Processing an Ha image the other day with my new ASI1600MM-C, I noticed a faint checkerboard pattern in the background. I have seen some discussion of it online, but nothing seemed conclusive. Any ideas how to handle it?
The attached ridiculously stretched processed stacked image shows the pattern clearly. It's not apparent in any individual sub as it's way below the noise level - which is already low. Not exactly sure at what level it is in the original data, since there's so much nonlinear stuff been done to this image. As it stands in the stretched version, the background is oscillating between say 12000 and 13000 ADU in that pattern.
Not sure if it's going to prove a problem in an image stretched more reasonably, but I think there are hints of it (bottom centre) in this image on Astrobin.
Comments/advice appreciated.
Thanks,
Paul
PS This was gain 202 offset 50.
PPS Subs were dithered to try to address any such fixed pattern noise issues. And obviously biases, darks and flats used.
EDIT: Just confused myself completely now. That image is a stack of subs from two nights. One night I dithered more than the other, so looked at the two separate night's stacked individually to see which had the pattern more badly. Neither do!! And yet it's there in the stack of the two nights together - even without any processing. Visible with just a Histogram stretch in PixInsight on the output from Pixinsight's ImageIntegration tool. So is it some bizarre ImageIntegration effect on images from two different nights (with two different noise levels as the moon was much worse on the second set) rather than anything intrinsic to the camera's noise pattern?
I have had some trouble in the past with a 694 CCD sensor that showed pattern noise sometimes and other times not. It cleaned up with a good bias. In that case it was the microlenses showing up I think. It wouldn't be that visible unless the image was stretched a fair bit. Regardless of what its an image of its safe to say its fixed pattern noise. It may be the circuitry setout on the sensor or the microlenses. It does not matter as it handles the same way.
Try making a better bias out of a large number of bias frames and see if that helps. Make sure your bias is at the same temp and gain as the lights. You can get odd effects if darks and biases don't match the lights exactly.
Try making a better bias out of a large number of bias frames and see if that helps. Make sure your bias is at the same temp and gain as the lights. You can get odd effects if darks and biases don't match the lights exactly.
Greg
Thanks Greg. Will give that a go. I also used PI's Superbias. I may then also go back to a straight master bias and see if that helps.
Quote:
Originally Posted by Camelopardalis
Nice Helix Paul
How many subs went into the stack?
I’ve not seen this in my stacks, but I’ll have a good stretch and see...
Thanks Dunk. That's 34 10 minute subs I think from 2 nights. The second night had quite heavy moon and so the stack wasn't much better with the 34 than the first 18 by themselves.
How have you calibrated them? Bias subtracted the darks and bias subtracted the flats?
Do you turn of Dark Optimisation when calibrating?
When taking your bias’, are they the shortest possible exposure or 0.3s? For some reason, exposures shorter than ~0.3s have some issues that cannot be calibrated out.
How have you calibrated them? Bias subtracted the darks and bias subtracted the flats?
Do you turn of Dark Optimisation when calibrating?
When taking your bias’, are they the shortest possible exposure or 0.3s? For some reason, exposures shorter than ~0.3s have some issues that cannot be calibrated out.
Thanks Colin.
Yes re bias subtraction and turning off optimsation. I even, for the flats, used some darks of the flat duration. But bias was short (0.01s). I will try with longer (and more, as Greg suggests).
As I added in my edit though, it's kind of weird that it seems to be the mix of the two nights that causes the issue. Each night by itself seems ok. To that end, I'm cheating and mixing the centre part of the image from the total stack and using the outer part from the first night alone (which was the moonless night). Will update the image on Astrobin with that. But will also try to update biases as suggested.
Paul, your running that 1600 at high gain, and shooting 10 minute subs. Normal practice with the 1600 at high gain is to reduce sub times. You get less well depth at high gain and shooting for longer doesn't make it any deeper. Have a look at Shiraz's narrowband exposure chart, which he has posted here. I shoot most narrowband at Unity (Gain 139, Offset 20) and that is with 5 min subs max,i would never shoot longer subs with that camera, even at max well depth of Gain 76. When i shoot high gain (200), i drop my sub times down to 90 sec.
Paul, your running that 1600 at high gain, and shooting 10 minute subs. Normal practice with the 1600 at high gain is to reduce sub times. You get less well depth at high gain and shooting for longer doesn't make it any deeper. Have a look at Shiraz's narrowband exposure chart, which he has posted here. I shoot most narrowband at Unity (Gain 139, Offset 20) and that is with 5 min subs max,i would never shoot longer subs with that camera, even at max well depth of Gain 76. When i shoot high gain (200), i drop my sub times down to 90 sec.
Thanks Glen. I was going by Ray's target ADU above bias from those tables. On the moonless night, even at 10 mins with gain at 202, I was only getting about 500-600 ADU above bias - i.e. less than the target of 650. So not quite shot noise limited. Seems surprisingly longer given Ray's estimates of the expected times, but I guess there's things like the large central obstruction in the RC scope which will add a bit of extra time.
On the second night, I was about 1100 over bias, and so should probably have wound back to 5 mins. But I was wary of mixing subs of different lengths in the final stack. That said, the mixing didn't work out well as it was!
Thanks Glen. I was going by Ray's target ADU above bias from those tables. On the moonless night, even at 10 mins with gain at 202, I was only getting about 500-600 ADU above bias - i.e. less than the target of 650. So not quite shot noise limited. Seems surprisingly longer given Ray's estimates of the expected times, but I guess there's things like the large central obstruction in the RC scope which will add a bit of extra time.
