Hi @Mitch Bailey @Stefan Schippers I tried to run LVS for the bandgap reference circuit between the bandgap.gds and bandgap_xschem.spice . However, both are not mateched due to pnp transistors. is there a good way to pass it?

First step is to get the pnp model names to match.

sky130_fd_pr__pnp_05v5 (2)                 |(no matching element)                      
...
(no matching element)                      |sky130_fd_pr__pnp_05v5_W0p68L0p68 (2)

This can probably be done with

sed -i.bak 's/sky130_fd_pr__pnp_05v5_W0p68L0p68/sky130_fd_pr__pnp_05v5 W=0.68 L=0.68/' bandgap_xschem.spice

@Mitch Bailey @Ahmed Reda it seems that for simulation purposes (ngspice) the right name is

sky130_fd_pr__pnp_05v5_W0p68L0p68

, while for layout matching

sky130_fd_pr__pnp_05v5

is the correct symbol name. What should the best way to go to avoid users dealing with these errors? with this difference either LVS or ngspice simulation will fail.

Might be able to add something to the netgen setup file. I’ll look into it. ngspice doesn’t have a generic

sky130_fd_pr__pnp_05v5

model with

W

and

L

?

@Mitch Bailey I see these 2 models related to 0.68 and 3.4 dimensions (both L and W):

sky130_fd_pr__pnp_05v5_W0p68L0p68.model.spice

sky130_fd_pr__pnp_05v5_W3p40L3p40.model.spice

@Tim Edwards I thought something like this would work, but it doesn’t seem to add the property to the devices.

property "-circuit2 sky130_fd_pr__res_xhigh_po_0p69" add w 0.69
    equate classes "-circuit1 sky130_fd_pr__res_xhigh_po" "-circuit2 sky130_fd_pr__res_xhigh_po_0p69"

It gives me a topology match with size errors

sky130_fd_pr__res_xhigh_po:24 vs. sky130_fd_pr__res_xhigh_po_0p69:R1:
Property w in circuit1 has no matching property in circuit2

@Mitch Bailey:

property ... add

doesn't mean to add a property, it means that the specified property is additive. There are definite complications with getting these fixed-width devices to extract right from magic. I made the resistors work reasonably well by defining various marker layers at different sizes. I never did that with the bipolars since there are only a couple of valid modeled device sizes. Magic will extract a PNP device marked with a

pnp

layer as

sky130_fd_pr__pnp_05v5

without the fixed-size suffix. It should work to use the

sky130_fd_pr__pnp_05v0_W0p68L0p68

layout cell from the library and give it a property called

primitive

that indicates that the subcircuit is supposed to be treated as a device and therefore extracted as a black-box. But if all else fails, I prefer your solution of just hacking one netlist to produce the right output with a

sed

script.

I may be confused. In the setup file there’s

property "-circuit1 $dev" series {l add}

which specifies adding property when merging, but in the code, there’s

char *options[] = {
        "add", "create", "remove", "delete", "tolerance", "merge", "serial",
        "series", "parallel", NULL
    };
    enum OptionIdx {
        ADD_IDX, CREATE_IDX, REMOVE_IDX, DELETE_IDX, TOLERANCE_IDX, MERGE_IDX,
        SERIAL_IDX, SERIES_IDX, PARALLEL_IDX
    };
...
            case ADD_IDX:
            case CREATE_IDX:

Looks like there’s also an

add

or

create

command used in the same place as

series

.

@Tim Edwards @Mitch Bailey @Stefan Schippers Thank you for you suggestions. The error has been fixed using <sed> script, just hacking one netlist to get right output.