The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
https://designers-guide.org/forum/YaBB.pl?num=1501674362

Message started by heavn007 on Aug 2^nd, 2017, 4:45am

Title: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by heavn007 on Aug 2^nd, 2017, 4:45am

Hello everyone,
I am trying to implement sccmfb based fully differential opamp for Delta Sigma Modulator. I am using two non overlapping clock signals (clk1 and clk2, clk1 is for sensing phase and clk2 for charge sharing phase). I want to know DC gain, bandwidth, loop gain of this circuit, so I am trying to run PSS + PAC and PSS+PSTB analysis.
While, I run the PSS analysis with following setup configuration:

Beat frequency : 20
No. of Harminis : 0
Accuracy : moderate
tstab: 40u

Below errors are coming in output log.

Zero diagonal found in Jacobian at `I6.net17' and `I6.net17'.
Zero diagonal found in Jacobian at `I6.net22' and `I6.net22'.

WARNING (SPECTRE-16266): Error requirements were not satisfied because of convergence difficulties.

Also, I am not getting any result from direct plot form.
In short, I want to know how can I measure dc gain, bandwidth, loopgain of SCCMFB based fully differential opamp. My clock frequency is 50kHz and signal bandwidth is 50Hz.

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by Geoffrey_Coram on Aug 2^nd, 2017, 5:25am

From the picture, and your statement that you have non-overlapping clocks, it's clear that, when both phi1 and phi2 are low, you have capacitor C2 and the two nets it is connected to (net17 and the unlabeled node on the other side of C2) just hanging out in complete isolation. The same problem happens with net22 and C5.

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by heavn007 on Aug 2^nd, 2017, 11:54pm

Hi Geoffrey_Coram,
Thank you for replying. I need to use non overlapping clock (NOC) for my design. Is there anyway I can remove the warning without modifying NOC and run PSS+PAC for analysis successfully. I would be grateful if you provide me any useful links regarding this matter.

I am attaching my test bench schematic.

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by Geoffrey_Coram on Aug 3^rd, 2017, 12:54pm

Can you put a very large resistor from net17 to ground (and another from net22 to ground)?

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by Andrew Beckett on Aug 5^th, 2017, 1:57am

Although there should be a gmin resistor inserted by the simulator from there to ground anyway (I believe).

The snippet of the log file you sent doesn't sound severe enough to cause PSS to fail - and even if it did, there would be something else which might give more detail. Can you please post the whole spectre.out file?

Regards,

Andrew.

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by Geoffrey_Coram on Aug 7^th, 2017, 6:07am

Interesting points, Andrew!

Andrew Beckett wrote on Aug 5^th, 2017, 1:57am:

Although there should be a gmin resistor inserted by the simulator from there to ground anyway (I believe).

Would that gmin be inserted, even if it wasn't needed at time=0? Suppose the switch is closed for the dc/time=0 solution, providing a dc path, but later opens up. Would Spectre know to insert a gmin?

Quote:

The snippet of the log file you sent doesn't sound severe enough to cause PSS to fail - and even if it did, there would be something else which might give more detail. Can you please post the whole spectre.out file?

Interesting point! The original post included only warnings:
Quote:

WARNING (SPECTRE-16266): Error requirements were not satisfied because of convergence difficulties.

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by Andrew Beckett on Aug 7^th, 2017, 10:15am

Geoffrey_Coram wrote on Aug 7^th, 2017, 6:07am:

Would that gmin be inserted, even if it wasn't needed at time=0? Suppose the switch is closed for the dc/time=0 solution, providing a dc path, but later opens up. Would Spectre know to insert a gmin?

Good point. I don't think it would.

Andrew.

Title: Re: Zero diagonal found in Jacobian error while running SCCMFB circuit simulation
Post by pdirac on Aug 7^th, 2019, 3:13am

Actually as Geoffrey_Coram stated the capacitors C2 and C5 are floating during the time clocks are off. Therefore what you should do is add some parasitic capacitors on the net 17 and 22. For example, connect a 10aF capacitor to the floating nets and you'll be done. Hope it helps though it's a very late answer) Of course, it might help future readers..