film -> video device

[FredThompson]

I'm developing a device to allow conversion of 8mm and 16mm film to video without the need for a projector. You WILL need a camcorder and an available printer port (gad, do I miss the Atari 8-bit game controller port I/O). There will be a little bit of device-specific software and some AviSynth scripting for post-processing. There will also be some script and filter chains for post-processing wrt image quality (noise reduction, scratch removal, etc.)

Hardware details are still being worked out and I'm not going to release the plans just yet. There's a chance this project might be attractive to O'Reilly's Make magazine, engadget, etc. so I'll start with a write-up similar to those from Radio-Electronics magazine. That means all the details will be shared, including the software, parts list, construction details, etc. Article will also promote doom9, neuron2's board, AviSynth, VirtualDub and freeware MPEG2/MPEG-4 encoders. If no one wants to print it, I'll release it to doom9, neuron2's board, etc. under some form of donation-ware (i.e., you use it, you like it, please send me a few bucks.)

If possible, I'll make all the parts available in kit form and, maybe, assembled units but I really don't care to do assembly if at all possible.

This is something I'm developing, regardless of anything else which might happen. There are between 600 and 700 reels of old family movies in my house. A couple of years ago I paid to have them converted to Hi8 and the people who did it screwed up the color balance. That was about $2000 down the drain. Tonight the gods of inspiration smiled on me and a "solution" for homebrew jumped into my head.

I know the obvious answer to this question is: "As cheap as possible" but I'd like to get an idea of what people might pay for such a hardware device. The goal is NOT speed, it's accuracy. Decent parts, especially optics, cost money and I'm wondering what people might be willing to pay for a kit. If the parts cost $200 and no one is will to pay that much, I won't waste my time figuring out how to put together kits. Kits would just a timesaver, really, not a way to make money. Having said that, though, It'd be nice to be able to gather a buck or 2 for eacdh full kit and send that to Avery, DG, the AviSynth team and doom9 to ehlp with bandwdith and show thanks.

[SeeMoreDigital]

Sounds interesting...

What type of optics/lenses do you require?


Cheers

[FredThompson]

Well, hey there, stranger!

Optics aren't that tough to find, Edmund sells them and I'm hoping to use generic lenses like the type Best Buy sells for camcorders. The real challenge is controlling the motion of the film itself, physical alignment and focal points.

[mustardman]

Perhaps the use of an old projector, then all the mechanicals of moving the film will be taken care of.

Of course, the motor would have to be replaced with a servo/stepper and the bulb would not need to be as bright.

But I get what you mean - feed the image direct into the camcorder, not via a screen, and control the frame rate of the film from the PC.

Nice idea - I'm currently getting about 3000ft of 16mm converted, and it is costing me about a grand. A "home remedy" would have been an attractive option, even at half the price.

MM.

[FredThompson]

There will be a screen but not like projecting on a wall. Right now I'm having a challenge finding appropriate cool light. There are commercial methods but $1000 for a light is a bit much. We don't need to throw it across a room, just a few inches.

I've thought about canibalizing old projecters. That would be an inconsistent supply. The only parts which would really be usable for this idea are the reels and possibly the sprockets. The whole projection assembly is wrong. It's designed to throw the image a distance and we'd have to counter the optics more extremely than I'm hoping. Still, we'll see...

[SeeMoreDigital]

Originally posted by FredThompson
There will be a screen but not like projecting on a wall. Right now I'm having a challenge finding appropriate cool light. There are commercial methods but $1000 for a light is a bit much. We don't need to throw it across a room, just a few inches. Are any of these high-brightness light emitting diodes any good now?

Some of them are bright enough to be used domestically as direct replacements for 20 watt halogens... but I don't know how "white" the light is?

I guess using a bunch of optical fibres to carry the light is not an option?


Cheers

[mustardman]

Another option to reduce the amount of heat that a bulb gives off (the real problem) is infra-red cut filters. They look like a simple piece of glass, usually found inside slide projectors. But just try burning somthing with a magnifying glass and put one of those in the way!

"White" LEDs are not really white, they just look it. There is a serious deficency of red (unless they've improved in the last 12 months).

Low wattage halogens would work best, perhaps even those new(ish) torch lights that are halogen. That'd keep the watts down without changing the colour temperature...

MM

[Arachnotron]

Right now I'm having a challenge finding appropriate cool light. There are commercial methods but $1000 for a light is a bit much. We don't need to throw it across a room, just a few inches.
What about the lamps used for those home build LCD projectors (http://www.lumenlab.com/) ? Probably way overkill for your purposes, but they are cheap.

[tedkunich]

Fred,

I'd be interested in what you have... $200-$500 is not out of the question... I have so many rolls of that doing them "profesionally" would cost an arm and a leg.


T

[tedkunich]

Originally posted by Arachnotron
What about the lamps used for those home build LCD projectors (http://www.lumenlab.com/) ? Probably way overkill for your purposes, but they are cheap.


Ummm... yea, a bit over kill ;) FYI, a 400W metal hallide is HOT... enough to cause a severe sunburn after 10-20 minutes with a foot or two proximity of the bulb. They put out a lot of UV.


T

[FredThompson]

I hadn't thought of fiber optics. That's an interesting idea but maybe something else will show up. I guess if that's the only option, they might work with a true-white bulb but the heat problem still exists.

There are diode arrays which claim to be proper white and adjustable. The camera-mounted rectangular ones are $1K. You'll see those ads in DV and other magazines. I'm really hoping to find something just a smidge cheaper, you know?

I'm looking at various microscope lights now. Maybe one of those technologies will be usable. Don't know how "pure" they are, though.

