Jan,
I was doing some initial measurements of a recently bought mount (an RST-135E that has a Renishaw encoder), and just decided to take a diversion to search for spectral spikes that are similar to the 2.9 second spike that you are seeing.
Turns out I did find two harmonically related spectral spikes (not related to the 430.82 second period of the mount, as far as I can tell) using PHD2 in ASIAIR to capture the data (the RA corrections have been turned off, so I was measuring the mount against the stars, given the star centroid detection in ASIAIR).
This is what a 4096 point FFT (Blackman-Nuttall window) of a 2-FPS PHD2 log file looks like:
The ordinate scale is logarithmic and in units of decibels. 256 bins from the FFT are displayed, starting with DC at the left. I have also used a very long FIR (9000 point) to remove the slow RA drift that is caused by any polar misaligment. Notice the DC term is some 60 dB down from the fundamental of the PE.
The fundamental error (430.82 second period) is that large 72 dB spike at the left. You can see the second harmonic a bit more than 20 dB below it.
To the right, you can see the spurious spike at bin 234 that is 15.6 dB lower than the periodic error's fundamental, and there is a related spike at bin 468 that is 17.5 dB down from the fundamental. The one at bin 234 corresponds to a period of about 9.2 seconds.
The spurii are small enough that they should not cause any guiding problem. But they are there (and I don't know the source of them for now). Especially since the periods are slow enough that a 2 FPS autoguiding can easily handle them.
FWIW, this is the PE curve that was measured (the ordinate scale is in arc seconds):
The peak error is smaller than the Renishaw specs, so I lucked out (the encoder is supposed to be a +/- 2.5 arcsecond encoder). Heck, assuming sinusoids, the RMS error of the RA axis is just 0.75 arc seconds, so I might be able to forgo guiding even with my primary 450mm focal length OTA, although guiding willl keep the target in frame with sloppy polar alignment.
And this is the finite difference (i.e., ∆x/∆t) approximation to the first derivative dx/dt (the ordinate scale is in arcseconds/second):
I have attemped to remove centroid estimation error with a running average of 16 ∆x/∆t values. Typical centroid errors from ASIAIR are much larger than the slope here would cause.
It looks like I can just turn on autoguiding and forget about it (as I did with the non-encoder version of this mount) since it is ASIAIR's centroid detection that limits the guiding performance for me (0.35"-0.4" region total RMS error).
Chen
