IIP3 simulations: qpss/hb vs hb

I would generally expect hb to match qpss (in hb mode). However, there may be some subtle difference in the setup which means they are different, particularly if you are trying to simulate a passive mixer using bsim4 models (or bsim3v3 but I'm sure it's not that at 28nm). This is a well known limitation of the bsim4 model (it's not a simulator issue; it's the model itself). There are several papers on this, which I'm sure have been mentioned before on these forums. Try searching (using the magnifying glass at the top of the page) for "passive mixer" and you'll get some hits that explain the challenges.

The solutions are typically either to alter the biassing of the circuit slightly (which isn't always viable), or to change the model to something like psp (also not always under  your control).

It may be something else, but I can see from your graphs that there's a bit of unusual variation in the curves. 

Your best bet is to contact customer support; it's really hard to offer much in the way of concrete help for this without being able to see the problem more closely.

Regards,

Andrew.

Hi Andrew,

Thank you. Yes, I am aware of the BSIM problems and the forum references. My models are utsoi (which are based on PSP). But apart from that, I am seeing these inconsistencies even without switching (e.g. just getting IIP3 of Ron of a MOSFET):

IIP3 curve still looks like this:

IIP3_se_spectre_hb.pdf

Am I am doing something significantly wrong (in this simple scenario) or can I be very sure that it's indeed the models?

The problem with bsim4 is not dependent upon it switching - it's dependent on the Vds=0 bias (see this post).

I don't know whether there's a similar problem with the utsoi model (as you say, it's based on the work in psp so hopefully should be OK) - I've not seen any reports of issues.

Best bet is to contact customer support so that we can get hold of the same models and check it. 

Regards,

Andrew