[SeeMoreDigital]

Have you had a look at the type of "back-light" used in scanners?

With regards to the "low-heat" LED approach, if the colour is off... maybe it could be "fixed in post" ;)


Cheers

[FredThompson]

yes, those are both good ideas. A custom filter chain is something which is already part of the mix. Color Mill (VIrtualDub) is part of the plan. I wonder if those cold cathod tubes sold for case mods might be useful. Hadn't thought of that. Maybe put the tube on the other side of some sort of diffuser like a translucent plastic.

[SeeMoreDigital]

Originally posted by FredThompson
Maybe put the tube on the other side of some sort of diffuser like a translucent plastic. Two or three years ago I visited a signage and photo software exhibition with DOP friend of mine...

Anyway, long story short. There was a guy their from Australia who had developed a new way of back illuminating certain types of signage. Normally such signs require mains voltage fluorescent tubes inserted into a metal or plastic light box. But his design used a thin (6mm/10mm) translucent board illuminated by dozens of low voltage white lamps positioned along the long edges of the board!

What was quite impressive was how his design had managed to maintain an evenly distributed quantity of light throughout the entire surface area. And some of these signs were as big as 1800 x 1200.

His portable light boxes for viewing "transparencies" were equally impressive ie: they were very bright and cool (cold even) to the touch!

I attempted to quiz him on how he had achieved this but he was understandably tight lipped (...something about a world-wide patent).

Sadly I can't find his details anymore but maybe his product is formally available now.

I had wondered whether he may have used some form of "prismatic" diffuser to transfer the light from the edge to the surface of the board...

Anyway... I hope this helps?


Cheers

[FredThompson]

Wonder what happened with that? Do you recall a business name?

I'll ask a buddy who collects arcade games if he has ideas. Those folks know a lot about decals and translucent panels.

[SeeMoreDigital]

Originally posted by FredThompson
Wonder what happened with that? Do you recall a business name?
Sorry no I don't :(

While looking through one of my electronics catalogues I found the following information about an "RGB Full Colour LED": -

http://www.maplin.co.uk/Search.aspx?criteria=n63ax&doy=24m5&source=15. Together with this PDF file (http://www.maplin.co.uk/Media/PDFs/N63AX.pdf).


Cheers

[Sergei_Esenin]

The goal is NOT speed, it's accuracy. Decent parts, especially optics, cost money and I'm wondering what people might be willing to pay for a kit. If the parts cost $200 and no one is will to pay that much, I won't waste my time figuring out how to put together kits. Kits would just a timesaver, really, not a way to make money. Having said that, though, It'd be nice to be able to gather a buck or 2 for eacdh full kit and send that to Avery, DG, the AviSynth team and doom9 to ehlp with bandwdith and show thanks.

I'd be interested, and willing to spend $200-$500 depending on the quality of results. I have a large collection of rare 1960's-1980's 8mm and 16mm erotica that I'm too darned embarrassed to get professtionally converted. :D

[FredThompson]

OK, guys. I think I've solved "half" the optical issues and eliminated the heat. The idea won't use a screen of any sort and the only optics issue is how to get a camcorder to properly focus at a short distance. Servo motors would be the best way to move the film. They need precise movement. I'm more than willing to take a second to move from one frame to the next if that is what is required. There's no sense in trying to "snap" the film movement for a short distance.

In the interest of time, I'm going to contact a projector tech who sells modified projectors for video conversion at ebay. Maybe it would make the most sense to find someone like that who can make the physical devices.

I'm very tempted to explain the entire idea here because I don't want to dedicate time to making the devices for other people. However, the initial comment about trying to make a few bucks and them to doom9, Donald Graft, the AviSynth team, etc. is something I'd prefer not to risk at this time. If we talk too openly about the method, there's always the risk it would end up on hackaday or some similar site which could weaken the ability to sell it as an article through Make magazine or something like that.

[rfmmars]

In the interest of time, I'm going to contact a projector tech who sells modified projectors for video conversion at ebay. Maybe it would make the most sense to find someone like that who can make the physical devices.

I was hoping that you were going to the whole thing from scratch, you don't learn much by using somebody's research and you sure don't break new ground.

I have built all my telecine units and learned alot. If you want to save and get a projector, then get a Sears 8mm and a Sears Super 8mm models built by B&H. Do not get dual format projectors. The top of these models have a machcanial speed control and a 4 blade shutter so the speed needed is 15 fps NTSC. Either shoot at the gate with one optic or a simple flat cardboard screen with two.

Scrap the camcoder idea and get either a machine vision camera or a Mintron star light color camera.

I was hoping you would stumble on something new by doing your own thinking but I guess that's not to be. I am planing to market a CD & DVD covering modifaction and procesiing for telecine ~ $19.95

richard
photorecall.net

[FredThompson]

You're either trying to throw people off the path of my idea or you're being presumptuous and condescending. I'll assume the former though it sure looks like the later.

[eb]

What about idea to use stroboscop lamp /that used by car service/, easy to control, cold and "cool" and acceptable price.

eb

EDIT
...and this connectyd to simple mechanism based on "elipse" wheel > up the glas > move film > down the glass > light shot > up the glas ...

[FredThompson]

Hmm...why use a shutter if you have a precisely controlled strobe? That's an interesting idea. It doesn't fit my idea but it would be an interesting way to do projection.

The only thing I'm still wondering about is properly controlled movement of the film itself. Servo would be best but I'm trying to avoid gear boxes. It will all make sense when the project is complete.

