Sprocket Registration


#1

Sprocket Registration

This is a big one. How do you tell when to take a new image?

This will depend largely on the optical/sensor solution we are using. In the case of a line-scan camera, the answer is almost certainly a rotary encoder. A roller’s rotation is monitored to determine the length of film that has passed. When the proper distance has been reached, a signal is sent to the sensor that closes the current scan image file and opens a new one.

If we go with a full-frame capture system (like a DSLR or traditional imaging sensor), then we will have two choices:

1 - keep the motion steady
2 - intermittent motion (like a projector)

Option #2 implies more parts and therefore more points of failure. If, however, we can design it well this is a valid option.

If we go with option 1, the concern becomes shutter speed of the sensor. Because the optical system is located very close to the film, blur becomes an issue with a constantly moving subject.

The solution is a balance of light (brighter = faster shutter speeds), processor speed (how many times per second are we checking for our sensor for a signal to take a picture), and sensor choice (dependability at high speeds and with various film stocks).

Right now, Kinograph uses a mechanical switch that is triggered by bumps on a sprocket. This insures that the signal to take a picture is always lined up with the film’s position in the gate (since the sprocket hole guides and bumps are on the same roller). It’s cheap and pretty reliable so far. It has not been tested at high , speeds however. And, I’m told that these mechanical switches can wear out (which makes sense).

Ideally, the solution would have parts that don’t wear out fast, are cheap, and can handle a variety of film sizes. In my mind, the rotary encoder system is a good option because it meets most of these criteria. I have not, however, tested the ones I have yet.

Many have suggested the laser or photo-interrupter option. This was tested (both with and without an IR filter) in the development of the first Kinograph. It worked great until I got a 335mm print whose base celluloid was clear instead of black. I could not get dependable results in this case and had to move to a mechanical solution due to time constraints. I would be interested to hear if there is a solution to that issue.

The other issue with a laser is that it would have to be moved when switching film gauges. This is yet another moving part and therefore another possible weak point in the system. If, however, the laser was incorporated in the swappable gate, then it might be a good option.

What does everyone else think? Anyone want to do some tests?


#2

Also, we have used these with great success on commercial scanners, perhaps you could build something similar:
http://www.mwa-nova.com/Products/Sound_Solutions/L._S._D..html


Film Advancement - moving the film
#3

I have had to replace the rotary encoders in systems where they reach end of life due to mechanical wear, the frustrating thing is that it doesn’t just “stop”. They tend to have little glitches and miss steps here and there until they finally die. Bummer about the clear celluloid for laser sensing! Would a laser proximity sensor be able to A) shoot through the tiny sprocket and B) stop at the celluloid?


#4

Damn, I wrote an extensive post here on this, but it disappeared, I’ll re-write it shortly.


#5

If using a continuous motion system, then a flashscan type light source is the easiest solution. The same trigger that signals the camera to take an image is used to flash the LED light source. The flash is fast enough to freeze the image on the film, and you can achieve 24fps quite eaily if the LEDs are bright enough and the camera is sensitive enough. I have been doing this for a while now with PGR cameras and a triggered light source, and a mechanical method of counting sprockets.

If messing with lasers, try using blue, I had better results in testing with blue lasers than with red.


#6

Cross-posting some posts from another thread that are better categorized here as Sprocket Registration. CLICK THEM TO EXPAND.

http://forums.kinograph.cc/t/film-advancement-moving-the-film/43/2?u=matthewepler
http://forums.kinograph.cc/t/film-advancement-moving-the-film/43/3?u=matthewepler
http://forums.kinograph.cc/t/film-advancement-moving-the-film/43/4?u=matthewepler
http://forums.kinograph.cc/t/film-advancement-moving-the-film/43/5?u=matthewepler


#7

This is great, thanks! Any idea what component parts are used, exactly? Maybe I can order one and do some tests. Do you have a link?

Does your Point Grey camera require a frame grabber?
What “mechanical method” of counting sprockets are you referring to?
What’s a “flashscan type light source?” Link?

