Digital cameras have been around for forty years or so, and the first ones were built around CCDs. These were two-dimensional CCDs, and if you’ve ever looked inside a copier, scanner, or one of those weird handheld scanners from the 90s, you’ll find something entirely unlike what you’d see in a digital camera. Linear CCDs are exactly what they sound like — a single line of pixels. It’s great if you’re into spectroscopy, but these linear CCDs also have the advantage of having some crazy resolutions. A four-inch wide linear CCD will have thousands of pixels, and if you could somehow drag a linear CCD across an image, you would have a fantastic camera.
Many have tried, few have succeeded, and [heye.everts]’ linear CCD camera is the best attempt at making a linear CCD camera yet. It took a fuzzy picture of a tree, which is good enough for a proof of concept.
The linear CCD used in this project works something like an analog shift register. With a differential clock, you simply push values out of the CCD and feed them into an ADC. The driver board for this CCD uses a lot of current and the timings are a bit tricky but it does work with a Teensy 3.6. But that’s only one line of an image, you need to move that CCD too. For that, this project uses something resembling a homebrew CD drive. There’s a tiny stepper motor and a leadscrew dragging the CCD across the image plane. All of this is attached to the back of a Mamiya RZ67 camera body.
Does it work? Yes. Surprisingly yes. After a lot of work, an image of a tree was captured. This is an RGB CCD, and at the moment it’s only using one color channel, but it does work. It’s a proof of concept rendered in a 2000 x 3000 grayscale bitmap. The eventual goal is to build a 37.5 Megapixel medium format camera around this CCD, and the progress is looking great.
Matthias Wandel had a pretty good one back in the day!
http://www.sentex.net/~mwandel/tech/scanner.html
I had forgotten how good his results looked – I’d be willing to bet his build (as it so often is) gets better results than this one!
How about doing it the VCR way? Mount 2-4 CCD sensors on a drum, spin it very fast and get the data the way VCRs did it? Just add a special lens or use post-processing to correct the image distortion. With 2 sensors one would need only 900 RPM to record movies with it…
Or mount the sensor on a disk, and spin it fast with a lens in front of it. It will work the way mechanical TV with Nipkow disks worked. There will be need for some processing to convert result into normal image, but it can be done with FPGA…
Remember: You paid for this https://airandspace.si.edu/collection-objects/camera-imager-viking-mars-lander #viking
Mount the sensor to something solid and spin a mirror.
Nice work heye.everts! I did this too with an old flatbed scanner in college back around 2008. But I never fiddled with the lens so nothing was ever in focus. Here’s a scan of me in my tiny college apartment bedroom (junk everywhere as always): https://i.imgur.com/wlY3Xw3.jpg . Props to Matthias Wandel as well, who’s project I definitely remember being an inspiration.
Canon’s LIDE scanners have very long, high resolution (up to 9600 dpi), linear CCD sensors and operate from power supplied by a USB port. They have been used to make scanner based cameras for years, example here: http://www.stockholmviews.com/diyphotogear/scannercamera.html USB power means you can build a portable camera that uses a laptop to scan images.
Unfortunately, Windows 10 killed most of the older, cheaper LIDE scanners’ support, but I believe you can still run them under linux. I think Canon makes newer LIDE-type scanners that have Win 10 support, but you’ll have to pay more for them.
The LIDE scanners work by sequentially cycling power to red, green, and blue LEDs as the CCD array is moved across the bed of the scanner. If you make a camera and want color, you have to use filters and make 3 scans (or maybe spin a filter wheel like those used in DLP projectors).
That type of camera is large format so you need a suitable lens that is capable of covering the scanned area with a sharp image (or learn to be satisfied by the fuzzy images rendered by a pinhole).
I have read good things about (but have not yet tried) the scanner drivers available for older scanners from Hamrick/VueScan. I have several deprecated LiDE scanners and some big lenses saved up for a rainy day.
Vuescan is great. On Linux, Mac OS X or Windows, I can use my 1998 Canoscan 10mpx 35mm SCSI-II film scanner or my 1999 Canoscan 600dpi USB flatbed.
