Home-brewed 35 mm motion picture film scanner

[Robert Knecht Schmidt]

Maybe a DV forum is the wrong place to ask this, but I'm thinking of trying some experiments with 35 mm motion picture film, and I'd like to get the best quality possible conversion to, say, 2K for post editing, digital color correction, etc. I'm wondering if it's possible to do this without paying big fees to a lab that has one of those fancy scanner machines. Can a home 35 mm still photography film scanner be fed 35 mm motion picture film? Might there be some other way to build-your-own film scanner to input 35 mm film footage into the computer?

[Andrew Smith]

I think it could be possible, but at 24 frames (scans) per second of footage it might also drive you nuts.

In the very least you'd have to use some software to (a) scan, (b) calibrate the scans to be correct colour, (c) align all the frame scans and (d) re-collate back in to a video file with the frames in correct sequence.

And the round-trip where you finally output to film again might be the part where you finally lose your sanity and dissolve in to a whimpering mess.

So naturally I'm curious as to how you fare with this. :-)

Andrew

[Mark Keck]

Robert,

How much time and money do you have??? Seriously??? I went down this road a couple of years back building an 8mm telecine. There is some good info on the web, but you have to search for it as it's not easy to find. It's a fun project, you will learn a lot, but be prepared to spend more than you want and for it to take way longer than you think. Also, do you enjoy building things??? This is not a throw-it-together-in-one-weekend type of project. I spent 4 years trying different approaches and experiments. By the time I was done I had built three different telecines, the first two with less than desirable results.

My first attempt was a flat bed scanner method. Some of the things I learned about flat bed scanner transfers. While it is possible to use a flat bed scanner to ingest film, there are lots of issues... consistent alignment of the frame, automation of the frame advancement, speed of transfer, and what I like to term "chromatic alignment" come to mind. Also, this method requires a lot of sw. Here's a few links:

http://8mm2avi.netfirms.com/
Flatbed Scanner Digital Telecine (FSDT) Project
Halfbakery: Simple 8mm movie scanner
http://avidtech.tv/16mmtransfers/8mm-Film-Scanner/
http://www.kaimio.fi/blogs/harri/ent...storms_old_8mm
home.comcast.net/~shay.mozes/8mm.html

My second attempt was to replace the flatbed scanner with a Nikon 35mm film scanner. My problem here was solving the alignment issue, film advancement issue, and speed issue. The quality of the transfer was far better than the flatbed scanner, and as you're looking at 35mm this might be an interesting choice for you. I don't have any links for you as this was my own gremlin. Sorry I never took any pictures of this setup as I abandon it with little more that testing the feasibility of it.

My third attempt, and the one that work well for me, was to use an 8mm projector modified to advance the film at about 2 fps. With this I used a 6in diameter field lens and my HV20 as a capture device. There are several websites that have similar types of telecines.

Construction of a home made Telecine machine
moviestuff entry page

There are several others, google is your friend, but these will give you the general idea.

Mine was built on an adjustable track and stable bed... The whole thing is about 6ft long. I'll have to dig around and find the pictures I took of it. The projector was obtained from ebay, gutted, and the motor replaced with a car windshield wiper motor with a couple of pulleys get the speed down to desirable 0.5-3fps... it's adjustable. I added a circuit that actually measures the speed in fps. The 100W projector bulb was replaced with a 3W white LED with an adjustable current source to adjust the intensity. A diffuser was added to even out the light. The film gate was enlarged so that the entire frame is captured. Most of the other telecines I've seen use a first surface mirror, mine does not. The HV20 is mounted so as to focus directly on the film substrate thru the 6in field lens. There is a circuit that detects the film/shutter position and signals the computer to grab a frame of video. The HV20 is connected to a macpro (but could be pc) via HDMI and a BM intensity card. It will capture a full 1920x1080 frame with no compression, the film frame does not fill the entire video frame (letterboxed). Capture is directly into Quicktime. The captures I've been able to get from 8mm film are about as good as you can get. You should be able to do much better with 35mm stock.

I've had thought's of re-doing the camera using one of the new Canon or Nikon SLRs with a macro lens, but I wouldn't really be able to get any better resolution from the tiny 8mm frame. But this might be something to look into for a 2k capture from 35mm.

Good luck, let us know what you do.

Mark

[Robert Knecht Schmidt]

Mark, thanks for all the information and links.

It just surprises me that with all the high-resolution, high-bit-depth still frame film scanners out there, none of them have a perf/pin feed system with registration-accurate alignment.

And it makes me wonder if there isn't some way to marry the pin feed mechanism of an old projector or camera to the sensor and electronics of one of these scanners in order to make a home scanner system.

My dream at the moment is to get my hands on a Super 35 3-perf camera (a refurbed ARRI, perhaps) and be able to play with the footage in the computer at a fairly good resolution and dynamic range without having to pay exorbitant scanner lab fees.

