Debayering a DSLR's Bayer matrix.

[Luis Campos]

Hello Gina,

Yes, this time was for good

I'm doing my first tests at this moment, full moon but I'm using a ha filter, It's a 1.25" but will have to do for now

when will you try your luck again Gina?

Cheers,

[pixueto]

Luis, do you think you could post a tutorial on your cooling mod?

Waiting for some mono narrowband pictures now!

Thanks

[alistairsam]

Well done Luis.

Look forward to seeing your Ha test.

Do also try an exposure with just a light pollution filter. this will simulate an "L" exposure.

I just bought a 600D, so am going to try that. only issue is that it has very small pixels, 4.7microns, so once I remove the microlenses, the sensitivity would drop.

I think for the B&W mod, we need to look at models with bigger pixels because the normal QE of these cmos chips aren't very high.

Cheers

Alistair

[Gina]

Hello Gina,

Yes, this time was for good

I'm doing my first tests at this moment, full moon but I'm using a ha filter, It's a 1.25" but will have to do for now

when will you try your luck again Gina?

Cheers,

Sorry Luis but I won't be. I've given up on DSLRs and I'm very happy with my Atik astro CCDs. The problem with the DSLR that appears best for AP, the 1100D, is that it has the cover glass welded on and it seems to be impossible to get access to the sensor without "fatal" damage

[pixueto]

I was getting impatient so I connected my 1000D batteries to the 350D battery with two wires, that charged up the 350D battery and i was able to take a few test images.

The results look good to me, although hard to say cause I didn't take the colour images.

looking closely, it obviously needs a bit more cleaning, but I'm happy with the way its turned out.

at this stage I still have the factory IR filter in place. I will replace it with the UV/IR filter and try out some images with my telescope, but since I don't have the timer for the 350d, i'll be stuck to 30s exposures. so might try a few globular clusters as they don't need the IR filter mod.

i've ordered a timer and will order the IR filter or a clear glass.

since this is more of a test, my next go will be with a newer camera other than the 1100D or the 20D.

I will also add a cold finger.

Maxim opened the CR2 files and I was able to save it as FITS and jpg's.

so no issues with processing the images.

Cheers

Alistair

Alistair, thanks for your tips; I might have a go with a spare 450D sensor I've got. I've already ordered the microscope.

Do you remember what stop, F, where those photos taken at? Just to get an idea about how dirty the sensor is. I thought about cleaning it with a blower, swabs, eclipse fluid and a specgrabber. Do you think that will work? Under the bayer matrix there is a glass/plastic surface covering the diodes and the connections, right? So it's safe to use wet swabs and cleaning fluid, right?

Thanks

Edited July 22, 2013 by pixueto

[Luis Campos]

Ok, here goes my "first light" with the mono 350D camera

This is a very poor image but nevertheless it shows what can be achieved now with th efull resolution working with a Ha filter, unfortunelly I only have a 1.25" filter and this causes some severe vigneting on the image, I croped heavilly the frame to get rid of it...

The image was captured with a full moon near by, on my back yard under a lot of street lights, poor transparency sky (high clouds), all conspiring against me to get an image

Next step will be waiting for the good old moon to go away and make same serious imaging on my dark sky spot with the 200mm f/4 newt....can't wait for that to happen!

More info on the image, don't forgte to see the full size version:

http://www.flickr.co...N08/9343677425/

Cheers,

Luís

[Luis Campos]

Luis, do you think you could post a tutorial on your cooling mod?

Waiting for some mono narrowband pictures now!

Thanks

Hi Pixueto,

I will make some work maybe next month, I have a lot to do but I will make a PDF with instructions for the cooling modd, that's a promisse

Cheers,

Luís

[Luis Campos]

Hi Pixueto,

I will make some work maybe next month, I have a lot to do but I will make a PDF with instructions for the cooling modd, that's a promisse

Cheers,

Luís

Oh...I forgot, you can see some images of the cooling modd on my Flickr gallery

Bye,

[Luis Campos]

Well done Luis.

Look forward to seeing your Ha test.

Do also try an exposure with just a light pollution filter. this will simulate an "L" exposure.

I just bought a 600D, so am going to try that. only issue is that it has very small pixels, 4.7microns, so once I remove the microlenses, the sensitivity would drop.

