Debayering a DSLR's Bayer matrix.

[Tf2k]

Some update from my work on the Nikon D40.

Last time, i thought i was not deep enough because of the varying colour information in the middle of the reddish tint image.

There was another layer which could be easily scraped away, too bad; that wasnt CFA. that was the first silicon layer of the CCD.

Now i am getting a black image with a littlecolour noise and a thin reddish border (where i didn't scrape away anything).

Man, we got too much time or money, lets see if i get it right with another D40 or EOS 350D Sensor.

Both Camera Bodys are here, so let them bring back to life again

[Gina]

Thanks folks

Clear sky forecast for tonight so I think I'll try to get the debayered sensor in use and give it "first light". I have a box for the camera but I haven't sorted out the cooling yet. Just have to see how it goes. I have a Baader 36mm Ha filter or I might just go for the Astronomik CLS CCD clip filter and do Luminance for the first test run (particularly if I don't get cooling working). I'll be using a vintage SLR lens - either 105mm f2.8 set to f4 or 55mm f1,8 set to f 2.8.

[Andrew*]

Forty-two white knuckle nail-biting pages!

Well done to all the debayering pioneers. It looks like you've overcome a lot of hurdles and found out a lot about the structure of DSLRs, which I'm sure will help a lot of others.

I'll be leaving my 450D alone for now, but maybe if CCD fever kicks in and a few people have done a successful debayer on 450Ds I'll reconsider it... Nowhere near as brave as you guys!

[Luis Campos]

Hmm... they look the same in this shot

I'm also getting a couple of lines in live view but they are in totally different places.

Hi Gina,

These lines are dead pixel collumns causes by too deep scratches that reached the photodiodes under the clear layer :/

I know this because i've just finished another 350D sensor, this time is working!!!! .....But I made ONE SINGLE little scratch but it was deep so I ended up with two dead pixel collums (they intercross), it's not the end of the world though, if you dither your subs you will get them out

I'll try this sensor tonight and see if that's the case....at least I think it is

Cheers,

[Gina]

Hi Gina,

These lines are dead pixel collumns causes by too deep scratches that reached the photodiodes under the clear layer :/

I know this because i've just finished another 350D sensor, this time is working!!!! .....But I made ONE SINGLE little scratch but it was deep so I ended up with two dead pixel collums (they intercross), it's not the end of the world though, if you dither your subs you will get them out

I'll try this sensor tonight and see if that's the case....at least I think it is

Cheers,

Thanks Luis Seems strange that the RHS problem starts with a few dud columns and gets progressively worse even underneath the epoxy resin where it's been protected from scratching. Yes, if it's just one row and one column, dithering should stop it affecting your image.

[Luis Campos]

I finished my camera, everything is working great!

I leaved some CFA remains on the edges because I didn't wanted to push my luck this time. But I can easily crop this.

I am going to dark location tonight, unfortunately Moon is big but I just want to see how this work.

Last night I did some test exposures to see if cooling works. I can get 30C below the ambient without any problems with condensation.

This is 100% crop, stacked 10 exposures at ISO1600 and with sensor temperature of -5C.

Details and resolution is outstanding

Filip....just WOW!!!!!!!

That 450D is one hellish camera now

Very well done, can´t wait to see it in action

Cheers,

BTW, definately the sides of the sensor are to be avoided, I made my second 350D sensor and it's working too, this time i didn't touch the sides, but one single cratch made a couple of dead pixel collumns, dither will take care of that though

Take care,

[Tf2k]

Just doing some research, what went wrong with my Nikon D40 before i scraped of the last bit of the sensor.

The Image quality was not what we have expected.

I figured out, the blue and g2 channel in my .nef file is nearly completely black.

On my way through the CFA, i killed these channels somehow.

for the debayering tools library (Diagnosis):

RAWDigger (http://www.rawdigger.com) Windows Binary, a fancy RAW analysis tool.

