Visible bands during processing?

[herne]

Hi all,

I recently bought a new (used) DSLR (Canon 750D), supposedly in "as new" condition (according to the seller's website).  Took it out for a little test run a couple of nights ago and did some basic runs on M13 and M81/2.  For context, as well as 50 or so Light frames taken on each subject, I also took 30 or so Dark, Bias and Flat frames.

During processing (Photoshop) I noticed horizontal lines or bands appearing across each image as I started stretching, particularly in the M81/2 image (see below).  I managed to process the bands out in the M13 image (just) but the image below is particularly bad and I stopped processing it half way through.  I've never noticed anything like this before and am thinking I'll need to return the camera as in my opinion the sensor is faulty.  Before I do I thought I'd ask opinions here just in case I'm missing something.

If it helps I use a William Optics Z61 ii + WO field flattener.  No filters in this particular set and the 750D is currently unmodified.

Edited July 24, 2021 by herne

[Budgie1]

Banding on Canon DSLR's is quite common, so much so that PixInsight has a scripts designed to remove it.

If you do a search of the forum there are a number of threads about it.

Edit: HERE's one which talks about the EOS 750D.

Edited July 24, 2021 by Budgie1

[herne]

2 hours ago, Budgie1 said:

Banding on Canon DSLR's is quite common, so much so that PixInsight has a scripts designed to remove it.

If you do a search of the forum there are a number of threads about it.

Edit: HERE's one which talks about the EOS 750D.

Well darn.

Thanks for the tip.  On doing further digging I may be trying to stretch the data too far.  I could return the camera and get something else, however the issue could still be apparent in another DSLR and I'd be back to square one.  The PixInsight script could be a solution and it seems there is also a PS plug in that could help (Astronomy Tools Action Set).

On the image of M13 I pretty much processed out the banding through gradient removal when filtering out some light pollution issues.  The M81/2 image I showed above had worse light pollution and the stretch was heavier.  I'm using the Arcsin stretch plug in for PS and I may have been a little over zealous with it.  The 750D is a decent camera and I don't really want to return it, so further experimentation is needed.

Thanks for the reply 🙂.

Edited July 24, 2021 by herne

[Clarkey]

As stated above, banding with the Canon DSLR's is common. Certainly mine has that problem. The only thing I would say looking at this image is the banding looks a bit 'different'. Whether it is a fault I'm not sure - the regularity almost looks like it might be electrical noise. Do you download the images or are they saved on the camera? Also, how is the camera powered - battery or dummy battery?

[alacant]

4 hours ago, herne said:

I'll need to return the camera

Hi

The 18mp eos' along with the more recent 24mp versions are pretty much band-free.

HTH

[herne]

3 minutes ago, Clarkey said:

As stated above, banding with the Canon DSLR's is common. Certainly mine has that problem. The only thing I would say looking at this image is the banding looks a bit 'different'. Whether it is a fault I'm not sure - the regularity almost looks like it might be electrical noise. Do you download the images or are they saved on the camera? Also, how is the camera powered - battery or dummy battery?

Images were saved to camera and transferred to PC via USB after the session (next day).  The camera was powered using a dummy battery plugged into the mains.  Digging around now I see electrical interference may cause an issue, will need to test using the battery.

5 minutes ago, alacant said:

Hi

The 18mp eos' along with the more recent 24mp versions are pretty much band-free.

HTH

The 750D is 24mp, I guess you refer to the more recent EOS line up (850D / 90D)?  Kinda thinking I should have gone with my original choice of a 60D, but this "as new" 750D came up, so I had a last minute change of mind.

[alacant]

11 minutes ago, herne said:

more recent EOS line up

We tested a 2000d but prefer the 18mp sensors which seem cleaner and more sensitive. The 60d may however be going just a bit too far back.

Unless you need the bells and whistles for daytime shots, just go with the most basic.

Cheers

Edited July 24, 2021 by alacant

[herne]

3 minutes ago, alacant said:

We tested a 2000d but prefer the  18mp sensors which seem cleaner and more sensitive.

Unless you need the bells and whistles for daytime shots, just go with the most basic.

Cheers

A flip screen rather than a cricked neck is the only thing I really look for.

[alacant]

6 minutes ago, herne said:

A flip screen

Much easier, clearer and practical viewing/focusing on your 'phone.

 

Edited July 24, 2021 by alacant

[herne]

1 minute ago, alacant said:

Much easier, clearer and practical viewing and manipulating the camera via your 'phone.