I think for the B&W mod, we need to look at models with bigger pixels because the normal QE of these cmos chips aren't very high.

Cheers

Alistair

Hi Alistair,

Thanks mate

Nope, I only did Ha, my LP filter is a very broad band (Baader moon & sky glow) it's great for mild LP and really gives a kick on colors but under heavy LP is not so good...

A 600D hu??!! Wow, that´s a lot of camera Yes I think you are right, the pixels are so small I think QE is a bit less with the loss of the microlenses, the 350D is a good compromise, ok it's 12 bits but it's not all that noticeabble in real life, I guess with th ecolling and amp-off modd, plus the mono sensor will be a serious bit of kit

So you will debayer the 600D?

Cheers,

P.S. Been checking you site, some serious DIY going on, that 10" f/4 is a beauty!!! Congrats mate!

[pixueto]

Looking good Luis.

So that's it? Has the DSLR monochrome mod been finally cracked?

[Luis Campos]

Looking good Luis.

So that's it? Has the DSLR monochrome mod been finally cracked?

Thanks!

That's it, the DSLR monochrome modd has been cracked finally, this all started about 8 years ago with a fellow that started to make experiences whith scraping the bayer out of digital point and shoot cameras, it was acheived with the input from all the brilliant minds from THIS FORUM!

What a great place!

Cheers,

Luís

[zahry]

Congratulations!

I'm very interested in this thread - I think mono DSLRs have huge potential especially for beginners and SCT owners with long focal lengths. I'm just working on write up about imaging and processing technique with ultra high ISO (12800, 6400) which requires minimum skills and it is showing very promising results. I'm getting photos with almost no noise, good color range and it doesn't look too bad either (considering that data acquisition rarely went over 60 min in total). http://www.flickr.com/photos/simpit/ all those photos we shot with ISO 12 800 or 6 400 without good polar alignment (just point the scope in right direction but guided) the shots from CGEM 925HD are usually 10, 60 or 120s and all photos from the 127SLT are usually 10s or less. The camera is canon 650D unmodified (just stock & no cooling - avg temp of the sensor was 15-20ºC) and my garden is in the middle of the city so I have to use Astronomik CLS filter due to bad LP. I think in Mono with Ha, OII and S filters the results would be even better. I'll share the technique as soon as I'll finish the write up - I'm trying to put there not just how to do it but how it works and why because there is a lot of myths surrounding this subject that I found (experimentally) to be completely wrong.

[alistairsam]

Alistair, thanks for your tips; I might have a go with a spare 450D sensor I've got. I've already ordered the microscope.

Do you remember what stop, F, where those photos taken at? Just to get an idea about how dirty the sensor is. I thought about cleaning it with a blower, swabs, eclipse fluid and a specgrabber. Do you think that will work? Under the bayer matrix there is a glass/plastic surface covering the diodes and the connections, right? So it's safe to use wet swabs and cleaning fluid, right?

Thanks

Hi, I'm not very familiar with photography as such so wouldn't know the stop, I think it was F5.6.

It should be clear out here on thursday, so it might get its first light then. i'll take some flats and post them.

the issue with liquid cleaners is that it almost always leaves a residue, so it'll be next to impossible to have it perfectly clean unless you have the tools, clean room etc.

main issue is with dust clinging on due to static.

I go through the surface with the microscope at max magnification, clean up whatever is visible and then semi-seal the front glass.

even maxmax mention that it won't be perfectly clean. Best bet is to always use Flats.

the photosites or the layer beneath the cfa looks like glass as I believe its a sandwich of multiple layers.

I think the blower and a fine paint brush would be the best as using a cloth would generate static and attract dust.

Good luck

Alistair

[alistairsam]

Ok, here goes my "first light" with the mono 350D camera

This is a very poor image but nevertheless it shows what can be achieved now with th efull resolution working with a Ha filter, unfortunelly I only have a 1.25" filter and this causes some severe vigneting on the image, I croped heavilly the frame to get rid of it...

The image was captured with a full moon near by, on my back yard under a lot of street lights, poor transparency sky (high clouds), all conspiring against me to get an image

Next step will be waiting for the good old moon to go away and make same serious imaging on my dark sky spot with the 200mm f/4 newt....can't wait for that to happen!

