Hello,
I am using LTspice to perform transient simulations of a complex coupled oscillator circuit that includes switching devices.
As the circuit becomes more complex, I have observed that the simulation results can differ depending on the LTspice version. When I compared the raw data files generated by different LTspice versions, I noticed that the transient time points are not identical. In other words, the internal adaptive time-step sequence appears to be different between versions.
This difference seems to affect the extracted oscillation frequency, phase relationship, and final state of the coupled oscillator system.
I understand that LTspice uses an adaptive time-step algorithm for transient analysis. I also understand that the maximum time step can be limited using the fourth parameter of the .tran statement. For example:
.tran 0 {time} 0u 1n uic
However, this only limits the maximum time step. It does not force LTspice to use a completely fixed internal time step, since the simulator can still use smaller time steps depending on switching events, convergence behavior, or local truncation error.
My questions are:
1. Is there any option in LTspice to force a completely fixed internal time step during transient simulation?
2. If a fixed internal time step is not supported, what is the recommended way to minimize version-dependent differences caused by adaptive time stepping?
3. Are there recommended .tran or .options settings for complex switching oscillator circuits where reproducibility of frequency, phase, and final state is important?
In particular, I would like to know whether the transient integration grid can be made deterministic and uniform across LTspice versions, so that the same netlist produces the same sequence of time points and more reproducible raw waveform data.
Thank you.
