Many years ago at Sony Music Studios we did some blind listening tests
comparing the original analog tape source with 96 KHz and 192 KHz material
at 24 bit resolution. The "Expert Listeners" were able to identify the 96
KHz material consistently, but could not reliably tell the difference
between the 192 KHz 24 bit and the analog source.
The company compromised at 96 KHz 24 bit, and actually I believe now
digitize at 88.2 KHz because of the ease of downsampling to CD. As far as
Nyquist, one of the flaws is using 2 X Max frequency is that this
theoretical approach is based upon a single frequency source which is
present while the samples are being taken. I know from many listening
tests that anything less than 88.2 KHz 24 bit is easily spotted. The point
about bit depth is correct, even 20 bits could be heard fairly easily.
Now regarding sampling rate, I think of this from the time domain
perspective. It's not so much the frequencies, but rather the phases and
spacial detail that is captured at the higher rates. We heard subjective
comments such as, "it feels warmer", "has more depth", and "punchier".
These are all qualities which would be captured by the faster sampling
Sound travels at 340 m/s so at approx 100 Khz sampling rate we would be
grabbing around 3.4 mm of spacial information each sample. This results in
exceptional clarity of a piece of music. This is observed even more so
when you listen to classical music recordings and choral works. The
spacial information produced creates some stunning examples.
Malcolm F. Davidson
Yes but as several have commented, a sampling rate of 44.1 or 48 WILL
> reproduce all audible frequencies, (unless you're a dog), accurately.
> When it comes to high resolution audio formats, (SACD or DVD audio), the
> bit depth plays a far more important role in producing the high quality
> audio than does the sample rate. 16 bit audio can only produce roughly
> 65,000 levels between noise floor and clipping; 24 bit audio can produce
> around 2.5 million levels. This is particularly important with classical
> music which is almost never at full amplitude, and, hence, almost never
> using all 16 bits on a CD. Any musical instrument produces a string of
> harmonics which are progressively lower level as their frequency
> increases; thus on a regular CD, the highest frequency harmonics, which
> represent the most complex component of the entire audio signal, although
> still well within the frequency range of CDs, are at such a low level that
> they are only read very poorly using a couple of bits. However when these
> audio signals are recorded using DSD or 24bit or higher bit depth, this
> harmonic content is
> far more accurately reproduced, giving the impression of an enhanced high
> frequency content, no matter where your personal hf cutoff point is, (I
> think mine is somewhere between 14k and 15k).
>> From: Mark Durenberger <[log in to unmask]>
>>To: [log in to unmask]
>>Sent: Friday, August 17, 2012 11:42:07 AM
>>Subject: Re: [ARSCLIST] Bit rates over 48 kHz
>>Best explanation I've seen for doing this!
>>-----Original Message----- From: Toby Seay
>>Yes, 44.1KHz and 48KHz cover the entire hearing spectrum. However, it
>>not cover it well. Resolution gets worse as you go higher in the
>>spectrum. Therefore, really high frequencies only get sampled a few
>>For instance, a 44.1KHz sample rate is only capable of sampling a 12KHz
>>frequency 3.675 times a second. That leaves you with a very poor
>>digitization at 12KHz. 96KHz does a much better job at providing a
>>representation of higher frequencies.