Debayering a DSLR's Bayer matrix.

[Steve789]

Hi frugal, I didn't bother attaching it to the scope but I did try the camera before I started. I borrowed a lens to see if I could solve the error 01 with a quick clean of the lens contacts, no joy there. I took a few shots with no lens at all, pointed at the TV. Couldn't see any evidence of dead pixels etc and it saved the image to the SD card ok. I haven't checked the USB so I hope that works.

I actually quite enjoy pulling these cameras apart and having a play with them. I just hope I don't get addicted to it like Gina!

[frostynixon]

Is it really addiction or more like a love hate relationship?

[Gina]

Touche

[Steve789]

Well, after applying the epoxy and allowing it to cure I reassembled the camera before doing any debayering. My worries about the corroded ribbon connectors were partly correct as the rear LCD and buttons appear to be dead (this was the worst affected ribbon connector). However the USB works and I was able to take some shots with APT and the images came out fine.

However after debayering the sensor with the melamine tool I now get Err99 showing through the viewfinder. The mirror flips up and the shutter does something and then the camera is unresponsive. When I open the battery door the shutter makes a second action and on closing the battery door the mirror flips back down.

My first thoughts were that I've killed the sensor but now I'm wondering if any of the corroded ribbon connectors are causing issues. Seems unlikely as I was able to get the camera working before I attacked the sensor but I think I'll pull the camera apart one more time to check.

Its worth noting that the micro-lens layer and the cfa came away much easier on the 1000D than on the 300D. If I try this again on another 1000D I'll try the polishing technique described by JTW Astronomy here: http://www.jtwastronomy.com/tutorials/debayer.html

[Cath]

Welcome to the Dead DSLR Devotees group

lol

[Herra Kuulapaa]

...

However after debayering the sensor with the melamine tool I now get Err99 showing through the viewfinder. ...

That's an unspecified error. It may or may not be the sensor. But who knows.

When you kill a sensor deviously with heat you'll get the err70.

One hundred pages of dead cameras

[Thalestris24]

I just had bad news about the repair to my debayered Rebel T3/1100d... After 7 months it seems JTW have given up on it now. Disappointed

Another one bites the dust!!

Louise

[Herra Kuulapaa]

Did you Gina do imaging with two cameras?

I have this kind of crazy idea to be tested. Two monochrome Nikon D5100 bodies with narrow band filters.

[Gina]

Yes, I did use two cooled 1100Ds with NB filters (Ha and OIII) that may be what you are thinking of.  These had their red filters removed but not the CFA so resolution suffered.  I'll see if I can find the images.

I have also used three CCD cameras.  Not yet debayered DSLRs but I hope to do that some time in the future.  I have future plans for widefield and standard imaging at the same time both with triple imaging rigs.  That would require 6 cameras and I can't afford that in astro CCDs so as a future project I would like to have three debayered DSLRs (probably 450D) for the widefield using SLR camera lenses and Baader 36mm unmounted NB filters.  However, an interim stage will be one debayered DSLR with filter wheel and I'm hoping to get onto that fairly soon.

[Fred_76]

Did you Gina do imaging with two cameras?

I have this kind of crazy idea to be tested. Two monochrome Nikon D5100 bodies with narrow band filters.

Fo you mean this ?

[Gina]

OK I give up - you win

[Herra Kuulapaa]

Heh, well, if I start of the mount first, it may be hard to believe, but I'm working with a slightly smaller budget, if it even can be called such..

And what a job it would be focus them all !!

[DragosN]

I have a question. Not for myself, I'm not a astrophotographer, but from a friend. 

If you count that a DSLR cannot be cooled in the same manner as a dedicated CCD camera, this operation, of debayering a DSLR sensor, it is worth the effort, even if we think that the result could be very good?

[Herra Kuulapaa]

I have a question. Not for myself, I'm not a astrophotographer, but from a friend. 

If you count that a DSLR cannot be cooled in the same manner as a dedicated CCD camera, this operation, of debayering a DSLR sensor, it is worth the effort, even if we think that the result could be very good?

Hi,

I depends if one can get it right first or second time. After that, hardly anymore if one is a sane person (it usually helps not to be). There is however a considerable risk of failure involved.

DSLRs can be cooled though and some interesting results can be achieved with that.

This is what happens in Nikon D7000/D5100 sensor when temperature decreases:

Dark current: "0.14eps at RT and 0.005eps at freezing."

Member: Astronomer at www.Nikonhacker.com, Writer of https://landingfield.wordpress.com/

And how it shows in real life:

My own picture, Cooled monochrome 5100 frame, ISO1600/10min, true dark current enabled.

Bigger image showing the noise characteristics better:

http://www.kuulapaa.com/Tahtidata2/Compare.jpg

[DragosN]

Thank you Herra Kuulapaa. So the risk is big. I have asked because I have no time to read 100 pages on this topic, to see if there is a guaranteed method of debayering. 

I saw the timelapse on landingfield. Very nice.

[Gina]

Sorry to have to tell you but there is NO guaranteed method of debayering that anyone has reported   Some models are easier and more successful than others.  I probably spent the best part of a decent astro CCD camera on used or non working cameras from ebay but it was fun

[DragosN]

Looking after the name of this group, ''Dead DSLR Devotees'' , I thought so, but I want to be sure. 

[Herra Kuulapaa]

Just to avoid confusion:

Although I posted the dark curreny table above, the Landingfield blog is not mine. It belongs to a member/writer "astronomer" in Nikonhacker.com

[Dave Lloyd]

Looking after the name of this group, ''Dead DSLR Devotees'' , I thought so, but I want to be sure.

