Tejas Nafde
01/07/2025, 4:54 PMMitch Bailey
01/07/2025, 6:43 PMM3
L=1.5
W=226.45
nf=1
The spice models are bin’ed, which means that they are only valid for certain ranges of parameters. I imagine that this very wide transistor does not match of the bins.
For example, the spice models for pfet_01v8
are in $PDK_ROOT/$PDK/libs.ref/sky130_fd_pr/spice/sky130_fd_pr__pfet_01v8.pm3.spice
There are over 60 bins - here is a sample with the parameter ranges:
.model sky130_fd_pr__pfet_01v8__model.0 pmos
* DC IV MOS Parameters
+ lmin = 1.45e-07 lmax = 1.55e-07 wmin = 1.255e-06 wmax = 1.265e-6
--
.model sky130_fd_pr__pfet_01v8__model.1 pmos
* DC IV MOS Parameters
+ lmin = 1.45e-07 lmax = 1.55e-07 wmin = 1.675e-06 wmax = 1.685e-6
--
.model sky130_fd_pr__pfet_01v8__model.2 pmos
* DC IV MOS Parameters
+ lmin = 9.95e-07 lmax = 1.005e-06 wmin = 9.95e-07 wmax = 1.005e-6
--
.model sky130_fd_pr__pfet_01v8__model.3 pmos
* DC IV MOS Parameters
+ lmin = 1.995e-06 lmax = 2.005e-06 wmin = 9.95e-07 wmax = 1.005e-6
--
Translating `M3`’s parameter units to the same scale gives l=1.5e-06 w=2.2645e-04
What you might try is increasing the number of fingers, nf
. This effectively splits the very wide transistor into multiple thinner transistors in parallel.Tejas Nafde
01/07/2025, 7:33 PM