I recently ran across a thread in another forum describing a group effort to figure out how to debayer a DSLR sensor for full-monochrome conversion. While the idea may be risky and not as good as a true, cooled monochrome CCD, it got me curious enough to give it a shot. It seemed that numerous efforts were met with success. And despite the pile of dead DSLR sensors scattered throughout the thread, I decided I was curious enough to try it myself.
I have a Canon T3i that's been astro-modified. However, it's not the same as a full-monochrome modification. The idea here is this: A DSLR has the Bayer matrix (microscopic colored filters) printed over the sensor, precisely-aligned to each pixel, so that some pixels get red filters, some get green, and some get blue. This means a color DSLR can only capture red on one out of four pixels, lower its sensitivity and resolution to the red spectrum. The same is true for blue. And green actually uses two out of four pixels. These four pixels (RGGB) basically combine to interpret the colors that we see in that portion of the sensor. If you remove the color filter array that's printed onto the sensor (Bayer matrix), this allows all pixels to gather all light. Now all 4 pixels in each set capture H-alpha, or OIII, or SII, or whatever pass filter you decide to stick in front of the sensor.
There's been some debate about the merits of doing this modification. Does it REALLY increase sensitivity to all light? Or just some light? If we scrape that layer off the sensor, it also removes the micro-lenses that improve the light-gathering efficiency of the sensor (QE). Won't that negate any gains from removing the Bayer matrix? And so on...
I was curious enough that I wanted to see for myself. Sure, it's a bit pricey if you kill a sensor. You'd better be tracking all of those tiny screws as you open up the camera! And, if you mess up? Well... get a replacement sensor and try again. Or don't. And just replace the sensor with a stock one (or astro-modified one).
I wasn't sure I wanted to risk the T3i. After all, they run $300+, even now. I ended up finding a cheaper Canon 7D to test this on. The sensors aren't cheaper than the T3i's, but the difference in price meant if I messed up, I'd still be out less money than the T3i after replacing the dead sensor in the 7D. Oh, and did I mention that I'd never opened up a camera before..?
DISCLAIMER: If you try this yourself, I'm NOT responsible for dead sensors or cameras! You do it at your own risk! I'd highly-recommend NOT doing this if you want to keep your camera in working condition.
The process... pull out the sensor, remove the IR/UV filters over the sensor, carefully remove the cover glass over the sensor chip, and then SCRAPE, SCRAPE, SCRAPE! (Sounds barbaric, doesn't it? That's not too far off, lol.)
Removing the filters wasn't too bad, but removing the cover glass right on the sensor had been tricky for some with other camera models. No one had mentioned testing this on a 7D, so this was new territory. I ended up using heat to loosen the epoxy around the perimeter (it turns white as it breaks free of the glass), and gently pried up with an X-acto blade. Very, very gently. Very, very slowly and carefully. With the heat, the glass popped right off in one piece. Nice!
After the glass was off, I removed the original UV/IR cut filters from the filter frame and used the frame itself to help guide my efforts. In particular, the efforts to avoid the ultra-thin gold wires around the two long sides of the sensor. Break one of those, and the whole thing is toast. There's also a blue perimeter around the sensor. I guess there's some super-tiny circuitry under there you really don't want to damage. So, avoid those things at all costs. The frame could help keep my tool from going into areas I wanted to avoid, as well as get a little closer to the perimeter.
I started out by scraping with a toothpick cut at an angle to make a flat surface. I figured the wood would be soft enough to avoid damaging the sensor. However, it was TOO soft. I couldn't even take off the micro-lenses. I had to upgrade to something harder... I tried the plastic handle of a sensor swab cut to a chisel point. That didn't do anything, either. I even tried adding a small amount of fine metal polish to the toothpick and it barely did anything. That's when I got serious. I filed a brass rod at an angle and used that to scrape. (This was based on someone else mentioning the use of an aluminum rivet in a similar manner. Brass, being softer, seemed a safer choice.) Boy, did it ever work. It went right through both the lenses and color filters without too much effort. It did take time to get things removed, but overall the process wasn't bad. I just scraped gently and things turned out fine. Once I had most of it scraped, minus the very outer perimeter, I cleaned up the debris using cotton swabs and rubbing alcohol.
Moment of truth... put the camera back together to see if it still works...
TADA! I see an image! It's all red now, because the camera has no idea how to interpret the data without the filter array, but it works! I need to run the RAW files through DCRaw right now (don't have Pix Insight) in order to interpret them without a Bayer matrix and convert them to 16-bit TIFF files. I need to learn the command line for DCRaw better, but this is what I saw worked for others doing the same thing, so I simply copied their command for now:
dcraw64 -v -r 1 1 1 1 -6 -T *.cr2
I took some test shots and a flat or two. Still a little debris to clean up (black specks in the image), but overall not too shabby! Here's how they turned out:
As you can see, I did end up hitting something in my efforts, causing a single line of dead pixels both horizontally and vertically. Fortunately, they're along the perimeter and can easily be cropped out of any images. They may even be removed with dithering during guiding. The corners also seem to show a little bit of something left behind. It's faint, but it's there. Perhaps an under-coating of some sorts? Anti-reflective? Not sure... Also, the removal of the UV/IR filters and cover glass threw off the auto-focus quite a bit. Even manual focusing with a lens doesn't work with far-away objects. So, I'll have to look into adjusting the sensor position to fix that, if I want to test this more in the daytime.
Now... will this be any better than a standard astro-modded DSLR? I don't know. But it was an interesting adventure. And I'm really curious to see how using a monochrome camera changes my imaging methods. This is so much cheaper than buying, say, an ASI1600MM cooled CCD that it's not a bad way to test the waters before fully-investing in a system and filters. Who knows? Maybe I'll go another step and learn about DIY TEC cooling! But, I'll save that for another time...
I also think I'll play around with imaging during daylight hours and using filters (IR pass, UV pass, etc.) to see how those types of photography work. I've never messed around in that part of photography, so it could be fun to experiment with it.
To end things here, the final purpose of this thread is to provide a place to document my imaging with this modded camera. It's been really cloudy for weeks now, so no idea when first light (first night?) will happen. But, rest assured, I plan to put this thing to the test! I have a 2" UV/IR blocking filter on the way that will sit in the imaging train. It'll start out basically shooting luminance frames. However, I also have a 6nm H-alpha clip filter to test out. So, things will get interesting!