@Tim Edwards slot-031 of mpw-8 uses the

sky130_fd_sc_ms

layout library. Looks like there is no

tap

layer in the layout cells - both

tap

and

diff

are on the

diff

layer. This may not be an actual mask problem, but I think it’s causing extraction problems in magic - the nwell connections are missing. Could this be caused by using a library not converted with

open_pdks

?

It sounds like something that might be caused by using the proprietary s130 PDK. It's worth investigating further. My impression is that SkyWater will be able to read both s8 and sky130 and convert both into s8, but it may not always work the other direction, and that may be what we're seeing here. But we need to make sure.

The user said that they remember changing the

tap

layer to

diff

. That won’t extract correctly with magic, will it?

@Mitch Bailey: I thought I had written the GDS input routines so that they would accept both

diff

and

tap

, but if layers are separated between cells, it could cause issues. And I could be wrong about the way I wrote the input routines. But the use of NSDM and PSDM implants should make it unambiguous; the split between DIFF and TAP is entirely redundant and I'm not sure why anyone decided to do that in the first place. The s130 PDK takes the more common approach.

@Tim Edwards thanks for the explanation. The problem (unconnected nwells) may be due to something entirely different. I’ll recheck.