Issues Simulating Duffing Oscillator In LTSpice

[Emery]

Hello. I am new to LT spice and I am having a lot of trouble simulating a duffing oscillator. I‘ve tried changing integration methods, changing tolerances, using .startup, using .uic, and lots of other things.

I get two different errors depending on my settings (see image):
1) time step too small without .uic
2) singular matrix with .uic

any help is much appreciated. I have included images of the errors, my .asc, as well as the paper that this is based on which includes the circuit diagram. Note that I am using an LF356 op amp. I know that my .include says LF3562, but the ’2’ is just due to me changing the name of the .sub file.

 

[throbscottle]

Well, I'm running that simulation. Changed the LF356 to one of the educational models and downloaded the model for the AD633J.

I ran into "unknown subcircuit" errors. So inspection of the AD633J.lib revealed that the fix for that was to change the "value" line for U5 and U6 from "AD633J" to "AD633". Maybe you have a different version of the model where AD633J is correct - check the ..SUBCKT line near the top to find out.

I have no idea if the LF356 model was causing your problem because I didn't download it.

The errors you are seeing can be loosely interpreted as "this thing is too complex". So you might want to try a different version of the LF356, if there is one. The LTSpice group on groups.io is very helpful in this respect.

A "trick" that sometimes works is to change components to strange values, typically odd numbers, so 1K might become 1.037k for example. Screws with your carefully calculated components but at least you get to see if the simulation runs!

 

[Emery]

Hey thanks so much for your quick reply!

The AD633J and AD633 are slightly different models, but no matter, because switching from an LF356 to an OP27 seemed to fix the issue.

This leads me to believe that the issue is likely due to the fact that I obtained the LF356 model from TI themselves, but it was labeled under "PSpice" models. I remember running into a forum post at one point talking about how PSpice models from TI are optimized for their software and can cause problems in LTspice. I have no idea if that's actually what's going on, just guessing.

One small issue now, is that this simulation takes approximately an eternity to run (10 minutes equates to about 1%). I know that LTSpice was almost certainly not designed to solve this sort of thing, but is there at least a way to give it access to more of my computer's resources?

The proper way to do this is to numerically solve the system of coupled differential equations laid out in the paper, which I have done in python. It would just be nice to corroborate that result with the actual circuit experimentally (which I have also done) and with LTSpice.

 

[Tony Stewart]

If you want Rail to Rail OP Amps , choose these R2R I/O type, as they go nonlinear often and act strange with inverting outputs when Vin=Vcc which is a topological fault of Vcm range and not a Duff or rookie mistake as in golf causing oscillations or and oscillator named after Dr. Duff. The Falstad .js circuit simulator has some others that are similar under Chaotic circuits.

http://tinyurl.com/26l7f39j SIM