Virtual Telecine of Super 8mm Sound (24 fps) cine film??

[doc_cyclops]

Hi folks,

This is a long post so bear with me.

Having researched this for a couple of years (and finding less than satisfactory information for what I wish to do), I have developed a theory on how to effectively copy my large collection of Super 8mm magnetic sound films @24fps to Digital Video (29.97fps).

Now I have far too many films (we're talkin' hundreds here) to pay for a "transfer" service and WorkPrinters do not handle sound.

Before I spend even more money purchasing an upgraded computer, a new minidv camera, audio adapter, vid capture cards and editing software, etc, I wish to run this theory by the AviSynth / VirtualDub / Decomb / etc. experts to see if I am of sound mind. :-)

The set-up: (hang on, I will get to the frame server bit in a second)

My Super 8mm sound projectors have a 3-blade shutters as well as as potentiometers to perform variable speed adjustments. I am able to adjust the projection speed of the film to 20fps (or 19.98) and with the 3-blade shutter, this would produce 3 images per film frame (20 X 3 = 60). This should sync nicely with a video camera locked at a shutter speed of 1/60 and would yield 3 NTSC frames per film frame with virtually no flicker. Yea!!!!!

Audio output from projector would be fed through a BeachTek audio adapater and into camera mic input. Flickerless movie now also has the audio track. Yea!!!!

Speed of recorded film is 20fps, the action is noticeably slow, and everyone (even the women) sounds like Lurch from the Addams Family. Boo!!!

My "mad" theory (with questions):

The frame pattern on the video should be something like below, depending on the relationship of the projector shutter cycle and the video fields.

(Each letter being a frame of film and each number the projector shutter cycle count for the frame).

A1 A2 | A3 B1 | B2 B3 | C1 C2 | C3 D1 | D2 D3 | E1 E2 | E3
or

A2 A3 | B1 B2 | B3 C1 | C2 C3 | D1 D2 | D3 E1 | E2 E3 | F1
or

A3 B1 | B2 B3 | C1 C2 | C3 D1 | D2 D3 | E1 E2 | E3 F1 | F2

Assuming the Frame pattern is:

A A B C C D E E F G
a b b c d d e f f g

can I use AviSynth with the Decomb filter (Telecide/Decimate) to do the following?

Telecide() # would replace the blended/combed video frames with a duplicate, 1 duplicate frame for every 3 frames.

A B B C D D E F F G
a b b c d d e f f g
0 0 1 0 0 1 0 0 1 0

Decimate(Cycle=3) # removes the 1 duplicate frame out of every cycle of 3 frames. As I understand it, this action would also change the frame rate to 20 (or 19.98) and the audio would remain untouched and thus in sync with the adjusted rate video?

A B C D E F G
a b c d e f g
0 0 0 0 0 0 0

AssumeFPS(23.97,1,true) # adjust frame rate back to 24 (well 23.97). I again assume from what I have read out here that this would speed both the video and audio up?

SSRC(44100) # resample audio

AssumeFrameBased # perform "telecine"

SeparateFields # split into 48 fields

SelectEvery(8, 0,1, 2,3,2, 5,4, 7,6,7) # perform 3:2 pulldown

Weave # create 30 video frames

AssumeFPS(29.97) # show at NTSC rate, audio still plays at last sampled rate

If that works, it would also be possible to transfer silent films (18fps) via this method as below:

Telecide() # replace combed frames with dupes
Decimate(cycle=3) # remove dupes
AssumeFPS(17.98) # restore frame rate
AssumeFrameBased # perform "telecine"
SeparateFields # split into 36 fields
SelectEvery(6, 0,1, 0,3,2, 3,2,5, 4,5) # perform 3:4:3 pulldown
Weave # create 30 video frames
AssumeFPS(29.97) # show at NTSC rate

If my understanding of all this is correct, then using this method, I should get better results than just aiming a camera at a projected image on a wall.

My overall question to you experts is:
"Is what I proposed from a AviSynth/Decomb perspective possible as I have described?"

Any advice would be appreciated.

Thanks!

Doc Cyclops

[jggimi]

Hello, and welcome to the forum!

After reading your post several times, I've come to the conclusion that it's better off being handled by the folks in our AviSynth forum. The Capture Forum is really for capture of analog video sources. Possibly, our DV forum members may want to take a peek at this as well, since there may be some issues with shutter speeds and/or DV fps that would need to be taken into account.

I'll post there and point to this thread for you, you may get some additional suggestions that way.

Personally, I've had some experience with Single 8 transfers, but only after they were transferred to VHS at 19.98fps. I kept them at 29.97, with 3 fields per film frame. Single 8 is silent, so audio wasn't a factor.

[doc_cyclops]

Thanks jggimi!!

I appreciate the help.

Doc

[jggimi]

You're welcome. I've posted a link to this thread in the DV forum. You might get some additional insights/suggestions from that community as well.

[Mug Funky]

a couple of notes:

best to resample to 48000 rather than 44100, as you'll have to go to that rate anyway for putting on DVD etc, but i can't envisage a situation where you'd be burning audio CDs of the same stuff.

for the telecine process, there will no doubt be a bit of flicker going, and also, every 3rd field will most probably be a blend. that's okay, because we've got deinterlace filters (tdeint springs to mind) that will seek out the best possible fields to interpolate from (you may not need to interpolate at all). this leaves you with 60 fields per sec, and in want of a sure way to get rid of the blends. the restore24 script should be able to handle this pretty well, with the advantage that it'll deal with it if the field-sequence drifts a bit which is very likely. the output from that will be 20fps if you used it right :). now all you need do is assumefps 23.976, and telecine it.

if you want a 3:2 pattern with a minimum of fuss (long selectevery strings melt my brain), just do this:

changefps(59.94).separatefields().selectevery(4,1,2).weave()

to get top-first output (suitable for DVD. use 4,0,3 for the bottom-first DV format).

apart from that, your method looks just like what i'd do. one thing - make sure you run all your equipment off the same wall outlet if you can - better to be safe than stuck with a ground-loop (i'm not sure if this would really be noticable though).

