Debayering a DSLR's Bayer matrix.

[Astroscot2]

Nightmare, sorry to hear that Gina

[Gina]

Nightmare, sorry to hear that Gina

Thank you

Yes, adding the epoxy was the problem. Whether it's the epoxy itself or whether I inadvertently got too close or even touched the wires I don't know. The epoxy is very viscous and I think it possible that it broke a connection on one (or more) of those very fine and delicate gold wires. I couldn't see any damage with a powerful magnifying glass but it only needs a break of a micron or two to stop it working. JTW seem to use epoxy successfully but there could be a difference between theirs and mine.

Edited August 30, 2013 by Gina

[alistairsam]

yes I dont think its worth risking it with the epoxy

if you start scraping from the middle with the lowest mag on the microscope you should be able to control the hand movements so you avoid the wires.

I just got my 600d replacement sensor so will be try the heat gun to get the glass off and a wooden tool.

the sensor unfortunately has a column defect so there's a line but it'll do.

pixueto, I'd suggest keeping subs to 8 mins max to keep noise manageable

cheers

Alistair

[StuartJPP]

As mentioned, it could be the shrinking of the epoxy during curing that is causing the bonding wires to break.

Some epoxy will not do this, such as chip on board epoxy (black blobs on some PCBs).

[Gina]

I used to have very steady hands and fine control of my fingers and used to work on PCBs with 50 mil spacing (0.05") but the irreversible passage of time has had it's toll on my body and particularly hand and finger control I'm not sure I'm up to this any more. "The heart is willing but the flesh ie weak"

[Luis Campos]

Thanks Luis. Could you elaborate on this?

'next input a generic name, this is what I do, I put "A" on input and "B" on output generic name and so on, so the previous output generic name will be the input generic name on next step, go for a, b, c, d ,etc...'

Can you explain those steps? Not sure what I'm supposed to do with input the previous output?

I have now the ED80 out with the 1000D debayered and a 7nm H-alpha 2" filter. I'm taking 20 minutes subs. We'll see what comes out.

Hi,

Means that the output name from the previous step (A for ex.) will be the input on the next step.

So you start with a input name like "IMG" on "Digital photo / preprocessing" and output name will be "A", so next step is "processing / stellar registration" in this step input name will be "A" and output will be "B", next step is "processing / add a sequence" so here input will be "B" and that's it from here you just save the data as TIFF and load into Photoshop for further processing.

Cheers,

[Gina]

I shall now have to think carefully where I go next. If I work on the second 1000D sensor and kill that, I shall no longer have a working 1000D to check which parts of the bought-as-duff camera are working and without being able to do that I couldn't sell any working parts as in working order. So I would prefer not to do anything to the working sensor that might harm it. I think what I'll do is to swap parts from the duff camera to the working one and set the working bits aside for selling later. Having destroyed one of the 1000D sensors I have no way of making up two working cameras other than to buy a used sensor from the USA.

There is something, CFA wise, that I can do without further commitment - try removing the CFA from the dead sensor. Practice working in the centre and staying away from the edges.

Edited August 30, 2013 by Gina

[Gina]

Been playing with the dead 1000D sensor. The abrasive simply won't touch the CFA unless it's gouged with a metal implement - I can't find any wood that will touch it. So this CFA seems just like the 1100D sensor - very tough. I've used soft copper but I think there's still a good chance of going too deep. The CFA does come off with the abrasive on a cotton bud once the CFA has been broken first. I guess I could try another abrasive to get things started.

Later... Tried various household cleaning products. Most did nothing but Ajax + water took the CFA off quite easily and with care I managed to stop before going right through the CFA. I was then able to remove more CFA with the paint cutback abrasive leaving a smooth surface.

Edited August 30, 2013 by Gina

[Gina]

I think I'll start a new thread for fault finding on the bought-as-duff 1000D as it isn't relevant to this thread and I apologise for going off-topic.

[Gina]

Having carried out various tests on the faulty camera I find there is more than one fault and concluded that swapping parts between the good and the bad risks damaging the good camera. That would defeat the whole purpose of buying it which was to use the sensor for these debayering experiments. Hence I have suspended fault finding and will be "playing" with the working sensor. If I destroy this one I still have the other which, while not taking a picture still lets the rest of the camera function and could be used for fault finding.

Now I have to decide on how I shall proceed. Some things I have already decided. I shall work carefully around the cover glass with the tip of a craft knife until the white has spread right up to the inner boundary of the bond. Then I think I'll stick a sticky pad on to stop and fragments of glass dropping onto wires or sensor in case it breaks up. OTOH I found that when I put the filter frame back over the sensor for testing and then took it off again, that bit of glass that I couldn't shift had stuck to the sticky on the filter frame and had come off the sensor frame. So maybe I'll just go as far as the whitening part and then put the filter frame back and see if the whole glass sticks to the filter frame.