More info on the image, don't forgte to see the full size version:

http://www.flickr.co...N08/9343677425/

Cheers,

Luís

Hi Luis,

Nice work. I think there is good potential there. The moon won't make too much difference if you were imaging away from it with a 7nm Ha filter, but high clouds and transparency will make a big difference.

are you able to post a link to a single raw frame? and also the stacked file? I might have a go at processing with startools.

yes, I'm definitely attempting the 600D once its delivered.

with my 10inchF4, the image scale is 0.89arcsec/pixel so is not ideal and i'll be sampling over the seeing out here.

which is why a sensor with a pixel size of between 6.5 to 8 um would be ideal.

the 5D Mk3 is a very good candidate, but pricey. so is the 1D, 6D, etc.

I've attached a spreadsheet showing sensor data for a number of sensors that I gathered from different websites.

just update the formula in the pixel scale column to match your focal length.

I'm trying to choose a camera to mod based on the pixel size, bit depth and if available, read noise, QE and dynamic range apart from AIQ.

I think its worth considering these factors before attempting a mod as otherwise the results may not be worth it due to the inherently low QE.

I think of it like this, a decent mono ccd is upwards of $2k. QE at Ha would be around 60%, 5.7um pixels, 16-bit, cooled.

a 14-bit DSLR with say 6.5um, 30%QE, would be between $400 and $800.

so even if you damage the sensor while attempting this mod, spare sensors are usually $70 from the US. so in the end, for an outlay of under $1k, you have a decently responsive mono camera and can do narrow band with it for subs upto 7 or 8 mins.

so makes a lot of sense.

Look forward to hearing more from the "brave".

Cheers

Alistair

dslr-canon-sensors.xlsx

[zahry]

Hi, I'm not very familiar with photography as such so wouldn't know the stop, I think it was F5.6.

It should be clear out here on thursday, so it might get its first light then. i'll take some flats and post them.

the issue with liquid cleaners is that it almost always leaves a residue, so it'll be next to impossible to have it perfectly clean unless you have the tools, clean room etc.

main issue is with dust clinging on due to static.

I go through the surface with the microscope at max magnification, clean up whatever is visible and then semi-seal the front glass.

even maxmax mention that it won't be perfectly clean. Best bet is to always use Flats.

the photosites or the layer beneath the cfa looks like glass as I believe its a sandwich of multiple layers.

I think the blower and a fine paint brush would be the best as using a cloth would generate static and attract dust.

Good luck

Alistair

Hi Alistair,

try ionized gas = İonizing Air Gun - it's used in industrial processes for eliminating static charges and getting rid off dust from surfaces. There are some ways how to make it DIY. Even the scraping will be generating quite a lot of static that will attract every piece of dust in vicinity. If the static charge will remain it could be causing problems with dust even later until the charge will be neutralized (for that quite a lot of dust is required ;-). Ionized air is the way how to get rid of the problem. Increasing humidity of air also helps to reduce the problem with static and dust although it doesn't eliminate it. Here is one ionizing air gun in action http://www.youtube.com/watch?v=jx2B6ZBm7X8.

I'm hoping to start working on my sensors soon (so far I'm bogged down with work)

[alistairsam]

here's a link to the same excel file in case the attachment doesn't work

https://www.dropbox.com/s/ea2jurl3lglh2hv/dslr-canon-sensors.xlsx

[Luis Campos]

Congratulations!

I'm very interested in this thread - I think mono DSLRs have huge potential especially for beginners and SCT owners with long focal lengths. I'm just working on write up about imaging and processing technique with ultra high ISO (12800, 6400) which requires minimum skills and it is showing very promising results. I'm getting photos with almost no noise, good color range and it doesn't look too bad either (considering that data acquisition rarely went over 60 min in total). http://www.flickr.com/photos/simpit/ all those photos we shot with ISO 12 800 or 6 400 without good polar alignment (just point the scope in right direction but guided) the shots from CGEM 925HD are usually 10, 60 or 120s and all photos from the 127SLT are usually 10s or less. The camera is canon 650D unmodified (just stock & no cooling - avg temp of the sensor was 15-20ºC) and my garden is in the middle of the city so I have to use Astronomik CLS filter due to bad LP. I think in Mono with Ha, OII and S filters the results would be even better. I'll share the technique as soon as I'll finish the write up - I'm trying to put there not just how to do it but how it works and why because there is a lot of myths surrounding this subject that I found (experimentally) to be completely wrong.

