>> While the truth is that higher sampling rates unequivocally result in =
>> greater bandwidth,
>
> I feel greater bandwidth is is not the primary reason for going to higher=
sampling rates. Doubling the sample rate - and this should be obvious fro=
m the term - doubles the number of times per second the original analog si=
gnal is sampled, giving a more accurate representation of the information i=
n that waveform. There is an article on the WSRS website written by Nagra'=
s John Owen that makes a similar point:
>
> "The sampling frequency and frequency response go hand-in-hand really, an=
d although using the Nyquist theory, 44.1 kHz is sufficient to record perfe=
ctly up to 22.05 kHz bandwidth, using higher sampling frequencies does appe=
ar to reconstitute the original sound more accurately."
John is a good friend, but he also is in the business of selling higher-sam=
ple-rate recorders...note the weasely "appear to reconstitute." No real dat=
a to support the claim.
The idea that higher sample rates are more accurate in the human audio band=
is a myth of digital recording that will never die. There is no difference=
except wider bandwidth. Wider bandwidth is a good thing for nature recordi=
ng, though.
> The benefit of recording at 24bit is largely dependant on the capabilitie=
s of the A/D used in the recorder. Many of the so-called 24bit recorders, e=
specially if the design is a few years old, have A/D signal to noise ratios=
of less than 96dB, so there is very little to be gained recording at 24bit=
. The FR2-LE uses an A/D chip with S/N of 95dB for example which effective=
ly makes it a 15-6 bit recorder. Raimund's recorder tests are performed at =
Message: 44.
Subject: 1khz/16bits (with one or two exceptions) which obscures the
performance =
advantages of those recorders which are capable of more than 16bit performa=
nce. The Olympus is a case in point, with substantial improvements to perf=
ormance when tested at 24bit.
Even with a noisy recorder, a 24-bit master has more room for processing la=
ter, no harm in turning it up in post.
It's a dirty little secret of audio engineering that even the best state-of=
-the art A-D converters only deliver about 21-bits of actual signal-to-nois=
e. That's more than we need, anyway.
-Dan
|