Debayering a DSLR's Bayer matrix.

[LarryC]

Herra,

I took those with my iPhone.  Here's a set I took tonight with my D800.  I probably won't get a chance to swap sensors in my 55D until this weekend, but I will post the results when I do. 

 

 

[LarryC]

Well,  Sadly It looks like I must have killed it.  Everything powers up, but I can only get black images.  Re-installed it a second time and same result.  Put he original LPF-2 modded sensor in and it works fine. 

Don't know what I might have done to kill it.  Truth is, though, I don't really know it  ever worked (though I assume it did).  I ordered it off the bay as a manufacturers refurb and didn't install it to check it's function.  Will do it differently next time!

Off to hunt for another sensor...

[Gina]

I test my sensors before work and between each stage to check all's well.

[Herra Kuulapaa]

But no err-message? Sounds like bonding connector failure (golden wire).

One failure source is scraping/polishing too deep. Canon sensors are generally more prone to that than Nikon ones.

 

[RFE3]

Hi Everyone,

    First Post, but a long time lurker. I've been following this Super thread for a while now, I have to say there are some real Pioneers in this bunch....My Hat's off to Gina, your determination and patience are borderline Awe inspiring..... After reading this thread through twice and all of Gina's Modding threads, I decided to give this a try. I have several Canon's laying around (350D, 2x 1000D and a 450D) and I decided to try this Mod. I picked the 350D as the Victim because it hasn't really been used in years. The two 1000D's are full spectrum modded (by me) and are what I currently use for AP, the 450D is my "normal Camera".

     The process went rather smoothly (to my relief), dismantled the camera, removed the cover glass (in two pieces), epoxied the "Golden Hairs of Satan", polished the sensor with a liquid plastic polish and two different Felt tips (Conical and Straight) in the Dremel. The hardest part was determining where to stop polishing. Every photo in this thread looks a little different, some more golden, some more copper, some greenish... I tried to polish down to the Golden layer but depending on how I held it in the light it reflected some green. If i tried to lightly polish the green away it became more copper in color. In the end I have some places that still have the Golden/Green tint and some with the Copper tint. After putting it all back together (Luckily) it all still works. The Sensor is still a bit dirty and there is no glass in the optical path (cover glas or IR filter). I've taken some photos with it and I quess it's Debayered. If I take a JPEG it's just Red much like my Full Spectrum 1000D's but different. If I open the CR2 file in PixInsight it's a lovely Wine Red color before applying the  STF function, then it turns sort of Orangy/Red. If I open the Raw (in PI) and save it as a Fits file, then open it in Fits Liberator, do a little streching and save it as a TIFF it looks completely different. I can then open the TIFF file in PI or PS and it's a nice Grey Scale image. If I convert the Grey Scale image to RGB and then run Channel Extraction, all of the Channels (R, G, & B ) are B&W.  I  assume this means it's now Monochrome.

     I've put up a copy of the CR2 and Fits files for the Flat IMG_1200 and another Carbon Fiber shaft IMG_1201 in Dropbox. If someone could please have a look at them and and confirm that they are indeed Monochrome I would be grateful.

https://www.dropbox.com/s/cia79i5wvu2nrqi/IMG_1200.CR2?dl=0

https://www.dropbox.com/s/umf3mrlcqnmcsyq/IMG_1200.fit?dl=0

https://www.dropbox.com/s/zk2hy878jwmuvuc/IMG_1201.CR2?dl=0

https://www.dropbox.com/s/743uhe4aer8p5n2/IMG_1201.fit?dl=0

     If this one is OK then I'll try one of my 1000D's. I've been really wanting to get into Narrowband stuff for a while now. Once again a huge BRAVO to all of you who have gone before.....

many thanks,

Rick

 

 

 

 

 

 

[Herra Kuulapaa]

There is something funny in the CR2 files. Please enable pure raw in Pixinsight and open files there zooming fully in. There is RGB matrix remaining if you see a pixel pattern.