[Gina]

I'm all set for "first light" on my debayered 1100D sensor tonight. I haven't got the cooling working yet but have camera in box and mounted on a dovetail bar. Since the moon is approaching full I've decided to use the Baader 36mm 7nm Ha filter. I've Selotaped it to my CLS CCD clip filter to hold it. I plan to run it from my netbook with the laptop providing mount control. Just unguided as, without cooling I don't expect to use very long exposures and up to 5m is fine unguided. However, APT isn't working properly or I have a camera problem as it's failing to do repetitive exposures. I'll try reinstalling APT. EOS Utilities is fine but doesn't have a sequencer so that would mean manual exposures - back to the old days

Edited August 18, 2013 by Gina

[russellhq]

Are you trying the latest update? I see version 2.30 was just released. Don't know if it will help though.

[Gina]

Downloading it now The version I have ATM ia 1.94 - pretty old

[JamesF]

Let's hope the promised clear sky actually happens. It should be clear here now according to the Met Office, but we're currently under thick black cloud

Thinking about those extra sensors, Gina... I wonder if they continued to work through the epoxy because if it's thin it actually transmits light reasonably well. If you've scuffed the epoxy since or caused it to scatter light more perhaps they're now struggling to get enough light.

Perhaps the way forward is just to keep well away from the sides of the frame and leave the CFA intact there. Limit the scraper movement with a couple of "fences" if necessary. You might lose a few hundred pixels off the width of the images, but I wouldn't have said that was a major issue.

James

[Gina]

Let's hope the promised clear sky actually happens. It should be clear here now according to the Met Office, but we're currently under thick black cloud

I thought it was going to clear but the clouds have got thicker and darker instead - in fact it looks more like rain I have reluctantly abandoned testing for tonight. The gear is playing up too. Camera, USB lead, netbook and PSU that were happily producing time exposures indoors have started giving err70 errors out in the obsy Tomorrow is another day, as they say...

Thinking about those extra sensors, Gina... I wonder if they continued to work through the epoxy because if it's thin it actually transmits light reasonably well. If you've scuffed the epoxy since or caused it to scatter light more perhaps they're now struggling to get enough light.

That could well be - because I have scuffed the epoxy. Maybe the reason for the err70 is that the light is that much lower now. Tomorrow will tell when I try it again in daylight.

Perhaps the way forward is just to keep well away from the sides of the frame and leave the CFA intact there. Limit the scraper movement with a couple of "fences" if necessary. You might lose a few hundred pixels off the width of the images, but I wouldn't have said that was a major issue.

Good idea. In fact there could be one advantage - I've checked the coverage of 1.25" filters and the coverage is almost the full width of the sensor so a narrower sensor would mean the 1.25" filters could be used with a considerable saving in cost. The filter easily covers the height. I think I'll do a test and cover the side with the extra sensors with insulation tape over the cover glass. I know some light will get in by scattering but it shouldn't be much. Edited August 18, 2013 by Gina

[pixueto]

I thought I didn't do justice to the camera with the photos I posted before so, here I go again, this time paying attention to light metering to avoid overexposure. The photos have been taken with a 50mm 1.8 nifty-fifty lens on a tripod. Again, the images were in raw format converted to jpeg without any demosaicing applied (apart from the overexposed) using DCRaw.

The image has been reduced to 26% (because of file size) but I'm showing too a crop at 100% showing the central region. I also show the original image that the camera took with its automatic light metering. You can see it was badly overexposed. The removal of the CFA has increased dramatically the sensitivity of the sensor. The camera estimated a exposure of 1/25 and ISO 800 but I had to go to 1/40 and ISO 400 to get the composition right. Thus, I had to go to manual, lowering the exposure time and ISO significantly in order to achieve a correct histogram. All photos were taken at F8.

I can't see any scratches coming up in the photo. As I said, I,ve also run a test for dead pixels with a number of dark frames at different exposures and ISO 1000. The test was both visually and using software which returned 0 dead pixels so the sensor seems intact despite the scratches created with the wooden tool?

Thanks for looking. Any comments welcomed.