Edited August 31, 2013 by Gina

[Gina]

Worked on the cover glass bond until it was all white. Did break a corner off though but not as far as the inside. Then put everything back together for testing. Had a bit of a fright at first - camera appeared dead as a dodo. So I took the back off again and went round the connectors making sure they were all properly seated and put back and battery back for another test. OK this time. Nice pic of my OH Phew!

[Gina]

It worked - the cover glass came off on the filter frame as I hoped. So I now have the sensor clear for CFA removal but OTOH those fine fragile wires are exposed.

I can apply abrasive in the middle of the sensor, staying well clear of the edges and as long as I leave a large margin I should be alright. Trouble is the cleaning off - I used a cotton bud and copious amounts of IPA, drying off with the other end of the cotton bud.

Later... I have now removed a small area of CFA in the middle of the frame using abrasives and cleaned up with IPA without going near the edges. Put everything back together and it's still working Here's the proof. The second is a flat stretched a lot in Ps.

Edited August 31, 2013 by Gina

[foundaplanet]

Hope you can crack it Gina....Good luck..

[Gina]

Hope you can crack it Gina....Good luck..

Hope you can crack it Gina....Good luck..

Thank you It's coming on - so far so good But how long will it be before I have a twitch and break a wire or two

[Gina]

Some more CFA removal.

[Gina]

Here's another pic and a Photoshopped version, stretched and cropped. The debayered area looks pretty good to me.

[JohnC64]

fingers crossed here

Edited August 31, 2013 by JohnC64

[Gina]

Thank you John

[Gina]

Last go for today - I'm not pushing my luck any further as I get more tired I've only cleaned the area where I've removed the CFA hence all the trash round the edge. The last image shows the flat cropped and stretched in Ps.

Edited August 31, 2013 by Gina

[pixueto]

yes I dont think its worth risking it with the epoxy

if you start scraping from the middle with the lowest mag on the microscope you should be able to control the hand movements so you avoid the wires.

I just got my 600d replacement sensor so will be try the heat gun to get the glass off and a wooden tool.

the sensor unfortunately has a column defect so there's a line but it'll do.

pixueto, I'd suggest keeping subs to 8 mins max to keep noise manageable

cheers

Alistair

I tried 10 minute subs and the histogram was to the left at the edge; even with 20 minute subs, the histogram was at about 1/4 to the left of the horizontal axes. It goes without saying that the noise was terrible.

This is new territory for me guys. I always try to get the histogram at about 1/3 to the left of the horizontal axis. Alistair, Luis, if I limit the exposures to 8 minutes, is that going to be good enough? As I said, the histogram was completely to the left.

Thanks

[pixueto]

Thank you Luis for your help with IRIS processing. I've eventually managed to process those images. I have processed three 20 minute subs (that's 60 mins total) with the debayered canon 1000D, the SW ED80 with focal reducer and a 7nm Baader 2'' H alpha filter. I've included 7 bias, 7 darks and 7 flat frames. This is the result. I'm new to all this narrowband thing so, I suppose the Tiff file could have been far better processed. I can leave a link to dropbox to those interested in stretching the file.

I went for the knot in the Heart nebula, as I wasn't sure if I would be able to frame the whole nebula with my partially debayered sensor.

Gina, I think the control frames have eliminated the imperfections in the sensor.

I would very much welcome your comments.

Edited September 1, 2013 by pixueto

[alistairsam]

Last go for today - I'm not pushing my luck any further as I get more tired I've only cleaned the area where I've removed the CFA hence all the trash round the edge. The last image shows the flat cropped and stretched in Ps.

Hi Gina,

How exactly are you removing the CFA? are you using a scraping tool at all or is it the ajax on a cotton bud?

I'd suggest caution with the bud cause you could easily knick those wires cause its broad, and you won't be able to get to the edges.

as for removing the cover glass, yes you use pressure with a small screwdriver or blade inward slowly till it gets white right to the inside of the frame and you contnue around.

one of the guys in IIS took a 350D sensor, scraped away the edges of the glass, apparently that removed the glue at the sides, and then was able to remove the glass the same way, that is pressure inward till it turned white and removed the glass in one piece.

i think that's the first time anyone's taken off a 350d intact. but I might try that as well although I know the glue doesn't bleed off the glass edges.

with testing the sensor at various stages, isn't that risking the ribbon cable connectors? i flipped one on the 350d and the clip came off. couldn't put it back.

wish all the ribbon cables were the slide in type like the 600d.

good luck.

