Hi,

I wonder if anyone has experienced a similar problem using the special
"sp1tswitch" provided with Spectre. This switch is useful if you want to configure
a netlist differently for different analysis and very useful.

However, if I have any such switch in my netlist (even one completely isolated from any net), Spectre refuses to perform a noise analysis citing a singular matrix as the source of error.

Any suggestions will be welcome.

Thanks
Vivek

that element is OK.. but I found Verilog A to be much more flexible..
for an example,  there is a listing for a loop-opener in the  paper below which includes an (outdated) example of its use.    
"Functional Verification of a Differential Operation Amplifier"
... a fully differential current mirror opamp with n-channel input devices ... support the development of a verification script. Figure 1 Fully Differential Current Mirror Opamp with CMFB The following table ...
http://www.cadence.com/whitepapers/FVofDiffOpAmp_wp.pdf - 309.8KB

Outdated because today one would use the STABILITY analysis to look at loop gain and phase margin..

JBD

sounds like bmslib sw_no would be a good alternative in that case..
jbd