Data suggests need for some Flat post-processing

[ChrisR Oz]

Hi Robin,

I generally look at generated flats overall, without zooming in on detail, as I don't expect there to be any. All correction for slight vignetting and some dust motes are all coarse scale features clearly seen in the Flat viewed as a whole. But recently, I did zoom in and found a strange "textured" pattern in the centre of the Flat frames as generated by SharpCap (both v3 and 4.1). Here is an example ...

Centre crop of a flat frame

Flat-UVIR-centre.jpg (111.14 KiB) Viewed 6502 times

Zooming in further ...

Zooming in further

Flat-UVIR-centre-zoom0.jpg (281.87 KiB) Viewed 6502 times

... and zooming in 4x further, reveals a crazy hieroglyphics pattern ...

Zooming in 4x more

Flat-UVIR-centre-zoom2.jpg (268.67 KiB) Viewed 6502 times

That scale of the pattern is about 10-15% of full range, so quite significant ...

Zooming in on Z scale with colour map

Zoom-colormap-scale.jpg (175.53 KiB) Viewed 6502 times

I guess using this Flat will not doubt "emboss" a texture pattern on all corrected frames. Thankfully, dithering will move this around, so we don't seem to have anything remaining in the final stack.

I use a flat tracing panel source, which seems to work very well. Perhaps this light source is the cause of this pattern, even though placed on the OTA entrance should be well out of focus. However, aside from debugging how it arises in my system, it does suggest a need for some post-processing of the generated Flat frame. I am finding now that a median filter (FWHM 5 pixels) does well to suppress this effect.

Perhaps a post-processing option in the SC Flat frame generation could be an option to remove any residual fine-scale detail in a Flat?

Cheers, Chris.

[ChrisR Oz]

P.S. The journey to apply the correction highlights some areas that could be improved in SC 4.1.

I do the correction in some software of my own, after importing as 32-bit float TIFF, exported from PixInsight. I export again as 32-bit float TIFF. But SC can't use the floating format TIFF, so I use PI to do some format conversions. I find that SC will only accept a Flat in 16-bit FITS format, as both 32-bit float FITS and 32-bit integer FITS are not accepted.

As PI tends to use floating point 32-bit (normalized 0-1) files for most data, it would be handy if SC could accept FITS (or TIFF) as 32-bit float (0-1.0) files.

Cheers, Chris.

[Menno555]

Hi Chris

Question: do you capture your flats during capturing? So with a cooled camera?
I ask because that pattern looks like you have ice forming on you sensor. This can happen in multiple forms. When it happens like this, you almost don't notice it in your subs but it will on flats.
You can check this by making flats with the same settings but with the camera not cooled. If the pattern is not there, then it's most likely is ice forming you have here.

Menno

[admin]

Hi folks,

interesting point from Menno that the concentration of the pattern near the middle of the sensor could be ice and it also suggests a physical cause rather than a software cause (if the software was doing something wrong it would probably be everywhere, or in the brightest or darkest regions, rather than having its own pattern).

It's worth checking the individual flat frame files (which you can save when generating flats) for hints of the same sort of pattern - if you see it in the individual files then the frame averaging code can't be causing it (whereas if it is in the averaged flat, but not in the raw ones, the frame averaging code is the place we have to look).

cheers,

Robin

[ChrisR Oz]

Hi Menno, Robin,

Interesting idea, as it does look like ice/condensation. This is a new 294MM, which I have not regenerated the moisture control tabs.

I’ll try your tip of doing this with Peltier off.

Thank you.
Cheers, Chris.

[admin]

Hi Chris,

with the 294MM there is also the consideration of not taking flats between gain 120 and about 190 - the 294 sensor (colour and mono) can fail to saturate the output signal under the settings used between gain 120 and 190 even when massively over exposed. This can lead the histogram to look like a nice flat (peak near 60-70%), even when the sensor is actually badly over-exposed.

I don't think that has anything to do with the patterns you are seeing, but worth being aware of.

cheers,

Robin

[ChrisR Oz]

Hi Menno, Robin,

I have tried many times, but have not been able to repeat this effect in Flats in SC. Although this is now using SC 4.1. With/without cooling and changing gain from low, through the dubious zone that Robin mentioned, and to higher gain, all produce nice flats using the same flat-panel display as previously.

It is still possible that it was ice, and the weather has not cooperated (a familiar story).
Cheers, Chris.