@jphfilm thanks for posting! I read through your wiki. You did a fantastic job of documenting your project - incredible that so many areas were ‘firsts’ for you. Bravo. Really impressed.
I was especially intrigued by your use of network streaming as a way to avoid slow file write times. I had never thought of that and it could be extremely useful for Kinograph. I’m thinking specifically of a situation in which multiple machines might be running in the same room. Instead of collecting cards or drives at the end of each scan, they could each be streaming to a centralized server. Awesome!
Some ideas for the transport and maybe some software could be gotten from it.
But I think jphfilm’s scanner already has this and even more. From a software standpoint I hope we can stick to something like opencv and don’t have to touch avisynth. Although that may be wishful thinking as jphfilm’s results for cross frame noise reduction and stabilization are just too impressive to get ignored.
Hello, I am brand new here although I have been examining the Kinograph instructable and this forum for about a year now. I have taken aspects of the Kinograph and have examined other homebrew scanners to develop something myself. While conducting a search today I found this listing for a Wolverine 8mm/Super 8 device. It seems too similar to their F2D series, but I am curious to see how a real life transfer looks.
Jitter doesn’t look to be the main problem. Digital noise / artifacts though is!
It seems indeed very similar to their other F2D devices. I own two of them and I’m not too impressed with the image quality. I like the speed of digitizing 35mm slides which is good for quick online copies. But not the best quality for archiving.
Drawbacks: Lack of dynamic range - of course - and quite a bit of noise in dark areas. The biggest problem with fading stock scanning on the F2D is the lack of active white balance. They do their exposure control on the camera instead of adjusting the lighting. Similar for format variations: 110 and 8mm formats are achieved by cropping. And dust is also an issue requiring good dust removal before any scan.
I assume they don’t do too much of the software cropping in the film version, but I don’t see why they would abandon their zoran coach platform. They already have a good firmware base. The normal F2D hardware is certainly not powerful enough to do H264 encoding, so they must have added a single chip encoder.
Great find! At 04:10 you get a shot of the gate. It seems it’s spring loaded from the top to account for shrinkage. There is also a reflection of laser dots in the film. It looks like a single column of laser sensors. That would do well for all gauges of film if the spots are close enough.
Anyone know where we might get a closer look at one of these machines?
I built this one a few years back. we use it for our regular 8 transfers. https://vimeo.com/43629295
And don’t forget the standard Spirit Datacine, still the workhorse for professional work-- here is ours www.electricpictures.tv
I like your swingarm design. Did that have a feedback mechanism to measure tension? Also, would love to see details of the gate if you have and close-ups. Thanks for posting!
We called them Dancer Arms. The feedback is from Allen Bradley 845 encoders, which are industrial robotic devices. they are basically rotary counters that emit pulses that measure position or speed. I used a third encoder as a tachometer to generate the pulses that trigger the light and the camera. You can see them in the video in the rear view.
Years ago, we build a gate for running super8 on our then-current Philips Quadra telecine, which predates the Spirit. This is the shoe that was on that gate. There is nothing too complicated about it. The lamp is a 100 watt led, and they are mounted in a modified lamp assembly from one of the Quadras.
Grace, I was thinking about using a tachometer or rotary counter for generating a trigger pulse, but have some conflicting thoughts. I wonder what you think about them:
Using a a device like that will require sprockets on at least one roller. If film is shrunken or fragile, it may not work at all if it can’t line up with the sprockets.
[counter-argument] - if the sprocket roller is designed correctly, it should be able to compensate for shrunken film by a.) minimizing the number of sprockets engaged at one time, and b.) a gentle sloping sprocket tooth that will guide the film into place.
What happens if there is slippage, or some error in the film path?
One forum contributor pointed out that using a phase-locked loop would effectively average out these anomalies.
Do we really need it? If we rely on optic sensors at the gate that are watching for sprocket holes, there is no need for sprockets nor tachometers/encoders. That makes everything cheaper and simpler.
[counter argument]: this would definitely require some sort of averaging, as films are very likely to have perf damage . If that damage is longer than just a few frames, the average could be thrown off significantly enough to affect the overall phase of the trigger signals.
CC’ing @VitalSparks because I think he’ll have some light to shed on this too.
I used the tacho with sprockets as a first step and it works, but you are right, it has trouble with some film types. You can do the same thing with a capstan if the film path is set up correctly, but that has a slightly different problem with damaged film. To make a capstan work well you need at least 50%–180 degrees contact. Capstan drive works well this way.
I have prototyped a laser sensor which will probably be the answer. Keyence makes a retro-reflective laser sensor that I tested and built into a circuit. The key properties of are that it is tune-able, and that it can distinguish between a clear material and empty space. So it can detect a sprocket hole on clear film. This is a difficult task, but they and other manufacturers have figured it out, so it is off-the-shelf tech.
The other thing is the sensors must be very fast. With exposures in the 20-40 microsecond range, the sensor needs to be able to switch at a tolerance of 1-2 microseconds.
I had the same hunch about refelctive sensors and have had good results in my tests. From a user perspective, you would have to align the sensor while the film is running when you change film gauges to make sure you have good alignment, then rewind and start over. Not ideal, but would still work.
In another thread, there is a video of the Kinetta which appears to be using a column of reflective or laser sensors to solve that problem. By using a vertically stacked array of sensors, you should be able to detect perferorations for any film type.
Again, this does require that the film be properly repaired if it is missing long sections of perforations on either side.
REWIND: yes, you need to be able to rewind. That is an interesting feature by itself. Any transport that will be useful in a production environment needs rewind capability. My transport is fully controllable for speed and direction. Any film you are going to transfer needs some setup for color and exposure, and then rewind to your start position for recording.
Sprocket position adjustment. If you look at the Muller, they used linear stages to move the sensor and the light. My plan was to build both the Keyence laser and the reflector into a platform that is mounted on a linear stage. Then motorize the stage with a stepper and control its position with presets for different guages or film types.
For missing sprockets, you can code an alternate trigger in software that fires the camera/light array if pulses are not delivered from the laser. if you keep track of the timing of the laser delivered pulses the code could be substituted at the same rate. Not that I actually know how to do this, but theoretically.
About the transport. One thing to note is the need for a rigid transport. The whole platform may not need to be rigid, but the core gate, light, sensor, and camera spine must be very rigid to isolate it from any vibrations. If any motion is induced from the rest of the transport, the images will be unstable.