Debayering a DSLR's Bayer matrix.

[Gina]

I did it! Easily remove the cover glass of my 1100D t3 x50 DSLR

The technique is very simple.

I'd recommend putting some tape in the center portion only of the cover glass so when it cracks it wont' fall and brake any gold wires.

1. Place sensor in a shallow pan of distilled water to insulate it thermal from the heat. (not sure if this is really needed, but I did it anyways)

2. Get a piece of 1x1.25 inch aluminum tube with very flat cut and place it on top of the sensor, stuff some rockwool or other high temp insulation inside. Put a few ouches of metal weights on top so it does move.

3. Take propane torch and heat up aluminum tube, at the very top, heat will travel down. Carefull it will start melting the aluminum tube after about a minute!

4. glass with probably start to crack and come free. You can see this because along the edge the bond goes from dark black to a lighter color as the glue bond breaks and you see a sliver of an air gap.

5. If you have sections that did not break free then heat up a flat tip screwdriver or a chisel end till red hot and then touch the glued cover glass, it will release

you will have to gently use exacto chisel to get off cover, angle sensor so glass falls away from it.

pictures below In random order showing how easy it is, took ten minutes.

In hind sight I think a red hot chisel or screw driver would work just as well. I was really hoping to get the cover glass off without any breakage, but I think that is doubtfull.

I also read that soaking the sensor upside down in a very very 1mm thick layer of acetone for a week would loosen the bond, but I'm impatient and I think if done wrong the acetone would get on sensor and destroy it.

This is an interesting idea but I would have thought it was very difficult to get the end of the aluminium tube really flat to get good contact all round.  Any air gap will result in uneven heating and consequent cracking of the glass.

[Gina]

Checked on my ebay Purchase History and find I have 4 x 1100D and 2 x 450D cameras - that'll do for now    Next step is to gather together all the bits and pieces to do the job, clear the table and get set up.  I plan to use my small hot air gun with DIY nozzle adapter to heat the cover glass round the edges where the glue is.  A craft knife with nice sharp edge to get under the glass will complete the cover glass removal kit.  Add to that holding the metal frame in a Mole Wrench so that the glass is angled slightly downwards so that if it does crack any loose bits will fall away from those delicate gold wires.

For debayering, two grades of abrasive I found worked before, firstly a coarse one to break through the top hard layer and then a fine one to gradually remove the Bayer layer.  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  The tool I used was a piece of plastic with a chisel edge and a few layers of cotton cloth held in place with an elastic band.

Apart from putting the remains of the camera after stripping off all unneeded bits for astro imaging, in a new box, the other mod wanted for astro use is to cool the sensor.  Nothing like as much as I used before - no benefit is gained in S/N performance below 0°C so cold finger and a small amount of Peltier TEC cooling.  In mid-winter a TEC may not even be needed at all   I might even try without a cold finger and just cool a chamber containing the sensor assembly and shutter.  Better would be to separate the shutter from the cold chamber with a UV filter (or if I get three camreas debayered for NB imaging - I could build in the appropriate NB filter in that position).

[Gina]

Had a go at removing the cover glass from an 1100D sensor.  Used hot air gun but without nozzle reducer and got it off without damage to sensor or gold wires but not in one piece and there's still a bit remaining along one side which doesn't get in the way.

Edited December 9, 2015 by Gina

[Gina]

Image testing has shown that there is damage to the sensor   Several hot lines and part of the sensor much more sensitive than the other though both provide images.

Two images to post - a flat and a view in the living room towards the Titan and UP 3D printers.

[Gina]

I shall go on to try debayering this sensor to get practice in the technique.  I should be able to tell if I create any more damage.

[Gina]

Used the coarse abrasive to get through the tough green layer and show the gold but now I get Err 70 and a black image   I shall investigate further but I guess I've now totally killed the sensor.  This is not a good start but it is only a start.  3 1100D and 2 450D sensors to go

[Gina]

I stripped the sensor out again and examined it carefully and also checked all connections.  The gold wires appeared to be intact but one row seem to have been bent inwards slighty so there could be an invisible break in one or more.  OTOH I could have gone too deep with the polishing.  At least I have some feedback from this experiment - I need to sort out a better way of fitting the filter frame and shield as it would appear that the shield touched the gold wires on one side.  I have to find out why it did it this time and not with an already dud sensor.

