PixInsight Forum

PixInsight => Image Processing Challenges => Topic started by: Greg Schwimer on 2016 January 01 11:40:44

Title: A7S banding/frame split help
Post by: Greg Schwimer on 2016 January 01 11:40:44
I have a Sony A7S, modified for full spectrum. I am running it with a HEUIB-II filter on a GSO 8" RC. A small handful have shown a problem with banding within the integrated images from this camera, as well as a "split" between the left and right side of the frame. I was previously only running short (30-90s) exposures, but now that I'm going longer (300s+) I am seeing this as well.

Any ideas what can be done to process this out? I've tried the CanonBandingReduction script and it sort of works, but it leaves other artifacts. I will continue to experiment with it...

This is 24x450s lights, 24x450 darks, master bias and 20 flats.

Here is the unprocessed integration:

Title: Re: A7S banding/frame split help
Post by: Greg Schwimer on 2016 January 09 15:30:18
I've been doing some experimenting to see if the optics were contributing. Best I can tell it is the camera. Here is a set of "flats" that I captured with no optics attached - just the camera body wide open pointed at a relatively flatly lit wall. I've found that splitting the integration into the RGB  channels, and applying an ABE helps make the result easier to see. It's mostly strong in the red, and blue is the quietest.

There is a thread over on Cloudy Nights where we're trying to figure this out. If anyone has any ideas as to how we can kill this in processing it would be very much appreciated.


Title: Re: A7S banding/frame split help
Post by: msmythers on 2016 January 09 18:20:00
I think more information is needed, a lot more. First what exact camera settings were you using for all files used to create this image. Also were you using any built in camera noise reduction. Second a sampling of the raw flats, darks, bias and lights. Your integrated image has numerous artifacts that could be caused by optics, camera and or poor integration.

Here is a section of your original image enlarged 3x. It is a section of the sky between the 2 galaxies. It's just full off all types of issues that would normally be taken care of with good calibration and proper integration. 

Title: Re: A7S banding/frame split help
Post by: Greg Schwimer on 2016 January 09 22:22:08

Thanks for the reply. Yes, there certainly is room for improvement in the image. I'm still dialing this set up in and it's improving rapidly. The real concern is the "banding" and "split" in the images. I did some searching around and see at least a handful of others having a similar issue specifically with this camera.

Camera details:
Sony A7S full spectrum mod
No internal NR
OTA: 8" RC
Other Optics: Innovations Foresight ONAG, HEIUB-II filter, TSRCFlat2 flattener

What changed recently is 1) the addition of the HEUIB-II filter, 2) the ONAG, and 3) the exposure length. Since I have been able to see the issue in both my flats and my lights, I decided to run a series of test flats today. This is a set of them:


These are flats of 1/250s at iso3200 with the camera directly attached to the ONAG and the HEUIB-II removed just to try to rule it out. The flat image I presented above in this thread is with no optics attached to the sensor. That is, I pointed the camera body with no lens toward a relatlively flat light source and captured. Again, what I was after was removal of the OTA and optics to verify the problem continues to exist, and it seems to.

Once I have the flats, I use ImageIntegration to integrate them as shown below. I then take that integration and run an ABE on it. I do this because this has been when the "noise" begins to show in the lights. From there I stretch the image by transferring the STF settings to HT, and adjusting. I also have been splitting the RGB channels out to try to determine if the noise is biased to a single channel. It seems to be more prominent in the red channel.

So, I'm coming to accept that this may be an idiosyncrasy of the camera. That said, I need to begin figuring out how I can take this repeatable pattern and calibrate it out of the final integration. Interestingly, in the Cloudy Nights thread, the OP states he too cannot get this "noise" to calibrate out.

Title: Re: A7S banding/frame split help
Post by: Greg Schwimer on 2016 January 10 14:51:36
Here is a set of RAW files from last night/this morning. I backed down the exposure time from 450s to 120s. It helps somewhat but it's still there. The flats still show it.


Title: Re: A7S banding/frame split help
Post by: Greg Schwimer on 2016 January 10 14:54:12
An experiment was done by the OP on Cloudy Nights. He varied the exposure time and made an animation of the banding change. The problem appears to be dynamic in nature:


Title: Re: A7S banding/frame split help
Post by: rgbtxus on 2016 March 22 10:40:14
Sorry, I was just quickly passing by and have not thought about this but you might want to read this http://diglloyd.com/blog/2014/20140212_2-SonyA7-RawDigger-posterization.html  Sony has been using a very poor way of storing RAW data which has been shown to generate visible artifacts in normal photography.  I am not sure if this contributes to your issue, but I thought I would point it out.  Now, I believe Sony has recently released firmware (not sure if it applies to the a7s) that allows you to avoid this problem by saving uncompressed RAW files.  Why they did not do lossless compressed is a mystery.  As I said, this may not be relevant to your problem, but I thought I'd pass it along.
Title: Re: A7S banding/frame split help
Post by: Greg Schwimer on 2016 March 22 11:24:31
Thanks for the reply and the info. I've installed the new firmware on my A7S but have yet to test it. As far as I can tell from the configuration menus, there is no uncompressed RAW option.

We did find that the issue I mentioned above can be remedied by pushing flat exposures beyond 50% of the back of camera histogram. Since doing this I haven't seen the problem.
Title: Re: A7S banding/frame split help
Post by: rgbtxus on 2016 March 22 13:57:13
Great.  Glad to hear it.  Yes, there is no lossless compressed RAW so you take a big transfer and storage hit choosing uncompressed over 11+7.  I have considered getting an a7s for AP (I currently use a Nikon D800 with the IR filter replaced and a QHY23 for mono work), but frankly could not get past living with 11+7 (which I could now do at a transfer time cost) and the fact that since there is not (I assume) an SDK so none of the programs like SGP support it.  I can't imagine not using SGP or something similar.  Anyway, great to hear you found a solution.  I look forward to seeing how the a7s performs for AP.