[doc_cyclops]

Thanks Mug for the information.

Some questions/comments:

The 48000 audio resampling is a great tip. My intent was to transfer the results to DVD.

About the flicker:
In theory (and I say "theory" because I have not experienced this empirically for myself), having the 3-bladed projector means 3 "hits" of the shutter per film frame. Having the projection speed adjusted to 20 fps (19.98 fps), this translates to 60 (59.94) shutter "hits" per second. With the miniDV cam shutter speed locked into 1/60, both the camera and projector shutter should be relatively in phase and flicker should be minimal.

About field blending:
Again, with the projector/camera set up above, the results would yield 3 NTSC fields per film frame. There would be blending, but it should be at frame level (top field is one film frame, bottom field is the next film frame) and not at field level (where the top field has information from two consecutive film frames, and the bottom field has information from the second film frame). I would imagine that field level blending would be the effect if I simply video'ed the film projected at 24 fps.

Hmmmm....just a thought here:

Since my projectors are not crystal-synched, the chances that they will remain at a precise 19.98 for the duration of the projection is probably not likely so some field drift may occur after all. :confused:

Actually, I was going to adjust the speed while video'ing some test film and continue to adjust until the flicker went away (inferring that with no flicker meaning shutter phases are synched, 1/60 to 1/60, my projected frame rate would be correct).

jggimi mentioned he had some film transferred to VHS at 19.98 fps (projection speed) yielding 3 NTSC frames per each film frame. I should ask him if there was any noticeable flicker.

TDeint() and Restore24()
I will have to do some more research on these then. I assume that these functions would be preferable to telecide and decimate because they can handle potential field drift better.

Ground Loop:
I will be sure to run both projector and camera off of the same wall socket. Thanks for that tip as well.

Thanks again for your help and for your time.

Doc

[Backwoods]

Even if you do have a correctly timed and flicker free transfer, it will not look the best it can.

If you have hundreds, I would purchase the Workprinter. Do your transfer to the computer with a nice DV camera then purchase a cheap projector that plays sound. Then transfer the sound from your projector through your camera to the computer, then sync the sound. This will give you the best quality you can get. Any other method will give you very poor quality compared to the method listed above. It really isn't worth all that trouble for the quality you will receive.


I can help you out if you do not want to do all that yourself for cheaper than anyone else would. I use the a workprinter. Shoot me an email bone785@aol.com and I could even mail you a sample on DVD.

[doc_cyclops]

Backwoods,

Thanks for writing to me and I sincerely appreciate your offer of help. I examined the workprinter option for quite a while. They even have a set-up almost identical to mine with the 20fps projection speed and everything. Although I think they are great machines, after weighing the pro's and con's of their's vs. mine, they are not my choice for my film-to-video project.

With either solution (a workprinter or my set-up), I still have to post process the captured film and therein lies the real focus of my information quest. Using AviSynth (and associated filters), how can I take captured NTSC miniDV video of a 20fps film source, remove the blended/combed frames, restore the frame rates to intended rates (18fps or 24fps), perform the appropriate pulldown process to get back to NTSC, and all while simultaneously maintaining audio sync and speeds.

Thanks again for your help.

[Backwoods]

Sure, no problem.

Now what will you be recording from? What will your projector be facing that your miniDV camera will record from?

While the workprinter runs at either 1 FPS or (roughly) 8 FPS, it records each frame perfectly, no combing or blending. With your setup you will see such either way, even after AVISynth filters. They will be greatly minimized but still there.

Best thing to do is just test and retest over and over til you are satisfied. You can compare your tests to a sample from the workprinter and judge from there. I can post a sample for you if you like.

[doc_cyclops]

I plan to use a 45 degree mirror/condenser lens assembly like the workprinters use, although I will have to fabricate one as they are no longer sold outside the workprinter packages (I wrote them and asked).

I have seen samples of the workprinter results and they are quite impressive. My set-up should be quite similar to their package that does real-time 20fps capture (not frame-by-frame). They have an enlarged gate and a low-wattage diffused light source, which I don't have. :( But I do have a higher reel capacity and I do have sound. :D

Most of my films are commercially produced prints taken from feature films (either abridgements or feature length). As some of them exceed the maximum 400' reel capacity of the workprinter, I would have to cut them down to fit on smaller reels and I do not wish to do that.

Also the idea of having to pass the film a second time to record sound and then attempt to post sync that sound does not appeal to me at all. As the sound projectors do not have crystal sync and will probably not hold an exact 20fps (varying slightly up or down and potentially causing the field level blending mentioned above), the audio would also have slight speed variances that could make post-synchronization a nightmare if not impossible. If the audio is "coupled" with the video during transfer, sync shouldn't be too much of a problem assuming the AviSynth filters work the way I think they do.

So, I acknowledge that my video reproduction will not be optimal. But, I will save wear and tear on my film, can do the transfers faster, don't have the post audio synch headaches, and I can keep my "project" costs down. So, I am happy for the time being.

Thanks again for your input.

Doc Cyclops
--Superstitious perhaps, baloney...perhaps not