h a l f b a k e r yBirth of a Notion.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
The simple 8mm scanner would consist of a small flatbed-style mechanism, two reel holders (one of them motorized; the other with a simple friction tensioner), and a couple of pulleys (one of which should preferably include a mechanical link to the friction tensioner).
To use the scanner, one would
load a reel of film onto the friction tensioner, a takeup reel on the motorized holder, thread the film through the two pulleys onto the takeup reel, and manually advance to the end of the leader. A scanning mechanism would then be able to scan a few inches of film. Once this was done, the motorized reel would advance the film slightly allowing the next few inches of film to be scanned. Software would identify the sprocket holes on the scan and use them to line up the pictures between scans.
By scanning several inches of film at a time and having software look for sprocket holes, one would avoid many of the mechanical complexities and potential problems surrounding conventional projectors or telecines. Because all the motive power for the film would come from the takeup reel, it would be virtually impossible for the machine to "eat" film. While there would be some danger of a bad takeup reel malfunctioning badly and dumping film all over the place, this could be mitigated by (1) using a good takeup reel design; (2) having the software keep track of how much takeup-reel rotation is necessary to advance the film a certain distance. If these two precautions were taken, the only danger would be when the film was rewound onto its original reel (which may or may not be of acceptable quality). Since the rewind would take comparatively little time, however, that could be done while the machine was attended.
To achieve optimal scan quality, the machine would have to operate at slower than real-time speed due to bandwidth restrictions on most interfaces. I would expect, however, that if the machine was equipped with fast electronics and a suitable amount of buffering, it could scan film about as fast as the scanned data could be sent to the PC. Were I designing the machine, I would probably have each segment of film scanned twice: the first time would be used to measure how far the film had advanced and judge exposure, and the second high-quality pass would be used to scan those parts of the film that were "new".
I would expect that such a machine could be constructed for quite reasonable cost, but the quality of scans produced thereby would be better than produced by all but the best telecines.
Digital Telecine Project
http://truetex.com/telecine.htm
experiments with scanning film [ben_krak, Oct 05 2004]
Film-to-Tape Goes Digital: Tackling Telecine Issues for DTV and HDTV
http://millimeter.c...otape_goes_digital/
In a nutshell, both the Philips Spirit DataCine and the Cintel C-Reality high-rez telecine seek to make possible what Philips cutely calls "film-2-data." The idea is simple on its surface: the telecine as a high-res 2K film scanner, with 2K x 2K film data frame buffers and 14-bit internal signal processing. [ben_krak, Oct 05 2004]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Annotation:
|
| |
Sounds like genius. Though, I know nothing of film, and little of scanners. |
|
| |
Are the sprocket holes always lined up with the pictures? If they are let the film move continuously at low speed and let the scanner do the corrections. In other applications I found that to give much more satisfying results because graininess from the resolution of the scanner was reduced. |
|
| |
This idea has a lot of potential. As the idea is now, it sounds like an attachment for a flatbed scanner instead of a standalone film scanner. I'm not sure how fast scanners can get, but for film, you'd have to scan at 2400dpi. I also dont think it should need a second pass more than once, as it could make things inconsistant. One pass would set it up, and then the rest of the roll would match. So anyway... lets say we can find a really fast scanner that can do a scan in 30 seconds. Let's do some math. Legal sized flatbed minus a bit of wriggle room: 1.2 feet. |
|
| |
-1 hour of film on 35mm. 0.8sof film per scan =37.5 hours of scanning at 30s per scan
-1 hour of film on 16mm. 2s of film per scan = 15 hours of scanning at 30s per scan
-1 hour of film on 8mm. 5s of film per scan = 6 hours of scanning at 30s per scan. |
|
| |
Obviously these times could be shortened if the scanner was specially designed to only scan the film area. Feel free, anyone, to correct my math if it needs to be corrected. |
|
| |
[ben_krak] - // Obviously these times could be shortened if the scanner was specially designed to only scan the film area // |
|
| |
Rotate film through 90 degrees? |
|
| |
ben_krak: Thanks for the link. Quite interesting. |
|
| |
I had envisioned the device I described as a standalone device rather than a scanner attachment for a couple of reasons: |
|
| |
-1- Scanners' optics are designed to focus on objects that are precisely on top of the glass. To avoid scratching the film, it should not be dragged while in contact with any surface. Thus, if the glass were in place the film would have to be some distance above it. This would lead to focusing problems which, given the sizes of images involved, could be severe. |
|
| |
-2- I would expect that a 4800dpi scanner that scanned a 16mm strip would be much cheaper than a 4800dpi scanner that scanned an 8.5" wide strip, in addition to being smaller. |
|
| |
I like some of what is proposed on the "digital telecine project" page, but for 8mm film where the image is about 1/4" wide, a 640x480 scan would require a resolution of 2560dpi; 600dpi or even 1200dpi just isn't going to cut it. |
|
| |
Also, I'm somewhat leery of using sprocket feed to advance film. On film which is in good condition, a sprocket feed should pose no problem, but on film which isn't so good, sprockets can wreak all sorts of mischief. To be sure, a well-designed system should be able to shut down before mangling film too badly, but it would seem better to have a system which was fundamentally incapable of doing anything bad. |
|
| |
Two more details: (1) my thinking with two-pass scanning would be that the scanning head would make the first pass in one direction and the final pass in the other direction, so as to minimize unnecessary movement; (2) performing color/exposure correction only once at the start of a roll is apt to be minimally adequate for cameras with automatic exposure control, and completely inadequate for film shot without. |
|
| |
Good point about the scratching.
