Somehow MOS in triode mode gives me 3 times less thermal noise than an equivalent resistor. This was not a case in older versions of spectre.
Anybody can reproduce this problem?

imagelimit.

I have a similar problem trying to simulate the effect of opamp noise while using an ideal OTA. I get exactly the amount of noise I expect when I use a CT circuit and the OTA, but when I try to use the OTA in a switched-cap circuit, I dont see any noise at all (almost none). I was just trying to check if the simulator would be able to help me compute the overall thermal noise in my SC circuit, accounting for effects that I might have overlooked. However, I am unable to make it behave and so have resigned myself to using my hand-calculations, which are not bad thankfully.

I even tried using hundreds of sidebands and simulating far beyond my clock frequency range to allow for enough noise aliasing.

You should try to do handcalculations as well. I think pac might work well in terms of predicting the NTF from any node to the output but for some curious reason, pnoise is not behaving well for OTA noise even though it works well for kT/C noise.

If anyone knows how to fix this problem, I would be glad to know. There are some bigger circuits that are harder to analyze at times.

Sorry, I cant post the modelcard for proprietary reasons, but I doubt if that has anything to do with this problem, especially since I see the exact amount of noise while using a CT circuit.

Vivek

Vivek, does this happen in older versions too?