I mentioned about tracking and there being two ways "around" the polar alignment accuracy requirement for unguided imaging.
One has to do with an experimental command in OnStep that has you center on a bright star, zeros the PZ and PE alignment terms, then does a goto back to the star. You then center the star with the mount's polar alignment adjustments.
The other way is to not bother aligning the mount accurately but use OnStep's dual axis pointing model compensated tracking. Again an experimental feature but since we're on the subject I did some tests of this feature Monday night and also last night. It seemed to have problems during the first tests and oddly (because I saw it working in simulation) showed tracking was drifting in Sky Planetarium also. The errors on the first images were in Declination and so I wondered about the sign reversal and pier side. After adding a little code to deal with that I did some testing.
For this test we want a really bad polar alignment so I just sat the mount down and pointing roughly at Polaris. After a 3-star align OnStep said I was 125' to the right and 3' below the NCP. It was windy out and focus was pretty bad but I was more concerned about the goal than taking pretty pictures. Also, one image set was East of the Pier and the other West. The camera was orientated so that the RA axis is mostly up/down and Dec left/right. Declination for both was about 40 or 50°N. These are all stacks of 4x 1 minute exposures (no registration so the equal of a single 4 minute exposure for showing drift.) The before/after images show tracking with no compensation to the left and tracking with full pointing model compensation enabled (dual axis) to the right.
Probably more alignment stars would help (as would being a bit closer to the NCP than ~2 degrees off) but obviously something good is happening.
For comparison to star trails in Khalids images my images to the left have star trails of about 100 arc-seconds length. The high frequency noise is not present from what I can tell (keeping in mind these are the same MPJA 18:1 NEMA11 stepper motors and LV8729 drivers on this LXD75.) The following shows the full 2 arc-seconds/pixel (blown up 4x)...
The latest Alpha branch OnStep on Github has the code I added last night.