Thought I may ask the question about a really weird result I'm getting from deep-sky-stacker. During astrofest I was shooting Corona Australis and got some wonderful shots from the D800 which stacked beautifully. I also took the same shot through the telescope and the results were garbage.
After a bit of digging I ended up stacking them in Deep Sky Stacker 4 times, once plain, once with bias only, once with darks only, and once with bias and darks as one would expect.
The results were exactly the opposite of what I expected as seen in the pictures below. The one without biases and darks looked by far the best.
In each case the picture was processed the same way, crop, rotate, histogram stretch, curves, and saturation. The only thing different in each picture was that the histogram is normalised to ensure they all had the same average brightness to start of with. The numbers I had to type in showed the ones with bias frames were significantly darker than the others.
Dark frame subtraction solved some green colour cast in the camera but ultimately added more noise to the final image...
Now I'm doing the same experiment with another picture that I didn't notice had a problem to see if this is isolated or just the first time I've noticed it. My Helix shot is now being stacked multilple times as well to see if I have had this problem elsewhere, I'll see the results soon.
Any idea what could be causing this? The bias frames are the same bias frames I've used for all my other stackings, only the darks are unique due to not normally shooting 300s exposures.
Any idea where else I can start looking for a solution?
Try a few different settings. Set drizzle on (2x or 3x). Align channels in the output. Try a median kappa clip of 2.00 / 50 iterations for ALL subs (light, dark, flat, bias). Try AHD debayer or Super Pixel (though not sure if this works with DSLR RAW etc).
It looks to me like a debayer issue - adjust the bayer matrix transformation settings under RAW/FITS DDP settings menu.
1) The obvious first: are the darks and biases taken with the same ISO, duration, and (broadly) temperature as the lights?
2) you don't mention flats: is that because there aren't any, or they processed fine? If the latter, then I would think you mt have light leaking in when you took your darks and biases. I had this when light was getting in the viewfinder, with similar results. The darks *looked* ok until I stretched them.
3) uncheck the "set black point to zero" box in the settings menu. I have no idea why this might work, but it did for me when I had a similar issue.
Looks like the issue was similar but very subtle in some of my other pictures. The biases came out no where near as bad in the case of my helix picture but it looks as though there's less noise on the ones with no bias / darks applied.
Quote:
Originally Posted by LewisM
I have had this happen previously.
Try a few different settings. Set drizzle on (2x or 3x). Align channels in the output. Try a median kappa clip of 2.00 / 50 iterations for ALL subs (light, dark, flat, bias). Try AHD debayer or Super Pixel (though not sure if this works with DSLR RAW etc).
It looks to me like a debayer issue - adjust the bayer matrix transformation settings under RAW/FITS DDP settings menu.
It's not a de-bayer issue as that would result first and foremost with a messed up colour scheme. That is actually ok. I use predominantly ADH and only ever use drizzle on small objects (drizzle has little to do with stacking and lots to do with light frame supersampling).
Quote:
Originally Posted by Jon
Could be a number of things:
1) The obvious first: are the darks and biases taken with the same ISO, duration, and (broadly) temperature as the lights?
2) you don't mention flats: is that because there aren't any, or they processed fine? If the latter, then I would think you mt have light leaking in when you took your darks and biases. I had this when light was getting in the viewfinder, with similar results. The darks *looked* ok until I stretched them.
3) uncheck the "set black point to zero" box in the settings menu. I have no idea why this might work, but it did for me when I had a similar issue.
1) Yes, yes and yes. Using a QHY10. Darks and Lights are taken at 300seconds. Temperature is -20degC in all cases.
2) No flats. I rotated the camera during the imaging session resulting in flats that do more harm than good. Light leaking in darks and biases does sound plausible. I may re-do them, especially the biases.
3) Sounds like a good setting to try, I won't get a chance till wednesday but I'll give it a go then.
I read on the PI forum about issues importing DSS files into Pixinsight for post processing. A thread on the startools forum highlighted that versions of DSS come with their own issues.
Alternatively, run your images through the PI BatchPreprocessing script and see what results you get. IMHO, BP produces superior results.
What is the mosaic pattern of your QHY, RGGB or something else?
How many bias and dark frames are you taking, Chris? If it is only a few then you will be adding noise when you calibrate.
It was 20 bias and 20 darks from what I recall, maybe 10 darks.
Quote:
Originally Posted by rcheshire
The first frame looks like fixed pattern noise.
I read on the PI forum about issues importing DSS files into Pixinsight for post processing. A thread on the startools forum highlighted that versions of DSS come with their own issues.
Alternatively, run your images through the PI BatchPreprocessing script and see what results you get. IMHO, BP produces superior results.
What is the mosaic pattern of your QHY, RGGB or something else?
