cannot solve milky way photo

[James Sun]

Hi, 

I tried various downsampling options, still cannot resolve this photo:

https://nova.astrometry.net/status/9853292

https://nova.astrometry.net/status/9853314

Should at least be able to see Sagittarius?

Thanks,

James

[Dustin Lang]

Hi,

Unfortunately, astrometry.net is quite bad at very wide-field images!  They should be the easiest, but the way the algorithm works causes it to not work well when there is a lot of distortion.

One thing you can do is crop out the middle section of your image, where the distortion isn't as bad.  For example, here's the middle one of your images:

https://nova.astrometry.net/user_images/10006699#annotated

cheers,

dustin

--
You received this message because you are subscribed to the Google Groups "astrometry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astrometry+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/4014673d-fdb2-4b15-b947-ed03e893ed86n%40googlegroups.com.

[Danko Dnevic]

Hi,

it seems that issue is not about lens distortion (AN is quite happy solving very wide fields around the zenith, or with bottom above 30 deg elevation), but with atmospheric refraction (particularly its gradient) below 45 deg elevation.

With narrow fields, above 15 degrees elevation, refraction is almost constant over the field, but as field widens in the vertical sense (especially those portrait oriented shots, like above) gradient might become extreme.

In the above shots, atmospheric refraction is approx.:

- 14 arcmin at 3 deg elevation

- 10 arcmin at 5 deg elevation

- 6 arcmin at 10 deg elevation

- 4 arcmin at 15 deg elevation

- 2 arcmin at 30 deg elevation

- 1 arcmin at 45 deg elevation
That makes vertical position error change, by cot() law, between 10(20) and 4(8) pixels in just about 12 degrees. Again, gradient is problematic, not refraction itself.

This is almost impossible to model by TAN WCS and quite hard to compensate by SIP. To make things worse, any previous camera calibration coeffs are useless.

For modeling, useful article to read is Bennett's: The Calculation of Astronomical Refraction in Marine Navigation doi:10.1017/s0373463300022037

Currently, I'm trying to make compensation, but have a hard time positioning compensator. Despite the fact that atmospheric refraction is extrinsic parameter, pre SIP placement doesn't seems correct, because forcing tweak on shifted coords and (I think) that invalidates SIP portability. Second option (the only one for me), is to try index tweaking, based on source elevation (yielded from catalog Ra/Dec and JD/lat/lon). Dustin, it would be of great help, based on your knowledge of AN internal flow, if you can give suggestion where to position that code (probably just set of vector multiplications)?

With Best Regards,

DD