The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> LTE tolerance
https://designers-guide.org/forum/YaBB.pl?num=1359737386

Message started by Shahriar on Feb 1^st, 2013, 8:49am

Title: LTE tolerance
Post by Shahriar on Feb 1^st, 2013, 8:49am

Hi,

I'm working on a circuit where I have a differential amplifier configured as unity gain and it has a small capacitor (6fF) at the output to ground. The signal to this UGA is 1GHz frequency and the UGA-capacitor should give me a sort of mean trendline of the input signal. When I simulate it, I keep getting a warning -
"LTE tolerance was temporarily relaxed to step over a discontinuity in the signal" pointing to the capacitor terminal that's connected to the UGA.

LTE is the error at a time-step between simulated data and low-order polynomial approximated from previous time-steps, right?

Could anyone, please, help me understand what the warning is about? Thanks.

Title: Re: LTE tolerance
Post by Ken Kundert on Feb 1^st, 2013, 9:50am

When Spectre emits this message it suggests you have a nonlinear capacitor model that is discontinuous.

The warning is emitted when Spectre has run into a convergence problem in transient. Such problems are often due to discontinuities in the device models. When there is a discontinuity in the device model it could be in the resistive part of the capacitive part. When the simulator encounters a discontinuity it will start by shrinking the time step. This generally helps if the discontinuity is in the resistive part of the model because as the time step shrinks the effect of the capacitive part of the model tends to dominate over the resistive part, effectively smoothing the discontinuity. If there is no capacitive part of the model, you can get the same effect by specifying cmin. However, if the discontinuity is in the capacitive part of the model, shrinking the timestep is the wrong thing to do. If Spectre is not able to achieve convergence by shrinking the step in transient analysis it will instead try growing the step. However, normally the stepsize is bounded by the constraints on the Local Truncation Error, thus it cannot grow the step without violating the LTE criteria. This could affect the accuracy of the result, so it prints the warning message if it is successful in converging by growing the step beyond what normally would have been allowed. If you look carefully at the waveforms at that point in time you will see the simulator took a large step and you can judge whether to trust the simulation results or not.

-Ken

Title: Re: LTE tolerance
Post by Shahriar on Feb 1^st, 2013, 10:41am

Ken, thanks for replying. I noticed that after the warning comes up, the timestep does become 3-4 times larger than the previous steps.

I also zoomed in on the waveforms at those specific timestamps and selected to mark all the datapoints. The datapoints should show the timesteps, right? After those specific times with warnings, the datapoints become much denser than the datapoints before the warning. Is this right?

Thanks for your help.