These would be nice: http://www.foveon.com/article.php?a=121

[eb]

when film is moving down it is centered by stable piramide tips
... and precise picture settings against black edge of the picture can be latter by computer

[Arachnotron]

I don't know the first thing about film conversion, but I do like tinkering :)

I was wondering: if high quality is your aim, why not go for a digital camera conected to a computer instead of a camcorder? You could attain higher resolutions for a low price, which in turn might come in handy if you want to do any frame-jitter (is that the correct term?) corrections later. And synchronisation would be easier. Also, you could use a macro lens. (I assume you will get some barrel-shaped distortion using a standard camcorder to film from a small screen in close-up)

[rfmmars]

Why are you reading into my post something that's not there? I would like to see your ideas, something new. It's ok to reinvent the wheel by paraell development, by not to try? I know that not you.

As far as my "Do it youself Package", it has been in the works for about two years, and not a threat to you or anyone. In fact I have shared a lot of it on this and other forums.

Read my post again with a different viewpoint. I was just again trying to point you in the right direction.

richard

[FredThompson]

Originally posted by eb
when film is moving down it is centered by stable piramide tips
... and precise picture settings against black edge of the picture can be latter by computer In theory, yes. I'm not sure the edges of the frames are sufficiently precise for this since the sprocket holes are what is used for alignment. There have been similar ideas for stabilizing shaky video. IOW, maybe the edges of the frames are more accurately considered analagous to overscan.

I'll look into it, though. Thanks for the idea.

edit: What I typed doesn't match what I was trying to say. The shutter repeat is hardware timed. Alignment is based on rate of flow of the film at the projection point. Accurate control of hte film movement is based on the sprocket holes.

[FredThompson]

Originally posted by rfmmars
Why are you reading into my post something that's not there? I would like to see your ideas, something new. It's ok to reinvent the wheel by paraell development, by not to try? I know that not you.

As far as my "Do it youself Package", it has been in the works for about two years, and not a threat to you or anyone. In fact I have shared a lot of it on this and other forums.

Read my post again with a different viewpoint. I was just again trying to point you in the right direction.

richard The first and next-to-last sentences of your post are the ones which irked me. I realized then, as I do now, that it was late for both of us so an inappropriate time to make a final conclusion, so to speak...er...type. It didn't seem characteristic of you, either.

I don't feel threatened at all. There is a lot of unique stuff in my idea for getting the images into the computer.

My interest isn't in building a machine shop and becoming a small parts fabrication expert. Basically, I think someone who already has a business selling tweaked projectors would be better equiped to make a couple of cutom pieces, especially the film movement mechanism. Canibalizing parts is far easier for someone who already has the sources for them.

Your idea of dedicated cameras has merit. A proper housing and fixed optics is definately more attractive than fiddling with a camcorder. There's also the distinct possibility of RGB instead of lossy compression. The trade-off is price, though. Still, I'll look into it. (edit: well, $1500 or more for a dedicated camera doesn't fit with the project goals. I'll have to stick with a camcorder for now. The idea is flexible enough, however, assuming whatever camera is used can make a file which AviSynth could read.)

I considered things like the Ambico prismed things and threw the idea away. The quality would be too poor.

[rfmmars]

The trade-off is price, though. Still, I'll look into it. (edit: well, $1500 or more for a dedicated camera doesn't fit with the project goals



Do you really think there is a cottage industry making telecine projectors? I know of only one company and they aren't going to stop production ($1200.00 to 2400.00) product to do some custom work at a reduced priced.

Instead of attacking me, you should have ask about the content of my post. You could have found, by asking, that you can buy a Revere 718 8mm projector for as little as $0.99 plus shipping which the whole mechanism comes out in one piece (gate,pressure plate, shutter, feed and exit sprockets,) and is so simple to do anything you wish, gate capture or project..

Here's the info on other projectors

The Sears projectors / Bell & Howell are the best for easy modifaction and built like a tank. Again so easy to modify which I have added audio tone bursts generators, so I know when editing, I can find the wrong exposures sequences in the final edit and delete. All this information was all for the asking but no all you did was to attack me for trying to help you.

$1500 for a camera? No, see links below.

http://www.polarisusa.com/cgi-bin/product_listings.pl?listing_category_id=45

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=5777317123&rd=1&sspagename=STRK%3AMEWA%3AIT&rd=1

http://www.polarisusa.com/cgi-bin/view_product_detail.pl?listing_category_id=&product_category_id=1&template_id=144

Your aptitude is typical of many at this and other forums. You think you know it all even when you never walked the walk!

If you want to do it right, do it “VideoFred's” way if you have all the time in the world.

Good luck,

richard
photorecall.net

[FredThompson]

Richard,

I did not attack you. I said your post was received as insulting and condescending and I would give you the benefit of the doubt that such was not your intetion. You later clarified there was no insult intended and asked for clarification as to what aspects of the message would have been received that way. I told you which two sentences broke such a reaction and that they seemed out of character and it was late for both of us. I do not take any ownership for your choice of words. I gave you the benefit of the doubt that what you meant to communicate is not what was communicated by these words:
I was hoping that you were going to the whole thing from scratch, you don't learn much by using somebody's research and you sure don't break new ground.I was hoping you would stumble on something new by doing your own thinking but I guess that's not to be.
Apparently, you refuse to comprehend how those comments would be received as insulting and will not accept the fact they were.

My comments were about your choice of words in those two sentences, period.

You are now blatantly attacking me and mixing your subjects.

I never said anything about cottage industries or telecine projectors. I used the word "business." "Business" does not necessarily equate to "cottage industry", it is simply a person selling the fruits of their labor. People who make candles as a hobby and sell them have a business.

