Debayering a DSLR's Bayer matrix.

[nytecam]

Well, since there is no way of removing the CFA without damaging the micro lenses, that begs the question: what's the purpose of the micro lenses? How important a role they play? The images with the CFA removed (and without the micro lenses) seem to be alright, don't they? Another possibility would be to find a camera model that has the micro lenses under the CFA (if any) so it can be modded for mono astrophotography.

The light sensitive area of each pixel is much smaller than the pixel itself and the pixel microlens redirects and recovers light that would be lost so the overall sensitivity would be much higher with the microlens intact. But it seems from experiments here the microlenses get destroyed with removal of the Bayer layer which is a shame when seeking supreme sensitivity Edited November 5, 2012 by nytecam

[Cath]

I didunt realise that Canon EOS sensors are CMOS and not CCD, but it seems with the small improvements Canon made to the CMOS sensor that it's no worse than a CCD. A CCD doesn't need microlenses as the pixel active area (the light sensitive area) covers most of the pixel area.

Canon sensor with microlenses

The new 1D X has what Canon call 'gapless microlenses' .. I presume they just mean the microlenses cover the entire square area of the pixel, rather than being round microlenses, and charge more for it.

Edited November 5, 2012 by Cath

[rcheshire]

That was an entertaining read and I'm sorely tempted to give it a go. Having moved on to a CCD camera, I might just let it slide.

BTW I have some 1000D spare parts including a brand new sensor and one modded. Additionally an almost complete camera except for a working main board and damaged view finder. After dismantling the cooling mod I put all the working bits together and gave it to a relative - their first DSLR.

[nytecam]

Saw this interesting reference on another forum...

"...modern CMOS image sensors (since 2002 or so) are taking the approach of using built-in integrated analog preamp and A/D converter(s). In another word, once the CMOS sensor is out of the maker, there is nothing that can be tweaked...." this may place limitation on their astro use.

[pixueto]

Sorry guys, I might say something dumb but is there any possibility that some suppliers will offer a CMOS Canon compatible with no bayer matrix? If so, would it be at a reasonable price?

Thanks

[Gina]

You can get debayered Canon cameras but they're very expensive and in the price range of decent proper CCD astro cameras which include cooling. I think there's a link earlier in this thread.

Edit.... Found it :- http://www.maxmax.co...amera_order.asp

Edited December 2, 2012 by Gina

[pixueto]

You can get debayered Canon cameras but they're very expensive and in the price range of decent proper CCD astro cameras which include cooling. I think there's a link earlier in this thread.

Edit.... Found it :- http://www.maxmax.co...amera_order.asp

Yes, I agree, for that price there is no much point in doing it as a CCD is cheaper but I was wondering if it would be possible to buy a canon compatible monochrome CMOS at an affordable price to assemble it into an existing DSLR such as the 450D. Is it possible?I was wondering.

By the way Gina, how do you find your astronomik 12nm H alpha filter? Does it work nicely with your DSLR? I am thinking about getting that filter to add a luminance layer to my images (and possibly imaging with the moon around). I also have bad LP in my area and I was thinking about using the H alpha filter to get the details and then the astronomik LP clip filter to 'colour' the picture. Will it work?

Edited December 2, 2012 by pixueto

[GlassWalker]

Using a narrowband filter on a colour sensor is always going to be a compromise. I also use Astronomik 12nm filters on DSLR, but of course you're losing a large chunk of the sensor when doing that (particularly hits resolution). Last night I had another go at the horsehead and flame, this timg in Ha only. The moon was still quite full and close, and I think that pushed the background level too high even in Ha. Even my colour images of the same region had better contrast, but the moon wasn't out when I tried that. Maybe narrower ones will help more...

[pixueto]

Using a narrowband filter on a colour sensor is always going to be a compromise. I also use Astronomik 12nm filters on DSLR, but of course you're losing a large chunk of the sensor when doing that (particularly hits resolution). Last night I had another go at the horsehead and flame, this timg in Ha only. The moon was still quite full and close, and I think that pushed the background level too high even in Ha. Even my colour images of the same region had better contrast, but the moon wasn't out when I tried that. Maybe narrower ones will help more...

