AM5 frequency analysis and drift corrected graph

tempus

I got the frequency analysis and drift corrected graphs for my AM5. The p-p error reported by Phd2 log review app are a bit bigger than the original plot provide by ZWO. I don't see a repeating pattern (unless I am missing the obvious) in the drift corrected graph. I don't think the predictive algorithm of Phd2 would be of any help for this mount (unless I miss the obvious again).
The frequency analysis reports a peak at 142 sec and 416 sec in the few imaging runs I did since I got the mount. I am not sure what they are exactly. Can anyone help me analyse those results?

w7ay

tempus I don't think the predictive algorithm of Phd2 would be of any help for this mount (unless I miss the obvious again).

You are right. The standard PPEC methods are pretty useless unless you can run for a sidereal day -- which is impossible because there is daylight in between, unless you spread the measurements out over a year.

The frequency analysis reports a peak at 142 sec and 416 sec in the few imaging runs I did since I got the mount. I am not sure what they are exactly.

The "416" probably corresponds to the fundamental frequency whose period is 430.82 seconds. The third harmonic thus has a period of 143.6 seconds -- that is the other strong spike that you reported.

I also see some very strong fifth harmonic on your mount (appears to be quite common with the strain wave gears that ZWO uses). It appears to come and go, and thus averages out in the long term FFT (either that, or the phase term of the fifth harmonic is time varying). This is where time series is better than frequency series analysis -- frequency analysis is only valid for stationary signals.

Chen

tempus

w7ay I also see some very strong fifth harmonic on your mount (appears to be quite common with the strain wave gears that ZWO uses).

Thank you Chen. Can you point out where you see those strong fifth harmonics? I am sure it would be helpful for others (and myself) who read this thread.

Thank you!

w7ay

tempus Can you point out where you see those strong fifth harmonics?

Places like this:

See that there are 2.5 cycles (or so) per 1/2 a period of the 430.82-sec fundamental?

Notice that it does not pop up everywhere, and that is why I mentioned the non-stationarity property to use frequency domain analysis.

https://en.wikipedia.org/wiki/Stationary_process

and you cannot use Fourier transforms with non-stationary signals:

https://www.quora.com/Why-is-Fourier-transform-not-suitable-to-analyse-a-non-stationary-signal

You really need to know the limitation of tools, and why and how they work, before using them blindly.

There is something called the Short-time Fourier transform (e.g., using very short FFT windows) that you can use to see these short term artifacts:

https://en.wikipedia.org/wiki/Short-time_Fourier_transform

Note down the hour angle (the angle where the OTA is pointed to relative to the Meridian) of the mount at the time it appeared on the graph above. In the future, if guiding suddenly gets erratic, compare the hour angle of the mount to what you have noted down, and if it appears at about the same time, you will know where the error is coming from, and know that there is nothing you can tune autoguide parameters to help, and thus not waste time.

There are multiple places with your mount, not just the place I screen grabbed.

Notice that the peak-to-peak amplitude of the fifth harmonic is larger than 1/5 the peak-to-peak amplitude of the fundamental itself. This means that the fifth harmonic distortion willl more than double the slope of the periodic error of a smoother mount.

Chen