I've already stated the approach is not the common method and I have no intention of explaining it in an open forum. I've also explained why this is the case. No amount of goading or verbage by anyone will change my mind. If the project hits a no-go point or I find I cannot continue it, I will share the details but we have not reached that point.

I've already acknowledged the value of your comments concerning cameras and did some research. In no way have I prevented you from doing anything you want to do. Frankly, I was quite surprised you thought such a thing.

Take some time out, cool off, and look at what has happened. You are taking things way out of context and accomplishing nothing positive.

[vhelp]

I had an idea, (perhaps building on Arachnotron's) ...

How about using (as a test platform / debugging) a USB cam. They
usually have full-frame. My old QuickCam has 640x480 resolution
and could be controled to take pictures. The only problem to this
is that I don't know how low (fps) you can go with(in-between) pictures.
But, that fait would fall just as Arachnotron digital camera then.

And, at least you have progressive frames and not interlaced to
taint your theories and things. I'm assuming that your home-made
telecine engine will be taking pictures of full-frame sources anyways.
So, a good progressive camera or quickCam usb device or something
would work. I don't know. I'm just trying to throw ideas around :)

Well, just trying ta help a fellow user :)

Good luck in your fun endeavor.

-vhelp

[FredThompson]

Just a quick note to say this is still in development. I'm just swamped with work and will be traveling for the next couple of weeks.

[JeremyIrons]

I would pay in the $500 range, no question

[Delphin]

When this thread is revived here are some observations . . .

For the Film Transport -
A good quality stepper motor should give quality approximately equivalent to the pull-down claw in a standard 8mm projector in advancing frame by frame. Standard small frame stepper motors on NEMA23 frames can be had on the surplus market for only about 15 dollars. Of course designing around a surplus motor could lead to concerns about the 'reproducibility' of the design. This need not be a concern. Low cost surplus motors can be used for prototyping without fear because these are 'commodity' type units and another type23 frame motor can be substituted with equivalent performance from any of a dozen vendors. These motors usually run 1.8 degrees per step (200 steps / revolution) and can easily be 'half-stepped' by a simple change to the drive sequence.

To be compatible with the above stepper motor, the film would be advanced using a simple roller having a number of sprockets around the circumference which evenly divides into 200(full step) or 400 (half step).

This lets us get to the next frame by just indexing the stepper forward by an integral number of steps (or half steps).

Lets look at a practical example. Check out this link . . .

http://www.lavezzi.com/PDFs/SprocketPDFs/SprocketCat.pdf

The 0.15 inch sprockets are for 8mm film and the .1667 are for Super8.

The 0.15 and 0.1667 inch sprockets with 16 sprockets around the circumference and with 0.25 inch shaft holes (with included set screw) will quite easily directly attach right to the 1/4 inch shaft of a typical NEMA23 stepper motor.

No custom machine work so far!

Now lets look at how that 16 sprocket number would work with a standard stepper motor running in 400step half-step mode . . .

400half-steps-per-revolution / 16 sprockets(frames)per revolution = 25 half-steps/frame

The steppers are not perfect so the could be an error equivalent to a small fraction of one half step, but with 25 half steps per frame, the errors should be small and manageable.

Several methods can be used to improve frame registration, such as a “periodic error correction” scheme based on 'micro-stepping' the stepper motors and a software calibration procedure which uses the detected edges from a dark-frame calibration film.

With a high-resolution pickup, software frame registration correction can also be applied simplifying the design of the stepper drive circuits (by eliminating any requirement for micro-stepping).

For the Light source -
Way too much emphasis has been given to exotic light sources. A single high brightness White LED or Tri-COLOR (RGB) LED working through a diffuser is enough! The Tri-color solution would allow for better color-balance compensation. One LED might not seem like much light but it is more than enough. This is because it is not necessary to 'project' the image. The best quality would be to back illuminate the frame (like in a slide viewer) and then directly image it with a high resolution CCD/CMOS sensor.

The Pick-Up device -
A high quality 1-Megapixel or higher (real pickup resolution, not BS 'interpolated' resolution) USB/USB2 Webcam. This webcam should preferably employ a CCD pickup and offer full 24bit frame dumps (some use CMOS and only offer 'already compressed' intel video formatted output).

Optics -
On many Webcams the built in lens can be unscrew so far out that the camera will work in 'macro' mode at very close ranges. If the lens is extended far enough the camera may be able to directly image 8mm/Super8 frames with NO other optics!
If that is not possible only a simple lens equivalent to an un-mounted Hastings triplet will be required. Remember: We are not 'projecting' the frame, just helping the web cam to directly image the backlit frame like a 35mm slide in a slide-viewer.

Software -
Custom software would trigger a single frame capture through the Webcams API (VFW or Twain). Then the software would change the 4 pins on the parallel port used to drive the stepper in half-step mode. Standard capture through the VFW interface would be slow enough to allow real time capture and compression (HuffyUV, DivX, XVID codec). The software could also optionally support a frame-source plugin for AVIsynth in addition to direct AVI captures allowing access to AVIsynth processing filters.

So, to sum up (lets not make this more complicated than it needs to be)

1. Simple stepper motor driven by PC parallel port drives the sprocket, which pulls the film through a simple friction-feed film gate (the film gate holds film flat at focus plane and prevents slack from causing uneven film advance).

2. Film is back illuminated by a LED light source through a diffuser.

3. Webcam captures frame using a custom software driver which then advances frame by advancing the stepper motor by 25 half-steps to return to step 1 (on next film frame).

So this does not need to be that complicated folks, it’s as simple as 1-2-3.