Image should look like this: Canon 1200D raw sample

[RFE3]

Thanks for having a look Herra,  Mine definatly don't look like yours, maybe I messed up the original CR.2 files with FL or PS. When I open the files they are either already Grey Scale or they have that funny Grain and are Blueish. I'll take some more photos and open them directly in PI with the Pure Raw enabled.  The FIT images look great, but maybe they are artifically Black and White. I expected the photos would look simmilar to the ones I take with my Full Spectrum Modded 1000d's but they don't. Is the photo in your link from a CMOS Sensor or a CCD Sensor?

[JanH]

I recently noticed that glass removal (soldering iron method) works much better when the sensor is frozen. At least on sony Nex5 14mpix cmos sensor it worked perfectly (it was struggle every time before on that type of sensor and glass from this last one came of in one piece and undamaged). This sensor is not suitable for polishing method though (leaving stains, waves... not sharp but disrupting structures visible in narrow band).

Now I'm having some trouble processing files. This time I managed to use DCraw -> DSS method (still don't know why it rotates some frames = I have to rotate them back manually before feeding DSS). Stacking went well (decently) but there are some FLAT issues. I have some lighter strokes visible on the sensor (and "waves" along longer sides) and flatening doesn't touch it. I must say that I don't  dither as I didn't work this out yet (in Lin_Guider).

And another issue - I can't open files generated by DSS (never had this problem before with colour images or super pixel grayscale or Bayer Drizzle). I tried to save it in different formats (tiff 32, 16, rational, integer) and some works at least in image viewer but none of them is recognized by RawTherapee or GIMP.

[JanH]

Here is 2hour exposure of horsehead taken with mirrorless Sony Nex5, 14mpix sensor, Baader 7nm Ha filter, smc A* 645 300/4 10min frames, 2hour total, flats, darks, bias, Processed in DSS as superpixel (thus 3.5Mpix = full resolution). Notice vertical lines along the left edge and slightly also on the right and there is massive inconsistency in bottom 20% of the frame. There is more detail in the files (I hope) just to get it. I overstretched the file to make it obvious.

It's pitty as right now I have just 2 cameras, much better Nex5N (very flat flats and overall better sensor) is on 8" newton and on 300mm lens is older nex5 but I'm constantly messing up with that newton so this telelens was piggybacked to tube ring and guiding with OAG on newton (shooting M42) but as I discovered after 2 hours of clear sky Newton was out of focus!? Shooting at 300mm FL and guiding at 800mm FL. Hmmm.

 

[Gina]

Very good

[JanH]

and matching flat.

[Stub Mandrel]

4 hours ago, JanH said:

And another issue - I can't open files generated by DSS (never had this problem before with colour images or super pixel grayscale or Bayer Drizzle). I tried to save it in different formats (tiff 32, 16, rational, integer) and some works at least in image viewer but none of them is recognized by RawTherapee or GIMP.

PSP and Corel Draw will open both 8 and 16 bit TIFF, Integer but not the 32 bit or rational types IME. Turn off compression too for maximum compatibility.

[Stub Mandrel]

Try prodigital astronomy tools vertical and horizontal line removal.

 

[Herra Kuulapaa]

2 hours ago, JanH said:

and matching flat.

Hi Jan, It looks like residue on your sensor surface. I think you can push your polishing further.

Exmor sensors tolerate heavy punishment on their surface..

[JanH]

I don't know. But I'm definitely not touching this sensor any soon. There are two working chips around with same issue (but worse, one or two pages back) and on one I tried more polishing  without making it better. I felt as I'm digging through layers, some are brighter, some dimmer, stains were moving but not disappearing. I can try more, no problem, it's gonna be cloudy for rest of the season anyway (dark sky departs within a month). Strange that newer sensor (16mpix in Nex 5N) doesn't do it, CFA is tougher but once it's gone there is decently flat surface (I can see some brighter areas but it's comparable to normal falloff).

 

And since there is little hope for clear sky any soon I processed Heart nebula also (the last image I have...). Unfortunately just  5hours Ha, 45min OIII and that's it. Ha=red, Oiii=blue, Ha 25%+Oiii75%=green.

