Debayering an 1100D sensor (Removing Colour Filter Array)

[Gina]

This thread takes my successful attempt at removing the Colour Filter Array from an 1100D sensor as described in the "Debayering a DSLR's Bayer matrix." thread, which has grown and grown "like Topsy" to form a much shorter, more manageable one. This was after a number of unsuccessful attempts.

Debayering consisted of two distinct parts both difficult with considerable chance of destroying the sensor in the process.

1. Removing the cover glass
   
2. Removing the CFA
   

First problem was removing the cover glass from the sensor. Both heat and pure physical force were tried before without success in keeping the sensor working. This successful run used heat from a small hot air gun applied in concentrated form with great care for short periods.

[Gina]

Removing The Cover Glass

Couldn't find the little copper nozzle I'd made so had to make up a new one. I think this one is a bit better.

The hole in the end is 3mm diameter and I applied the hot air right on the glass above the bond for a couple of seconds. I started with the corner and let it cool then I proceeded slowly along the edge for about 3 seconds. By the time I got half way I saw a change in the bond. At that point I stopped. I checked the sensor frame and cold finger and it was barely warm to the touch.

This photo shows the sensor assembly and cold finger placed on a 5mm thick aluminium plate. The heat treatment was on the cover glass at the top and towards the left. I started with the top right corner (as in the pic) and then from the corner towards the left. The unit was arranged to show the change in the bond to it's best - in the sunshine.

My 1100D sensor is still working well I put it back in the camera and replaced main board etc. Tested with netbook and EOS Utility. So far so good

A bit more and still working

Further round and still working "Slowly, slowly catchee monkey"

Bond apparently broken all the way round but still stubbornly attached to the frame! Sensor is still working fine

Put some adhesive tape over the cover glass to hold any possible loose bits and while lifting the edge of the tape, worked gradually around the bond edge with the tip of the craft knife. I could see varing diffraction patterns as I went round then after going round several times the glass cracked and half of it was free. So with tape and free part of glass still in position, I continued and eventually the other half came off and tape plus the glass could be lifted off.

The sensor is still working SUCCESS the glass is off and the sensor still works :) I'm taking a rest before tackling the CFA

[Gina]

Removing The CFA

Now all prepared for CFA removal I used the frame that originally held the IR cut and colour correction filters to protect the fine gold wires that connect the actual sensor to the carrier. Here we see a filter frame without any filters, screwed in place onto the sensor assembly, to shield all those ultra fragile fine gold wires.

Started CFA removal on my 1100D sensor. Sensor still working Not doing any more tonight - not pushing it

Here are some images captured on my netbook by EOS Utility. 0.1s at ISO 100 and somewher around f4. The first pic was taken looking out of the lounge window with the lights on hence the reflections Then we have flats taken with a tissue over the lens. As taken, then stretched and finally cropped to the cleared area.

[Gina]

The CFA was removed entirely using abrasive as previous attempts with a soft copper scraper caused scratches right into the pixel layer. I started with a coarse abrasive to break through the surface then finished off with a very fine abrasive. The coarse abrasive was ordinary kitchen scouring powder in a little water. That was removed and all residue cleaned off before gradually polishing away the CFA with a very fine abrasive designed for polishing scratches out of car paintwork - Meguiar's ScratchX 2.0 :- http://www.amazon.co...0?ie=UTF8&psc=1

After each CFA removal session the polish was removed with IPA on a cotton bud with the other end of the cotton bud used to dry off the IPA and pick up any remaining residue.

[Gina]

More CFA removed Not very well cleaned up

1. View south out of window 1/4000s ISO100
   
2. Flat as taken
   
3. Flat stretched histogram
   
4. Ditto, cropped
   
5. Curves applied to flat
   

[Gina]

One thing that needed an explanation in the main debayering thread was the apparent lack of colour in the yet to be debayered area around the edge of the frame. The outer area is not completely devoid of colour just a bit lacking due to the lack of red killing filter. Hence the greens aren't very green. The blue in the sky (when we have any) shows as a rather magenta shade of blue. Outside the debayered area is an area where the micro lenses have been removed hence being a bit darker than out at the edges.

Here's another photo taken through an open window to the east. I have annotated it within Ps.

And this is a photo of the sensor and filter frame from the front. The matt area is where the micro lenses have been removed but the CFA is still intact.

[Gina]

More CFA removal and still OK

Earllier I used a DSLR sensor cleaning swab cut down a lot to provide a narrow polishing tool but this proved too soft and the cloth used had a habit of shedding fibres so for this latest session I cut the tool down further and shortened it to make it stiffer. I dumped the original swab - it was full of green CFA anyway - and used some cotton cloth folded several times over the end of the plastic tool. Unlike the cloth that was on the original cleaning swab, this doesn't shed fibres. It's also thinner so provides a finer edge. Here is an image captured with EOS Utility and processed in Ps to return the colour balance to "normal" around the edge. The debayered area still looks red but less so. I think the monochrome area looks a bit magenta because the colour area has twice as much green as red or blue.

