I design a common passive mixer-first receiver(passive mixer +TIA) in TSMC65LP CMOS process. In the test bench, I drive the receiver with a port , then a idea_bulun , then an ac coupling capacitor connect to the receiver. I use HB method to simulate the IIP3 of the receiver(Lab 10: IP3 Calculation (hb with Three Tones) ). I have two questions:
1. why the IM3 curve is not 3dB/dB ?
2. when I set different "number of harmonics"（from auto 3 3 to 3 2 2）, the IM3 curve change greatly, but the slope is still non-monotonic.
Why is the difference so significant and how can I simulate the correct curve ?
I'm not sure reducing the number of harmonics is very likely to help, but there are known challenges with simulating IP3 with passive mixers due to limitations in the bsim4 and bsim3v3 models. See this article: Difficulties simulating IP3 on a switched-FET mixer (which references a paper from IBM describing the problem too). It's related to the fact that the derivatives in the model are not smooth at Vds=0 - but the paper includes more details.
thank you. It seems that I don't have the access to support.cadence.com.
Here's a link to the IBM paper: https://briefs.techconnect.org/papers/linfet-a-bsim-class-fet-model-with-smooth-derivatives-at-vds0/
Thanks again Andrew!