Max_5478 For the guide I use an Orion 60/240 with Asi 290 mini mm (exposure 0,5 sec).
That looks fine.
However, if you don't consistently get 12 stars in ASAIR guiding, you may want to use a larger sensor; but make sure you can get 2 FPS when using 0.5 second exposures. FPS is more important than exposure time. The reason for using 0.5 second exposures is to get that 2 FPS. For larger sensors, ASIAIR may not keep up with 2 FPS even when set to 0.5 second exposures. Real computers have less of a problem in that area.
Nope, you are correct.
The reason for doing that is to try to find the smallest max pulse durations. Most mounts, running at 0.5x sideral rate, don't need pulses that are longer than 80 ms to fully keep up corrections. Crummier mounts may need 200 ms pulses; but 2000 ms that ZWO defaults to is pure rubbish. Unnecessarily large limits allow the autoguider to make "corrections" from things that are not related to the mount (wind gust, centroid measurement error, bright satellite passes through a guide star, etc -- those are momentary and will correct themselves).
Fortunately, you have one of the 430 second period (approx 1.8 degrees) mounts. If the gears have the same percentage harmonic distortion, a mount with 288 second period (~ 1.2 degrees) would have increased the worst case slope by 50% - which is not a small amount -- a 0.4" RMS mount becomes a 0.6" mount. I would have asked for a 50% price decrease.
If the max pulse duration is too small, what happens then is that the guide pulse is insufficiently long to counter the slope of the periodic error.
I can have an almost vertical tangent for a tiny section of the curve
Yes, that is typical of strain wave gears -- and it also depends on manufacturing precision of the gears (there is a flexible (yes, flexible!) spline gear inside a strain wave gear, and the sloppier it is, the worse the problem is. That problem can even cause the axis direction to change -- i.e., an RA error then couples into declination error. If you see both RA and declination charts moving in sync sometimes, that could be a cause; it can also come from a mount cone error (from ZWO's own posting, they are not modeling the mount and calibrating out manufacturing imprecisions; I pointed that out in a different thread).
The key to looking for the worst case slope is to find the max duration needed to be large enough to correct those times.
I calculated an error of 0.16 arcsec and therefore, if I understood correctly, I should lower the Ra and Dec values from 2000 to 100-150 ms (values similar to Dave's, maybe even less) with an aggressiveness between 20 and 30%.
You probably mean 0.16 arcsec/second. It is a slope. That is how fast (not how much) the mount is deviating from a perfect tracking rate. How much would be the PE curve itself, how fast is the slope of the PE curve. Once you grok this, that sawtooth curve should become obvious.
If you are using 0.5x sidereal rate, the correction pulse moves at 7.5 arcsec/sec. So, to be able to keep up with 0.16"/sec, you would only need a pulse that is 0.16*1000/7.5 milliseconds = 22 milliseconds long. If you are using 0.25x sidereal guiding rate, then the pulse needs to be 43 milliseconds long, etc.
If you now assume that the (useless!) part of the chart that ZWO gave you is not showing the worse case slope (it is a segment that has worst case PE amplitude, not slope), and apply a 2x factor, then 50 milliseconds at 0.5x sideral guide rate, or 100 milliseconds at 0.25x sidereal rate should be sufficient for the RA max pulse duration.
The only problem with setting such small limits is that dither (with todays software) can take a long time. I.e., you can only recover at a rate of 0.16"/second (or whatever). So, if you had dithered by 6 arcseconds, it could take 38 seconds to complete a dither. So, try to dither as little as you can get away with. If you write your own guide code, then just don't limit the pulse duration when you dither -- just keep on applying the pulse until the guide star(s) has(have) moved the amount of time.
Chen