[Gina]

So far all the images have been captuired and displayed as JPEG but since the whole idea of this mod is to produce a mono image, it makes sense to display the results in monochrome from the original raw .CR2 file, developed without debayering into proper linear 16 bit TIFF format that we use for astrophotography. Hence, I've been reading up on DCRaw which does precisely that (if told to).

I've put dcraw.exe in the Windows directory and a batch file in C:\Users\user\AppData\Roaming\Microsoft\Windows\SendTo containing the following code :-

@echo off
:begin
if _%1_==__ goto end
echo Processing %1...
c:\windows\dcraw -D -4 -T %1
shift
goto begin
:end

Putting the batch file in the SendTo directory means that DCRaw can be accessed easily using the "SendTo" right-click option on an image file. So right-clicking on the image file name and choosing to SendTo "dcraw" will bring up a command line window where the selected file will be converted to TIFF (called "developed"). After a while the image will be saved as a 16bit TIFF file in the same folder and with the same file name but with the extension changed to TIFF and the command window will close. This makes using DCRaw a simple operation without needing any knowledge of the workings of the command line.

[Gina]

It seems to be working Converted CR2 file to TIFF, stretched it in Ps and here's the result - whole frame resized and cropped to debayered area and resized. With so much glass removed the kit lens will no longer focus on infinity so these are slightly out of focus.

[Gina]

Here's the latest flat - CR2 to TIFF in dcraw then well stretched in Ps to show unevenness of the sensitivity. No sure if this is all due to eneven polishing or partly due to some residue.

[Gina]

More CFA removal and I think I might stop now while I'm still winning Here is the latest flat, stretched in Ps till it squeaked I did my best to clean it up using IPA on a cotton bud then dried using the other end so the darker area in the middle is where I've gone a bit too deep with the polish. It would appear to be still useable if corrected with flats.

More photos - mono image and colour photo much stretched to show sensor surface.

[Gina]

Here is what it's all about. At the bottom of the image the CFA is more or less intact and you can see the quads of pixels and the fact that some of them are quite dark. The top part has had the CFA removed and all the pixels contribute to the brightness, in fact showing more sensitivity than any in the region below. This would be even more apparent if the view were a red colour rather than greenish as only one in four of the pixels with CFA would be responding to the red. This demonstrates that not only does this make the sensor more sensitive but the resolution is also improved. This is a close up view of part of the image above right.

[Gina]

And that Ladies and Gentlemen brings this story right up-to-date. I have now copied all relevant information from the "Great DSLR Debayering Thread" I believe, adding a few extra comments as seemed necessary.

The next stage will be "first light" on the night sky and probably a series of images of a nebula in Ha. I have the debayered sensor set up in a test camera with slight cooling ready for a clear night. Meanwhile, I'm working on Peltier cooling of another 1100D. This I will be describing in the companion thread - "Adding cooling and filter wheel to a debayered 1100D mono DSLR" :- http://stargazerslounge.com/topic/194790-adding-cooling-and-filter-wheel-to-a-debayered-1100d-mono-dslr/

[SnakeyJ]

My apology for interrupting mid thread Gina, but extremely useful post/summary from your original thread. Will be waiting with baited breath to see how this performs in real use.

I'll ask one of the mods to remove my earlier post!

[Gina]

Thank you Jake Glad it's of interest

[badgerchap]

Great stuff Gina. Eagerly awaiting some first light results. So glad we have you to guinea pig all of this, don't think I'd be quite brave enough to give it a shot otherwise!

[Gina]

Great stuff Gina. Eagerly awaiting some first light results. So glad we have you to guinea pig all of this, don't think I'd be quite brave enough to give it a shot otherwise!

Thank you I enjoy doing these experiments and passing on the results I've always liked research.

[badgerchap]

Thank you I enjoy doing these experiments and passing on the results I've always liked research.

Do you ever feel that the struggle is more rewarding than the goal? I sometimes feel that the mods themselves are more satisfying than the results they produce!

[Gina]

Yes, I do sometimes There is definitely a joy to be had from achieving something difficult and the more difficult the more the satisfaction when it works.

[Gina]

Here are some images with the mono 1100D on the MN190 when it was reasonably clear for a while - with the Baader 7nm Ha filter.  Focussed up on the moon and got some shots at the same time then slewed to the Eagle Nebula and took a few at different exposures.  The one below is 10m at ISO 6400.

1. Moon cropped just to remove the CFA border
2. Cropped more
3. Actual image size pixel for pixel
4. Eagle Nebula
5. Flat.

 

   

It needs a nice long clear spell to test properly, where many subs can be stacked in DSS with darks and flats.  I'm also working on a proper cooling system and I think that might be ready before we get any more clear night skies

[Gina]

Here is the 10m Eagle Nebula sub taken with the MN190 and mono 1100D processed in Ps subtracting the single flat frame and applying various tweaks to improve the image.  I din't get the flat subtraction exactly right but here's what I've managed.  Firstly cropped to remove CFA and cropped more to show the nebula.  At least the Eagle is recognisable.  Not too bad for a single 10m sub with virtually no cooling.