Note: In regard to the capture device in step three using a high-quality digicam could give very high quality, but most of these use a hybrid electro-mechanical shutters (in addition to pure electronic shutters). It is doubtful that these mechanical parts were designed for the kind of duty they would see in video capture (tens-of-thousands of single frame exposures). Any high quality digital camera having a purely non-mechanical electronic shutter would be fine.

Hope this helps your project, let me know if you have any questions.

[FredThompson]

You're pretty close to the ideas I've had. I did some experiments with a filmstrip backlight from an old flatbed scanner.

There are some folks who are doing this sort of thing with flatbeds. I'm not convinced the frame alignment is good enough but there are some definite advantages wrt resolution, full RGB per pixel and economy of scale.

It really doesn't matter HOW we get the sequence of images, scanner or camera, as long as the light is consistent. A big challenge I saw with a camera was focal distance and ensuring the camera was aligned properly. A relatively simple control of film transport would small but the tolerances for manufacture are quite tight.

There are some other things I considered to reduce noise and variability. (Forget the magazine article idea, better to get the project going.) I was originally thinking of backlight and a standard video camera. Non-linear feeding of film would allow multiple images per frame which would be averaged together to reduce rounding variability. AviSynth script would do that as well as throwing out the video frames in which the strip is advancing.

I'm now quite partial to the idea of a modified flatbed scanner to produce a series of images because it's an economical way to get full RGB, proper focal distance and alignment with the film. A few weeks ago I contacted some of the people who have been doing this to see if we could combine efforts as I started approaching the problem from the film processing aspect and worked towards the physical transport and they're doing it from the other end.

http://www.jiminger.com/s8/index.html
http://www.public.iastate.edu/~elvis/8mm/film_scanner.html
http://truetex.com/telecine.htm

Having typed all of that, and re-reading your post, I like the idea of a webcam. That would allow multiple reads per frame and a fairly small package. The big challenge is the machining and precise control of the filmstrip. Yes, parallel-port driven stepper motor was another aspect of the original idea. I thought a short tube with a rectangular opening for the filmstrip and a backlihgt behind it combined with stepper-driven film reels would work just great. Then the big problem became alignment, optics and control of the filmstrip. The film needs to be moved a set linear distance at the "projection" point so some form of distance-traveled sensing must be used. I'm hoping there are parts which could be scavenged from old projectors.

[Delphin]

You're pretty close to the ideas I've had. I did some experiments with a filmstrip backlight from an old flatbed scanner.

There are some folks who are doing this sort of thing with flatbeds. I'm not convinced the frame alignment is good enough but there are some definite advantages wrt resolution, full RGB per pixel and economy of scale.


Without a transfer lens system of some sort (tricky to design to work with the moving linear CCD pickup) a flatbed scanner will be a little marginal on resolution.

This is because it is rare to find REAL resolution higher than 2400dpi on a flatbed.

Though some claim 3200 or even 4800 few really live up to those numbers.

The Epson scanner used by the first fellow claimed 4800 dpi but the images still looked fuzzy (like the scanner was delivering an effective resolution of 1200 or less)

The guy at the second link is using an Epson with a claimed Resolution of 3200 and his captures are also a little fuzzy.

The third guy concludes that his flatbed is not doing the job and uses a film scanner and STILL gets fuzzy scans.

My brother and I converted the family’s old 8mm films to video by just projecting them on a wall and shooting them with a hi-8 camcorder and got better results than the ones I saw. True, it was very tricky finding a setting that would reduce the flicker problem you get in this setup, but my projector has continuously variable speed which helps a lot (and now there are Virtual Dub and Avisynth filters to help fix flicker).

I think the reason these guys got fuzzy frames is that trying to grab 8mm frames from a flatbed scanner requires it to ACTUALLY perform to a pretty high DPI spec (which is rare given all the mechanical variables).

I think that the 2400dpi number quoted on these web sites is a bit too low for a ultra high quality capture if you want to do the software frame alignment talked about.

Even 35mm camera lens were frequently rated to as much as 50 line-pairs/mm

To resolve a black/white line pair actually requires 2 pixels so that’s about 100 pixels per/mm or about 2540 pixels/inch JUST FOR A SHARP IMAGE AT 35mm.

At 8mm you want EVEN SHARPER OPTICS, say at least twice that number (equivalent to 5000dpi REAL scanning resolution).

This would be what was needed to guarantee that film grain is the only limiting factor, assuming a source from a camera with good 8mm optics and finer grained film stock like Kodachrome 25, or Ektachrome 64.

A scanner uses a linear CCD pickup and associated linear 'lens' optics to capture a (theoretically) very finely focused image, but where are the precision adjustments to guarantee perfect focus?

The slop in the mechanical scanning mechanism, and all the other mechanical tolerance issues, conspire to make REAL resolutions of even 2400 DPI hard to achieve.

From the information I found on the web, the active part of a super8 frame is only about 0.22 x 0.16 inches at the most, so a good solid 2400dpi (honest resolution) would only give about 528 x 384 pixels in the sampled frame

This is actually marginally acceptable, but a little low if we also want to do software frame alignment, because we need to do sub-pixel adjustments to align the image, and, in that process, we will lose a little sharpness (in shifting and re-sampling the image).


Having typed all of that, and re-reading your post, I like the idea of a Webcam. That would allow multiple reads per frame and a fairly small package. The big challenge is the machining and precise control of the filmstrip. Yes, parallel-port driven stepper motor was another aspect of the original idea. I thought a short tube with a rectangular opening for the filmstrip and a backlight behind it combined with stepper-driven film reels would work just great. Then the big problem became alignment, optics and control of the filmstrip. The film needs to be moved a set linear distance at the "projection" point so some form of distance-traveled sensing must be used. I'm hoping there are parts which could be scavenged from old projectors.