100 pages and still the will it / won't it uncertainty.

[Herra Kuulapaa]

100 pages and still the will it / won't it uncertainty.

Mostly doesn't  And here's one 550D that does not either.

But something we can learn nevertheless: There should be no shortcircuit risk with cold finger as far as one keeps away from the sensor connectors. Space is really limited with 550D, there's still enough space for a thin element though.

[spokeshave]

As a recent devotee to the DSLR Killer Klub, I though I would report my progress. I posted a couple of months ago about an idea I had to use abrasive blasting to remove the CFA. I successfully removed the glass using a simply soldering as posted in a previous thread. I bought a cheap "air eraser" which is much like an airbrush but is intended to spray abrasive instead of paint. It can be very precise and I though it would be ideal for removing the CFA if I could find an abrasive that was harder than the CFA but softer than the silicon dioxide optical pad underneath. The abrasive I chose to try is sodium bicarbonate (baking soda), and I started off at very low air pressure (15 psi). The baking soda  is quite soft. So I tried it on a spare, partially broken but functional 550D I got off of eBay for $60. The results were truly amazing. It removed the microlenses and CFA almost instantly and didn't appear to harm the substrate at all. That was very promising. Sadly, when I put the sensor back in the camera, I got the dreaded Err 70. I figure that the abrasive action induced some static electricity on the face of the sensor. So, I tabled the abrasive blasting idea until I could think about it some more. I used the dead sensor to experiment with some of the "old school" mechanical removal methods.

I settled on using aluminum as the scraper material. I took the shaft out of an aluminum pop-rivet, sharpened the tip, and chucked it in a pin vise so I would have very good control. I didn't bother with trying to epoxy the gold wires since epoxy shrinks as it cures and I think that has been the cause of some of the failures posted here. Used with a light touch, the aluminum tool seemed to do a decent job of removing the microlenses and CFA although I was having some difficulty with the CFA material smearing and then becoming very difficult to remove. Still, I thought it was worth pursuing. I found a lot of 5 broken 550D cameras on eBay for less than $200, and the seller said that all had intact sensors. So I took a chance and bought them think I would have 5 sensors to work with.

After removing and testing the first sensor to ensure it was good (it was - they all are), I went to work on the scraping method. To help keep me from damaging the wires, I removed the hot mirror filter and re-installed the plastic holder. This covers the wires and also gives me a surface to hold the tool against while carefully working the edges face of the sensor. The results were less than satisfactory. I did remove most of the CFA, but some of it was smeared on the sensor face. I needed a method to remove the smeared material. Taking a cue from JTW, I tried some automotive paint polishing compound. I cut the tip off of a toothpick at an angle and flattened the tip a bit in some very fine sandpaper to make a polishing tool. The polish worked beautifully (3M Machine Polish Blue). I was able to polish off all of the smeared material even very close to the edges.

I put the sensor back in the camera, and IT WORKED!. No obvious scratches, no dead pixels, rows or columns, just a clear monochrome image. Pic is attached.

That's not the end of the story, though. During the time that I have been thinking about this, I have done a fair amount of research on the CFA. Information is limited, but from what I have read, I believe that the CFA is composed of photoresist material that has been dyed. The photoresist material is an epoxy-like material which explains its chemical resistance that has been documented here. However, in the semiconductor manufacture industry, they use a solution called "piranha solution" to remove photoresist. This solution is typically three parts sulfuric acid and one part 35% hydrogen peroxide. This solution will essentially dissolve any organic molecule, which the photoresist is. So, since I have 4 more sensors to play with, I intend to try it. I should point out that piranha solution is very dangerous to work with, but I fortunately have access to a fume hood and appropriate protective gear. I will only mix up a milliliter or so of the solution at a time, so the risks will be small. If this works as I anticipate, I should be able to completely debayer a sensor with zero damage to the sensor underneath (it is completely non-organic) in a matter of minutes. I ordered the hydrogen peroxide (it is conveniently available in 35% strength as a nutritional supplement) and picked up some sulfuric acid from the auto parts store (battery acid).

So, the experimenting will continue. My hope is to develop an effective method to easily remove the CFA while leaving the sensor undamaged. I will continue to report on my progress.

Tim

[RAC]

Very good Tim.

[Gina]

Yes indeed

[spokeshave]

If anyone is interested, here is a link to the full-size image. The image in my first post is uncropped but downsized and jpegged. This is the RAW file that has only been run through DCRAW with the -d, -6 and -T switches. Those switches force no demosaicing, 16-bit output to a TIFF file respectively.

http://www.timsroadster.com/pics/mono.tiff

Tim

[Xplode]

If anyone is interested, here is a link to the full-size image. The image in my first post is uncropped but downsized and jpegged. This is the RAW file that has only been run through DCRAW with the -d, -6 and -T switches. Those switches force no demosaicing, 16-bit output to a TIFF file respectively.

http://www.timsroadster.com/pics/mono.tiff

Tim

I see there is still a bayer pattern in that image with uneven values for the pixels, i wonder how that would be fixed best?

I tried to debayer the image and it turns out looking good, but i guess resolution is lost in the debayering process.

Is there a way to change the pixels values so they get the correct values so it doesn't have to be debayered? Could DCRAW fix this or do they need to be processed by some new software?

500x500 crop from original image

VNG debayered by pixinsight

Changed to grayscale (extract luminance seems to give exactly the same result)

Could you post the RAW file so i could experiment more?