Ideas on this "noise"

[ppeake]

I am using a D7500 with a 70-300mm zoom lens (mostly at 300mm).
I know the limitations of this lens, and that it could be a lot better  ... before anyone goes off on that tangent.

However, I don't think this problem is really attributable to the lens.
I get a lot of diagonal "noise" in the integrated images. Example below is 35 integrated 3 minute subs.
The only other processing is having run ABE.

This doesn't appear to be "noise" in the normal sense. If I look at the individual wavelet layers (ExtractWaveletLayers) with the default 5+R layers, this doesn't begin to show until looking at the "R" layer.

I can hide it by attenuating the color information and burying it in the black level, but it still shows over features I want.

Any ideas what it is? How to prevent it? How to deal with it?

I can't see any trace of it in the individual subs.

Philip

[ngc1535]

This is a "fixed pattern noise" that comes from repeated pixels of similar brightness/value/response in each image. You will not see the pattern in a single frame but instead only when you integrate everything together. This gives away a little information on your acquisition methods. It appears you took multiple images likely using a sequence that your camera software provides. However, the telescope was not moved between exposures... at least not intentionally. There is a tracking drift- perhaps a polar alignment error, that is constant during your acquisition.

So one thing that can help is to randomly move your telescope a few pixels between exposures. Many camera/mount controllers that talk to each other can do this. You want to dither your data and when you integrate the process of averaging and rejection will result in a more pleasing result. DSLRs notoriously have this issue when looking at low signal objects. Better calibration can help, but I am not an expert on these cameras so I am not certain to what degree you can minimize the fixed pattern you get.

-adam

 

[ppeake]

I think you are probably right. I went through the subs again, looking for any sort of pattern.
There is none. But there is, as you wold expect, colored (random) noise.

But I suspect that there is some coherence in this noise when averaged over multiple frames.

Unfortunately, I am doing this as cheaply as possible, which means an external intervalometer, PHD2 running on a R-PI4 and a SGP tracker.
PHD2 can talk to the SGP (obviously...),  but it has not communication with the intervalometer to sync random movements, and doing those during an exposure sort of defeats the whole point of guiding ...

This wasn't quite so bad before now I think about it. The SGP has some significant periodic variance @ 300mm. That forced me to keep the exposures short, and probably acted to to minimize this effect.

Rather than just setting things up and leaving it to run for 40 exposures, I suppose I could do four runs of 10, with a bit of movement between, and re-syncing PHD2.

[albireo00]

Hi Philip
Have you stacked the frames with DSS?
This is a common problem when using a too low star detection threshold and DSS confuses noise with real stars.

Jose

[ppeake]

No, this is all done in PixInsight. If the images were aligned on noise, the star registration would be off, and it is fine.

This looks just so regular that I suspect that it wouldn’t be super hard to write a filter to remove it.
Unfortunately, my programming skills are not up to that.

If it is a common problem with DSLR cameras, maybe someone will get around to it one day...

[pfile]

one thing thats worth a try is to load one of your calibrated frame and inspect for hot/warm pixels hanging around. if you see them you can try using CosmeticCorrection to remove them. this may help mitigate the streaky "walking noise" in the integrated image.

rob

[ppeake]

I did use CC on the subs.
The RA guiding was pretty good (for a SGP), and the polar alignment was good enough to keep the DEC graph pretty much in sync with the RA graph for most of the acquisition time, so any hot/warm pixels are not going to move much, certainly not as much as the length of those "streaks".

Looking around, I see other people with similar problems, so it does seem to be endemic to DSLRs.

This might be just due to trying to stretch the image too far, bringing up the noise, which happens to be somewhat coherent, making it visually more obtrusive than random noise would be. Maybe the real answer is just collect more data and be less aggressive in processing.

.....

Hmmm...maybe not. Took another look at the integrated image with auto stretch, it's visible there.