Mesliz I tried 3 methods but with strange low results below 0.1
The numbers that look "strange" are probably correct. Especially when two different methods yield similar results.
You are being misled by ZWO's 2000 ms number, which are so out of whack that I had repeatedly called it utter nonsense. I really want to see what mathematical analysis led ZWO to recommend that number.
Good strain wave geared mounts need a pulse duration of less than 100 ms for a guiding rate of 0.5x sidereal, and for many of them, you don't need more than 150 ms even for a guide rate of 0.25x sidereal.
The problem is that you cannot tell from that sheet of paper that ZWO gives you, because the expanded view is for the location where peak-to-peak PE is the greatest, not where the slope of the PE is the greatest.
That is why I usually just recommend doubling the max pulse durations from what you can see on that sheet of paper. And even then, you will find some mounts not needing pulses that are longer than 40 ms to keep up with the mount's error.
However, it would be a disaster (not being able to keep up guiding) if the particular mount needs more than 40 ms @ 0.5x sidereal and you had limited it to 40ms -- thus it is crucial that you make that determination yourself for your particular sample of the mount.
Now, if the number is as low as 40ms @ 0.5x sidereal rate, you can get another huge improvement by using 80ms @ 0.25x sidereal rate. I.e., an even gentler and smoother guiding. I hadn't realized that myself until recently when I stared at the sawtooth guiding dynamics!
Just depend on what you know about slopes, and not let the 2000 ms number from ZWO mislead you about how low that number should be. If the pulse needs to be greater than 150 ms @ 0.5x sidereal, something is wrong with the mount and it should be returned.
Your piece of paper shows lots of wiggles, but notice that they are not strong enough to change the worse case slope by as much as the poster just before you. I.e., you can see a lot of harmonics, but they are not strong (i.e., the amplitude of the higher ordered terms of the Fourier Series is not large). His/her worst case slope is probably 1.5x to 2x your worst case slope.
Chen