That's really interesting. Could you post any pictures to have an idea about the kind of resolution it's lost? To be honest, I don't think the DSLR images are that bad compared to CCD ones (particularly in this cold weather). Some imagers here are doing wonders with them. If I could add that luminance layer and get away with a bit of light pollution that would be brilliant. I know the H alpha filter will only use 1/4 of the available pixels in the sensor but do you think the trouble is worthwhile?

Thanks

Edited December 2, 2012 by pixueto

[pixueto]

Can the lack of resolution be solved with more or longer exposures?

[GlassWalker]

By loss of resolution, I am talking about only 1/4 of the pixels being effective when using Ha or SII. A nominal 12MP colour camera is effectively 3MP when you put that filter on. It just seems a bit of a waste that 3/4 of the sensor is doing nothing useful. The reason anyone does that is any decent mono CCD is going to be quite a lot more expensive.

However, we might be looking too hard at resolution here. Most astro CCDs are relatively low MP count too. And of course the resolution will also depend on your optics and even the conditions if you push hard. Regardless of the sensor, if you feel lacking in resolution you can always go longer focal length, and if needed also mosaic frames to cover bigger subjects.

As a side note, I like using OIII more than Ha because more of the sensor is active in that case... but it does rather depend on the subject being rich in OIII in the first place.

[Gina]

Yes, I agree, for that price there is no much point in doing it as a CCD is cheaper but I was wondering if it would be possible to buy a canon compatible monochrome CMOS at an affordable price to assemble it into an existing DSLR such as the 450D. Is it possible?I was wondering.

No idea I'm afraid but I very much doubt it.

By the way Gina, how do you find your astronomik 12nm H alpha filter? Does it work nicely with your DSLR? I am thinking about getting that filter to add a luminance layer to my images (and possibly imaging with the moon around). I also have bad LP in my area and I was thinking about using the H alpha filter to get the details and then the astronomik LP clip filter to 'colour' the picture. Will it work?

I haven't yet tried using Ha as a luminance layer but others have I believe with good results. I have had quite good results with the Ha clip filter for wide-field imaging using an old 200mm f3.5 telephoto lens both alone and combined with OIII as a bi-colour NB image.

Here's some examples :-

1. Cygnus Loop in Ha with image converted to greyscale
   
2. Heart Nebula in Ha with colour retained
   
3. Cygnus Loop bi-colour with Ha and OIII combined in Photoshop
   

[pixueto]

Thank you guys

[Gina]

Yes, the resolution is 3MPx as only a quarter of the pixels are used as has been said. But that's still 2061 x 1383 which is not bad. The sensitivity suffers too so you need more exposure time. Fortunately many nebulae are very rich in Ha and that helps.

NB filters help reduce the affects of moonlight and light pollutuion but with 12nm this is limited. Still a bit helpful though. Another benefit is reduction in star brightness. I think Astronomik clip filters are worthwhile for wide field with lenses particularly with a twin setup with two cameras and two lenses like I have. With a scope I think filters in a filter wheel are a better bet and, in fact I have been working towards this. I decided to reduce cost by going for the minimum size of filter that will work witth a DSLR viz. 36mm unmounted. But I'm still waiting for the OIII filter which has been on backorder for some time now.

[pixueto]

I think you will find this article interesting:

http://www.stevehuff...patrick-clarke/

So, no need to keep the microlenses then?

Edited December 2, 2012 by pixueto

[Gina]

Yes, interesting - thank you I'm sure the cover glass on the 1100D sensor is epoxied on. The only way to remove it is to break it.

[Cath]

One thing I've found out about in the past week about DCRAW is that it totally discards the 2nd green pixel, just deletes it .. Never like throwing data away like that.

I sometimes hear/read people mention that their are no green objects out their in the universe, but that's very wrong. Anything that has a grey/white/yellow/orange etc colour to it contains green.

Edited December 3, 2012 by Cath

[jnp]

Excuse the late interest in this thread, I haven't been on here for ages.

This thread is indeed an epic, but I am unclear of the conclusions as this goes off topic in the 11 pages.

