Cine transfer and rolling shutter

[chrisk]

Does anyone have experience of cine transfer using a video camera with a CMOS rolling shutter? I ask because I will probably need to use a Canon HV40 next time I make some copies. However I would expect some serious exposure artefacts across the frame unless the open period of the projector shutter can be held precisely in phase with the camera shutter throughout the transfer. I have only previously used a camera with CCD global shutter and have never been able to maintain exact synchronisation throughout, usually resulting in slowly varying exposure variations, but have been able to largely correct these variations using the Premiere 'Equalize' filter.

Any experience or comments would be appreciated.

Chris

[tom hardwick]

Interesting Chris, and like you I await the replies. I too have only ever used CCD cameras for capture but thinking about it logically I'd say that the CMOS chips won't have any adverse effects as long as the three-bladed projector can be set at 16.7 fps.

You say you've, 'never been able to maintain exact synchronisation throughout, usually resulting in slowly varying exposure variations'. Do you mean the image you get on your PC and output TV flickers, or varies slowly up and down in brightness? If so this is easily cured by varying the projector speed. All films must be projected at 16.7 fps and either slowed down in post (Standard-8) or speeded up (Super-8 at 18 and 24 fps). Premiere's equalizer is not the best way to cure this flicker.

tom.

[chrisk]

(If a similar post to this turns up it is because I thought I had submitted a reply before being called away, but it has disppeared, probably my finger trouble)

Hi Tom,

Thanks for your reply, I guess no one on this forum has yet had experience of a rolling shutter camera for cine capture. When I have the opportunity to set up a projector again, I will try it out and report findings.

I appreciate the need to set the projector to 16.7 fps and I normally use an electronic tachometer sensing a reflective tag on the shutter shaft to set the speed to 16.67 fps. However in my experience the speed does not remain constant during the projection of a reel of film and without manual intervention I have noted short and long term variations between about 16.3 and 16.8 fps as the load changes over the period. The projector that I use is equipped with a shaded pole induction motor for the drive and the speed is reduced by a variable resistor in series. I think that this is a quite normal method, but the slip of an induction motor using such a simple technique is bound to vary with load. I believe that some high end projectors used a proper electronic speed control method to maintain correct speeds at 18 or 24fps without mechanical switching, and no doubt some users have got into the innards to set the required speed. (If I remember rightly, a Beaulieu projector that I had some time ago used such a method). Although I attempt to manually trim the speed variations using up and down paddles during transfer, I have not been able to keep it constant, and the changes in phasing of the projector and camera shutters result in small exposure variations and this is where I find ' equalizer' to be useful if applied over a selected clip.

As I understand it, a rolling CMOS shutter acts rather like a still camera focal plane blind, so making progressive parts of the CMOS sensor active over the period of, I presume, one field or 1/50sec. During this period the projector gate aperture will vary from fully covered to fully open and the averaging imposed by a global shutter will not apply. I would therefore expect shading to occur over each field and this would not be easy to filter out. I am likely wrong in this theory and stand to be corrected, otherwise only a test will tell.

Chris

[tom hardwick]

My Elmo GS1200 has electronic speed control so holds rock steady speeds regardless of the spool size and projector load. My Eumig MK DL (for Standard-8) uses an asynchronous "shaded pole" motor (it has copper "loops" in it's laminated core) so can be speed controlled using a ordinary thyristor dimmer circuit. I speed control this machine mechanically by carefully positioning the drive on the rubber plate that's fitted to the shutter shaft. The rubber plate needs cleaning every now and again to keep friction levels high.

The CMOS shutter does indeed act very like a vertically travelling focal plane shutter in a still camera, such that the top of the frame is photographed before the bottom, giving the characteristic 'elliptical wheels' to moving vehicles. So I think you're right - CMOS cameras are about to introduce more transfer problems, even if they're used in their progressive scan modes.

Test it and see's the answer.

tom.

[Dave Watterson]

Why is everyone so impatient with film to video transferring?

Is it not simpler to work frame by frame? That is: move the film on one frame, photograph that frame with a modest stills camera or a camcorder passing images direct to computer. Move on to the next frame ... and so on. It would be slow but who cares? An automated process could run overnight.

That would banish flicker problems caused by mismatched projector/camcorder.

Even a modest stills camera would generate images much better than either a standard of high definition video frame - so HD transfers ! That would give you some leeway for working on individual frames if necessary to clean up the images.

What am I missing?

-Dave

[tom hardwick]

What are you missing? Enough time and money Dave.

A 30 minute reel of film contains 32,400 still images (give or take a few). If you can automate the single frame re-shoot process that's good news, but I'd expect a DSLR's shutter to last in the region of 100,000 cycles, so maybe it would transfer 3 reels of film before seizing solid.

I've just transferred over 30 hours of ciné film and realise that you can't leave the recording camera on auto exposure as

a) it tries to remove the fades
b) it underexposes church spires against the sky
c) it fluctuates as the black car crosses the scene

and so on. Your overnight camera would have to be constantly monitored.

tom.