Re-read my post and follow the link to the PDF catalog (and check the math) related to sprockets with 16 teeth for use with a .25 shaft motor. Advancing the motor by EXACTLY 25 half-steps of 0.9 degrees (for a standard 400 half-steps/rev motor) will move the sprocket wheel around by exactly ONE FRAME (within the accuracy limits of the steppers which should be pretty good). The half-steps are just that STEPS so position sensing should not be necessary.

So you don't need position sensing, you can just hit the stepper with a 25 half-step phase stepping sequence, between each frame to advance the film to the NEXT frame and everything should be hunky-dory.

One reason I specified a 1Megapixel Webcam or higher (again REAL sensor resolution NOT 'interpolated') was to guarantee that we could resolve to the film-grain limit, and also to provide a margin for 'software frame alignment' done later in post processing.

You had a good idea to do the capture at a higher frame rate and then decimate down with beneficial averaging.

I would design the PC parallel port with a 5th line in addition to the 4 lines providing phase drive to the stepper. This line would be able to shut down the LED backlight source, if you want to do this trick. This could be used to force a 'dark frame' between valid AVI frame capture groups.

Say we capture with a USB2 webcam at a full 30fps . . .

Then we budget capture frames like this ;

2 frames @ 30fps dark (guarantees at least one fully dark frame with async frame advance timing)

8 frames @ 30fps sampled (guarantees at least 7 frames for averaging with async frame sampling)

2 frames @ 30fps dark (guarantees at least one fully dark frame with async frame advance timing)

8 frames @ 30fps sampled (guarantees at least 7 frames for averaging with async frame sampling)

Sounds slow, but this is still 3fps or about 1/4 to 1/5 ‘real time’ for the slow 12-15fps 8mm frame rate (not too bad).

Then the valid frames could be extracted and averaged with a fairly simple AVISYNTH script.

Some Webcams (like my Phillips Vesta) also give the option to capture using a TWAIN driver (Thought I saw a Twain plugin for AVISYNTH somewhere). This can give better quality by transferring a full 24 bit frame, so this would be a good alternative if a plugin can be found or custom driver written.

The Webcam would definitely give a less cumbersome and more compact ‘professional’ looking assembly.

Given today’s low prices on computer related items a suitable Webcam should certainly be in the price range from 50-100 dollars.

As I mentioned before the surplus stepper should be only about 15 (to perhaps 30 dollars worst case) on the surplus market.

I don’t know how much LaVezzi gets for it’s sprockets (where is Spacely Sprockets when you need them?) ;)

[FredThompson]

Yes, I agree with you except I was planning for a full minute per frame. There will be some variability due to A/D conversion and I figured it would be a good idea to have plenty of frames. I was hoping to twist Donald Graft's arm to get a verion of his frame dupe replacer which doesn't have a max number of frames. The older version didn't have a forced reset period. The idea was to let a camcorder run then remove the junk frames (film advance) then run the remainder through dropout/scratch removal then averaging and keep only 1 of x frames which would be the "original" frames that would go through the final part of the filter for color restoration. The goal should be to do it "properly" then worry with optimizing speed. 8 frames might not be enough for averaging. I was thinking 60 frames so aberant pixels would have very little influence on the result. I'm not the most talented statistician and that was just a safe number.

8mm and 16mm should be supported, not just 8mm.

Audio, well, that's a whole other problem...

One of the guys I've talked with who did some of the scanner method told me he has a new machine shop setup. What it consists of, I don't know. Someone with a CNC setup who is willing to donate their time to our obsession would be greatly appreciated, wouldn't they? :P

A pure image webcam would be fantastic. I assume it would have to shoot more than the image portion of the strip and the user would mark the crop lines. Maybe the case of the webcam would be removed so the CCD could more accurately be positioned inside the device.

[Delphin]

Yes, I agree with you except I was planning for a full minute per frame.


Uhhh ... 1 minute could beeeee jeeeusst a leettle bit on the steep side.

I am an engineer so here's the skinny on the 'averaging' thing . . .

When we apply multiple noisy (but correlated) signals to a summing node (like by adding them together mathematically and averaging) the noise power doubles for each two signals we combine (no advantage so far, twice the noise is twice the noise)

The advantage comes in because the correlated signals add more efficiently and so the ‘signal’ power QUADRUPLES. So we end up4:2 or 2:1 or 3dB ahead. This has to do with the fact that voltages of correlated signals will add in phase and the resulting signal power is proportional to VOLTAGE SQUARED. (Noise signals only add linearly)

Double the Doubling and you can only get another 3dB

So here’s the breakdown (Where N equals the number of combined signals and S/N is the signal to noise improvement).

N = S/N
------------
2 = 3 dB
4 = 6 dB
8 = 9 dB
16 = 12 dB
32 = 15 dB
64 = 18 dB
128 = 21 dB
256 = 24 dB

Or generalized : LogBase2(N) x 3dB

Going to your example:
1 minute at 30fps = 1800 combined frames

LogBase2(1800) = 10.8 x 3 dB = 32.44 dB

This would just about double the S/N of a typical cheap Webcam except . . .

Sadly, it just isn’t going to happen because we will just start to see nasty 'correlated noise’.

This will happen because parts of the noise (for example due to interference inside the camera)
are also correlated with each other (like the signal) and therefore don’t cancel.

