Noel Borthwick [Cakewalk]
Another way of wording this is by showing how latency is calculated.
 
At 44100 samples per second:
A 1024 (1K) sample buffer corresponds to 23.21 msec of latency
 
At 88200 samples per second:
The very same buffer size corresponds to 11.61 msec of latency
 
So you can see that doubling the sample rate halved the latency for the same buffer size.

Anderton

cclarry
So, no, it's not necessary...this debate has gone on for years...
If you go by the science...NO ...not required....

I don't think there's much debate that theoretically, higher sample rates aren't necessary. But there are so many wild cards. For example, one reason why some people might hear the difference between 192 kHz audio and 44.1 CDs in tests may have nothing to do with the sample rate, but instead be due to the 192 kHz signal being played from a hard drive, which has less jitter than something played back from an optical drive.
 
I couldn't hear a significant difference between 44.1 and 96 kHz until I started doing lots of ITB work with amp sims, virtual instruments, and dynamics processors. But it had nothing to do with human hearing, it was all about technological limitations that caused foldover distortion in the audible range at lower sampling rates.
 
Filtering has always been a consideration too, although filtering technology has improved dramatically since the CD was introduced. So the reason I'm curious is because some people swear they hear a difference with 192 compared to 96. In the case of the Be6 speakers, the response is only up to 40 kHz so in theory, 96 kHz and 192 kHz should have the high frequency components reproduced equally well.
 
The whole debate reminds me of cables. I was in a studio in Chicago and there was a vehement argument going on about whether cables made a difference. It was the old "it's just wire, you moron" vs. "but I can hear a difference." I finally stepped in and asked what the outputs and inputs feeding the cable were...and yes, with a tube amp and a long cable, capacitance can affect pickup tone...but with a high-output synth going into a mixer, "it's just wire."
 
It would be nice to determine once and for all whether people can hear a difference with double-blind testing that goes beyond Meyer-Moran, but it would be even nicer to find out why people hear a difference if there is a technological reason. I'm not ruling out sample rates per se, but I tend to think it might be something that's a byproduct of sample rates.
 
And I STILL think DSD sounds better than CDs...but in the immortal words of Herman Cain, "I don't have facts to back me up."

sharke
As an aside, am I right in assuming that if your intention is to slow audio down (time stretch), the end result will sound better for audio recorded at 96kHz than audio recorded at 48kHz? In much the same way as film recorded at a higher frame rate looks smoother when slowed down.

sharke
As an aside, am I right in assuming that if your intention is to slow audio down (time stretch), the end result will sound better for audio recorded at 96kHz than audio recorded at 48kHz? In much the same way as film recorded at a higher frame rate looks smoother when slowed down.

mike_mccue
It seems like a time stretch system could use oversampling when it's time to time stretch.
 
Just saying.

deswind
I am confused as to why a higher sample rate does not provide higher resolution?  If I take 12 samples per second.  It will sound very grainy and choppy, like looking at a very slow fan.  At one point does that graininess and choppiness go away?

lawp
here ya go:

(from http://cs.iupui.edu/~aharris/n351/notes09/MM01-Analog-digital.html via google)
what i know is that if i have 100 audio tracks to mix, i will get a better mix with higher resolution :)
and whether craig meant it or not, it really is in the maths - if not then why, for example, bother with the 64bit mix engine?

lawp
here ya go:

(from http://cs.iupui.edu/~aharris/n351/notes09/MM01-Analog-digital.html via google)

1. Unless effects, synths, etc. are scripted to process the data (the software/process comments made above), nothing can be gained from feeding it more. Most do not disclose what they process internally anyway, so I personally doubt many (if any at all) are written to process >96kHz.
2. The system resources consumed can get out of hand quickly, especially chaining processes which can process every sample.
3. What is to be gained from such? The core frequencies (and even harmonics in most cases) are <11kHz for most instruments, with pianos and synths being a notable exception. The human ear responds best to ~3KHz, and even a piccolo's fundamentals straddle this (http://www.independentrecording.net/irn/resources/freqchart/main_display.htm). An interesting game is to high pass something at 10K and listen/play with just that region (sort of eye-opening actually).
4. The ultrasonic discussion has validity to some extent, and cells can respond to higher frequencies provided enough power is delivered (as this is done in a medical application, but with very specific equipment); however, 1) I do not believe programs are designed to process such frequencies in the first place, and 2) speakers capable of playing such (at the proper power) are required.