The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> trans analysis convergence difficulties in spectre
https://designers-guide.org/forum/YaBB.pl?num=1068541196

Message started by tulip on Nov 11^th, 2003, 12:59am

Title: trans analysis convergence difficulties in spectre
Post by tulip on Nov 11^th, 2003, 12:59am

my circuit is a high frequency mixer(LO:1.5G, RF: 1.3G) circuit.

when simulated with transient analysis, spectre report " convergnence difficulties resulted in error requirements being unsatisfied."
"zero diagonal found in Jacobian at 'I0.I51.C0.1" and 'I0._net3",

I tried serveral methods, such as using gears method,
set gmin to 1e-18, set cmin to 1f, and set the maxstep,
but convergence difficulties still exit.

How can I slove this problem?

many thanks! :)

Title: Re: trans analysis convergence difficulties in spe
Post by Ken Kundert on Nov 11^th, 2003, 9:08pm

Spectre is telling you that there is a problem at 'I0.I51.C0.1" and 'I0._net3". There is some problem such as a floating node or loop of shorts there. When you look in that area, what do you find?

-Ken

Title: Re: trans analysis convergence difficulties in spe
Post by tulip on Nov 12^th, 2003, 12:54am

I changed my circuit a little. Now it do not report"zero diagonal found in Jacobian at 'I0.I51.C0.1" and 'I0._net3", but it still report" convergence difficulties resulted in error requirements being unsatisfied".

when I change the simulation option from conservative to moderate, the warning message disappears.

when I run the PSS analysis, the convergence problem still exits, and the circuit simulation stops before it get the results.

Title: Re: trans analysis convergence difficulties in spe
Post by Ken Kundert on Nov 12^th, 2003, 9:15am

It seems as if the basic problem still exists. There appears to be some problem in the circuit or with a model near 'I0.I51.C0.1" and 'I0._net3". When you look in that area, what do you find?

-Ken

Title: Re: trans analysis convergence difficulties in spe
Post by Venkatesh TS on Dec 29^th, 2005, 7:15pm

Hello,
i'm also facing exact same issue. Spectre says
"Warning from spectre at time = 2.5016 ns during transient analysis `tran'.
Convergence difficulties resulted in error requirements being unsatisfied."

What to do?

Title: Re: trans analysis convergence difficulties in spe
Post by Ken Kundert on Dec 30^th, 2005, 9:45am

The message "Convergence difficulties resulted in error requirements being unsatisfied." indicates that Spectre could not take a time step small enough to maintain accuracy because of convergence problems. This implies that there is some problem in the circuit that only becomes exposed with small time steps. Typically, this happens when there are discontinuities in the capacitors. With large time steps (long time periods) the capacitors are diluted with the resistive portion of the circuit. But with very small time steps the capacitive part of the circuit dominates. In the extreme, with very small time steps, the effect of the resistive portion almost disappears. In this case, the capactive part of the circuit must be simulatable by itself. Thus, the capacitor models must all be continuous and there must be capacitive paths from all nodes to ground.

You can often circumvent this problem using cmin, which adds a capacitor from every node to ground, which provides the capacitive path to ground and tends to smooth the discontinuities.

-Ken