Alistair

[alistairsam]

Thank you Luis for your help with IRIS processing. I've eventually managed to process those images. I have processed three 20 minute subs (that's 60 mins total) with the debayered canon 1000D, the SW ED80 with focal reducer and a 7nm Baader 2'' H alpha filter. I've included 7 bias, 7 darks and 7 flat frames. This is the result. I'm new to all this narrowband thing so, I suppose the Tiff file could have been far better processed. I can leave a link to dropbox to those interested in stretching the file.

I went for the knot in the Heart nebula, as I wasn't sure if I would be able to frame the whole nebula with my partially debayered sensor.

Gina, I think the control frames have eliminated the imperfections in the sensor.

I would very much welcome your comments.

Hi,

Could you please post links to the raw files and the darks?

I use maxim to convert the raw files to tif and use DSS to stack them, then startools to process.

I'd like to try out your data.

as for subs, yes, 8 mins at 400 iso would be manageable and about the limit for a 1000d i'd think.

also depends on your location.

if the site is dark, then you could keep the subs within 4 mins at 800 iso but take more of them so you increase the snr after stacking.

one other way to determine is to measure the background ADU.

so take a 3 min sub, measure the background ADU value in maxim or other software, enter the value in the ccd tools website and it'll give you your ideal exposure length after which only background values increase without significant increase in actual signal of the DSO.

this works very well for CCD's and I use it very often. i've also experimented with subs of different length and right enough, after a certain point, there is hardly any difference in longer subs.

with dslr data, i'd suggest you test subs of various length and iso for a given night, and try and determine the optimum length where noise is not too bad. noise is mainly dependent on iso, temp and length.

Cheers

Alistair

Thanks

Alistair

[Gina]

Hi Gina,

How exactly are you removing the CFA? are you using a scraping tool at all or is it the ajax on a cotton bud?

I'd suggest caution with the bud cause you could easily knick those wires cause its broad, and you won't be able to get to the edges.

I haven't used a scraper at all. I started with an aqueous solution of Ajax in the middle of the sensor to break the surface of the CFA. I just barely went through it in a small area. Then I washed that off with IPA and continued with the finer abrasive of the "Scratch X 2.0" fine scratch and blemish remover for car paintwork. Both were applied with cotton buds. I shall need to sort out a narrower and more square ended device for applying abrasive to get nearer the edges and ends. I think I have some sensor cleaning swabs that I got for cleaning DSLR sensors and I'm thinking of cutting one of those down to form a narrow, square ended tool.

as for removing the cover glass, yes you use pressure with a small screwdriver or blade inward slowly till it gets white right to the inside of the frame and you contnue around.

one of the guys in IIS took a 350D sensor, scraped away the edges of the glass, apparently that removed the glue at the sides, and then was able to remove the glass the same way, that is pressure inward till it turned white and removed the glass in one piece.

i think that's the first time anyone's taken off a 350d intact. but I might try that as well although I know the glue doesn't bleed off the glass edges.

Yes, that's the technique that I've been using on the 1000D and the 350D. It certainly helps to gouge out the glue from under the edge.

with testing the sensor at various stages, isn't that risking the ribbon cable connectors? i flipped one on the 350d and the clip came off. couldn't put it back.

wish all the ribbon cables were the slide in type like the 600d.

Yes, it is but I take great care with these as I know they are rather delicate. OTOH the enormouse amount of practice I have had at this helps but I'm still careful

good luck.

Alistair

Thank you Definitely getting to the dangerous part now Edited September 1, 2013 by Gina

[Gina]

one other way to determine is to measure the background ADU.

so take a 3 min sub, measure the background ADU value in maxim or other software, enter the value in the ccd tools website and it'll give you your ideal exposure length after which only background values increase without significant increase in actual signal of the DSO.

this works very well for CCD's and I use it very often. i've also experimented with subs of different length and right enough, after a certain point, there is hardly any difference in longer subs.

I haven't heard of that before - seems a good idea - thank you I take 5m subs and see what the stars and DSO looks like as well as the background and guess at suitable length subs from that using what experience I have gained so far. This has worked fairly well so far though sometimes I've had to adjust the exposure. eg. with my widefield rig imaging of the Heart & Soul Nebulae, I found I had to increase the exposure to an hour using 105mm lenses at f4 and Astrodon 3mm and 5mm filters for OIII and SII. Plenty of Ha and the camera and lens for that (135mm at f2.8 and Atik 460EX) gave good results at the estimated exposure of 20m.