|Hardware: Laser T-962A TVM802A «FilmScanner» ePump Classic CAN|
Building a 16mm Film Scanner (Telecine) from scratch
Digging through my Dad's office and archive, I found a great number of films in 16mm format (1938 to 1964!). He has many more at home. It was hard enough to get a working 16mm projector, but watching the films was complicated and time consuming. It was smelly, too, especially after the projector blew a taar capacitor in the middle of the living room.
Obviously, I need to transfer those films to the PC and run them on an adequate HD screen.
I digitized all films by pointing an old video camera at the silver screen, but the quality is terrible (as expected). Having them professionally digitized is out of the question (it would cost me thousands). So what's more obvious than building my own film scanner? Nothing, right? ;-)
There is space for big reels at the top left and right. At the bottom left and right are two arms that keep the correct tension on the film automatically. In the middle are four more rollers for transport (one roller is connected to a stepper motor). An LED light shines from the top through the film into a digital microscope.
The microscope hast better than full HD resolution. Controlling the LED brightness, I can easily get a full HD image with HDR image data (12 bits per color!).
At the heart of the machine is an Arduino. It controls two quarter-scale servos that turn the reels, a stepper motor for transport, and a bunch of LEDs. Two potentiometers tell the microcontroller about the film tension. FOur buttons and a 2x16 display help the user to control the machine.
Inside the case is a power supply and a USB hub to hook up the Arduino and the Microscope. A host computer is needed to record images, but winding the film will be possible with the built-in controls.
The back side of the case. There are some strips for reinforcement. The material is PVC-covered HD particleboard that was laser-cut. If the machine will work, I will publish improved CAD files for everything.
Before mounting the reel drive servos, they must be modified for continous rotation. This was very easy for the servos I used. The built-in potentiometer is controlled by a clip in the top gear that can be removed with a small screwdriver. This gear has a bar that keeps it from rotation all the way and destroying the pot.
Remove the plate and the bar. File down the edges, or the bar will grind on the stoppers.
How does the tension thing work? Below the reels are arms that hold a loop of film. A spring pulls the arm down. If the loop get too loose, the reel motor will slow down to feed less film. If gets too tight, the motor will increase the feed. The receiving reel increase speed if the loop is too loose, and decreases speed when it is too tight.
This is a nicely self-controlling system. Running the film transport stepper at any direction or reasonable speed will change tension on the arms, which in turn will update reel motor speed.
This system is much nicer on the film than the slipping clutch in the film projector which puts an unpredictable tension on the film, especially when the clutch has aged and may stick or slip unpredictably.
Back of Reel
Back of Arms
Tension Arm Slot
Here is the complete machine. Large reels would hang over the sides of the machine, but stay clear of any mechanics. One pushbutton is mounted at the bottom left. The bottom middle slot is for the microscope. The middle slot is for the 2x16 display (nothing fancy here at all). Three axles for the wheels are visible. The bigger hole is for the stepper.
Arms with Roller Rods
Testing with Arduino
Printing the Rollers
I am printing the first roller that I designed in OpenSCAD. This takes 2 hours in best quality, so no testing until later this afternoon. The stepper motor is also in place, but needs a different hub on the transport wheel.
These rolls make sure that the film is flat and horizontal over the camera. The top right wheel is driven by a stepper motor for precise transport over the camera. At a first inspection, it seems that friction is too low to reliably transport the film strip. I will either have to make the roller turn more easily (use oil instead of grease), or connect two rollers with a chain (or all four), or add sprockets to the transport wheel. This would require another set of rollers. So glad that ABS extrution is relatively cheap.
Tensioning the film
Magic Magnetic Microscope Mount
Arduino hooked up
This is the Arduino Micro hooked up to the servos, the Polulo, and the pots. The Polulo is set to 16 microsteps to make operations as smooth as possible. Why doesn't any of my screwdrivers fit on the tiny pot to adjust the current?
Find the wiring error and win some free beer.
Wire it up.
After calculating and poking around in the docs, I was able to write a tiny firmware that self-regulates the tensions and the feed via the arms. Very nice! The stepper cotrol needs more thinking. We need to ramp up and ramp down the speed and make the rotation as smooth as possible to make the transport as reliable as possible.
One button must disable the automatic tensioning, so that the reels stop spining when inserting the film. The scanner must reenable it softly without jerking the reels around. The reels should have a maximum acceleration coded into the firmware.
Nothing yet. But I am happy to announce that my microscope works with OpenCV on OS X. I am very happy about that!
It's time to create a little protocol between the machine and the PC. I need to be able to send relative positioning commands and receive a status report, so I know when the end of the film is reached or any other event happened.
Bold text items have priority.
Also how about timeout commans, button assignments, menus for buttons and display?
Integrating OpenCV into an FLTK app to capture frames and find the index hole was very easy. Now I need to add code to send adjustment commands to the machine via serial port, read the machine status, and finally add the last grabbed picture to a movie file.
Easy as pie... .
A First Test!
How exciting! I managed to read the first frame out of the machine. Focusing correctly is tricky. I may have to add a guide window to keep the film flat. Quality and resoluion however is amazing. And with a black base color film, finding those index holes is child's play. Other film strips may be transparent around the index holes though, making registering a lot more sensitive.
Lighting the Scene
The new light source is a simple 12V LED bulb that I had laying around. This is good enough for B/W film, but may not be sufficient for color film. I am contemplating an RGB LED array to do color correction on the fly. But that is for version 2.0 or even 3.0.
A simple paper is used as a diffuser to spread the light more evenly. I will have to build a box, a film guide, and a better screen here.
A First Movie Clip
And here wa have it. The very first clip of 2 seconds, semi-automatically scanned by my film scanner. There is obviously a lot of room for improvement. More importantly though, the initial concept works.
Room for improvment:
Yes, yes, it's still chaotic in the back. I added a single power cord that runs a 3A/7.5V power supply for the motors and electronics, and a 5V supply for the USB Hub. The Hub connects the microscope and the Arduino to a single USB port and is working well. Wire managment is still an issue.
Pin Wheel Transport
Transport was an issue from the beginning. I tried to get away with rubber rings, but there is always some slip and the film starts drifting. Using optical recognition to find index holes is nice, but quite involved. So I finally decided to print a transport wheel with sprockets. To my surprise, it works quite well and does not seem to damage the film. The sprocket distance should probably be corrected by 0.2 milimeter or so. ABS is always shrinking an a bit unpredictable.
A Second Clip
Here is a second clip. This one is scanned with the new pin wheel. The result was run through a stabilizer.
After scanning these two seconds, the film came apart where it was glued together. I ordered some old-school editing tools, but I will also have to update the software to put as little stress as possible on the film.
I also improved the host software by adding communication, and a record button that tells the Arduino to advance by one frame. With this simple setup, I was able to record a short clip automatically. Yay!
Here is a scan as it comes raw from the software. The image is pumping up and down which is an indication that either the transport wheel is not exactly round, or the rollers have too much play. Waiting desperately for the new 3D printer... .
And here is the same clip after stabilization:
Here is the scanner at work. The firmware need a lot of improvement to keep the reel motors from twitching.
Remarks for Version 2.0
With a laser cutter in reach and less than 10 Euros of wood required to build the case, I am quite seriously thinking about building a new case for the machine. The current wiring is horrible, and transporting the machine is pretty much out of the question. Reliabilty is quite important, because the scanner will sometimes have to run unattended for several hours.
(c) 2012-15 Matthias Melcher