So an issue that we encounter quite often in tinytapeout when hardening with the

hd

library is that it fails the

npc.2

klayout FEOL precheck. After investigating a bit the cases that happens and how the various cells are laid out, I think the main culprit is actually the

conb_1

cell. To make sure cells can abut properly, the

npc

geometry of the cell must have the "central" band which is common to all cells and if it needs more geometry to contact a gate outside of that area then either that geometry needs to be more than 0.27u away from the edge, or it must extend right to the edge. This avoid forming a small gap between two abutting cells. The current geometry of

conb_1

doesn't respect this. I have written a demonstration python script using

gdspy

that corrects the

npc

geometry inside the

conb_1

cell and this has fixed the issue in the various cases we encountered ( https://pastebin.com/uQqhi3eY ). The geometry still doesn't strictly conform to the requirements above, so it's not 100% fool proof but AFAICT it's enough to fix compatibility with all the other cells. (I'll need to write a quick script to verify that though). So how would one go about getting this fixed in the PDK ? If we're following the same path as for the

decap_12

issue, I think a new

sky130_ef_hd__conb_1

should be added and then pointing yosys to use this instead of the one in the library ?

@Tim Edwards we discussed this a few months ago, and this seems to fix the issue well

Proposed fix : https://github.com/RTimothyEdwards/open_pdks/commit/923cae1c09a239a97f5ed9f34b8690f41a66f79a

@David Lindley fyi

@tnt Thanks! It's now on my todo list. I have some additional filler/decap work to do, too.

So wait, we're dynamically replacing them in user submissions or should we fix this in the PDK?

Right now I plan to fix things in the submissions. It needs to be fixed in the PDK for 2411.

@David Lindley Will all the

decap

be gone, or just enough to pass new density ? ( like replace

decap_12

with

decap_4

and

fill_8

maybe )

@tnt More like 60% of the decap_12 will be swapped with a combination of fill_4 and fill_8. We have to leave the diff in there for the fom density requirements. That's what I did for the harnesses. But, I'm still working on the solution.

👌

This patch has been applied to the sky130_fd_sc_hd repository, and I have updated open_pdks with the new commit hashes for the standard cell library (and the I/O library, for which I applied another patch to a DRC error yesterday).