George's calibration was done at 85.4° DEC, thats very close to the north celestian pole which could possibly explain the issues. [Ah, Kevin spotted it already...]

George: as Kevin said, guiding near the pole can be problematic. You certainly don't want to calibrate there and re-use that calibration elsewhere. The reason is that the guider assumes that (a) the RA and DEC directions are orthogonal, and (b) that the directions are constant throughout the image. These assumptions are roughly true through most of the sky, but near the poles the RA direction will get more and more curved.

Peter: Ekos calculates RMS in arcseconds, and thus needs to convert to arcseconds from a displacement in pixels, which is the thing that's actually measured. The equation used is:
arcseconds-per-pixel = 3600 * (180/Pi) * pixel_diameter_millimeters / focal_length_millimeters.
Should the system, for some reason have the wrong pixel width or focal length, either due to a bug or a user error, then it will report wrong values for RMS error, though it may still guide well (as long as that same error was in place when it calibrated). Whatever the issue was, I'm hearing you say that it is fixed now.

Hy

Still PHD2 did much better. I'll check this on the next clear night and I'll take care to calibrate close to the meridian and horizon; I'll let you know how this goes.

I had only a little time to try last night; the results so far:

- Internal guider calibration looks worse than the PHD2 one ... why?
- Internal guider ends up reporting higher error that PHD2 does.
- I get an error anywhere from 0.7 to 1.5 with PHD2, up to 3 with internal guider.

I'll have another clear night in a couple of days so I'll try using both methods and I'll compare 2-5 min guiding exposures as the two tools might be reporting things a but differently.

Note that for this test I calibrated as close to the horizon as I could (about 20 -25 degrees)

Why did you calibrate close to the horizon? In general folks seem to recommend calibrating near the meridian and the equator, ow where you’re imaging.

Because the south (where the meridian and celestial equator cross) is not available to me. I can try getting closer by pointing east or west, but I don't think I'll be able to get within 20 degrees of the equator and I'll certainly 90 degrees off the meridian.

Maybe you already know, but while Ekos reports guiding error in arcseconds, PHD2 reports it primarily in pixels, and in arcseconds in parenthesis; so be sure to check arcseconds with arcseconds.

Maybe you only wrote it wrong, but to be clear: it's not "close to the horizon", it's "close to the celestial equator" (i.e. close to Dec = 0).

HTH
Matteo

I wanted to come back to this thread to report back on what I have found out.

Thanks everyone for your explanations and advice, it was useful.

So basically, things are as follows:
PHD2 does do it 'better', but this is actually irrelevant because it partially masked the real issue.
The entire system was horribly unbalanced. While I am aware of 3D balancing and I had thought I had done this correct, this was incorrect. I had RA and DEC balanced, but any movement would cause both axis to run away in one direction or another.
The internal guider had a much harder time coping with this than PHD2, but again .. irrelevant. Both results were bad.
While I have not managed to perfectly balance the system, it is now much, much better behaved and the internal guider is managing reasonably well.