After installing the latest QSpice update, a previously working flyback converter simulation now crashes instantly at startup.
The simulation immediately fails the initial operating point analysis (Gmin, source) and throws a fatal error on the very first transient step:
PlaintextFatal error: Timestep too small(1.25025e-19) at t=2.01231e-10
The issue appears to be caused by tighter discontinuity or matrix convergence thresholds introduced in the updated solver. It chokes on the sharp mathematical edges of the macromodel’s internal behavioral components—specifically the ideal digital logic subcircuits (COUNTER, SRLATCH, DFFSR) and clamping diodes with steep emission coefficients (N=0.1) right at $t = 0.2\text{ ns}$.
Standard convergence options like .options cshunt=1p gshunt=1n and .options method=gear do not bypass the lockup. Looking for advice on native QSpice flags or macromodel modifications to resolve this solver conflict with behavioral/logic components post-update. flyback.qsch (71.7 KB)
I navigated through my TI model library and observed this same issue in a LM5002 model. Comparing the versions between 22-May and 24-May, the 24-May version fails to resolve the circuit. Just sent an email to Mike and hope the problem can be identified regarding what was updated to cause this issue. Not all TI models have this problem.
Apologies for the confusion there; the .tran directive in my previous message was accidentally left commented out. It was intended to be active for this simulation run to capture the transient response.
Hello all, I checked this morning; there was an update that was around a day newer. This seems to have fixed some of the models that were not working. I suppose maybe with another update or a few tweaks to other models it should be fine.
Yes, Mike uploaded a fix for that. Lucky that you observed this and immediately report it, therefore, this issue can be identified and fixed relatively easier.
I think Qspice more stable in handling TI Pspice model in past few months. But still, TI Pspice model is challenge sometime. But a schematic that work continuously but suddenly stop working, it always worth to report. This year I remember two cases where an update that broke the TI model.
Yes, I think the case is settled now; the other models will in time be fixed with newer updates, or minor tweaks might get them up and running. Thanks for the support. I am marking your reply as the solution if that works.
The best approach is to upload the model you are experiencing issues with for the community to review. We just had a case where @sergio.pesenti narrowed down a convergence issue, and it eventually turned out that Mike could make a change in QSpice. We have talented people in the community to help. JFET op-amp circuit import unsuccessful - QSPICE - Qorvo Tech Forum
It is also common in SPICE to tweak certain parameters to run a TI model. From my point of view, TI models are quite complex. In particular, PSpice does not support digital devices and they have to implement logic with B-sources, which makes things more complex.
Well, my experience is that simply sending a TI model to Mike and saying it is not working generally won’t help. But if the community can narrow down the issue and troubleshoot it to a certain level, sometimes we can contribute to helping QSpice improve. Just like your case, where you reported it to the community once you observed the Qspice performance changed.
Thanks for the guidance and for the continued support from the community. Yes, the uploaded model (flyback.qsch) is the one I had encountered issues with.
I also appreciate the point about narrowing down issues within the community before escalating them further. It is really valuable to have such a supportive group around Qorvo and Mike, helping troubleshoot and improve QSpice together.
Thanks again to everyone in the community for always being supportive and willing to help.
If I may know, what exactly was the issue since an update did fix things around?