On the second night, I was about 1100 over bias, and so should probably have wound back to 5 mins. But I was wary of mixing subs of different lengths in the final stack. That said, the mixing didn't work out well as it was!
Oh the RC explains it i guess. I am used to doing everything at f5. I can't offer advice on the pattern, never seen it before.
After quite a lot of reading about the best use of calibration for this camera it appears it’s best to skip the Bias altogether. Just do a master dark subtraction on the lights and do a flat dark (dark with the same duration as the flats).
I have in the past toyed with the super bias but found better results from just adding more to the master bias anyway. This was when I was using my QHY22 and not bothering with darks at all.
After quite a lot of reading about the best use of calibration for this camera it appears it’s best to skip the Bias altogether. Just do a master dark subtraction on the lights and do a flat dark (dark with the same duration as the flats).
I have in the past toyed with the super bias but found better results from just adding more to the master bias anyway. This was when I was using my QHY22 and not bothering with darks at all.
[QUOTE=DiscoDuck;1339086]Thanks Greg. Will give that a go. I also used PI's Superbias. I may then also go back to a straight master bias and see if that helps.
Intersting. I have no experience with their superBias but I am under the understanding it will only give a gain if the ADU of your bias varies from bias to bias. In other words you get bias drift.
I would be interested to know if the straight bias works and the SuperBias does not.
After quite a lot of reading about the best use of calibration for this camera it appears it’s best to skip the Bias altogether. Just do a master dark subtraction on the lights and do a flat dark (dark with the same duration as the flats).
I have in the past toyed with the super bias but found better results from just adding more to the master bias anyway. This was when I was using my QHY22 and not bothering with darks at all.
Could you explain this Colin. I've seen it mentioned elsewhere (I think it was John Rista on CN?) that its best not to use bias frames for these CMOS cameras.
Unlike some cameras (thinking some of the Sony sensors) you MUST use darks with a good number of the CMOS sensors. So using a bias on your lights becomes redundant as the dark contains the bias. Subtracting the bias from the master dark is only useful if you want to use dark scaling which also doesn’t work with the CMOS sensors as the amp glow and banding issues don’t appear to increase linearly.
There is also an arguement to be made that these glows start with relatively short exposure, maybe as short as 1s. So with this in mind it may be better to create dark flats as opposed to using a bias which also shouldn’t be shorter than 0.3s.
The QHY manual mentions not going below 0.3s as it starts creating some banding issues in the master bias of which I’ve noticed myself. So when I’ve mapped my flat times I’ve used the Lum filter at 0.5s and referenced the others from there.
Doesn't the QHY version have a DDR? The original versions of the ASi1600 do not have a DDR. Is this a factor to be considered when comparing the bias and dark performance? Right side AMP Glow certainly is a factor in the ASI1600 subs at long (>=300S) narrowband lengths. If your shooting mostly short (<60s) sub broadband then it is not a concern. I agree it's not linear in my experience. Shooting Dark Flats, seems to me, to be an un-necessary overhead, when i already have a great library of Bias and Darks. I have read some of Jon Rista's commentary. I have seen no evidence of banding in the Master Bias of my 1600MM-C.
The DDR buffer is mostly for anyone using a USB2.0 connection so that you can download into the buffer really quickly (reducing some noise) and then download from the buffer at whatever speed your USB connection will handle. With a faster computer (one that can handle higher USB bandwidth) and USB3 I don't think the DDR buffer makes any difference as you can download direct from camera as fast as from the buffer.
I don't remember seeing banding in my ASI1600 but I do know that the QHY163 had it with exposures shorter than 0.3s, that may have changed since some of the early drivers but as I haven't used a bias with them since those early drivers I'm not sure. And when you're looking at a small background difference in 30,000 ADU (referring to flats) it is only very minor. It may not be an issue with medium sized stacks (50-200) but above that and you may start running into some non-uniform background gradients that are difficult to remove. The larger the stack size the closer to perfect all of your calibration frames have to be before you limit your SNR by calibration noise.
Paul, your running that 1600 at high gain, and shooting 10 minute subs. Normal practice with the 1600 at high gain is to reduce sub times. You get less well depth at high gain and shooting for longer doesn't make it any deeper. Have a look at Shiraz's narrowband exposure chart, which he has posted here. I shoot most narrowband at Unity (Gain 139, Offset 20) and that is with 5 min subs max,i would never shoot longer subs with that camera, even at max well depth of Gain 76. When i shoot high gain (200), i drop my sub times down to 90 sec.
Yeah, Jon [Rista] over at CN has argued pretty much the same - keep subs short, and lots of them, and don't go ballistic with the bias/offset. I personally think you're bias/offset settings are too high and you would do better to drop them.
Yeah, Jon [Rista] over at CN has argued pretty much the same - keep subs short, and lots of them, and don't go ballistic with the bias/offset. I personally think you're bias/offset settings are too high and you would do better to drop them.
I haven't bought my ZWO 1600 - yet. I'm gonna start with 100/21 and see how things go when I do get it. A friend on YouTube shoots @ 139/21 from memory and gets great results too. From my understanding, you get no real gain from bumping up the gain (no pun intended!) - it's far better to preserve SNR.
I'm very new to imaging, so take what I've said with a grain of salt please! I've been doing so much reading on it all, and noting what others say/do, to better help me adapt when I get the camera. :-) ymmw.