This looks like a bug but it's so basic that I'm almost surely going to end up facepalm. Almost.
Repeated observed behavior: Ekos plate solve will claim the telescope is aimed at the requested coordinates (within, say, 30 arcseconds) when in fact it is not. This is a problem with Ekos and not the external solver, astrometry.net in my case, because the correct coordinates of the test shot are written to the FITS header.
Example from last night: requested slew to Rho Capricorni by coordinates, 20h20m57.52s, -17:48:49.43, in the control panel. After settle, requested "capture and sync". Repeated until "capture and sync" reported error less than 30 arcseconds. This was clearly in error because there was no mag 4.8 star in the field of view ever. Captured the field anyway and checked the FITS header; the header reported location of center at 20h27m32.81s, -17:49:33.17. That's 1182 arcseconds difference in RA and 44 arcseconds difference in Dec!
Additional notes: system field of view somewhat narrow at 21' x 14'. Aperture somewhat wide at 25" so plenty of stars were captured in 1 second exposures. In another example with bright star target, Eta Pegasi, the star was captured in the first test shot and Ekos homed in on the star without difficulty. So the problem only seems to occur when the test field lacks a bright star for Ekos to home in on even though the plate solver has no difficulty.
Thoughts? I'm happy to provide access to FITS files. I'll try turn on logging tonight to see if that helps.
Thanks for reading.
No pertinent information in the logs. But I did follow double star STF2205 for a couple hours this evening -- interesting in that the discrepancy in reported coordinates between the plate solver and the (correct) FITS header stayed pretty much constant.
I'm left wondering where does INDI get the coordinates written into the FITS header? Why would they be correct when the plate solver is wrong?
Last edit: 2 months 1 day ago by Mark Copper. Reason: more thoughts and data
Partial facepalm: The Ekos platesolve and mount modules and the telescope tab in the control panel both assume JNow coordinates.
In the mount module is a mount control button that launches a virtual handpad which contains a radio button that provides a choice of JNow and J2000 coordinates. But choosing J2000 on the handpad did not seem to propagate to the control panel or platesolver in my setup.
I think it would be helpful to include coordinate system options in the platesolver module. Conversion routines have already been written in the INDI library, libastro.
Another reason to make J2000 coordinates an (explicit) option in the alignment module is that the astrometry.net index files are keyed to J2000 and the WDS coordinates written to the FITS headers are therefore also J2000. And though other catalogs are available, astrometry.net is the system documented in the Ekos tutorial.
Or maybe coordinate conversion between systems could be a separate entry?
Or am I missing an option somewhere? The INDI written headers include the equinox as a separate field, indicating, perhaps, that it can be changed.