@User I added

.option RSHUNT 1e20

, but ngpice log still shows singular matrix at the same exact circuit node location (Net17).

In that case you will need to find out what's going on around Net17. In my experience, a persistent singular matrix error points to a bad circuit description in some way.

bad circuit description ?

@promach , i have tried adding a

.options rshunt=1e11

(and even lower values) this seems to be ignored for the operating point calculation. Adding 1e11 resistors to GND in the schematic attached to the floating nodes removes the singularity.

@Stefan Schippers See the ngspice forum thread : https://sourceforge.net/p/ngspice/discussion/133842/thread/e27767b336/

@promach @Tim Edwards the very strange fact is that on a small test circuit with 2 series capacitors the .options rshunt=1e11 is effective. I thought that, yet another bug....

@Tim Edwards @Stefan Schippers but

.option rshunt=1e11

still give me singular matrix