One more observation about the LVS: It seems to me that the diode polarities read from the cdl file are reversed. Look e.g. at this test case:

.SUBCKT dantenna
D1 net1 GND dantenna w=780.00n l=780.00n a=608.400f p=3.12u m=1
D2 net2 GND dantenna w=800.00n l=780.00n a=624.000f p=3.16u m=1
D3 net3 GND dantenna w=780.00n l=700.00n a=546.000f p=2.96u m=1
D4 net4 GND dantenna w=780.00n l=780.00n a=608.400f p=3.12u m=2
D5 net5 GND dantenna w=780.00n l=780.00n a=608.400f p=3.12u m=4
.ENDS

These diodes are pieces of n+ in the common p- substrate. The common node (GND) must be the anode (p-) for this structure to pass LVS. But, the spice diode format is the opposite of the above: "D anode cathode." I found this while running LVS on the IHP pad cells. They only pass when I reverse the diodes (and thereby make the netlist incompatible with standard spice).

@Farag Elsayed Could we take a look at this?

I have followed the xschem term order in the extracted netlist. Happy to change if xschem will have the

anode cathode

order. @Krzysztof Herman

@Farag Elsayed so it seems that xschem does follow the

anode cathode

pin order. The spice models implemented in

diodes.lib

also do.

Okay, thanks. I have updated the LVS runset.