[Gina]

The biggest problem with any sort of shield is getting it in position without touching the gold wires.

Edited December 10, 2015 by Gina

[Gina]

I think if the filter frame is fitted first and secured with its two screws, the shield can then be clipped it without touching the thin gold wires.  Maybe last titme I just took the filter frame complete with shield and placed that onto the sensor unit.  In that case the shield could easily touch the wires if the filter frame wasn't moved into place straight down.

[Gina]

I thought I'd try removing the cover glass on a 450D so I've been trying to find my previous posts on this but without success so I decided to press ahead and see if I could do it without heat.  I could and did   Cover glass removed in one piece simply by carefully working round the edge and glue with a craft knife   Hooray   Now I need to find where I put the filter frame from this camera (or another from a 450D) as it's different from the 1100D one   I wonder if all 450Ds are this easy...

[Gina]

Found the 450D filter unit and removed the two filters.  The first filter with the piezoelectric "shaker" just lifted out and the craft knife made short work of removing the blue colour-balance filter and plastic mask.  Both filters came out in one piece.  I then checked that the shield I had made for the 1100D fitted this filter frame, which it did - perfectly.  Next I carefully placed the filter frame on the sensor unit and attached it with the two tiny screws.  Finally, I lined up the shield and clipped it into the filter frame.  I reassembled the camera and tested it to be sure all was well.  Yep - fine. 

Since the 450D has live view I had a thought - maybe I could use live view to watch the effect of debayering as it proceeded   I took the image sensor unit back out and reattached the main board without it.  Then I just plugged the sensor unt into the MB and connected everything else needed and powered up.  Set EOS Utility to Live View and displayed the image on the laptop.  Thus the laptop was displaying live view as a flat on the screen with the sensor totally available for "action"   When I actually do the debayering I shall remove the lens and place the camera body on the table rather than on the laptop and use a longer USB cable so that I can have the camera in front of me with the laptop behind the camera.

This first photo shows the sensor unit removed from its usual position and sitting on the main board (just held away from it).  The shutter is closed and shows the remaining damage after I had repaired it and refitted the blades into their slots.  I bought the camera with a broken shutter unit.

This next photo shows the camera sitting on my astro laptop running EOS Utility in Live View with the live image displayed on the laptop screen.

Close up of the camera with sensor ready for CFA removal and shutter open in live view mode.

[Gina]

Another of the 1100Ds I won in ebay auctions arrived this afternoon.  Bunged a battery in it and connected it to my astro laptop, switched on, ran EOS Utilities and the camera works fine

[RobRob]

Hi Gina, I think that is really smart debayering while watching the screen for damage signs. This will provide a much better idea what is causing the damaging (instead of guessing for a root cause)!

[Gina]

Thanks Rob

I've started CFA removal using kitchen cleaner dissolved in water - a pretty coarse abrasive but the fine abrasive won't break the top layer.  Three photos, first shows the sensor viewed from above while the second and third show photos of the laptop screen displaying the live view "flat".  Not that the incident light is perfectly flat but it's light from a cloudy sky through a big window so not bad   Surprisingly,  the coarse, scratchy CFA removal seems to have left a reasonably flat image though I see I have already caused the odd hot line and several hot pixels so I'm not continuing with the coarse abrasive - I'll see if the fine will get at the edges to remove the remaining CFA.

At one point I thought I'd clean up using IPA but this caused half the sensor to go dark - like my previous attempt with the 1100D sensor - but this time it wasn't permanent and the sensor has recovered.  I think I'll just stick to water and leave solvents alone

I would now like to test the sensor with an Ha or SII filter to see how much improvenent CFA removal produces before removing the CFA from all the sensor.  I shall need to design and print a holder for the filter.

