wrote in message ...

Does anyone know if it is typical for an ADC to distort when driven near
full
scale (0dB)?

Yes.

I've got an Edirol UA5 USB audio interface. I put a 1600Hz sine wave in
the
line input (played from the line out of a Creative Labs Nomad Jukebox 3),
turned the level up to just under 0dB and recorded a few seconds of audio.

When I looked at the spectrum (Adobe Audition 1.5!) I see several
harmonics
(multiples of 1600Hz), and the largest one was only -40dB down or so. I
noticed these extra peaks went away once I lowered the input to say, -6dB.

Yes. That last dB or so before FS is a problematical area for many
converters. The better ones can be very clean until they are just a few
tenths of a dB below FS.

So, it seems to exhibit some kind of clipping/overloading behavior, but
the
input signal was not over. It was something like -1 or -2dB.

What you seem to be describing is some kind of "soft clipping" effect.

I looked at the circuit in this box. It is based on the AK4524 ADC. The
circuit seems to follow the standard design spec. The chip operates on
+5V.
The driving opamps run on +/-5V. The input signal to the ADC seems to be
about 0.7V RMS at full scale.

Shouldn't be a problem. With modern op amps, it should be possible to get a
clean 2.8 volts RMS out of a circuit running on +/- 5.


So, could it be that these ADCs are really not very good?

Typical of cheap circutry, or someone is going down the garden path to soft
clipping.

Or is the
supporting circuitry poor? Should I expect a more expensive box, like
MOTU or
Presonus, to do it right? I guess it pays to check these things out.

If you want to see an audio interface done *right* check out the LynxTWO.

If you look at the AES standards for measuring computer multimedia audio,
they specify (off the top of my head) that measurements be taken at 3 dB
below the point that results in 1% THD or obvious clipping.

http://www.aes.org/publications/standards/

AES17