[quote author="Rochey"]I've asked a few of my customers about this quite a bit recently. Some of my high performance soundcard customers mentioned that they'd had similar concerns, however, they'd managed to get 120dB between signal and noise by using regular thick film SMD resistors.
Just food for thought.
/R[/quote]
Be aware that static-stationary noise versus full output signal is not the problem here. If you have no current through a resistor, most of the time your noise will be canonically thermal, unless there is something really wrong going on (like ionic contamination* [yech!] leading to built-in electrochemical potential drops and their own evolution). Now start putting current through your resistor and see what happens to the noise. Good thin film will have a small increase. Lousy thick film can be awful--trust me on this! Then realize that, those currents are what is flowing at any given instant, based on the instantaneous level of the signal you impress across it in actual applications. The noise rides with the signal, but unless you are doing some very fancy measurements you can't readily separate it from other distortion mechanisms. But you might be able to hear it :roll:
Also, while you are at it, measure the temperature coefficient and the voltage coefficient. The latter requires a very careful pulsed measurement so that it is not contaminated by the tempco effect.
Again, not all thick film is created equal! I found batches of Rohm parts to be lousy, Yaego and Pana to be much better, but as JR says, YMMV.
*a certain Thai vendor of metal film axials had some contamination in their process. It even led to some of the resistors going open-circuit under d.c. load and their disqualification as a supplier to Harman years ago. It was a crisis situation until understood, and produced some very awkward moments when a QA consultant to the resistor company alleged that the problem was localized self-heating leading to failure, because they had loaded the wrong cores on the spiral-cutting machine, and the current distribution was less uniform than the design target. This was in a part that dissipated a couple of milliwatts---absolute rubbish of an explanation.
Eventually a huge batch of that vendor's parts and another vendor's were subjected to high dissipation-voltage-humidity-temperature in a long experiment. At the end none of the Koa Speer parts failed, and three or four of the other vendor's had. No more biz for them!