[Fred_76]

I shall need to design and print a holder for the filter.

2015 CFA Rmoval 01.jpg2015 CFA Rmoval 02.jpg2015 CFA Rmoval 03.jpg

Hi Gina,

Here is what we made to hold 1"1/4 filters as a Clip EOS :

http://www.webastro.net/forum/showthread.php?t=123946

Fred

[Gina]

Thank you Fred   I have 36mm unmounted filters for DSLRs and printed a 4 way filter wheel but ATM I want to use just one filter and have printed a ring to take the 36mm filter in the throat of the camera.

[Gina]

I have now reassembled the camera without shield and with sensor unit back in its normal place so that I can use the shutter and take actual exposures.  As a quick test I have taken a flat with Ha filter and sheet of paper to flatten the incident light from a fluorescemt table light, as JPG colour, but clearly I need to save as RAW in .CR2 format and then use DCRAW to convert to something usable.

[Mick J]

Another great thread to follow, good luck with these Gina.

[Gina]

Thanks Mick

[Gina]

I've been Googling for differences between the 450D and the 1100D and apart from 1100D doing video and higher ISO plus that the 1100D is later with a later processor, there seems to be negligible difference.  I came across a post I made saying that the 1100D was far lower noise than the 450D but I don't remember this and really don't know where it came from but I shall be doing tests anyway.  I can do some tests before debayering since it's dark performance that's affected.  I've done extensive noise tests on the 1100D in the past but don't remember doing anything like that for the 450D.  As I recall, I only started looking at the 450D since starting to try debayering.

Edited December 12, 2015 by Gina

[Gina]

Wired up PSU to camera so no need for battery (I couldn't find a charger though I think I've got a 450D one somewhere).  Set image format to RAW + LargeJPG so I can have both formats and taken an exposure with Ha filter.  Ran .CR2 file through DCRaw to concert to TIFF then through Photoshop to convert to PNG for upload here.  Also zoomed in greatly to show mosaic and selected two areas to snip - with and without CFA and microlens removal.  No level shifting or brightness/contrast processing applied - purely scale.  I have increased zoom to maximum and snipped a smaller area to compare directly with the debayered area.

Subjectively I would say that the increased sensitivity to Ha obtained by removing the CFA has well made up for the loss due to the removal of microlenses.  I put this down to the shape of the red sensitivity curve and the filter cutting the Ha wavelength.

[Gina]

Now the same with SII filter...

[Gina]

I would do the same with the OIII filter but I can't find it   All I can do ATM is surmise that the response to OIII spectral lines around 500nm is well off the peak of both blue and green spectral responses and that the CFA will reduce the sensitivity to OIII in both blue and green pixels.  Whether this would be less or more than increased sensitivity with microlenses I don't know.

Wavelengths :-

OIII  501 nm and 496 nm
Ha   656.28 nm
SII   672.4nm

This is the quoted spectral response of the 450D sensor without IR cut filter.

Edited December 12, 2015 by Gina

[Gina]

Done some more CFA removal with fine abrasive.  Gone pretty well except that I've collected another hot line.  Other defects arose from using the coarse abrasive - next time I must be more careful but I think this sensor might be usable.  It could be relegated to SII if I do a better one for Ha.

Here is the result - a flat taken with Ha from my desk lamp with paper to disperse the light.  Exposure 1/25s at ISO 1600. 

Curves applied in PS to stretch the histogram to show flat.  The scratches are clear as are the lines but apart from a couple of bits of thin remaining CFA I don't think it's too bad.  The black bits round the edge are polishing paste that crept under the shield and which I haven't risked trying to remove.  The shield does cover some of the image area but well protects the gold wires and I think the loss of image area is worth it.

This is a small area of the debayered area blown up to show the pixels.  The variation in sensitivity is now more apparent.

[Gina]

Taken a series of darks culminating in 10m at ISO 1600.  Here's the latter highly stretched in PS to show the noise.  I seem to have a very hot pixel on the bottom edge that's somehow spreading light around

Edited December 12, 2015 by Gina