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 at certain net
https://designers-guide.org/forum/YaBB.pl?num=1121783526

Message started by reilly on Jul 19^th, 2005, 7:32am

Title: Zero diagonal found in Jacobian at certain net
Post by reilly on Jul 19^th, 2005, 7:32am

Hi everyone,
Currently I am attempting to run a dc sweep simulation by sweeping the power supply from one voltage to another voltage using Cadence IC5. I used the default setting in ADE (Analog Design Envinronment) and start to simulate using spectre.

Eventually, the spectre simulator found zero diagonal at certain nets and trying to reorder Jacobian. Anyone know what is this reordering Jacobian means and how to solve this problem since spectre simulator took more than half hour to simulate. Any setting to change in order to speed up or avoid zero diagonal? Is it due to the design, simulator, models? In certain cases, the simulator do not manage to get a solution. Hope to get explaination and ways to overcome this problem. By the way, how to ensure simulator manage to get the convergence using homotopy gmin/source which always give faster result compare to other convergence method. Thanks. :)

******************
DC Analysis `dcOp'
******************
Important parameter values:
reltol = 1e-03
abstol(I) = 1 pA
abstol(V) = 1 uV
temp = 35 C
tnom = 27 C
tempeffects = all
gmin = 1 pS
maxrsd = 0 Ohm
mos_method = s
mos_vres = 50 mV
Trying `homotopy = gmin'.
Continuation method failed. Step size too small.
Trying `homotopy = source'.
Trying `homotopy = dptran'.
Zero diagonal found in Jacobian at `net172' and `net172'.
Reordering Jacobian.

Title: Re: Zero diagonal found in Jacobian at certain net
Post by svensl on Jul 19^th, 2005, 8:35am

What circuit are you using?
I have had the same problem in SC circuits with feedback. Often there was a floating node or loop of shorts there. This was caused by the wrong clocking scheme or non-delay feedback paths.

Sven

Title: Re: Zero diagonal found in Jacobian at certain net
Post by Ken Kundert on Jul 20^th, 2005, 12:14am

In particular, did you look at or near net172 to find out what the problem was? As Sven says, in particular look for nodes that are floating or a loop of shorts.

The problem looks as if it occurred on the first sweep point, which is another clue.

-Ken

Title: Re: Zero diagonal found in Jacobian at certain net
Post by reilly on Aug 22^nd, 2005, 9:22pm

Hi,
Thanks for all your reply. I managed to solve the zero diagonal once I removed an inductor near the net. Since I do not know which node to declare for the initial condition, I run a first simulation using a simple stimuli with the circuit. Once an initial spectre.ic being generated, I changed the stimuli again to the stimuli I required the circuit to run with but defined with readic=spectre.ic in order to established the initial condition. This helped to solve most of the convergence issue I attempted most of the time. By the way, most of the circuit consists of current mirror with some feedback circuitry. Thanks again.