Still disagree about doing colour correction more than once, if it did it differently from time to time, the footage could turn out inconsistant. |
|
| |
Or you can do things like my uncle did: Project film on wall, videotape wall. |
|
| |
[supercat] - I had assumed that if a conventional flatbed was used, the glass would be removed. However I agree that a film scanner is a better choice than a flatbed. Perhaps I haven't looked hard enough, but I've yet to encounter a flatbed that can scan at 2400+ dpi without lots of noise (especially in darker areas), where as film scanners will happily give you the resolution without the noise, and generally at 12-16 bits/channel too. Speaking of which, at 16 bits/channel, you could just do one pass and perform exposure correction later in software. |
|
| |
[benjamin]--My expectation would be that one of the passes would be at much higher speed than the other, and one of its functions would be to identify which areas needed to be scanned at the slower speed. If the high-speed scan could be done in one direction and the slow-speed scan in the other, the speed penalty for doing the high-speed scan should be pretty minimal. |
|
| |
There is, though, another reason I thought of that one may want an "extra" scanning pass though this would have to be in high quality as well. If film has scratches or gunk on it, it may be helpful to scan it once backlit and once frontlit. If the film wasn't moved between the two scans, registration should be perfect, and having the two perfectly-registered scans should allow scratches and gunk to be identified and "filled in". |
|
| |
What good is frontlight/backlight on a transparency? To illuminate the scratches better for removal? |
|
| |
It seems to me that a high-quality frame-size CCD sensor with a gate and electronic "pin" registration is going to be a lot better than a flatbed anything. Ooops, I just described a telecine. (And the pin registered part is pointless, I guess, as there are no pin reg 8mm [or 16mm for that matter] cameras.) |
|
| |
Maybe a simple version could just be a modified digital camera, nothing fancy, and a film transport and illumination source. I suppose you could go the Kodak Lightning route and use a laser for the illumination. |
|
| |
For scene-to-scene color-correction, you could scan the film in it's entirety to disk recorders, at neutral settings in an HDRI format, and perform the primary/secondary corrections from disk. That way you only have to haul the film once and you can version easily. |
|
| |
Also, remember that almost all telecine and scanning is done from negative stock as positive, or reversal, transfers suck. For anything decent you'd need to be shooting with a stock like Kodak's Vision T neg. |
|
| |
(By the way the Spirit and the Rank are half-rez devices. 4K is considered full rez and 8K is, nominally, at grain rez.) |
|
| |
Using a single-frame capture device would probably work, but would require a mechanical means of advancing the film a frame at a time. This would require some means of "pulling" other than the takeup real, and would thus entail more risks of things going wrong and damaging the film. |
|
| |
I'm well aware that filming stuff in negative is better than using reversal film, but most home movies weren't shot that way so the positives are the only copies that have ever existed of the material on them. |
|
| |
As for frontlighting the transparency, the idea would be precisely to illuminate and identify scratches, gunk, etc. A bright or dark spot on the backlit scan which shows up bright on the frontlit scan is a scratch, dustspec, or other such defect. A bright or dark spot on the backlit scan which shows up dark on the frontlit scan is more likely to be something "real". |
|
| |
//For scene-to-scene color-correction, you could scan the film in it's entirety to disk recorders, at neutral settings in an HDRI format, and perform the primary/secondary corrections from disk. That way you only have to haul the film once and you can version easily.// |
|
| |
What is the single-shot lattitude on most CCD modules? |
|
| |
Depends on the CCD. If it's a DALSA it is very wide. Besides, since it won't be a real-time transfer, there's nothing to prevent you from generating multiple latitude exposures and blending. Just a matter of code ;-) |
|
| |
Home movies? Hmmm. Maybe some sort of adapter for a digital SLR would be better (cheaper). A box with a light source, film handling mechanics, a mount point for the camera and software to control it all from a PC. |
|
| |
I disagree that all film handling should be done from the take-up reel. I think it should be handled by the gate in consort with the take-up and supply side reels. Certainly seems to me to be the least hazardous for the film stock. |
|
| |
Deluxe versions could have a wet gate for subduing scratches. |
|
| |
What is the hazard in driving the film from the takeup reel, if the supply reel's drag is dynamically adjusted so as to maintain constant film tension? Were there no tensioning I could see potential problems with uneven winding, but what's the hazard if tension is controlled? |
|
| |
