Thanks in advance for helping me with this issue:
I have a problem with the extracted view of a simple NAND circuit. I am using IC 6.1.8, running the LVS with PVS (19.10) and the extraction with Quantus (19.13). The PDK I am using is the GPDK045 provided by Cadence as a generic 45nm process design kit.
The layout of my NAND has 2 PMOS with a shared drain as shown in Figure (the central node):
The shared drain diffusion has an area of 38.4E-15m2 (240nm x 160nm). After running the extraction I checked the extracted netlist which looks like this:
Both the drain area of M3 and M4 (the 2 PMOS transistors) is calculated as being equal to the area of the shared drain diffusion. In this way, the corresponding parasitic capacitance is counted twice in simulation, as far as I understand.
Is there something I am missing or some checkbox I should tick setting up the extraction?
Thank you in advance and best regards
After additional checks, I concluded that the problem should be in the GPDK045 extraction rules file since replicating the same procedure with a commercial design kit from an actual foundry, the drain area of each of the 2 transistors is half the area of the shared diffusion, as it should be. Any idea on how to correct the problem? I am using the gpdk for teaching purposes and this kind of problems undermine the insight I would like to give to my students. Thank you!
I did some checks with this yesterday on a flight, and I agree - this looks wrong. It’s also wrong in the Assura rule deck too (that has a second problem where the drain area is used for both drain and source, and doesn’t share it).
I will file an issue against the team that produced gpdk045 and also see if we can issue a patch to the rule deck.
Dear Andrew, thank you so much. I will wait for possible updates.