Hi, we have npc violations in klayout DRC while our magic DRC is clean. Can someone please explain what is

npc.2

?

Looks like npc (nitride poly cut) spacing rule. Can you show just these layers

npc 95/20

,

poly 66/20

,

licon1 66/44

? There may be a corner case where standard cell placement causes an error. In magic, it looks like npc is a generated layer and does not have a drc check in the normal rules. #shuttle-precheck @Tim Edwards This may be way off base, but AFAICT, the pre-check uses the normal tech file for DRC which doesn't appear to have a NPC check. From

mpw_precheck/checks/drc_checks/magic/magic_gds_drc_check.py

magicrc_file_path = parent_directory.parent.parent / 'tech-files' / 'sky130A.magicrc'

From

mpw_precheck/checks/tech-files/sky130A.magicrc

tech load $PDKPATH/libs.tech/magic/sky130A.tech

There is another tech file in the pdk,

sky130A-GDS.tech

, that does have an NPC check. Should this be the tech file used with pre-check?

Thanks, @Mitch Bailey. It looks like the following with just those 3 layers.

Looks like a "lucky" corner case due to cell placement. I think you could add NPC to fill the space (being careful of minimum width), or in this case, swap the

conb

cell with the

fill

cell to the left and adjust the wiring.

How can you tell which is the

conb

cell and which is the

fill

cell? Would this be a manual swap I do in Innovus?

Not familiar with innovus. I was thinking magic or klayout. Your original post had the standard cells. You can somewhat tell what the cells are by the poly.

This looks like standard cells that were placed without filler cells between them---those can and will cause all sorts of errors. The standard cells were not designed to have gaps. The NPC layer stretches across the width of the standard cell so that it (also NSDM and PSDM, done the same way) makes a continuous unbroken row across the entire digital cell. Magic auto-corrects this kind of error in analog layout, but where it is using standard cells, it does not see all the layers that it would normally auto-generate itself. That said, the error above looks strange because as far as I know, all standard cells are designed with a single line of NPC in the middle. What cell is it that has, apparently, poly contacts on both the top and bottom of the cell?

@Tim Edwards The fill cells are there, it's just only the poly is displayed (see the first screen shot). The npc space error occurs when the

conb

cell is placed next to the AND? cell.

@Mitch Bailey: My mistake(s); I really thought that the fill cell has a strip of NPC across it to maintain continuity across a digital design. But I looked at it, and nope, it doesn't. And yes, the conb cell has poly contacts top and bottom and has a 2nd, unconnected area of NPC. The cell on the right is not an

and2_1

. I'm trying to figure out what it is. I'm curious if there is really a standard cell combination that violates DRC rules when abutted as designed. If so, the offending cell might be able to be redesigned, or just added to the "don't use" list (which is already too long, in my opinion).