Forgive me if I’m being nosy. You have a lot of great advice on the forums, just trying to figure out what machines you’re using/building/etc. If you’re keeping some trade secrets for your business, I can understand, however.


#8

No, the camera just streams via USB, so no frame grabber required.
The mechanical method is using tiny neodynium magnets on a sprocketed wheel with a hall effec sensor.
A flashscan style light source refers to an LED array that is pulsed for a few milliseconds up to about 300 milliseconds for each frame, rather than staying on constantly. You can flash each colour for varying amounts of time to adjust the colour mix.

It might be worth talking to these guys re sprocket sensing:
http://www.balluff.com/balluff/MDE/en/products/selector-photoelectric.jsp#/PHOTOELECTRIC?SAP_ZD_017=Slot%20sensor&Range_max.=up%20to%2050%20mm&page=0
The slot or angle sensors would most likely work.


#9

Yes Matt, by dumbed down I meant a cheaper low res sensor/camera just used for the openCV frame boundary registration. Although the reflective sensor control parts you attached could provide this functionality in a simpler solution, only requiring film base density calibration on each newly loaded roll.


#10

Hey @matthewepler I was reading about acetate blocking IR light. Apparently US soldiers with IR night vision cameras couldn’t see through acetate signs. If you were to get an IR-only laser of the right wavelength, you might be able to use it as you first thought, even though lots of film appears clear to visible light around the perfs.


#11

I have a scientist friend who can do a spectral analysis of a film sample very easily. Maybe there’s a specific wavelength of laser that will be blocked inherently by all acetate film regardless of whether it appears clear in visible light. I’ll keep you posted.


#12

Hi @MikeThibault,

do you already have any news about your analysis of the film sample and what light might be blocked by acetate film or a film with polyester base?

Thanks!


#13

I’ll be giving samples to my friend this week; she says that it’s a pretty straightforward process. I’ll be out of town until after Labor Day but I imagine I’ll get results fairly soon!


#14

I’m sending some film samples to a sensor company in Melbourne Australia to test, hopefully they can give us some feeback on the best solution for sprocket sensing.


#15

@MikeThibault and @Peter
Great, thanks a lot! Really looking forward to hear what you find out!


#16

Long time lurker, first time poster. Please be kind !
I have built the machine ( in my mind’s eye ) from end to end many times and the " Frame " accurate positioning remains the main challenge for a given budget.

I am considering the digital options. The first contender would be " mocha " by Imagineer. Mocha is a powerful, frame accurate tracking software and more. We could give the " eye " of mocha something to cling onto for registration each frame and let it go to work.
I understand the issue of dependency to third party resources, but hey this will get the cart out of the barn sooner.
I am planning a short test with a few off registration scanned frames that tacked with mocha. I’ll post the results soon.


#17

Hi Claw, welcome on the forums!
I’m experimenting with a photoresistor for frame counting at the moment. So far I get good results.
Thanks for the link!
Please keep us updated! Thanks!


#18

Looking forward to your test results! We’ve had someone else do tests with Blender and Fusion 7. Would love to see the workflow for Moca in a similar video if possible. Thanks for doing this, @Claw_No_More (nice name btw).

Links to the tests:
Fusion
Blender


#19

I can confirm that IR is significantly attenuated by transparent film stock. I have an IR thermal temp sensing gun, which cannot see a soldering iron through it.

My guess a simple ‘clipper’ circuit, used to detect zero-crossings in waveforms would do the trick if the IR detector was AC-coupled. Also, I would suggest an IR led as the source, like those used in surveillance cameras.

Jeff


#20

The IR absorption may vary across film stocks, I did a quick measurement between the sprockets of a film I had on hand and it is pretty transmissive in the IR:
so you would have to be careful about the diode/wavelength selection.

I think a visible wavelength would be fine and would make alignment easier, even if the film stock is transparent there would probably be a detectable difference between the sprocket hole and film.

I came across this system which uses a red laser (3rd pic in the slideshow):
http://www.filmfabriek.nl/muller-hds.html

Will