Removing Banding

[georg.viehoever]

Hello,

I am slowly working through the pictures I took this summer with a Canon 40D attached to my Newton. A problem that I see with most pictures is a banding structure that also does not disappear when stacking with flats and darks. Certain horizontal rows appear to be different. While the general area is similar on all pictures, the intensity and exact location seem to move from picture to picture. Since the problem appears on pictures taken between 3 and 30 degree Celsius, it is probably not temperature related.

An example (a raw converted to jpg and histogram-pushed) is


http://cid-c56e60845cfa6790.skydrive.live.com/self.aspx/Pixinsight/ngc7000_banding.jpg (4.2 MBytes)

I wonder what the cause for this problem is. Am I pushing the raw pictures too much? Can the problem be corrected using Pixinsight?

Any hints are welcome,

Georg

[Simon Hicks]

Hi Georg,

I use a Canon 400D and get the same lines if I push the stretching too far. I have been told on other forums that it is due to the readout noise (or bias) and that stacking say 50 or 100 bias frames (rather than say 10) to create a master bias frame has helped others get rid of it. Bias frames are quick and easy to create, so it might be worth a try.

However, if the PixInsight guys have a good way of processing it out of the image then I would be really interested as well!

Cheers
              Simon

[Philippe B.]

Hi Georg
I have a 40D moded with baader filter and I have the same problem like you.

At 320 iso, it is good (but could be seen when extrem stretching)
At 400 iso, problem is here and can be a problem
At 800 iso, it is a disaster

I made a composition of 60 dark and 60 offset. But if only a small amount of images, master dark don't really save the situation. More you have image more the quality will be better. If you stretch near the background, work at lower iso.

So, now, I limit at 320 iso. Maybe less...
But this week-end, I worked at 400 iso   :roll:

philippe

[David Serrano]

A quick google search for <<40d banding>> yields a promising first result: http://forums.dpreview.com/forums/read.asp?forum=1019&message=26357028 . The third reply in the thread mentions a "CMOS adjustment" that can be carried at your nearest Canon service centre.

[georg.viehoever]

Thanks for all the hints. The "banding problem" indeed seems to be widespread with Canon EOS cameras, as a Google search reveals. Unfortunately, there seems to be no consensus on how to fix it (and Canon usually claims that the camera is working within specs, even if it shows banding). Also, I would hate to reduce ISO to something like 320, when the general recommendation seems to be to use IS800-1600 for best efficiency.

I tried if using more bias frames would help. I used 120 instead of my usual 20, and the problem basically remains the same. The banding is visible in the bias frames themselves just as in the dark and light frames, but in the master bias frame (produced from 20-120 individual bias frames) it almost (but not completely) vanishes.

Some Pixinsight procedure to at least reduce the problem would be very valuable. It may be possible to do that because the banding produces fairly specific artifacts, not just random noise. I think something that adjusts the current row based on the average values of the neighboring rows may help. But I have no idea how to do that in detail.

Georg

[Philippe B.]

I'm not sure best is 800-1600 iso because you loose dynamic range.
For me, the best is near 400 iso. 320 iso is very good compromise between sensitivity and dynamic range.

Now I'm thinking to buy early next year a CCD camera. I was looking for a moded 5D MkII but in this price range and little more, I can access to cooled monochrome CCD with filter wheel ( with KAF8300 device) which can be ideal for my astrophysics 140EDFS.
 
best,
Philippe

[georg.viehoever]

There is a discussion regarding this topic in a Yahoo group: http://tech.groups.yahoo.com/group/digital_astro/message/120252 . Maybe they find a way...

In the meantime: Sending my Canon to the Canon Service did not help. They exchanged the card slot, but that did not change the (non-)quality of the pictures,

Georg

[David Serrano]

Here's a JPEG shot (not raw) taken with my shiny new 40D. 30 seconds, ISO 800:



You can see some banding, although not a "disaster" as CCD1024 said earlier. The firmware version of my camera is 1.0.8.

[Philippe B.]

Before my 40D has been moded, 800-1600 iso were nice !
It is during the filter removal where you must desoldering some parts, there should be something "wrong". Electronic noise seems to be very delicate to set then if you touch some parts you can make a little shift in voltage or shielding so noise appears.

I will try to find a JPEG of the banding I got. I don't have my 40D anymore because I have ordered a CCD camera.  During delivery time, I will use my Nikon D3 which is very promising...

I see I am not alone to have hudge banding on 40D. I won't say "I am glad to see" but "desappointed to see" that problem.

Work at 320 iso maximum. It is not so bad. :  M42 ,  and   Veil nebula

At 800 iso, after processing, banding is HERE  :  M81-M82

[georg.viehoever]

Hi,

if you are plagued by this problem: I have created a script that appears to reduce the problem. See http://pixinsight.com/forum/index.php?topic=1159.0 for more.

Georg