6 replies to this topic

[grsotnas]

Hello everyone,

 

I ask the help of this awesome community aind its experts, from which I've learned so much, on some issues I’m having regarding my Canon 80D’s astrophotos. When strongly stretching the data for faint details, I ran into severe color banding, which ultimately limits how far I can process the image for faint stuff (before other sources of background noise kick in).

 

First, some background: I currently use 2 DSLRs. A Halpha modded T5 and the stock 80D. The 80D was purchased last year, and I’ve been using it ever since. Only recently however I had the chance to deeply integrate with the 80D. The T5 (being modded) became primary camera for most targets this year, and I’ve been using it almost exclusively with my 135mm.

 

I use AstroPixelProcessor for pre-processing and currently do not own PixInsight - I especially consider this software for its analytical capabilities, which could help track down the problems and try to solve them, not to mention its extensive processing power. I’m getting a bit carried away here - let’s get back on topic =D.

 

My calibration workflow is based on about 20 Darks, 30-50Bias and ~20 Flats. I also use a Bad Pixel Map. All calibration is done in APP, and has always worked wonderfully [especially with my extensive T5 use]

 

So, as for the banding story: in July I captured some data on Barnard’s Galaxy. I captured some 108 minutes integration simultaneously with the T5/135 and 80D/200mm. The subs were all 240s ISO400, dithered every 2 subs. Ambient temperature was ~+5C, and sensor temperature ~26C for the T5 and 18C for 80D. Both were calibrated (10darks, 25flats, 30bias) and integrated in APP. The 135/2.4 is great for capturing photons quickly, and I got the result below after a quick APP strong stretch - overall OK-great.

 

  

 

The 200mm data was integrated separately. When extreme stretching is applied, I got the bands visible below.

 

 

That’s when the issue first became apparent, and prevented me to merge both sets of simultaneous data (which would effectively double my integration time).

 

I made some investigation myself, which is presented below, but first I straight-forwardly state the points I ask your help with:

 

1 - Identify the issue. What is this banding? Is it normal or expected? Please share your experience if you have had something similar! I am almost certain it is some fixed pattern noise (or is it the usual “Canon banding” I always hear about?).

2 - Potentially find solutions to mitigate or solve my issue - where should I take more care to get rid of it.

3 - In case this issue becomes impossible to cope with, should I consider switching cameras? My 80D is a great workhorse for daylight work, but be the banding impossible to solve, I may look for other alternatives for astro work. I ask that because my trusty cheap T5 has never exhibited the issue, even when extreme stretching.

 

 

MY INVESTIGATION THUS FAR

 

Once again, I’ve never seen such issue with my T5 data, despite the price and (supposedly)
performance gap. My T5 is a trusty performer, despite its older tech -  80D’s sensor is a newer generation low-read-noise, ISOless sensor. I have used my 80D for images before, especially with the Chamaeleon Molecular Cloud [https://astrob.in/345170/0/]. But as this image was shot unguided and with short exposures (30s vs 240s), and with over a half frame differential due to tracking drift and field rotation: a huge offset. I figure this would be a apples to oranges comparison) . So, here’s what I came up with:

 

My hypothesis: this is fixed pattern noise, uncorrected because of dark frame issues. I checked a single light, and the pattern is faintly there. Below is a single 240s sub overstretched, uncalibrated.

 

 

I checked my dark frames. In the master dark, the pattern is also there (below is a overstreched, oversaturated master dark).

 

 

I believe this could be a result of dark frame temperature mismatch. Indeed, the dark frame averaged some 5C above the lights. That’s just too much. So I took a new set of darks in my fridge, that much better matched the light temperatures - within 1-2C on the sensor. That is as good as I can reasonably get with a camera without set-point cooling. The result was not much improved. The banding is still here.

 

I started to look for the banding in past images with the 80D, and I found it is present - with various degrees. It is visible in my M31 and SMC cloud images of October 2017. The ambient temperature that session was higher - 14 C (sensor 23C). On my B228 image from July 2018 (the same session of the Barnards Galaxy), the issue is somewhat reduced, but faintly visible [image here: https://astrob.in/357657/0/]. The B228 lights are 120s (vs 240s). On a recent Magellanic Clouds image, with a different 40mm lens, the issue, however, was not apparent. The temperature of this set was a freezing 1C ambient (10C sensor).

 

I can provide the calibration masters (or even the subframes) and a set of any aforementioned light frames if someone is willing to have a look at my raw data.

 

So, I ask the help of you experts to figure out this intriguing and frustrating problem. Could it be a calibration issue (mismatched Darks, or too few Bias)? Is it a result of the 80D noise performance - be it such a (supposedly) less noisy camera (than my T5), is the issue caused by the even more extreme stretch that this low-noise is capable of handling? I’ve been a happy 80D user, especially for daylight work (even recommended the camera to other people), but after this issues, I am discouraged and reconsidering this DSLR for astro work. Please help me if possible!

 

If you have read this extensive topic thus far - you already have the sincerest thank you from me!

 

Clear skies to all

Best regards

[sharkmelley]

Did you see my review of the Canon 200D?

https://www.cloudyni...view/?p=8903230

 

The Canon 200D’s sensor is also the newer generation sensor.  But no-one should make the mistake of calling it ISO-less.  The 200D showed similar banding and the position of the bands changed from exposure to exposure so it cannot be calibrated out.  It's disappointing if the 80D is behaving the same way.

 

What ISO were you using and where was the skyfog peak on the back of camera histogram?