[Dave Watterson]

OK but what about a camcorder feeding constantly into a computer ... the images would be captured by the computer ...?

[tom hardwick]

That's the way - project the movie film at about 4 fps onto a chip and feed that to a computer. I know people who are doing this (diffused LED as a cool light source) but again, every frame has to be watched because the receiving chip can't be left to its own (exposure) devices, and you can't leave it in automatic for reasons outlined above.

Then there's the mechanical side of things. The films I'm transferring have splices that are c. 40 years old, so none of them can be trusted. Then there are the ones that can be trusted but they generally lose the loop because the accurate perf pitch has been interrupted.

You can only lift off the audio at a smooth linear speed, so at some stage the film must be moved past magnetic heads at a constant 3" or 4"/sec. And the audio can be main, balance, main + balance or stereo. And this week it's open reel tape.

tom.

[chrisk]

Probably been done before, but would this work with no need to control the projector speed:-

The projector runs at normal speed with sound being recorded by the computer.

The computer continuously monitors the video out from the camera, without the camera recording.

A pickup added to the cine projector shaft sends a signal once per frame when the aperture is fully open to the computer and this cause the current frame to be grabbed and stored.

We end up with one frame (or field) per cine frame and these can later be rebuilt into a properly timed video with audio

This would need a simple bit of engineering and some programming.

The mechanical and other problems outlined by Tom would still apply.

Chris

[granfer]

The well established method is to run the projector at about 5 to 7 fps (not critical) with a switch fitted to the shutter shaft sending a signal to a computer once every frame at a time when the film is stationary and the shutter is open (there does not even need to be shutter; most people remove it).
The specialised Software (and there are many free ones) then captures the first available COMPLETE frame of video from ANY video camera, and that frame ONLY. The succession of video frames is then automatically reassembled into an AVI file exactly representing the original frames, and it can be reproduced at ANY speed, the Software adding or removing (or even blending) frames as required. Then normal editing software takes over.
Sound is seperately recorded at a covenient fixed speed to the computer, corrected for length to match the video file and to the correct pitch (which should be correct if the length is correct) and recombined with the video in normal editing software.
It is really quite simple and certainly NOT beyond the capabilities of anyone used to handling video/sound editing software. The degree of sophistication in the projector modification varies from very simple to complex, depending on the wishes of the experimenter...
But one clear fact emerges.... the "frame by frame" method produces transfers totally free of flicker. The only limitation is that of speed of transfer, and that is limited by the handling capability of the transfer software and the speed of the exposure control response of the Video camera employed.
Some experimenters are now achieving transfers "frame by frame" at speeds approaching 24 fps.

[chrisk]

Hi Granfer,

Thanks for that explanation, more or less what I was thinking of. As I have not done any cine capture for some years, I guess techniques have moved on, but having recently moved house I have found a few more reels that I would like to transfer.

A few thoughts and questions, probably already answered by those using the method - With the projector shutter removed, even a CMOS rolling shutter camera would probably be OK. I expect the video input to the computer must be via the analogue port as firewire would introduce too much delay after the projector sync pulse, although this could probably be compensated. As the projector and camera run asynchronously, I am assuming that a low projector speed is necessary to ensure that full frames of interlaced video are captured but wonder how is it ensured that captured frames are always in the same field order - does the grabber software detect this? How is the once per frame signal interrogated by the computer, does it utilise the standard parallel or other digital I/O interface?. Can you say what frame grabber software (preferably free) is available that allows triggering by an external signal? I can't find anything that specifies external triggering on my first search.

As you say, once the file is assembled the rest is easy.

Thanks

Chris

[granfer]

i currently use a program called CineCap, but this is no longer available on the open market.With this, the "trigger" is effected with a microswitch in the projector wired across the "left" button of a computer mouse (mine is a USB mouse) Video input from the Analogue output of the camera is via an external USB "dongle" (I use either CameraMate or Pinnacle devices). But you can use an internal capture card with an analogue input.
The software has a slight delay (partly due to the USB trigger input), then detects the next full video frame (even field followed by odd field, then stop until the next trigger), so without a shutter at the slow transfer speed there is oceans of time to actually make the actual capture, which gives the camera time to adjust exposure between shots.
It is sometime since I tried any of the then available free programs; it would appear that they are now difficult to find as most have developed into sophisticated versions with a price tag! But trial downloads are often available... for example a popular one is http://www.capturix.com/default.asp?tar ... n=features
which they claim can work up to 30 fps (do not forget that other factors may effect attainable speed) and offers several trigger methods.

[chrisk]

Hi Granfer,

Thanks for the additional info. I see that Capturix can be purchased at a reasonable price and looks interesting, a trial version is also available so when I am set up again, I'll download and give it a try. My mouse uses bluetooth so may be a problem there, but the normal mouse port is fitted so should be able to wire it in. I also see from the Capturix overview that triggering through RS232 or joystick port is also supported and RS232 would be particularly useful as it is available on my main computer and by ExpressCard on the laptop.

Thanks again

Chris