So, if we suppress random noise well enough, pretty soon this gets to be the limiting factor. :(

At a guess, the advantage of combining frames could ‘max-out’ between 32 and 64 frames (maybe a little more for a REALLY good webcam (but a REALLY good web cam will have less noise to start with):)

Some Webcams are great (look like DV video with enough light) others are junk. I would definetely read Web reviews then buy from a store that will let you make a return if not happy ;)


One of the guys I've talked with who did some of the scanner method told me he has a new machine shop setup. What it consists of, I don't know. Someone with a CNC setup who is willing to donate their time to our obsession would be greatly appreciated, wouldn't they? :P

A pure image webcam would be fantastic. I assume it would have to shoot more than the image portion of the strip and the user would mark the crop lines. Maybe the case of the webcam would be removed so the CCD could more accurately be positioned inside the device.

With some care, probably the only part that would be so critical that it would need to be 'machined' would be the sprocket roller, and there seems to be a company that sells the ready made part (wonder what there price is?). I think other parts could be fabricated with careful work using common hand tools.

To support multiple movie formats, multiple rollers could be put on the same stepper shaft but some provision would have to be made for changing the film path and image scale (this could be as simple as repositioning an external triplet magnifier used with the webcams lens).

With the right optics the film frame will fill the Digital frame with only a tiny 'justification' border (which could be used for digital frame alignment correction). On the other hand with, say a 2megapixel camera there would be enough resolution that the frame could just be sized (magnification wise) for the 16mm format. Then we would just put up with the smaller frames used for 8mm and super8 using proportionally LESS of the frame (this would certinly be the simplest).

[Arachnotron]

A pure image webcam would be fantastic. I assume it would have to shoot more than the image portion of the strip and the user would mark the crop lines. Maybe the case of the webcam would be removed so the CCD could more accurately be positioned inside the device.How about modifying a 35mm slide viewer and point a camera at that in a fixed setup? Something like this (http://stores.tomshardware.com/search_getprod.php/masterid=743798//).

[Delphin]

How about modifying a 35mm slide viewer and point a camera at that in a fixed setup? Something like this (http://stores.tomshardware.com/search_getprod.php/masterid=743798//).
I used a simalar rear projection unit with my movie projector to capture 8mm but the idea here was to avoid the problems of having to project an image.

The 8mm or Super8 frame is tiny, and so is the CCD in a typical web cam, so why enlarge the image just to shrink it again?

With only a small external lens, a webcam (or other ccd camera) should be able to directly image a backlit 8mm frame with no problems.

We just need a reasonably precise 'film gate' and a little hardware to hold the frame in place and keep the webcam optics in alignment and a good way to advance the film by exactly one frame at a time (which is what I was proposing the stepper motor for).

Your idea is closer to the tried and true method that some folks have used for home movies, but what Fred and I have been working on is closer to how a professional Telecine machine works and therefore would hopefully give better quality.

Thanks for your interest, and the suggestion :thanks:

[Arachnotron]

I used a simalar rear projection unit with my movie projector to capture 8mm but the idea here was to avoid the problems of having to project an image.Fair enough :)

I just have two last suggestions, and then I'll butt out :)

First, check out amateur astronomer websites. The typical rig involves a home made computer controlled setup with the telescope being positioned by stepper motors, while a cheap CCD (often a webcam) is used to collect multiple exposures to be averaged for optimim picture quality. So most of the software modules and interfaces you are going to need should already be available over there.

The second is to use a printer or pen-plotter as a source for the stepper motors and interfaces. Maybe not for the final product, but for development they have the advantage of being accurate while you can position them using basic printer commands so you don't need to develop any interfaces or software just yet.

Good luck with your project!

[Delphin]

Fair enough :)

I just have two last suggestions, and then I'll butt out :)

First, check out amateur astronomer websites. The typical rig involves a home made computer controlled setup with the telescope being positioned by stepper motors, while a cheap CCD (often a webcam) is used to collect multiple exposures to be averaged for optimim picture quality. So most of the software modules and interfaces you are going to need should already be available over there.

The second is to use a printer or pen-plotter as a source for the stepper motors and interfaces. Maybe not for the final product, but for development they have the advantage of being accurate while you can position them using basic printer commands so you don't need to develop any interfaces or software just yet.

Good luck with your project!

Yes, those are very good suggestions . . .

It looks like you and I were thinking almost along exactly the same lines.

I was one of the first, several years ago to take the lens off of my Phillips Vesta webcam and hook it to my Takahashi Refractor. There were a few issues to work out to get really good quality (for example you need an IR filter with a Refractor) but it’s amazing how well this works.

Like some others, I found out you could get better image quality by ‘stacking’ images This just made sense based on my work in electronics, were adding correlated signals to reduce noise is commonplace (as I described above). At first I just used Photoshop to combine the images, but then some clever folks came up with about a half dozen programs (like Astrostack) to automate the process.

Then a guy named Steve Chamber's figured out how to modify the Vesta webcam so you could take VERY long exposures (several minutes), and things REALLY started to take off. That's one of the great things about the web, someone tosses out an idea, and someone else really runs with it!!! Steve deserves the credit for turning the webcam into a serious imaging tool for amateur astronomers. Now someone working with very low cost equipment can get images rivaling the beauty (if not the scientific accuracy) of some of the early Palomar astro-photography.

I also fiddled with the Mel Bartels Stepper controller (one of the telescope controllers I presume you were talking about), though I now use the commercial SkySensor 2000.

Mel's site was a great source of links with information about steppers. By comparison this project should be fairly trivial (though Bartles code might come in hand if it does turn out the we need micro-stepping and PEC to get perfect frame alignment).

