<@U016HU5HK8V>: FYI: There are various other thi...
# sky130
t
@User: FYI: There are various other things you need to do to make the cell work properly within a standard cell set, such as declaring the abutment box with "property FIXED_BBOX". Also change the FETs from "nfet" to "scnfet" and magic will automatically generate the "standard cell ID" layer over the whole cell. Plus the rules for "scnfet" are slightly relaxed so you can move contacts closer to the gate.
✔️ 2
t
Ah thanks for the tips !
Is there a way to force the area.scid zone to be generated if I don't have any
scnfet
? I need my own tap cell (which has no device inside).
t
I added a new feature recently which is the cifoutput operator "mask-hints _name_". I have used this in the most recent tech files for the THKOX (HVI) layer, and the P+ and N+ implants. It can and should be extended to cover the identifier layers. If you use "mask-hints STDCELL" in cifoutput, then you can have a corresponding property in the cell that works like the FIXED_BBOX property: You give it the property name "MASKHINTS_STDCELL" and the same rectangle coordinates as the FIXED_BBOX, and then magic will generate the layer in the GDS output. This is to meant to avoid having such layers be visible or selectable, and it works best for things like the ID layers.
👌 1
It is probably also necessary to use the mask hints to apply a stripe of NPC that fills the center of the cell from edge to edge, as this is what is done in the other HD standard cells; otherwise you'll probably get NPC spacing errors from abutting this cell with other standard cells. I notice, though, that you have N-diffusion on the top of the cell. Isn't that going to cause spacing errors with the N-wells in adjacent cells if the adjacent cells are from the HD standard cell library?
t
Yes, you can't just put the ROM cell anywhere. I am making special 'boundary' cells that will surround the actual ROM area to transition from the region with nwell to the one without and ensure DRC works.
MASKHINTS_*
stuff is working. Submitted a PR with a couple of cosmetic fixes. (It prints some spurious errors ...)