[quote author="Rochey"]Guys,
I fired off the question (as it was of interest to me too) to our Pro Audio Apps Engineer --- he came back with the following.
Resistors are the least of your worries for lower frequency noise.
Typically, 1/f noise from an IC amplifier or from within the ADC itself is going to dominate.
We typically use 1% or 0.1% metal film surface mount resistors (0805 sized) in and around input buffers, ADC front ends, DAC output filters, and mic pre circuits, and have found them to work well. Panasonic's are orderable from Digi-Key and I use them constantly on my EVM designs. Since I'm measuring a pretty flat noise floor from a few tens of Hz all of the way up to Nyquist, I would say that they pose no real issue, IMHO.[/quote]
Humble opinion or arrogant opinion, he is neglecting the presence of d.c. (when it
is present) or similar low frequency noise riding on signal swings. And, note that he is talking about metal film (i.e., I believe, thin film) resistors, not the bog-standard cheapest thick film parts. These were anything but "the least of [my] worries for lower frequency noise." :razz:
The app that brought the terrible performance of Rohm thick film parts to my attention had lots of d.c. as standing current in a discrete component front end. Audibly and measurably the noise was terrible. And I can assure your engineer that the 1/f noise from the JFET and bipolars in the circuit were negligible by comparison.
It was tamed by a combination of reducing the standing current and changing vendors (Yageo and Panasonic were both a lot better), but it never got anywhere near as good as ~1/4W standard metal-film axial parts, even with those running the higher currents. I'm sure the parts Fred F. mentions would have been much better than the Yageo/Pana parts, but they appeared to be too expensive for this very cost-sensitive design. And no board layout changes could be made by customer edict at the point these issues surfaced---no larger footprints for bigger parts (which helps) and no place to be made for axials.
The subtlety here: one could say "Well, isn't it obvious? Don't run d.c. through your resistors!" But realize that large low-frequency signal swings are tantamount to d.c. and will indeed have excess noise riding on them. If you are only concerned about passing a spec then this may be satisfactory. But at least it is something to be wary of.
See also in this connection the Gilbert article and his remarks (see Note 4 in particular) about active circuit biasing.
Note to self for future controversial threads: stochastic resonance