Cheers,

Definitely a mono DSLR is a great tool for the low budget amateur astronomer, you can buy a used DSLR body very cheap these days, havind two identical cameras will enabble you to image from the city with narrowband filters with the MONO camera and then swap for the color camera to make your chrominace, then blend the result in your editing program, or simply have a set of narrowband filters and work just like a mono CCD...

The increased sensitivity of the mono sensor will also benfit long FL imaging with RC's and SCT's, no more shooting 10 hours to get that signal, I'm convinced that you can take out at least 30% of integration time VS a color camera, even with the filter modd.

You have same great images on your Flickr gallery, nice work

Greets,

Luís

[Luis Campos]

Hi Luis,

Nice work. I think there is good potential there. The moon won't make too much difference if you were imaging away from it with a 7nm Ha filter, but high clouds and transparency will make a big difference.

are you able to post a link to a single raw frame? and also the stacked file? I might have a go at processing with startools.

yes, I'm definitely attempting the 600D once its delivered.

with my 10inchF4, the image scale is 0.89arcsec/pixel so is not ideal and i'll be sampling over the seeing out here.

which is why a sensor with a pixel size of between 6.5 to 8 um would be ideal.

the 5D Mk3 is a very good candidate, but pricey. so is the 1D, 6D, etc.

I've attached a spreadsheet showing sensor data for a number of sensors that I gathered from different websites.

just update the formula in the pixel scale column to match your focal length.

I'm trying to choose a camera to mod based on the pixel size, bit depth and if available, read noise, QE and dynamic range apart from AIQ.

I think its worth considering these factors before attempting a mod as otherwise the results may not be worth it due to the inherently low QE.

I think of it like this, a decent mono ccd is upwards of $2k. QE at Ha would be around 60%, 5.7um pixels, 16-bit, cooled.

a 14-bit DSLR with say 6.5um, 30%QE, would be between $400 and $800.

so even if you damage the sensor while attempting this mod, spare sensors are usually $70 from the US. so in the end, for an outlay of under $1k, you have a decently responsive mono camera and can do narrow band with it for subs upto 7 or 8 mins.

so makes a lot of sense.

Look forward to hearing more from the "brave".

Cheers

Alistair

Thanks buddy

My filter is quite "broad" it's a 12nm one still get a lot of "contamination"...

Yeah, darn clouds all over, it's Murphy's law I guess I gess tonight will be clearer but still there's a lot of moon, will try another run but this time with the 200mm f/4 newt.

I will upload a RAW frame latter and also the callibrated image, no probs.

Can't wait the see the result of that monofied 600D, good luck with that mate

About the more recent canon models, they claim better QE, but I think it's because of microlensing caracteristics and not the sensor itself...not sure though :/

You thinking about the DSLR vs CCD's is absolutelly correct, I sure hope many more will folow this "new trend" as this is like a small "revolution" underway Thanks to the brave ones!

Cheers,

Luís

[pixueto]

How do you think the lost of microlenses will affect the images? Do you think the resolution will be similar to that of a camera with the Bayer matrix with its colour pictures interpolated through algorithms? Will it be even better considering the anti-aliasing filter is removed? If the image quality is similar, that would be good enough as we would still be using 4 pixels for the narrowband sessions.

Luis, any chance of you posting a flat frame to finally convince us all that the paint brush mod is the way forward?

Thank you

[Luis Campos]

How do you think the lost of microlenses will affect the images? Do you think the resolution will be similar to that of a camera with the Bayer matrix with its colour pictures interpolated through algorithms? Will it be even better considering the anti-aliasing filter is removed? If the image quality is similar, that would be good enough as we would still be using 4 pixels for the narrowband sessions.

Luis, any chance of you posting a flat frame to finally convince us all that the paint brush mod is the way forward?

Thank you

At your request Sir.

http://www.flickr.com/photos/37419943@N08/9344497729/

The resolution is certanlly improved, not only because there's no CFA but also because the stock filter is gone (mine has a Baader filter so it's gone for a long time) the micro lenses loss is no problem, the removal of the CFA more then makes up for this, it's like gaining maybe a full f/stop...that's not a little thing

But yes, the BIG advantage is beeing able to work with NB filters with FULL RESOLUTION.