[LarryC]

Well,  I have my second experimental T2i sensor on it's way to me.  I'm not sure what I'll try different this time - scraping or polishing- but I will test it at every step as Gina suggested. 

[Stub Mandrel]

> I'm not sure what I'll try different this time - scraping or polishing

The popular approach seems to be scrapping.

Sorry...

[Gina]

Hmmm...

[Gina]

Regarding to scrape or to polish, I've tried both and found polishing better (on Canon EOS sensors) with less chance of damage to the sensor.

[RFE3]

I had finally got a chance to do a little more work on the 350D, there was another color layer under where I stopped last time. It went from a Goldish layer to a Copperish layer and the image was pretty uniform but then I went back and tried to even it out a bit more in one spot and welll.....I should have left it as it was.....I went a tad too deep and got into the sensor layer. The Chip sitll functions but has an area that is not registering.  Luckily I was able to purchace on Ebay two more 350D sensors from the States for 10USD each.  One down, two to go.

Very overexposed but easy to see the non functional area of the sensor.

 

I have another question about shooting with these Mono Modded DSLR's and Filters. For shooting Luminance Frames do I need to use a UV/IR Cut Filter as the Luminance Filter?

Something like this: http://www.teleskop-express.de/shop/product_info.php/language/en/info/p572_Baader-2--Infrarot-Sperrfilter-fuer-CCD---UV-IR-Sperr--und-Luminanz.html

Would I also need the Cut Filter in place when using a NB Ha Filter (7nm) or just the NB Filter?

I have removed the Canon Low Pass Filter from the Camera and was planning to reseal the Sensor with only the Edmund's Optics AR Glass.

Edited March 22, 2016 by RFE3

[LarryC]

On 3/22/2016 at 02:42, Gina said:

Regarding to scrape or to polish, I've tried both and found polishing better (on Canon EOS sensors) with less chance of damage to the sensor.

My plan was to scrape but with a different type of wood or plastic tool, but I did notice that it was fairly easy to clean up areas by hand polishing that I was not able to scrap clear, so maybe I'll start with polishing and see how that goes, then revert to scraping if necessary. 

 

I have a large biologics laboratory and we have some interesting plastic scrapers intended to remove cells from plastic surfaces without killing them.  I was thinking of tying one of those.

 

 

[Herra Kuulapaa]

Yes. you need a IR/UV cut, but not when you're using already a narrow NB.

Edited March 23, 2016 by Herra Kuulapaa

[LarryC]

Well, it's a sad day indeed.   I received the new ebay sensor module  and replaced the module in my current 550D.  After reassembly and power on, I got a menu and could navigate around fine, but when I pressed LV I got an image for a second or so then the screen went to the first image below and after 5 or so seconds, went black and the red activity light came on and stayed on, and would not shut off even after I turned off the camera.  I had to pull the battery.   Turn it back on again and there's the menu just fine but it says something like "failed to properly shut down, not loading module".  If I try to take an image in any mode, the shutter and all seems to work, but there is no image the the red light again comes on permanently and I cannot turn off the camera.

So, I take a close look at the new module and notice that one of the grounding straps is broken off (second photo - notice the upper left screw, you can see the end of the broken piece).  Looking closely at the ebay photo, sure enough the ground strap is broken.  It was easy to miss because the break is clean and in general the two modules look the same. 

I put my "good" sensor module back in and I get the same results.  I looks like I might have cooked my camera?  Could the missing grounding strap have killed it? 

 

 

Edited March 26, 2016 by LarryC

[Herra Kuulapaa]

I dont think it's the "missing" grounding. There should be a connection from the sensor metal frame to camera body without it as well. Error sounds to me somekind of connector issue. Is the sensor main connection cable intact?

[Stub Mandrel]

My bet is ribbon cables needing playing with. Very much doubt that you could cook the main camera with a dodgy sensor.

Make sure the connections are fully home, it's so easy to not have those flip-ribbon connectors not quite right.