 

You mean via wifi and the Canon Connect app?  (Sorry I've come from a very old Sony to Canon, so trying to catch up with the 21st century lol).

[wxsatuser]

Never had an issue with banding on both my 60Da, 7dmk2 or 6d, noise yes but no banding.

[michael8554]

There seem to be two types of banding on Canon DSLRs.

The first is most noticeable during hot weather, and isn't like the OP's image, it's in many more regularly spaced horizontal bands.

That was my experience with a 600D, it only showed up on hot summer evenings.

The fewer bands in the image may be due to the second theory :

Newer Canon DSLRs (700D onwards ?) have many more autofocus sensors, now set in horizontal bands buried in the sensor.

Not noticeable on daytime snaps, but very visible in stretched long exposure astro images.

Michael

Edited July 25, 2021 by michael8554

[wxsatuser]

Can I just add, although your sort of correct there are two autofocus systems in Canon cameras.

The sensor autofocus only works in Liveview mode or in mirrorless cameras, I believe the whole sensor is used in dual pixel AF.
In view finder mode there is a dedicated autofocus chip, I would assume that one wouldn't use Liveview in AP.

[alacant]

10 hours ago, herne said:

via wifi and the Canon Connect app

Not necessarily. You could use an OTG cable. Even better, control your astro sessions with a dedicated image capture program on a laptop or NUC.

 

[alacant]

15 minutes ago, michael8554 said:

700D onwards

750d onwards. 850d and newer are fine.

[herne]

1 hour ago, alacant said:

750d onwards. 850d and newer are fine.

I seem to have landed just in the wrong spot lol.

Appreciate everyone's responses.  I think that as I'm still within the time I'll simply return the camera.  I could keep it and mess around in processing but simpler to get rid of the issue in the first place as it won't cost me anything (except postage).

If the 18mp sensors seem the better option then something along the lines of a Canon 1300D seems fair.  Unless something from Nikon might be a better option as I'm not tied to any manufacturer?

[Budgie1]

Nikon are good cameras but for Astrophotography then I would stick to Canon as there seems to be more upgrades/kit/software available for them.

I have an astro-modified EOS 1300D that I was playing with last night and I've just processed this of NGC7000 from 220 x 30s exposures at ISO400 using the Samyang 135mm F2 lens on a Skywatcher EQ5 mount, unguided.

There is still a little bit of banding, I can see a couple of bands at the top-right of this image now that I look closely at it, but it's not that bad and can be processed out.  

 

[wxsatuser]

If your doing short exposures,  longer exposures may help, expose so the camera histogram peak is at least 25% from the left.

[herne]

1 hour ago, Budgie1 said:

Nikon are good cameras but for Astrophotography then I would stick to Canon as there seems to be more upgrades/kit/software available for them.

I have an astro-modified EOS 1300D that I was playing with last night and I've just processed this of NGC7000 from 220 x 30s exposures at ISO400 using the Samyang 135mm F2 lens on a Skywatcher EQ5 mount, unguided.

There is still a little bit of banding, I can see a couple of bands at the top-right of this image now that I look closely at it, but it's not that bad and can be processed out.  

May I ask what your processing workflow was (a brief overview)?  I know you mentioned Pixinsight above so I assume that was involved?

For my image I used DSS + PS along with the Arcsin plug in which may be simply too much for the data captured and light pollution issues I had.  I don't really want to give up so soon on the 750D as it is a nice camera but will do if need be.  As I mentioned above, the 1300D may be a good alternative and at c.£200 from MPB is almost half the price paid for the 750D (unless you'd consider putting your modified 1300D back on the market of course 😉).

[Budgie1]

36 minutes ago, herne said:

May I ask what your processing workflow was (a brief overview)?  I know you mentioned Pixinsight above so I assume that was involved?

For my image I used DSS + PS along with the Arcsin plug in which may be simply too much for the data captured and light pollution issues I had.  I don't really want to give up so soon on the 750D as it is a nice camera but will do if need be.  As I mentioned above, the 1300D may be a good alternative and at c.£200 from MPB is almost half the price paid for the 750D (unless you'd consider putting your modified 1300D back on the market of course 😉).

I stacked in DSS, saved a FITS file and then did all the processing in PI, so all the tools a PI based.

The workflow on this image was a little different to what I normally use as I was experimenting a little. So, I did an Autostretch (which doesn't actually stretch the image, it's just you can see the detail) then cropped the image to remove the stacking artefacts.  Then PhotoColourCalibration so I could see any vignetting and ran AutoBackgroundExtraction to flatten the background. This worked really well on this image and normally I would do the background extraction before the colour calibration. There is no noise reduction on this image and I didn't think it needed it.