I've been using PI Preprocessing elsewhere, however I'm new to it and I haven't been able to get either my helix nor my corona australis pictures to work in PI using batch pre-processing due to a lack of stars when stacking. PI throws an error.
Can't remember exactly what pattern the QHY had, but it shouldn't be relevant as all 4 pictures have the same debeyering method applied.
I've been using PI Preprocessing elsewhere, however I'm new to it and I haven't been able to get either my helix nor my corona australis pictures to work in PI using batch pre-processing due to a lack of stars when stacking. PI throws an error
You could try using the standalone StarAlignment module and reduce the number of stars required by changing the Star Detection or Star Matching setting - I don't have PI in front of me right now. Look for the default 2000. Reduce it to 200 or less.
I would open a dark frame and possibly post that here, it will tell you more about what is happening. It is possible that the dark frames you are using picked up some signal from somewhere.
I would open a dark frame and possibly post that here, it will tell you more about what is happening. It is possible that the dark frames you are using picked up some signal from somewhere.
PI looks like it may be waaaaay to complicated if I use standalone modules. But oh well if that's the only way then I'll spend the rest of the year trying to figure it out.
I'll do some comparative stacks in images which from PI as well where the data was good. That will rule out the dark and bias frame quality.
Quote:
Originally Posted by Jon
3) uncheck the "set black point to zero" box in the settings menu. I have no idea why this might work, but it did for me when I had a similar issue.
Just checked this setting and it appears that it only applies to RAW files not FITS files. I'll uncheck it anyway and do a stack but I'm not expecting anything.
Hmmm.... there is a definite crosshatch in the areas of low SNR. If your calibration frames are introducing artifacts, ditch them and capture a new set. Try 100 bias frames and see if that adds less noise to your images - should be the case. Have you investigated the colour mosaic setting for your camera - could be the problem or part thereof? Are you dithering?
The PI procedure is correct. BPS does all calibration in monochrome, so if you do manual processing, see that you type 'raw cfa' (without quotes) in Format Hints of the image calibration module. That will override any Format explorer settings. BPS does not subtract a bias pedestal and avoids manual settings when using ImageCalibration stand alone.
You can follow the updated PI DSLR_RAW workflow in the tutorials section for guidance if you wish. That way you will have correct settings.
Not sure what you mean by colour mosaic, if you mean the beyer pattern then yes I've played with that. I'm not intentionally dithering but guiding was quite crap that day so there was a bit of drift between frames.
I've got the camera re-doing bias and as many darks as possible right now. I'll up my bias to 100 frames and we'll see how I go from there. Thanks.
Keep in mind that I am thinking DSLR/OSC. It's a start. Sometimes drift between frames can exacerbate things - better to do in clean steps. The image is very good otherwise. Nice data.
The way I see it drifting between frames shouldn't affect the noise as calibration of darks and biases is done pre registration and stacking and after calibration the noise should still be uniform and random in nature.
I'm using a CCD by the way. Anyway to continue making bias and dark frames I have the camera sitting in the fridge this morning (the deltaT on my camera isn't getting down to -20C at the moment ). With any luck by this afternoon I'll have enough new bias and darks to try again.
The way I see it drifting between frames shouldn't affect the noise as calibration of darks and biases is done pre registration and stacking and after calibration the noise should still be uniform and random in nature.
I'm using a CCD by the way. Anyway to continue making bias and dark frames I have the camera sitting in the fridge this morning (the deltaT on my camera isn't getting down to -20C at the moment ). With any luck by this afternoon I'll have enough new bias and darks to try again.
Yes. And calibration will be more reliable than a DSLR. Still, any calibration mismatch is covered up as well.
Fixed!!!! There was totally a problem with my bias and dark frames. I'm not sure if it was the number of them but I re-did them all and took more of them.
The results with 100 bias frames and 60 dark frames speak for themselves:
Top left: Old Bias and Darks (DSS stack)
Bottom Right: Old Bias and Darks
Bottom Left: No Bias and Darks
Top Right: New Bias and Darks
Also what speaks for itself was the way PI stacks the poor data vs DSS. Excuse me while I drag DSS to the recycling bin.
Agree totally re PI calibration and stacking. Its anywhere from subtly better to shockingly better against anything else I've ever trialed. People always get hung on up how the interface works etc, but for me its about quality of outcomes that matters. The difference between DSS and PI on stacked widefield DSLR images is pretty frightening too when you zoom down to the pixel level. Such comparisons are always heavily dependent on how well the users optimises settings in each piece of software of course, but I'm yet to see anyone show an example where PI was inferior.
Couldn't agree more. Glad to see you've got it sorted. When you get poor results with PI, look to your data. 9/10 it's calibration. I have a suspicion that your old bias and dark frames may have seen prior processing. Further subtraction just doesn't work.
). With any luck by this afternoon I'll have enough new bias and darks to try again.