I thought it may be worth recalling an earlier post I made with monochrome images with my partially debayered 1000D sensor. As you can see in the pictures, there is no CFA left on the left and right sides (apart from the bits close to the golden connectors). I was quite ruthless removing the CFA in both sides completely unaware that those where supposed to be sensitive areas. The camera is still working; no problem whatsoever.

My point is that there seems to be an unusual amount of dead or damaged sensors since we have started applying epoxy on the golden connectors -or the heat used to remove the glass. I know the sensors worked immediately after the procedure but it may be something that takes some time? I think I'm going to skip that step with my 350D and take my chances trying to avoid the golden connectors.

Edited August 18, 2013 by pixueto

[Gina]

I tried scraping without covering the gold wires and was quite unable to avoid them. Why? you ask - because my muscle control is not what it was and I'm just not steady enough nowadays to do this job without having the gold wires protected.

Edited August 19, 2013 by Gina

[JamesF]

I tried scraping without covering the gold wires and was quite unable to avoid them. Why? you ask - because my muscle control is not what it was and I'm just not steady enough nowadays to do this job without having the gold wires protected.

I can't see that the epoxy is likely to be a problem. Electronic components are sometimes embedded in epoxy to protect them from moisture or vibration, for instance.

James

[Gina]

For astrophotography I think a squarer format is better than the "widescreen" shape so I would prefer to lose width rather than than height. ATM I'm really not sure what to do about the sides - whether to cover with epoxy resin or not.

Edited August 19, 2013 by Gina

[Gina]

I can't see that the epoxy is likely to be a problem. Electronic components are sometimes embedded in epoxy to protect them from moisture or vibration, for instance.

James

Yes, precisely.

[fwm891]

Great thread - is there a 'short' version detailing the salient steps as I've only just started thinking about doing something with my 450D....

[Andrew*]

Great thread - is there a 'short' version detailing the salient steps as I've only just started thinking about doing something with my 450D....

"The Abridged Debayering a DSLR" doesn't exist I'm afraid. TBH, I think it is worth reading through the thread as there are lots of projects followed through with detailed pics and lots of examples of pitfalls to avoid.

Only took me about 2 days to read!

[JamesF]

"The Abridged Debayering a DSLR" doesn't exist I'm afraid. TBH, I think it is worth reading through the thread as there are lots of projects followed through with detailed pics and lots of examples of pitfalls to avoid.

I'm inclined to agree. There's a lot of information on what's worked and what hasn't, but it's quite haphazard. Until it's a successfully repeatable process I think it will probably remain that way. Just don't let your camera see the stuff that Gina has done to some of hers. It'll be running for the door before you have a chance to grab it

James

[fwm891]

"The Abridged Debayering a DSLR" doesn't exist I'm afraid. TBH, I think it is worth reading through the thread as there are lots of projects followed through with detailed pics and lots of examples of pitfalls to avoid.

Only took me about 2 days to read!

I may have retired but...

I'm inclined to agree. There's a lot of information on what's worked and what hasn't, but it's quite haphazard. Until it's a successfully repeatable process I think it will probably remain that way. Just don't let your camera see the stuff that Gina has done to some of hers. It'll be running for the door before you have a chance to grab it

James

Having followed some of Gina's other projects I can well believe it!

Thanks both

[Gina]

Have to say - it's great to be at the "cutting edge" of R&D again

[Gina]

I've just checked the 1100D and time exposures are working again so it would seem that the sensor is not working in low light conditions So the assumption that if the sensor is working for "normal" light levels it will be alright for astro use has been proved invalid

THIS IS SERIOUS!

[Gina]

SO... Seems another test is needed to check that the sensor is still working for astrophotography - A DARK TEST.

Another thing occurs to me. I'm wondering now if temperature will make a difference. So far all my tests have been at ambient temperature - it could be better or worse when cooled to -10C or more. This project has just got BIGGER!

[JamesF]

Does it fail again if you try to run an exposure with the lens cap on, or is the camera not in a suitable state for that?

James