Here is a test for all you 'experts'.

[avandonk]

This image is the curvature output of CDDInspector for RGB and HA images taken sequentially with the RH200. Nothing else was changed apart from the filter. Tracking was better than one arc second.

Please give me you best evaluation or opinion. Remember neatest correct entry always wins!

Large image here 1MB

http://d1355990.i49.quadrahosting.com.au/2014_01/CCDI_F.jpg


Bert

[Geoff45]

I await the results. I've always found the curvature analysis of CCDInspector to be highly variable, even with images taken directly after each other.
Geoff

[RickS]

That's been my experience as well, both with curvature and tilt measurements.

These days I prefer to check over my subs with PixInsight using the tools in Mschuster's Scriptbox.

Bert: do you want to leave copies of those subs somewhere? Could be interesting to see how they compare with different tools.

[avandonk]

Does it help if my inspection of the corners as far as star distortion correlates perfectly with the curvature output of CCDI.

Why CCDI is all over the place is because our images are!

Whether this is due to flexure or any other contributing aberration is a matter for elimination.

This is effect is far more insidious than what I have presented here. You will be very surprised what the problem/solution really is!
Perfect hindsight is a wonderful thing.


bert

[MrB]

Filters?

Edit:
haha, obviously it is the filters when you mention it's the only thing you had changed ;)
One word answer when I couldn't think of the word I needed; 'perpendicular'

Anyway, here goes:
Is it due to the filters not sitting perfectly perpendicular to the optical axis in the filter wheel?
Or maybe the filter wheel is tilted?

A stab in the dark :D

[cfranks]

My experience with exactly the same effect is that the CCDI results depend greatly on the density/brightness of the star field being imaged. Try a large image of Omega Cen. vs Crux for example. Possibly your different filters also effected the density/brightness sufficiently enough to give you the variation you see.

Charles

[mill]

I am not a expert but i think the filters are not 100% parfocal and/or perfectly flat compared to each other (tilted).
Also the FWHM value between filters is different.
Also the refraction thru the filters might be different.
I might be totally wrong here :shrug:

[avandonk]

I will state the evidence again. My Astrodon R,G,B and HA filters are parfocal. The HA is 8 micron different to the RGB which are identical.

I took four sequential images of a non descript area with no bright stars or nebulosity. NOTHING changed apart from the four filters. Here are the four unstretched fit images only corrected for flats and darks.

75MB

http://d1355990.i49.quadrahosting.com.au/2014_01/RGBHA_TEST.zip

This data should give you the same numbers in CCDI.


It is a shock to the unprepared mind that the best filters on the planet also need to be very accurately orthogonal (better than one second of arc!) to the optic axis of a high quality fast astrograph such as the RH200 with the large chip on the PL16803 camera to get the 'best' stars in the corners.

Below is a small image of the test field.

I was hoping I was wrong! I just needed confirmation from others to see if I had made any logical errors. It is very difficult to change one's erroneous thinking when it is set like concrete.


Bert

[MrB]

I'll take that as a yes :D
Two words for the same thing.

I'm no expert, was the only thing I could think of.

[RickS]

Bert,

Have you taken repeated images with the *same* filter and seen consistent results from CCDI? It never happens for me even with a refractor (FSQ-106ED) and a very solid imaging train.

I certainly wouldn't be surprised if a very slight tilt on the filters caused problems at f/3. I'm amazed your filters are parfocal at f/3. My Astrodons need significantly different focus offsets at f/5.

Cheers,
Rick.

[Peter Ward]

The shift of the center wavelength of a filter can be accurately quantified by a model of the center wavelength λ vs. angle of incidence θ .

With a steep F3 light cone, I suspect you are seeing an intensity transmission effect, which attenuates the light transmission the more you go off axis.

CCDI is probably interpreting these dimmer (hence smaller) stars off axis as some for of field curvature....which it isn't.

QED.

[Peter.M]

I would have thought that this effect would only be justifiable with narrowband filters as with broadband colour the wavelength shift is such a tiny fraction of the light allowed to pass through the filter.

If curvature was measured by brightness then putting a star cluster off axis would manifest itself as curvature. (I don't know if this would happen but I assume it dosn't)

[Peter Ward]

They are all multi-layer filters....interestingly the H-alpha has the most layers and throws up the odd result....and smaller stars....:question:

Yes, putting a star cluster off axis shows up as curvature in CCDI. (trial copies are available as free downloads....good fun to play with)

[Peter.M]

Hm interesting, I am happy to be proven wrong might give it a go for myself next time im out.

[multiweb]

Yes I have experienced that as well. For CCDIS to be effective you have to pick a homogeneous field devoid of any bright stars or nebulosity. A cluster for sure will bugger things up. I did such a field centered on NGC104 once and the field was bell shaped.

[MrB]

Pardon my ignorance, I have never used CCDI, but I thought the idea was to use a light box or similar, like taking a flat field?

[Geoff45]

CCDI is used to measure stuff in a pic you've taken. It doesn't actually do anything except diagnose your pic. It is supposed to tell you things about how round your stars are, focal plane tilt, focal plane curvature, FWHM, vignetting and more.
Geoff

[MrB]

Thanks Geoff, I had thought it was only for tilt/vignetting/curvature etc which I figured only needed a light box.
Had no idea it was for star roundness etc too.
Always learning.