Skip to content

gf180 mimcap: real model name, and emit a capacitor rather than a subckt call - #109

Open
carloscl03 wants to merge 2 commits into
ReaLLMASIC:mainfrom
carloscl03:gf180-mimcap-model
Open

gf180 mimcap: real model name, and emit a capacitor rather than a subckt call#109
carloscl03 wants to merge 2 commits into
ReaLLMASIC:mainfrom
carloscl03:gf180-mimcap-model

Conversation

@carloscl03

Copy link
Copy Markdown

Two small things that kept the gf180 mimcap out of every LVS run.

The model name did not exist. gf180_mapped.py mapped 'mimcap' to
cap_mim_1p0fF, which is not in the PDK. The three real models are
cap_mim_1f0fF, cap_mim_1f5fF and cap_mim_2f0fF. Any netlist emitted for
a gf180 mimcap named a device the simulator and the LVS reader had never
heard of.

X is a subcircuit call, not a capacitor. The netlist emitted

X1 V1 V2 cap_mim_2f0fF l={l} w={w}

so the LVS reader built a subcircuit instance and never a capacitor device.
The extracted MIM then had nothing to pair with and was reported as a
layout-only device, no matter how the schematic side was arranged. The fets
go through an equivalent X->M rewrite in the gf180 runner for the same
reason.

In the opamp this was the last mismatch standing: with C1 it pairs and the
cell matches. Verified with tests/lvs/run_cell_lvs.py --pdk gf180; needs
#108 for the runner to extract the MIM at all.

'mimcap_1p0fF' no es ningun modelo de gf180: los reales son
cap_mim_1f0fF, cap_mim_1f5fF y cap_mim_2f0fF. Prefijo invertido y 'p' por
'f'. Cualquier netlist de referencia que glayout generase para una celda
con MIM nombraba un dispositivo inexistente, asi que el LVS no podia
emparejarlo nunca con el cap_mim_* que sale de la extraccion.

Se pone 2f0 porque es lo que asumen las cuatro variantes del runset de
LVS del PDK (mim_cap=2 en todas). Cual corre la fabrica sigue siendo una
decision de proceso.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant