Hello, I have a weird problem when I run my Preche...
# openlane
c
Hello, I have a weird problem when I run my Precheck. The magicDRC gave me 66,333 errors, but they do not correspond to the layout. I'll show you a few examples. One DRC error says "under HVI or LV nwell," but I don't have this kind of layer. The DRC provided by KLayout says I have 400 errors in total, and they correspond to something
m
@Charly Meyer maybe share you precheck result directory?
c
I did a local precheck and reran a new precheck. I encountered DRC errors provided by KLayout, but many were from MagicDRC. I am sharing my results with you.
m
@Charly Meyer Looks like there’s a problem here.
Copy code
Reading "inv4_hvl".
Warning: direction reversal in path.
This is where a path bends back on itself. I think the fab refuses to manufacture designs with these errors. From the
precheck.log
Copy code
2024-05-22 09:55:08 - [ERROR] - Violation Message 'N-Diffusion spacing to N-well < 0.34um (diff/tap.9)' found 27216 times.
2024-05-22 09:55:08 - [ERROR] - Violation Message 'All nwells must contain metal-connected N+ taps (nwell.4)' found 13 times.
2024-05-22 09:55:08 - [ERROR] - Violation Message 'poly contact spacing to P-diffusion < 0.235um (licon.9 + psdm.5a)' found 36492 times.
2024-05-22 09:55:08 - [ERROR] - Violation Message 'N-diffusion contact width < 0.17um (licon.1)' found 1276 times.
I didn’t see any klayout drc error reports with errors. Do you know which one it is?
c
I'm not familiar with this log interface. I didn't see this error: "Reading 'inv4_hvl'. Warning: direction reversal in path." However, regarding the other issues, I corrected the design to have no DRC errors with KLayout, but I still have many with MagicDRC, as I showed previously.
m
@Charly Meyer the
direction reversal
message is in
logs/magic_drc_check.log
The magic errors may be due to more checks in magic than the absolutely required manufacturing checks. They may also be caused by cross hierarchy diffusion/implant/poly which magic does not always process as expected. Do the DRC errors appear to be true errors when cross referenced with the pdk? You can find the explanations and related diagrams at https://skywater-pdk.readthedocs.io/en/main/rules/periphery.html. For example, search for
nwell.4
You can view the magic errors in klayout with this file
./outputs/reports/magic_drc_check.xml
.
c
In fact, the major problem is that it shows issues in the middle of nowhere. I suppose Magic DRC uses another layout or a combination of several layouts. I can't explain it otherwise
I tried a solution by writing the GDS file with Magic, and the Magic DRC errors have vanished. However, I have 5 DRC errors that I can easily correct. I loaded a GDS file with Magic and wrote a GDS file of the same layout without making any changes.
m
Magic can change gds data that it reads and then outputs. In your case, it seems that it fixed most of the errors, which is a good thing!