Cheers,

Luís

[pixueto]

Brilliant Luis! the flat frame seems perfectly fine and pretty clean too! What F was that picture taken at? Only a bit of extra work is needed to remove a few bits left of the array. Congratulations! I still can quite understand how the bayer array can be scrapped away without scratching the glass/plastic protecting the connectors and photodiodes but that flat frame is the proof!

Any suggestions on how I should start scrapping the bayer matrix? I suppose once you have some bits removed it will be easier but I'm scared about the first steps.

I noticed that you got some scratches in a tiny area. Was it there where you started removing the array? I have a bricked 350D with an intact sensor that I can give you if you want to have another go with a new sensor.

How did you remove the glass covering the sensor?

Edited July 23, 2013 by pixueto

[Luis Campos]

Hi Luis,

Nice work. I think there is good potential there. The moon won't make too much difference if you were imaging away from it with a 7nm Ha filter, but high clouds and transparency will make a big difference.

are you able to post a link to a single raw frame? and also the stacked file? I might have a go at processing with startools.

yes, I'm definitely attempting the 600D once its delivered.

with my 10inchF4, the image scale is 0.89arcsec/pixel so is not ideal and i'll be sampling over the seeing out here.

which is why a sensor with a pixel size of between 6.5 to 8 um would be ideal.

the 5D Mk3 is a very good candidate, but pricey. so is the 1D, 6D, etc.

I've attached a spreadsheet showing sensor data for a number of sensors that I gathered from different websites.

just update the formula in the pixel scale column to match your focal length.

I'm trying to choose a camera to mod based on the pixel size, bit depth and if available, read noise, QE and dynamic range apart from AIQ.

I think its worth considering these factors before attempting a mod as otherwise the results may not be worth it due to the inherently low QE.

I think of it like this, a decent mono ccd is upwards of $2k. QE at Ha would be around 60%, 5.7um pixels, 16-bit, cooled.

a 14-bit DSLR with say 6.5um, 30%QE, would be between $400 and $800.

so even if you damage the sensor while attempting this mod, spare sensors are usually $70 from the US. so in the end, for an outlay of under $1k, you have a decently responsive mono camera and can do narrow band with it for subs upto 7 or 8 mins.

so makes a lot of sense.

Look forward to hearing more from the "brave".

Cheers

Alistair

Hi Alistair and friends,

Last night I had a go at M27 under a full moon with the mono 350D, good transparency this time but very windy spoiling autoguiding :/

I also have a dear neighbor that left his 300 watt stadium light on during most of the capture time...what can i do...the guy was grilling fish at allmost midnight!

Anyways...I'm literally BLOWN AWAY with perfromance of the mono sensor, with 1 min. subs at 400 ISO with the LP filter I was near the sky glow limit, ok it's a full moon but I'm quite certain that comparing with the color sensor I have at least doubled the sensitivity, only future tests will tell

This M 27 image is a stack of 37x1 min. subs @ 400 ISO callibrated only with BIAS and FLATS, no darks were used, I'm attaching the RAW stack and a quick and dirty stretch under PS just to show what's there, there is no noise reduction whatsoever on the image, the stars are a bit weird due to the wind.

The M27 RAW stack:

https://www.dropbox.... (NO DARKS).tif

M 27 JPEG quick processing test (needs a run of gradient Xterminator):

https://www.dropbox.... (NO DARKS).jpg

This is NGC 6888 RAW STACK first test with the ED80 and a 1.25# 12nm Ha filter:

https://www.dropbox....00 ISO ED80.tif

This is a single sub NGC 6888 with the ED80:

https://www.dropbox....00 ISO ED80.tif

And this is some fruit

https://www.dropbox....3tdg/220713.TIF

Please have a go with this and let me know your oppinion

Many thanks,

Bye,

Luís

Edited July 24, 2013 by Luis Campos

[Luis Campos]

Brilliant Luis! the flat frame seems perfectly fine and pretty clean too! What F was that picture taken at? Only a bit of extra work is needed to remove a few bits left of the array. Congratulations! I still can quite understand how the bayer array can be scrapped away without scratching the glass/plastic protecting the connectors and photodiodes but that flat frame is the proof!

