Hello all,

I just ran a few tests on my new Edirol UA-20, which supposedly
samples with 24-bit resolution. I say "supposedly", because yes, it
does create samples of 24 bits resolution, but it also creates noise.
So much noise that when viewing the spectrum, 16 bits and 24 bits
looks almost exactly the same (3 dB difference here and there).

So as far as I can tell from my own device, the 24 bits thing is
nearly a hoax. It's a lot of trouble to handle these 24 bits samples
(softwarewise), and all you get from these extra 8 bits is the
machine's own noise.

Has anyone bothered to check this up except me?

The test I did was to take the "Input Level" knob to minimum, and
sample with a quiet amplifier attached. I know the noise doesn't come
from the amplifier, because the noise level doesn't depend on the
"Input Level" knob's position (to be precise: it vanished about in the
middle of the knob's scale).

Now look at the spectrum. It should be 48 dB lower than 16 bits
capture. By far it isn't.

Or another, simpler test: If one looks at the digital samples, as a
list of numbers, with no desired signal, the number should be
constant. If it changes just a little (say, 45 plus minus 4) then the
samples are fairly quiet. But the 24 bit samples jumped thousands.

In short, is it only me, or is the 24 bit concept nothing more than a
marketing ploy?

Eli