Sony A7S Split Sensor And Banding

[sharkmelley]

I've finally got to the bottom of a very weird problem with my full-frame Sony A7S - coloured banding on the left hand side of the image.  It is made visible by stacking hours of data and it won't calibrate out.  Here's the banding as it appears in a flat taken at a low signal level (I've subtracted a blurred version of the flat to make the details more visible):

The problem arises from the fact that the full-frame sensor is split down the middle causing some kind of left/right discontinuity which somehow seeds a horizontal band.  That would not be a problem if the band appeared in the same place and calibrated out.  Unfortunately the banding changes according to signal level.

For the following video I created a series of daylight master flats using successive increasing exposures from a shutter speed of 1/30sec to 1sec:

https://drive.google.com/file/d/0B3Ky5pyZvsINTTQwNllXTXFpWm8/view?usp=sharing

Watch how the banding expands and becomes more complex as the signal level increases.  This explains why the flats do not calibrate out the stripes found in the lights - it is simply impossible to match them.

If you are interested, there is a long thread I initiated on Cloudy Nights:

http://www.cloudynights.com/topic/519941-sony-a7s-split-sensor-and-banding/

I am not the only Sony A7S user to have encountered this issue.  It's not yet clear if every Sony A7S is similarly affected.  It's also not yet clear if any other cameras (e.g. full frame Nikons) that use full-frame Sony Exmor sensors are similarly affected.  I suspect full-frame Nikons are not affected because surely the problem would have come to light by now.

Given this problem and given the bulb-mode star eater issue it is impossible for me to recommend Sony for deep-sky astrophotography.

Mark

[Mick J]

Wow, cheers Mark, just did the cloudynights thread, great read and stunning reply's.  Hope you find a solution to this, seems you may be getting closer to finding a way to work around or control it.

[sharkmelley]

I should have added a link to the methodology I used to highlight those coloured bands - taking an underexposed flat at low ISO and then subtracting a blurred version of the flat from the original:

http://www.cloudynights.com/topic/524352-which-full-frame-sensors-are-split-sensors-with-banding/

I am very interested in results from other full-frame DLSR cameras.  Do they have this "stitched" split sensor?  We already know that the Nikon DF and the Canon 1D MkII have a split sensor - probably most full-frame DSLR cameras do.   That in itself is not the problem - it is the weird banding "seeded" by the discontinuity that is the problem.

If you try this on your own full-frame DSLR then please post the result so the astro-community can begin to build up a knowledge-base of other cameras that might be affected.

Mark

[StuartJPP]

This isn't good news Mark but it is good to see your continued investigation in to this (as seen on CN). Unfortunately, like most things, everything is not always what they first appear. It is a shame since the Sony sensors are really, really good. Unfortunately Sony probably don't care or are not even aware of the few consumers who are using these in the way that we would.

Not long before this thread started I had purchased a Canon 6D with the sole intention of modifying it. I had wondered if I had made a mistake, but I am quite happy with it so far. If I really want to go further I would probably get a dedicated CCD, but I am a long way off from that due to the cost per use ratio. Hopefully the next generation of Sony sensors will address these issues.

Have you managed to have an imaging session on a single target collecting as many 30 second exposures using your f/2.8 Tak as you can? It would still be interesting to see the end result.

I will have a go at your above banding experiment when I have a chance.

[sharkmelley]

Hi Stuart,

The Canon 6D is one of the cameras I'm considering as a replacement for my Sony.

I haven't gone much beyond 2 hours worth of 30sec exposures with the Sony A7S because banding becomes such an issue. 

Here is one with 240x30sec on the Tak Epsilon 180ED:

Larger version here: http://www.markshelley.co.uk/Astronomy/2015/horse20151208.jpg

It is not processed optimally because I was trying to keep the banding in the dust clouds at the top left under control. Even so, it is quite noticeable to me, maybe because I know it is there. I do have one or two images with more than 2 hours of data but they are sitting on my hard drive waiting (in vain) for some way of removing the bands.

Mark

[StuartJPP]

Hi Mark, although I haven't really played around too much with your image it doesn't appear as bad as I thought it would...it is noticeable but to be honest I don't think I would have spotted it if you hadn't said where it is.

Doing a harsh stretch on it does make it more visible but even then it doesn't jump out at you.

I will have a look on my other monitor which usually shows off more defects.

Will the 6D not bring other issues to the party?

[sharkmelley]

The 6D certainly has a higher dark current than any of the Sony Exmor sensors - that could be an issue on slower scopes or imaging in warm conditions.  From my current understanding (see for instance the tests by Roger Clark at http://www.clarkvision.com/reviews/evaluation-canon-7dii/), the APS-C size Canon 7D mkII is the first Canon to have dark current down at Sony Exmor levels.

According to Roger Clark, the Canon 6D has the usual Canon banding issue well under control.

My shortlist for a full-frame camera is currently:  Canon 6D, Nikon D800, Nikon D810.  The obvious choice, the Nikon D810a specifically created for astrophotography, is rather too expensive.

Mark