Any suggestions on how I should start scrapping the bayer matrix? I suppose once you have some bits removed it will be easier but I'm scared about the first steps.

I noticed that you got some scratches in a tiny area. Was it there where you started removing the array? I have a bricked 350D with an intact sensor that I can give you if you want to have another go with a new sensor.

How did you remove the glass covering the sensor?

Thanks mate

The flat was taken with the ED80 (Sky Watcher) at f/7.5 with a white t-shirt.

The scraping method with the wooden tool was discovered by fellow Alistair, and this works beautifully, the first few passes do nothing to remove the CFA but I used a small hipodermic needle to gently crack open the CFA and begun sraping with the wooden tool from there, it's easy once it starts to crack, what happen is that once there is some CFA material embeded in the wood tip itself acts as a abrasive to remove the rest of the filters, it's very much efective!

The most dificult part is to remove the cover glass, on my practice sensor it was easy and only broke in half, but on the working one it was a nightmare, I folowed the tip from one of the friends here, and put some tape over the glass before I cracked it into pieces, worked fine, but you have to be VERY carefull not to scratch the sensor under it!

Are you serious? You have a spare working sensor you can give me?

I will PM you soon...

Thanks mate!

Cheers,

[pixueto]

Sure Luis, the 350D is yours if you need it. I bricked it when a screw got stripped and tried to force it a bit. You may need some good pliers to release the motherboard and free the optical train but the sensor should be intact.

Well, I'm now waiting for the microscope but I had a go at removing the glass covering the sensor in the 450D and managed to release it pretty much intact. I think I have now a procedure to do it more or less safely but it will take no less than 30 minutes. Also, I cleaned that piece of glass with swabs and eclipse fluid and it worked as expected so it's ready to be reattached when the bayer array is removed (if successful). We'll see how difficult is to remove that array in the 450D. By the way, I hadn't realised how tiny the sensor is. Which paint brush did you use guys? Have you got a picture or a link? I'm terrified at the prospect of hitting those gold connectors!

Edited July 24, 2013 by pixueto

[Luis Campos]

Sure Luis, the 350D is yours if you need it. I bricked it when a screw got stripped and tried to force it a bit. You may need some good pliers to release the motherboard and free the optical train but the sensor should be intact.

Well, I'm now waiting for the microscope but I had a go at removing the glass covering the sensor in the 450D and managed to release it pretty much intact. I think I have now a procedure to do it more or less safely but it will take no less than 30 minutes. Also, I cleaned that piece of glass with swabs and eclipse fluid and it worked as expected so it's ready to be reattached when the bayer array is removed (if successful). We'll see how difficult is to remove that array in the 450D. By the way, I hadn't realised how tiny the sensor is. Which paint brush did you use guys? Have you got a picture or a link? I'm terrified at the prospect of hitting those gold connectors!

Many thanks mate, it's very kind of you

I'll get in touch so we can arrange things

I'm very curious to know about the 450D you have, glad you could remove the cover glass intact, it's a good start.

I have some videos showing the scraping procedure and showing the wooden "tool" I'll upload it tonight on You Tube for you to see

FYI I've done my sensor with no optical aid, just a pair of glasses lol, yeah those tiny gold wires are nerve wrecking...please be xtra carefull when you pass near them, dont drink any cofee before you start!

What I used to check for CFA leftovers was a 25 mm eye piece inverted it magnifies quite a lot if you look through "the other side" and place it near the sensor just like a watch technician does.

Yesterday I had another test run, I reassembled the TEC cooling unit and heating on the sensor chamber, here's what I learned...as expected, the sensor surface was covered in moist after a 10 min. run with the TEC on, this tells me that when removing the cover glass the sensor is no longer air tight, and when reassembling the filter even if you place some silicone all around, like I did, it will not function with the cooling method (dew problems), the cure for this will be removing the sensor again, remove the filter and seal the unit with hot glue, then insert a needle in the CMOS chamber and purge with Nitrogen, luckily I have a Nitrogen supply (I'm a bicycle mechanic and I use this to service the rear shocks) so I will make a small modification on the filter holder to purge the thing and hopefully end this problem.

Cheers,

Luís

Edited July 25, 2013 by Luis Campos