magic extraction cannot discern nfet_03v3 from nfet_03v3_dw in GF180MCUC

[spnadig]

@RTimothyEdwards - attached is a gds file with 2 kinds of inverters with nfets. One with DNWELL and the other without - which should correspond to nfet_03v3_dw and nfet_03v3 respectively. The extracted netlist (attached with extraction script used) however shows same devices (nfet03v3) for both cases.
GF lists them to be seperate devices - as they must be mapped to different models -

inv_dnwTest.zip

[spnadig]

@proppy

[proppy]

Looks like the spice deck also use nmos_3p3 and not nfet03v3 as the naming convention, see:
https://github.com/google/globalfoundries-pdk-libs-gf180mcu_fd_pr/blob/main/models/ngspice/smbb000149.ngspice

@mithro can you link to the document that specify authoritatively the name of the primitive.

/cc @atorkmabrains

[atorkmabrains]

@proppy It's 'nfet_03v3' in here:

https://github.com/efabless/globalfoundries-pdk-libs-gf180mcu_fd_pr/tree/main/models/ngspice

Google repo is not up to date.

[mithro]

https://bit.ly/gf180mcu-names

[RTimothyEdwards]

The model name is the same whether or not the device is in deep nwell or not, so why even make the distinction?

[spnadig]

Oh I was under the impression that models change with DNWELL given GF classifies them as different devices in their list?

[RTimothyEdwards]

@spnadig : The first table that you posted has a column for "Model Name" that indicates that the same model is used for both the device inside and outside of deep nwell. I have not looked at it carefully; maybe the device is just the underlying FET device? There might be additional components in the subcircuits (i.e., does the deep nwell device have additional terminals and/or diode components representing the deep nwell and substrate connections?).

[spnadig]

No there are no additional connections - it just has DNWELL layer and guard ring (which has 2 additional layers) around a typical nfet

[RTimothyEdwards]

@spnadig : I've used PDKs in the past where the deep nwell device models have additional connections representing the deep nwell and substrate in addition to the pwell and diffusion. That makes them 5- and 6- terminal devices. But in this case, I just looked at the GF SPICE models; there are no subcircuit models for the devices in deep nwell.