Next I Binned the image x2 (I didn't need it full size and it makes the PC run faster ) and did the stretch. After that was Starnet to separate the stars from the nebula so I could work on them separately, starting with reducing the saturation on the stars using Curves to remove the red & blue hue from them, also the lightness to bring them down a bit. 

Now on to the nebula, using Curves, I used saturation to bring out the colour to where I wanted it and tinkered with the RGB & Lightness (CIE-L) curves to get what I was happy with. I didn't use any sharpening or HDRMultiScaleTransformation because these just made the image blotchy. 

Finally, the stars & background were recombined using PixelMath and resized.

If you want to post the DSS stack of your image (unprocessed) then I can run it though PI and see what I can do.  

I may put the EOS 1300D back on the market in a month or so but it's the only thing I currently have that I can use with the Samyang 135mm lens. When the mounting collar for the my ASI294MC comes back in stock with FLO then the 1300D will probably go up for sale again.  

[herne]

6 minutes ago, Budgie1 said:

If you want to post the DSS stack of your image (unprocessed) then I can run it though PI and see what I can do.

Appreciate the offer but I won't waste your time.  It was a first light experiment with a new camera, the full moon was up and I was shooting over the local town without a filter.  Frankly I'm amazed I captured anything.

I'll post below the M13 final processed image which I captured on the same night.  Here I processed out the banding via gradients in PS and I had better data to work with than the M81/2 images.  Better data will always have an impact.  I probably will return the camera even though banding issues may still occur in any replacement.  I could go for one of the latest Canon D90 or D850 but I don't see the extra value in that when a used D1300 or D2000 (or even a D60 which I know gets a lot of love even if it is 11 years old now) would give me pretty much the exact same results for much cheaper.

With hindsight I think the D750 was an unlucky choice which, as it came with a 14 day return policy, can be rectified.

[Budgie1]

Nothing wrong with that!

If you decide to send the 750D back then maybe have a look at Cheap Astrophotography if you fancy a modified DSLR. That's were mine came from and it was here in about 5 days from ordering.

[alacant]

1 hour ago, herne said:

I don't see the extra value

You don't need most of the facility a DSLR offers. No autofocus, exposure meter, flash, the myriad of useless stuff on the PASM dial...

So long as B works and the shutter opens; a faulty DSLR would be fine. 

[herne]

Thought I'd make an update after playing around more with stacking (DSS) and processing in PS.  Attached image below, compare to the first image in the OP.  Data from the lights is still poor (excuses in early) but I'm more concerned with the banding at present anyway rather than the image itself (which is a bit pants).

Anyway, this is what I changed:

- The number of Bias frames when stacking in DSS from 30 to 110.  Probably overkill but all part of the experiment

- Just did some basic stretching, adjusting the shadows and midtones using Levels in PS at first.  No Arcsin at this point

- Masked out the stars and galaxies, subtracted the background from the original to reduce noise and light pollution, and some of the residual banding

- Stretched a little more using some gentle Arcsin curves and Levels

- Mucked about with Camera Raw Filter

Some faint banding issues remain and further processing could perhaps eliminate them, but overall much better than in the original.  I think the 2 largest impacts were using additional Bias frames and going easy on the initial stretching.  I still wonder if using a dummy battery plugged into the mains may cause some issues, but clouds have rolled in so can't test that yet.

Edited July 25, 2021 by herne

[Budgie1]

That's looking much better.

I've just noticed that you mention in the OP that you're using Darks & Bias calibration frames with Flats. As part of your "experiment", the next time you get a chance to image, shoot Darks, Bias, Flats & Dark-Flats (same exposure settings as the Flats but with the lens cap on, I put it into Tv mode to take these). Then try a stack using all of them and a second stack without the Darks and see if there's an difference, good, bad or none.

I read that for DSLR's, Darks could be replaced with Bias & Dark-Flats because the Darks are temperature related and, unless you have a cooled DSLR, you can't get a set temperature for the session. The ambient air temp will be higher at the start than at the end and sensor temp will be lower at the start than at the end (that's my understanding but I'm more than happy for those with more knowledge to correct me )

I tried it with the 1300D and I didn't notice any difference in the stacked images, so I stopped taking Darks with the DSLR. It means I can use all the time to take more Subs, instead of stopping the session early to take Darks.

I still have to take Darks for my ZWO OSC camera but the set-point cooling on that means I can build a library of darks for the different exposures & temperatures and just reuse the Master files until something changes.