Can someone correct my understanding below and/or provide a management type summary (i.e. aimed at 4 year olds)?

1) A 1000d has had the bayer matrix + micro lenses ground off. The sensor works but the quality is poor due to uneveness of the sensor?

2) Gina performed a heroic quest to debayer a 1100d, but was unable due to the differences in construction, expecially the fixed glass plate.

If point 1 is true then has anyone considered using a transparent gel (ideally non electrically conductive) under a glass plate over the sensor to 'fill' the scratches? Perhaps optical cement will do it? (Aplologies if this has been suggested but I didn't see it)

I am close to getting a CCD but wondering if the 'spare' 1000d can get a new lease of life?

cheers

John

[12dstring]

This looks like fun

I have a water damaged 350D I bought a while back for spares, the main board is toast but the sensor worked fine... so it's now volunteered itself to science.

The cover glass didn't want to come off nicely in one piece, but got it all of eventually whilst holding it upside to avoid any tiny bits of glass falling on the sensor.

I'll have a play with it under a microscope next week, and see if any solvents have any effect before attempting to polish it off.

[Gina]

Good luck I suggest testing the sensor after removing the glass but before going any further.

[astromerlin]

This looks like fun

I have a water damaged 350D I bought a while back for spares, the main board is toast but the sensor worked fine... so it's now volunteered itself to science.

The cover glass didn't want to come off nicely in one piece, but got it all of eventually whilst holding it upside to avoid any tiny bits of glass falling on the sensor.

I'll have a play with it under a microscope next week, and see if any solvents have any effect before attempting to polish it off.

How about asking if there are any chemists on the forum that might suggest a solvent or chemical that would remove the matrix?

What's the substrate under the bayer matrix made of?

[Chris]

Good luck Dave:), my modded 350D is a bit long in the tooth so if you find something that works I might follow suit.

[12dstring]

Good luck I suggest testing the sensor after removing the glass but before going any further.

Thanks Gina, and testing first was a good idea.

I popped it back in and got the dreaded err 99 after taking a picture. This didn't happen with the sensor unplugged (which I know just gives a black image, but no errors).

Taking the sensor out and inspecting closely I think one (maybe two) of the gold wires aren't quite connected any longer.

I'll check with the microscope next week. It's not the end though, I've soldered smaller things...

[russellhq]

How about asking if there are any chemists on the forum that might suggest a solvent or chemical that would remove the matrix?

What's the substrate under the bayer matrix made of?

I'm sure I read somewhere that some are made from organic dyes, in which case an organic solvent cleaner like Dichloromethane or Toluene might work.

[12dstring]

I've got it under a microscope now, definitely two of the gold wires no longer connected due to the broken glass, so will attempt to reconnect with some silver glue at some point and try and bring it back to life.

In the meantime I've been plating with surgically removing the bayer filter. I tried a few solvents; IPA, acetone and xylene had no effect at all. I don't think this is the way to go here in any case.

I did some probing with the rounded edge of a dental pick. There's several distinct layers, the top one is the microlenses (visible on the left in the image below). This is very soft and easily removed by just brushing the pick along the surface, it almost feels gel-like at this scale.

With that gone you can see the darker filter array underneath (upper centre in the image). This filter layer feels glass-like, and it doesn't take much force to break a hole in the surface, and once you've done that it's easy to remove more as the edge crumbles away quite easily with further force.

Once this is removed you can see the yellowish photosites underneath. They seems to be covered by a glassy substance (planerizing layer?), it's very smooth and isn't easily damaged - though can be scratched with the sharp edge of the pick.

My feeling is that polishing will have the problem that you won't know how far you've gone through the filter array, or if you've started polishing away the layer underneath (causing the uneven flat fields?). I'll keep scraping away the filter under a microscope. Sure it'll take ages, but I can easily see what I'm doing and if I damage anything, and can be much more precise around the edges. If the filter removal goes ok but can't bring this sensor back to life, I could probably persuade myself to try on another one. I'm pretty confident I can remove the cover glass in one piece if I tried again, and for a backup would put tape over the glass first in case it breaks.

For reference the image below is around 320 microns across, or ~1/70th of the sensor width