Dual-pulse test

There was a discussion about the speed of Qspice64.exe and Qspice80.exe, refer to the link. Actually, fastmath (Qspice64.exe) is not necessarily faster than Qspice80.exe. We had a counterexample in that post, but you know, Qspice keeps updating and I don’t verify.

I personally prefer 80-bit math over 64-bit math. In general, I have found that with ideal switches or diodes, it is better to use 80-bit math (your simulation always with ideal device, I suggest always 80bit). Mike recommends using fastmath (Qspice64) unless you encounter a problem. However, I always set my Qspice to run with Qspice80 (disable fastmath in options). Simulation time is not generally a concern for me. Both Qspice64 and Qspice80 use 80-bit math, but Qspice80 has more 80-bit math implemented.

I expect you disable fastmath to solve the convergence problem… but in weird situation, I did observe thing can run with Qspice64 but not Qspice80, but it very rare.
And I expect crash means timestep too small or something like that, not really program crash, right?