.
Fourier and Nyquist
Nyquist tell us that to accurately get a sampled wave you must use a sampling frequency equal to twice the frequency you wish to sample. With 44.1 khz you can sample up to 22.05 khz bandwidth signals. Fourier tells us that every wave is but an addition of sine-waves at different frequencies and amplitudes. So, at any given point, what you sample is an addition of sine waves.
Ok, now lets introduce and examine false argument #1 towards 96 khz being more precise than 44.1 khz.
Maximum frequency
Each sample is a "photo" of the sound wave position in time. A sound is but a "wave" that wobbles up and down at a certain frequency and we hear it. Many will say that between a 44.1 khz sample and the next one there will be a lot of "unsampled" variations. Agreed. But now here's an issue...
All the things "in between" the 44.1 khz samples HAVE to be signals at a frequency SUPERIOR to 22.05 khz, hence, we don't care anyway! (Our ears, for that matter.)
Precision
So here's a shocking truth: the precision at 20 hz-20 khz of a sampled signal at 44.1 khz, 96 khz or even analog is EXACTLY the same, all are equally precise in the below Nyquist limit bandwidth.
Why 96 khz then?
Now things get complex, but first it's because it's cheaper to produce a more accurate sampler /converter using 96 khz, whereas to achieve the same degree of fidelity with a 44.1 khz one would require more expensive design.
But I digress, as I said, a sound wave is a wobbling signal that goes up and down. This up/down can vary up to 22.05 thousand times per second to be sampled by 44.1 khz rate accurately. But allow me to try...
Aliasing
Now imagine you have a signal that is running at 30 khz... The signal is a wobbling sound that goes up and down 30 thousand times per second... And this is a problem. Why?
From those 30 thousand times, Around 15 thousand are bound to e sampled erroneously by a 44.1 khz rate. It wont catch the 30 khz wave, but it will catch parts of it that will sound audible as a somewhat 15 khz or something sound wave (whistle, "chirping"). This is known as aliasing. The only way IS to STOP frequencies higher than 22.05 khz from ever getting in the sampled signal to begin with - using a brick wall filter - and HERE is where digital has its major handicap, and here is where the PRECISION starts to fail.
Brick wall
These brick wall filters are what make 44.1 khz signals lack in fidelity what 96 khz can provide, the "brick wall" filters are not "brick walls" but rather progressive filters that induce a lot of distortion, like phase shifts that can go up to 1000 degrees (unless they are really expensive)... At 96 khz, the brick walls are set at 48 khz (not 22.05 khz) and their effects (distortion) are NOT audible, hence why it's better (hence, cheaper).
Digital age
With sophisticated oversampling and digital filters, brick wall filters are not as distorting as they used to be so, with an irony twist, 44.1 khz is starting to sound as perfect as 96 khz, and as cheap (96 khz is more expensive on the long run, more CPU, more storage, and so on).
Psycho Acoustics
Nonetheless, ultrasonic material DOES alter audible one (know it for sure, if you modulate two ultra-sonic beams against each other, sub multiple audio waves come out of nowhere, or should I say, the interception point), ultrasonic waves may produce "beats" that carry down to audible frequencies coloring it.
But it's rare in what applies to sonic music material (the importance of having ultrasound recorded), an example is at an orchestra, when you catch the sound far away, ultrasonic content has already interacted between instruments and you get an accurate picture, even at 44.1 khz (the ultrasonic sounds already did the interaction work, before hitting the microphone).
But, for the sake of argument, if you close mic the instruments and later mix them at 44.1 khz, the ultrasonic part - that causes the coloring "beats" - will be missing, so no "coloring" will exist, as opposed as it would with a 96 khz sample/mix, whereas you have the ultrasonic info and it WILL color the mix with audible results - assuming your sampling did not cut ultrasonic frequencies to begin with (many do, even at 96 khz).
Although in the end, mind you, if you down sample the 96 khz ("beats" doing the work) mix down to 44.1 khz, it will sound the same. So we are back to 44.1 khz is enough, again.
So why go 96 khz?
Simple, or is it?
At 96 khz there is much bandwidth for plug ins to work in, if they're poorly engineered, it may make a difference. The real achievable difference for having high sample rates would be re-pitching or non-linear processes like compression (and I mean really heavy one, like 20:1).
The number of bits (24 vs 16) is actually the most important thing regarding fidelity, spreading rounding errors across such a vast space they wont make the a big difference in distortion (the more bits, the better). Time-stretch and pitch-shifting will benefit a lot on native 96 khz samples, but the rest won't a bit, and I mean volume control, EQ, reverb, and so on...
Does DVD-A makes sense then?
As far as a listening experience, the 24 bits make more of a difference than the 96 khz. But as said above, having native ultrasonic content alters the sound (although you can record that "altering" in 44.1 khz), and it will alter it according to your room acoustics IF you can play it back (and was sampled/mixed) at 96 khz.
96 khz will sound more analog (actually it feels more than it sounds), but its a very subtle effect and only shown off with very good mixes always done and preserving the full 48khz audio bandwidth. Today that's still hard to do since most mics don't go beyond 20 khz, as most synths (44.1 khz based) and guitar rigs (typical 15 khz bandwidth).
I believe and advise to always use 96 khz for the "cheaper and better" reason when it comes to converters (but expensive with computing resources), BUT only take it seriously if you use gear that goes up that high (48 khz) when capturing sound (produced by instruments with ultrasonic components captured by ultrasonic capable mics/devices).
Please mind this is a rather "simplistic" explanation and there are other math and physical phenomenon I've not talked about, but are not really that relevant and more technical oriented (limits on analog to digital converters and path stage electronics, to name one). 44.1 khz is enough, and if you don't use A/D converters, for example, you only use sound generators and virtual instruments, you can use ASIO 48khz, instead of ASIO2 96 khz for it saves you CPU and the end result is exactly the same when it comes to audio quality of the render - if it seems it does not (with some rendering synths, not samplers) it's because of internal reasons, not the sample rate (and typically 96 khz sounds warmer and fat - but it's actually just the "octave higher" effect, usually, try hitting the higher octave at 48khz and it will probably sound similar/the same), less than 96 khz sounds cold and thin (the octave lower effect), but you may want the later)...
So, unless you're writing some tune for dogs and bats, or working on ultrasonic experiences, I'd stick to 44.1 khz with no second thoughts, just don't up sample it and sell it as 96 khz, I really don't like that. My advanced resolution DVD-A content is all mostly based on content sampled/processed on native 96 khz resolutions or higher (and 24 bits and higher), so you will have that added value that can't be putted on 44.1 khz/16 bit, or can it?
Article featured at: http://www.alvaromrocha.com/en/2007/10/sample-rate-96-vs-441-khz.html
? 'lvaro M. Rocha 

http://groups.yahoo.com/group/sonicapro/