Closed source, but…
Unfortunately, there are some scanners (like the Opticfilm series) for which it only works as a frontend for the OEM drivers, so doesn’t work on otherwise unsupported platforms.
There is a guy who did a lot of this quite a while ago. He did entire moving freight trains and printed them out as 10 or 15 foot long prints. The Internet is getting too crowded. I can not find any keywords that find his work.
Would you be thinking of Adam Magyar?
https://medium.com/matter/einsteins-camera-88aa8a185898
No, retirement age engineer or professor. He did quite a bit of experimental and specialty photography. Searching some more…..
://elm-chan.org/works/lcam/report.html
This one?
Got it. Andrew Davidhazy https://people.rit.edu/andpph/text-strip-basics.html
CCD sensors are superior anyhow, because you can shift the entire image into an adjacent dark frame in analog mode. Then you can read it out at your leisure, unlike with CMOS where you have to scan the image in-situ on the sensor. That’s why CCD doesn’t suffer from the rolling shutter effect – the bucket brigade shift into the dark frame can happen extremely rapidly – and it gets better dynamic range because your ADC doesn’t need to be so fast that it gets noisy.
But, the sensor is essentially two sensors in parallel and takes a larger chunk of silicon, so it’s more expensive to implement, and physically larger, which is why all the consumer cameras went to CMOS.
FLIR systems used linear infrared detector arrays scanned by a mirror wobbling on a cam from the late 1960’s forward. A second mirror tilted to provide a 2:1 interlace. Such systems were used on the B-52, B-1, A6E and the original M1 Tank FLIRs. The M1 Thermal Imaging Systems interlaced both the horizontal and vertical scans. Each detector channel had its own analog chain. The detector channels weren’t digitized, but analog multiplexed for a CRT display.
Michael Golembewski was working on something like this back in 2006. He attached a scanner to a Horseman 4×5 camera. The wayback machine has an archive of some of his stuff. https://web.archive.org/web/20060301230906/http://www.scannerphotography.com/cameras/current/index.html
RW:”Ok google, hey Siri, Alexa, Chinese IOT surveillance bureau and NSA, do I still have that scanner CCD I got from allelectronics like a decade ago because it was dirt cheap?”
Regional Waste:”It wasn’t in your garbage”
Canada Post:”You didn’t mail it to anyone”
ISP:”It’s not installed on any of your connected devices”
RW:”Okay guys stop, wow.”
Interservice Cybernetic Insect Development Group: “It’s in sector 34 X-ray Charlie Delta 93 of your basement, under the 3rd spider web to the left of the washer.”
RW: “WTF, who are you guys?
You also forgot Bixby.
:)))
So this is akin to a TDI line scan camera?
No, its akin to a regular linescan. A tdi linescanner has multiple lines to increase sensitivity
What, you mean like the Better Light digital backs for medium format cameras? They seem to have gone bust (unsurprisingly) as scanning backs are hopeless for moving subjects. Meanwhile 2D sensor resolution has reached a point that scanning doesn’t have any advantages at all.
Hmmm… I wonder how well that would work for photogrammetry. “Hey Meshroom! Here are 6,000 photos for you to process. They are all 1×8,000. That’s fine, right? Right? … Meshroom, are you okay? Aw, don’t cry, little guy; I’m sure that you can do it! Tell you what; I’ll give you a whole day to figure it out before I ditch you for Photoscan, m’kay?”
I wondered about doing astrophotography this way keeping the telescope still and letting the earth’s rotation do the moving.
I’m reminded of a time long, long ago, back in the 1980s, when DRAM chips came in ceramic packages with little Gold tops, which could be pried off. Once pried off, the chips where photosensitive, and made reasonable camera sensors. Usually.
There were also analog “Bucket brigade” CMOS delay line chips, which were usually linear, and could also be used as camera sensors.
Hmm, what was that? Everything old is new again?
I am just seeing this. Did you see this scanner camera https://hackaday.io/project/12058-spud-a-self-contained-scanner-camera