Need help interpreting CCD Inspector results

[vlaiv]

I've recently taken a shot of M27, and noticed rather funny shaped stars in the corners.

Scope in question is TS (GSO) RC 8", and star deformations are quite unexpected. Some being out of focus displaying astigmatism that I would expect out of RC due to field curvature, others are out of focus without apparent astigmatism (no ellipsis shape, still round with offset center), and some display sort of coma appearance (different corners have different aberrations).

Given that I Roddier tested that scope to 0.94 Strehl I think that optics are ok, which would leave camera tilt + collimation (one, the other or probably both). I decided to download trial version of CCD Inspector just to see what results it would give me.

Results gave me bit of a scare, to be honest. I ran test on 60 calibrated frames, each 1m long exposure, guiding was spot on (OAG, almost no shift between the frames, I think that largest shift when aligning was sub pixel), I did not use dithering, my guide RMS error was in range 0.5-0.8" (very good given HEQ5 and seeing), seeing was good to fair (FWHM measures between 1.9" and 3" with most of the frames being around 2.2") but results for curvature map differ quite significantly between the frames.

I've created a mosaic of curvature maps for first 25 frames (did not want all 60 due to size, mosaic is large as is), and results for individual frames fall in range:

Curvature: 47.7 - 62.5"

Tilt X: 0.2 - 0.7"

Tilt Y: -0.1 - 0.1"

Total tilt: 2% facing left - 34% facing right (with changing direction, and angle)

Collimation: 1.0" - 4.4"

Here is the mosaic to visually show how field curvature is dancing all over the place between the frames

Now the question is what does this all mean?

I would expect results to differ between the frames (don't know, but I would not be worried if values jumped around 5-10%), but range of values just seems too much.

Don't know robustness of CCDI algorithms, but it is possible that target is not suitable for analysis - very rich star field of milky way, it might be that better results can be obtained with in sparser areas of the sky (200-300 stars instead of thousands).

Or it might mean that something is seriously loose in my optical train and shifts with each guide correction?

Could the camera jumping around be the cause (better), or is it one of the mirrors (probably not so good situation)?

Can anyone please help with this?

[vlaiv]

To answer my own question - it is down to CCDI.

I did have problem with collimation, and something loose in optical train (don't know what it was, but I suspect one of the extension rings for RC 8" that comes before focuser (between ota and focuser) and brings focus in 50mm), so I tightened everything, did a round of primary collimation (I might need a second round, don't think I nailed it 100%, still have a bit more curvature on one side, just by eyeballing subs), fixed secondary collimation after, and star shapes improved considerably.

I then checked it with 5 shots of NGC 6940 and CCDI, and still got results all over the place, so my conclusion is that CCDI is probably not reliable unless one uses high SNR stack instead of the subs, and even then only as a guide.

[xthestreams]

I, glad it’s not just me! I don’t u der stand CCDI at all despite really wanting to be able to trust it. 

[tomato]

This thread may be of use:

 

[Adam J]

On 20/07/2017 at 19:22, vlaiv said:

I've recently taken a shot of M27, and noticed rather funny shaped stars in the corners.

Scope in question is TS (GSO) RC 8", and star deformations are quite unexpected. Some being out of focus displaying astigmatism that I would expect out of RC due to field curvature, others are out of focus without apparent astigmatism (no ellipsis shape, still round with offset center), and some display sort of coma appearance (different corners have different aberrations).

Given that I Roddier tested that scope to 0.94 Strehl I think that optics are ok, which would leave camera tilt + collimation (one, the other or probably both). I decided to download trial version of CCD Inspector just to see what results it would give me.

Results gave me bit of a scare, to be honest. I ran test on 60 calibrated frames, each 1m long exposure, guiding was spot on (OAG, almost no shift between the frames, I think that largest shift when aligning was sub pixel), I did not use dithering, my guide RMS error was in range 0.5-0.8" (very good given HEQ5 and seeing), seeing was good to fair (FWHM measures between 1.9" and 3" with most of the frames being around 2.2") but results for curvature map differ quite significantly between the frames.

I've created a mosaic of curvature maps for first 25 frames (did not want all 60 due to size, mosaic is large as is), and results for individual frames fall in range:

Curvature: 47.7 - 62.5"

Tilt X: 0.2 - 0.7"

Tilt Y: -0.1 - 0.1"

Total tilt: 2% facing left - 34% facing right (with changing direction, and angle)

Collimation: 1.0" - 4.4"

Here is the mosaic to visually show how field curvature is dancing all over the place between the frames

Now the question is what does this all mean?

I would expect results to differ between the frames (don't know, but I would not be worried if values jumped around 5-10%), but range of values just seems too much.

Don't know robustness of CCDI algorithms, but it is possible that target is not suitable for analysis - very rich star field of milky way, it might be that better results can be obtained with in sparser areas of the sky (200-300 stars instead of thousands).

Or it might mean that something is seriously loose in my optical train and shifts with each guide correction?

Could the camera jumping around be the cause (better), or is it one of the mirrors (probably not so good situation)?

Can anyone please help with this?

Its probably really tiny changes in tracking between each shot and changes in background levels resulting in veriability  / noise in how CCDI is calculating your curvature.

[xthestreams]

Thanks for the links! It now makes a LITTLE more sense, I didn't realise I could "stack" images within CCDI to get an aggregate view of the distortions - the new run of data suggests my "notorious" GSO RC12 scope is okay but image tilt is the issue.

More work to do but feeling less paranoid.