(Sorry for posting this here. But I forgot again where to report bugs.)
Resistors can now carry the KF property, which is the 1/f noise coefficient.
The coefficient of the 1/sqrt(f) voltage noise spectral density produced by the resistor appears to scale as (as currently implemented):
vn = sqrt(KF * Vappl * R)
So that means the flicker voltage noise density contribution produced by a resistor is:
flicker(f) = sqrt(KF * R * Vappl) / sqrt(f)
This is unphysical/incorrect behavior.
Flicker noise voltage spectral density scales proportionally to the applied voltage or current, not to the sqrt of applied voltage.
Futhermore, in most usually used notations of the flicker noise coefficient, the noise also does not scale with R at all. So by scaling it with sqrt(R), the KF parameter gets an awkward and unintuitive dimension.
Usual notations to express the flicker noise of a resistor are either the “noise index” (produced noise scales only with the applied voltage), or perhaps more intuitively the precision limit of the resistor. When flicker noise is specified in terms of a relative precision limit of the resistor, then the flicker noise voltage contribution scales proportionally with R, but again, not with the sqrt(R).
I think not a problem to leave it here, you can post the follow up. Sometime I help to verify the bug and submit to Mike. But noise is not my expertise and I just worry your concern may not get addressed. That why I sent this reminder.
Mike replied with an example that uses AF=2 as another parameter beside the KF parameter. With that setting for AF, all the scaling issues seem to have been resolved. Everything works exactly as it should.
I don’t understand exactly what the AF parameter does, but the default value of AF=1 doesn’t seem to provide useful results, at least not with respect to excess noise of resistor components.
Yes it helps to clarify how AF affects the overall behavior, and why AF=2 is the “correct” setting. However, I am puzzled as to why other exponents are allowed (I guess options are better than no options) or at least, why AF=2 isn’t the default.
Is there a line I can throw into schematics that changes the AF to 2 for all R instances globally?
Just checked the Hspice manual, and the default values are AF=2 and KF=0 (flicker noise disabled). So, maybe you can check with Mike to see if AF should follow the default value in Hspice as 2 instead of 1.