Your telecine setup worked at 2 frames/second. That seems pretty fast. For my purposes, I wouldn't mind having a system that could only scan a frame a minute--just as long as I could be somehow assured of accurate interframe alignment.

[Mark Keck]

Originally Posted by Robert Knecht Schmidt

It just surprises me that with all the high-resolution, high-bit-depth still frame film scanners out there, none of them have a perf/pin feed system with registration-accurate alignment.

You have to be careful with the marketing numbers for resolutions… About the best you can do is a little less than 4kdpi. Anything quoted above that uses pixel shifting and software approximation…. Not really true resolution. This is true of all scanners that mere mortals can afford. Now if you want to run with the big dogs, I'm sure ILM would probably be glad to help you out.

Look at a Nikon film scanner (my second attempt)

Nikon | Imaging Products | SUPER COOLSCAN 5000 ED

It arguably has the best resolution on the market for the price range. I have the older Coolscan 4000 model and have tested it’s true resolution with 1951 USAF targets to be about 3600-3700 dpi. The reason I tried this is exactly what you said… make up the pin/perf mechanism and it should be good to go. One of the real funny issues with all scanners is what I call “chromatic alignment” When you look at a scan at the pixel level you will see that the RGB layers don’t align quite right, and this actually affects resolution by causing a slight blur to the image. My hypothesis on this is (1) that because the light is ever so slightly diffracted as it passed thru the glass on it’s way to and away from the film. Normally this is not an issue but when your source material is so small it adds up. And (2) the scanner was not well aligned from the factory. I was able to take images into gimp and manually adjust the Red and green channel to realign the image. It was usually just 1-2 pixels in both directions. It does help. I’ve seen this on 3 different scanners… 2 flatbeds and the Nikon film scanner. FYI… I saw about a 100-150 dpi increase in resolution doing this. To be technically accurate the resolution was always there, just blurred. Now how much resolution do you need??? For the 8mm film I was working with I tested the grain to be about 1800dpi, so anything more that this was a waste. With 35mm stock being so much bigger, you really shouldn’t need to push beyond the limit of available scanners.

I remeber an interview with George Lukus a few years back talking about the technical aspets of remastering the original Star Wars movies into digital. They were using a 4k scanner... but that's 4k across the entire negative. Not sure what that would work out to buts it's somewhat less than 4kdpi.

Originally Posted by Robert Knecht Schmidt

And it makes me wonder if there isn't some way to marry the pin feed mechanism of an old projector or camera to the sensor and electronics of one of these scanners in order to make a home scanner system.

This is what I was trying to do with the film scanner.

Originally Posted by Robert Knecht Schmidt

My dream at the moment is to get my hands on a Super 35 3-perf camera (a refurbed ARRI, perhaps) and be able to play with the footage in the computer at a fairly good resolution and dynamic range without having to pay exorbitant scanner lab fees. .

I haven’t really touched on dynamic range, but anything better than D=3.6 is pure vapor. I was actually considering using liquid nitrogen to try and improve on this much like is done in astronomical spectrophotometers (what was I thinking???).

Originally Posted by Robert Knecht Schmidt

Your telecine setup worked at 2 frames/second. That seems pretty fast. For my purposes, I wouldn't mind having a system that could only scan a frame a minute--just as long as I could be somehow assured of accurate interframe alignment.

If you have infinite time, one frame a minute would be acceptable. Assume you have 1 minute of film to convert…. That’s 24fps x 60s or 1440 frames. At a frame a minute it’ll take you 24 hours to scan that minute of film. If you’re working with the Zupruder film, I’m thinking that would be ok, anything else… well I don’t have that much time.

I would suggest that you look into using a digital SLR with a really good macro lens. Pull the image right off the substrate in RAW format. It’ll be way faster and probable better resolution that a scanner.

This was an incredible fascinating project for me, I hope it is for you as well. But honestly, if I had to do it again… I would. In my day job I’m an engineer… I like to build things. And I’m not kidding when I say it took me 4 years.

Best o’ luck,

Mark

[Robert Knecht Schmidt]

Originally Posted by Mark Keck

I was actually considering using liquid nitrogen to try and improve on this much like is done in astronomical spectrophotometers (what was I thinking???).

I don't know what you were thinking, but if I wanted more dynamic range out of a scanner sensor, I wouldn't approach it by trying to cool the sensor to reduce noise, I would do it by varying the intensity of the light source.

[Mark Keck]

Originally Posted by Robert Knecht Schmidt

I would do it by varying the intensity of the light source.

That'll only maximize the dynamic the sensor already has.... it won't increase the dynamic range of the sensor.

Here's a good read on the matter:

Dynamic range, 24 bit vs 36 bit

Mark