 

It's possible that this problem can be mitigated by using an even higher ISO e.g. 1600 or 3200 and ensuring the back of camera histogram peak (due to sky fog) is 1/3 from the left hand side in an attempt to make the signal swamp the banding. 

 

Mark

[balu01]

I am shooting with my 80D (stock) as well , I never experienced such issue with the camera not in lights or dark frames. In fact I am super pleased with this camera, so far it seems it performs better than my 5D3 actually.
So other than telling you to get an other one or return if under warranty I cannot be of much help.
Rather than trying to process it out I’d take the route I am suggesting if I was in your place. That seems like constant headache if you are sure you did everything right.

Edited by balu01, 28 October 2018 - 02:11 PM.

[grsotnas]

Guys, I thank you very much for hearing my story and answering my questions.

 

I apologize for not responding yet. I was away and quite busy to write a proper response. I'm really sorry for that - and in my own topic! I hope you can understand.

 

 

Did you see my review of the Canon 200D?

https://www.cloudyni...view/?p=8903230

 

The Canon 200D’s sensor is also the newer generation sensor.  But no-one should make the mistake of calling it ISO-less.  The 200D showed similar banding and the position of the bands changed from exposure to exposure so it cannot be calibrated out.  It's disappointing if the 80D is behaving the same way.

 

What ISO were you using and where was the skyfog peak on the back of camera histogram?

 

It's possible that this problem can be mitigated by using an even higher ISO e.g. 1600 or 3200 and ensuring the back of camera histogram peak (due to sky fog) is 1/3 from the left hand side in an attempt to make the signal swamp the banding. 

 

Mark

Mark, thank you very much for the response! I've read your SL2 topic, and it seems the issue is at least similar, more likely identical. However, I'm not sure about the bands being time-dependent. Given my preliminary analysis they appear to be somewhat fixed, but not properly calibrated out. Anyhow, I also find very disappointing this kind of color banding is arising in the newer Canons.

 

I called 80D isoless as 80D's 24MP sensor is somewhat of a newer generation than earlier Canons (18MP sensor). Data showed it performed better than the previous generation, with higher DR, lower read noise and a flatter read noise curve. It is still lagging behind Nikon's D5300+, though. However 80D is still not ISOless - Thank's for pointing my mistake! I learnt a new thing!

 

The lights were ISO 400 240s (at f/3.5 200mm). The BoC histogram was a bit on the dark side. I thought this would be OK (since the camera has little RN, and by shooting with a lower ISO (I usually shoot 800 with my T5) I could benefit from higher DR). Below is a screenshot from Lightroom Histogram [therefore the streched RAW histogram, not the linear histogram]. I also added a T5 histogram from that same night, shot simultaneously (ISO 400 240s 135mm but f/2.4)

 

 

Would it be OK if I uploaded some FITS Masters files so that people can better see the data per se? Later I will upload some frames to Google Drive and share them here.

 

Mark, our analytical ability is something I always learn from here on this forum! Thank you for that (not only on this topic, but many things you test, such as the recent star color/AA filtering)! 

 

 

 

I am shooting with my 80D (stock) as well , I never experienced such issue with the camera not in lights or dark frames. In fact I am super pleased with this camera, so far it seems it performs better than my 5D3 actually.
So other than telling you to get an other one or return if under warranty I cannot be of much help.
Rather than trying to process it out I’d take the route I am suggesting if I was in your place. That seems like constant headache if you are sure you did everything right.

balu01 thank you for answering.

 

I am also very pleased with my 80D as far as daylight performance is concerned. And for night time use I was very happy as well! However, this banding intrigued me. It might be I'm shooting with less than perfect settings - perhaps too dark... It might be something about my processing - to really stretch for the faint data....

 

What settings do you usually use? Could you share some frames with your 80D? It would be very interesting if this issue is present on my, but absent from your camera - if the banding is camera-body-specific, or is camera-to-camera variable. My reason tells me this should be unlikely, though - but we must see the actual real world data before any conclusions.

 

Clear skies to all,

Gabriel Rodrigues Santos

[cosmo59]

 I am not sure but I think Gary Honis reviewed the Canon 80D a while back and nixed it for astro after testing several different copies of the camera. If I remember correctly banding was the issue.

 

Mark

 

PS the Witch Head image is beautiful.

Edited by cosmo59, 05 November 2018 - 12:22 AM.

[sharkmelley]

 I am not sure but I think Gary Honis reviewed the Canon 80D a while back and nixed it for astro after testing several different copies of the camera. If I remember correctly banding was the issue.

 

Mark

 

PS the Witch Head image is beautiful.

 

I don't think banding was the issue Gary found on he 80D - just high levels of dark noise:

https://www.cloudyni...t-of-canon-80d/

 

The Canon 200D I tested had banding in addition to high levels of dark noise.

 

Mark

[james7ca]

I think you need to blink your calibration frames and see if this "banding" is consistent or whether it varies between frames (I suspect the latter). I'm not sure what tools support a screen transfer stretched blink (kind of like watching an animation or video playback of all of your frames), but as you might expect PixInsight has a very nice blink tool (under Process -> ImageInspection).

 

Perhaps you should download PixInsight and use their 30-day (45-day?) free trial.

 

Other than that, you might want to omit the bias files and make certain that the master dark isn't applied with any form of optimizations or scaling (you can't do the latter without bias files).

 

In any case, "banding" like this isn't uncommon with one-shot-color CMOS sensors, it really comes down to a question of how bad and whether you can detect it given your conditions and processing.

Edited by james7ca, 05 November 2018 - 06:12 AM.