As far as the obtaining the steppers goes, I found a big pile of very nice 'new surplus' NEMA 23 frame 6-wire Vexta motors at a local surplus yard so I am pretty well situated in that regard, but you are right, old plotters, printers (and some VERY OLD full-size 5inch 10-80MB hard drives) are all excellent sources of FREE stepper motors.

Another great source of Stepper driver info is to do a search on amateur CNC (Computer Numerical Control) milling machine links. I learned this when I helped someone do a Sherline mill conversion.

Again, thanks for your input, it will be interesting to see where this goes.

[FredThompson]

This is very similar to what I had envisioned except there would need ot be motors and gears and such behind the device. There's also the issue of aligning the camera perfectly parallel with the film path.

[FredThompson]

Got an email from someone who rekindled my interest in this project and saw this on Slashdot (http://games.slashdot.org/article.pl?sid=05/10/16/1544219&tid=100&tid=207&tid=137)

Craig writes "PSP Owners have long been interested in watching the UMD films and playing games on the TV, well now according to a report from Lik Sang they can, the new PSPTV being produced eventually by Gametech will be a no modification addon. From the article: 'The TV Adapter for PSP lets you hook up your PSP to your home television (NTSC and PAL) via Composite or S-Video and Stereo connectors. This adapter requires no modification of your PSP console. This new peripheral takes a completely different approach and clips on top of your PSP screen, with two screws to fit at the back of the handheld (in these two holes you can see on the top of the UMD drive). Some sort of pyramid grows from the base, with a precision lens and mirror system at the top, capturing the image and light, in a similar way a scanner or camera would. It then converts it into a video signal that is sent through video leads going from the adapter to your TV set.'"
Could a prism could be used to connect the "light table" portion to an optical sensor? Diffused light would go through the film strip, through the prism and into the sensor. No need for a bulky tunnel and, hopefully, easier to make with proper alignment.

There's still the issue of stepper motor control to move the film strip.

[Arachnotron]

Maybe you could use the (penta)prism from an old SLR?

[FredThompson]

Hmmm, maybe the body could be used as a mount, too. Might make it easier to get the prism/camera element alignment proper. Fo that matter, maybe some kind of macro lens. Hmmm...you just might be onto something there, webhead!

[FredThompson]

Here's stepper motor control through parallel port code: http://www.codeproject.com/vbscript/Stepper_Motor_Control.asp

[dokworm]

I was just wondering what the problem is with stripping down a $30 second hand projector?
You already have the gate, the drive system, (effectively) the trigger mechanism, the pressure plate, the sprocket feed and the reel holders - Just widen the gate out a bit.

All you need to do is replace the motor with a slow powerful motor like one cannibalised from an old cassette deck.

A piece of opal glass and a high brightness white LED (coupla bucks) provides the light source, and a webcam like the ToUcam for 640x480, or one of the new logitech true 1024x768 webcams (<$100) on the other side focussed directly on the film in the gate.
Run the projector at a few fps, have the PC triggered either by a cam and microswitch on the driveshaft, or an optosensor firing at the shutter and use cinecap (free) and you are done.

Half a days work, and maybe $200 and you have a true 1024x768 telecine rig. Done.

Spend a bit more and you can have a USB2 or Fireware CCD camera that can do true HDTV resolutions at up to 6-8fps.

Seems easier to use an already precision machined projector as your feed rather than try and build a feed mechanism from scratch.

To summarise.

Get a projector, slow it down, widen the gate, attach a trigger, stick an LED light source and diffuser in it, point a webcam with a decent lens on it at the film gate and run cinecap.

Now I know that 'old projectors' don't make for an easy 'kit' type setup, but there are plenty available cheap, and you could probably amass plenty of a particular type if you went hunting.

You could still make up a kit of almost all the parts, and people just bought their own projector to finish it off.
It could consist of the light source, the webcam and appropriate lens, the opal glass diffuser (also standard size for most projectors), the microswitch or optical switch, a low rev high torque motor, and the software etc.

The light source (a led on a board with plugs etc. to be a slot in replacement for standard 12v projector lamps) would be asy enough to churn out. Most of the other stuff is off the shelf, but many people would prefer to buy it in a kit rather than trying to track stuff down from everywhere.
Basically people could then just get a suitable projector and then retrofit it with the kit.

[FredThompson]

I've been thinking about your post. Seems like a good idea. I wouldn't mess with a trigger for the capture, though. Just use a lossless capture and an AviSynth script to throw out the junk frames which should have a consistent pattern. The original idea was to use hard drives/lossless compression/scripts to eliminate some of the physical components and allow averaging of multiple images.

A good idea, indeed. I notice VideoFred hangs out here. Think I'll point him at this thread and if he's interested.

[Fizick]

FredThompson,
Do you know about my GetDups plugin (for AviSynth) to select needed cuptured frames?
Only small 8mm projector modification is needed (removing obturator).

[FredThompson]

No, I hadn't thought about using that filter. I thought it would be something like keep 10, discard 10, repeat and the "kept" frames would be denoised and averaged. I do love your filters, though.

If we got VideoFred involved, we'd have 3 continents involved :P I guess we could add an AviUtl part of the filter chain to make it 4 continents...

There is a man who has been doing a number of experiments with flatbed scanners who returned my email a few months ago. He said he has some new machine shop tools. Maybe we could get his help for the physical portion and we could work on the filter chain. I'd be more than willing to send you parts if it is safe now. A few years ago it wasn't safe to send things into Russia. I guess an American return address was too tempting for theft. I could carry them to Italy and send from there.