Perhaps nothing at all but, if I were to design the system, I would want the take-up side system to be devoted wholly to just moving the film from the output side of the gate and providing the very best tensioning for gently wrapping the film on the reel as well as isolating the film from shock transmitted from the rest of the path. I'd leave the frame-to-frame advancement up to the gate mechanism as it's the smallest length of film in the path that I have to handle positively and with the greatest degree of precision. I'd have no drag at all on the supply side, rather it would positively feed the film, dampened but lock step, to the input side of the gate. |
|
| |
If you put an optical rotary encoder in the tensioner for positional information than you could use a linear CCD and run the film past it continuously as reasonably fast as you can. To avoid scratching maybe a glass or lucite tube as a pulley/tensioner with a light inside to provide a "flat" drum and backlighting for the CCD optics to focus on. Sounds like a good idea. |
|
| |
Bristolz: The idea with having the long scanner do 4" or so at a time would be that there'd be no need for particular precision in starting and stopping the film. Provided there was enough overlap for some simple pattern-matching software to line things up (and the sprocket holes would be sufficiently big and obvious that such matching should not be difficult). As to the question of whether it's better to put the drag in the film gate or the source reel, I can see arguments for the filmgate, including the question of how one photographs the end of a roll with no trailer. My concern would be with avoiding scratches, but there are probably ways of dealing with that as well. |
|
| |
BTW, a couple of related notions would be whether it would be practical to make the optical path long enough to minimize sensitivity to focusing distance (thus easing requirements for supporting the film) and how flat film would tend to be if kept under reasonable tension but not otherwise held flat (perhaps hung vertically). |
|
| |
Spare parts: Doing a straight linear scan of the film and getting decent results would require that the film travel at a well-regulated constant speed. This would require using either sprockets or powered rollers. Sprockets have the potential to really mess up film if something goes wrong; powered rollers by their nature generally have to be in contact with a flat part of the film and pose the risk of scratching or other damage. |
|
| |
If the optics could be constructed to be tolerant of small variations in focusing distance, and if the film could be scanned without reliance upon powered rollers or sprockets, this would also facilitate the scanning of film whose sprocket-holes are damaged beyond usability. |
|
| |
supercat: film unsupported would tend to twist and curl, distorting the image. If your tensionor had a rotary encoder in it like one of the two inside a regular ball type mouse, that could give the software the position data to at least the resolution of the scanner. Then the software could be set to call 6.4532 rotations a frame, or whatever. To support the film and prevent scratches, one could have a 1 inch diameter glass tube half an inch wide or so mounted on one side with a bearing and in the other side a white LED or whatever your appropriate backlight source is. That way as the film unrolls off the supply reel it comes in contact with the glass wheel (no sprocket needed at all if the tension is right) which rotates freely along the film path passing the images precisely in the focus plane of the optics all illuminated and everything. If the encoder were reading the glass pulleys rotation then the computer would know at exactly what speed the image was traveling (and what direction). Just like the old hand scanners that could be rolled across a page. In fact I bet one could cannibalize one off of Ebay and make it work fairly easily. Oh and one more thing, your first pass could roll from A spool to B and your final pass could roll back and thereby get both ends, too. I do like the idea of not relying at all on the sprocket holes. |
|
| |
Spare Parts: I do like the glass-cylinder idea, and it was one of the various variations I'd thought through. I'd be worried, though, that if the film had some foreign matter on it (e.g. adhesive left over from using the wrong sort of tape) that it could come off on the glass drum and ruin every few frames of films. I suppose this isn't any worse than the hazards associated with film gates, but I'd still worry about running such a machine unattended. Perhaps if the glass drum had some sort of cleaner at the bottom it might be workable especially if the cleaner used were one which would prevent film from sticking to the glass. |
|
| |
Thinking about it, if the light source were inside the glass, and if the reels were designed so that the base side of film could be against the glass for both original and duplicate films, having the film roll against glass probably wouldn't be too bad. The only danger would be with films that were hand-painted on the base side. Except for such films (and I wouldn't expect there to be many of those in 8mm or even 16mm), the base side is much less prone to suffer serious damage than the emulsion side. I think for optical reasons the drum would have to be larger than an inch--probably more like 4 inches or so--but that would help reduce film flexing anyway and so probably wouldn't be a bad thing. |
|
| |