16mm double head to video

Serena wrote on 11/13/2009, 8:54 PM

I'm transferring 16mm film shot at 18fps in 1977 to DVD and have had the visuals transferred to MiniDV PAL. I understand that the film is run at 16.8fps for this. The film has two sync audio tracks (on separate 16mm sprocketed tape) and my plan is to lay all this onto timelines (Vegas) and do all the usual post work. The visual and audio tracks have matching head & tail sync marks, so it seemed a simple matter to align these (contracting the video track to match and restore the original speed). But there is a snag and maybe others know the answer. Aligning the tracks by sync marks (video and track #1 is all I've tried) doesn't get things in sync and if I manually match the start and end of the film (10 minutes) the centre is out by 2.5 seconds. There is a non-linearity that I don't understand. The audio tracks were transferred to digital 48KHz 24 bit by me. This "double heading" process always works well when the visuals begin as video, but then each begins life at the correct frame rate. I've tried leaving the video track "as is" and stretching the audio to match, but same result. Be nice not to have to go in and recut the whole thing; the only available material is the the final film cut. I have set "smart resample" for the video (playback speed 1.0705).

Back to post

Comments

johnmeyer wrote on 11/15/2009, 1:12 PM

I'm not convinced that changing the playback speed f the AVI is such a wise move. Vegas will still have to conform it to 50i, that means either it'll interpolate (if enabled) or add some form of pulldown.It is definitely true that Vegas (or some other program) is eventually going to have to do interpolation or pulldown to get everything to play right at 25 fps interlaced. However, the reason why I recommend this workflow (and use it myself all the time) is that it avoids having to do this until the last step of the workflow. This is important because if you let Vegas conform to the project setting which, in this project Serena is working on, is apparently 18 fps, then you already have duplicated and/or blended frames. Then, when you finally render to 50i, you're going to get a second set of duplication/pulldown/blending. So, you're going to end up with one strange pattern of duplicates.

And, all of what I said above assumes that you don't have a blended telecine to begin with (i.e., that the film was scanned by a Cintel or equivalent). Since that is NOT the case here, the final result will have THREE different blend: the blending done by the telecine process; the blending you get when Vegas stretches the video to conform to 18 fps; and the blending done when the project is rendered to 50i.

If you use AVIFrate, you completely eliminate one of those degradation steps. In addition, since we still haven't figured out what is causing Serena's problem, it has the added benefit that it lets her get on with her project (although it would still be very interesting to figure out why Vegas is causing sync to drift away, and then magically go back to perfect sync by the end of the video.

So, I would still recommend AVIFrate even if she wasn't having a problem, but since she does have this problem. it is probably the only way to get things finished.

farss wrote on 11/15/2009, 1:29 PM

John,
you're entirely correct however much of what I had to say was to justify leaving the vision alone and to try retiming the audio.

Bob.

Serena wrote on 11/15/2009, 2:11 PM

Thanks John. Understood. Clearly the best starting point would be to have the film scanned (giving also the best image quality) and then adjust frame rate using AVIRate. Using the cheaper process it may be better to transfer at 25fps rather than at "silent" speed.
Bob, I'm convinced the problem of non-linearity is a Vegas issue, not anything to do with the transfer system. I also encountered it a while back (without recognition) when I did a quick and nasty transfer using the EX1 to film the flatbed screen (again 18fps material). In that case I ran the machine at 25fps and shot at 25fps progressive. I was puzzled that in Vegas sync of the separately digitised audio had to be manually adjusted, but time and the film was short so I didn't investigate.
How did you test Vegas? I suppose you could video a clock and stretch it and plot out the time shown as a function of time. In this current project my audio track provided the standard against which the visuals could be compared. I got the same result applying the process to the audio instead of the picture; I suggest that all mechanisms for this external to Vegas can be neglected. I'd say that my plots of the error trend, with and without stretching, are definitive. I did say I found it hard to believe and hence my request for guidance. An interesting result, I think.

farss wrote on 11/15/2009, 2:28 PM

"How did you test Vegas?"

See my other post regarding a "Serious Bug in V9.0c"

I created a file consisting of 24frames of black and one frame of white repeated a zillion times. I rendered this out and changed the plaback rate by a very large amount, placed a marker on the timeline at each white flash and measured the time between them. It was either 15 or 16 frames, there was no consistant ramp or any such.

I've done oodles of long 25fps to 24fps frame to frame conversions and sync is perfect, the only gripe with Vegas doing this process has been the poor pitch shift algorthm Vegas used to use. If Vegas was doing some oddball ramp then the imact on the pitch shift in the audio would be obvious. No one has ever mentioned this problem.

I cannot for the life of me see how what JM is suggesting will change anything. Vegas lets us change the fps playback rate, over riding the value in the file header. Never found a problem doing this and I still cannot. I suspect if you manage to change the playback rate in the header then the outcome is going to be exactly the same.

I'd love for you to simply try changing to the audio, it'll only take a minute to ctl+drag the audio to get it in sync. If doing that creates the same problem then the problem must lie in the telecine process.

Bob.

johnmeyer wrote on 11/15/2009, 3:31 PM

message deleted because my reply didn't make sense.

apit34356 wrote on 11/15/2009, 4:02 PM

Bob, I agree with direction about testing rendering, done a few test ;-) looking for adding or missing frames that appears. Now, not to get to deep into number theory, there are large set of whole numbers the don't "round" well, causing offset errors when this numbers are used in computations that use multiplication/division in their "expressions". Fixed frames, ie 24, 60, etc, can use preset values, that work well....... but if the clip starts on a "bad" whole number, a small error can slide into the calculations, not as bad as computing each unique time/frame.

But I have seen this audio problem more than once, ;-(

Serena wrote on 11/15/2009, 4:59 PM

Well now, I've done my proposed "film clock" experiment (darn you Bob!), stretched it by 175% and plotted the visual time vs timeline time. The relationship is linear (ie I was wrong).
That's good to know, but I'm more puzzled than ever. Also in regards to previous rough transfer (mentioned above). As I said above, changing the audio got me the same result as changing the picture.

farss wrote on 11/15/2009, 8:29 PM

Sorry I missed your post about already trying retiming the audio.
I'm REALLY out of ideas and I have a looming deadline that I'm way behind on.

I have seen things get a smidge out of sync, I can understand how with resampling off Vegas can get out of sync by one frame, shortly after that point it will get back into sync though. How it can accumulate 2.5 seconds escapes me.

Bob.

Serena wrote on 11/15/2009, 11:55 PM

This has been an curious exercise and AVIRate is a very useful and easy to use tool and great for adjusting frame rate. The final conclusion is that the "telecine" wasn't constant in speed and I should have looked more closely at my linear plot of the difference between the audio and picture. Looked linear until I fitted a linear curve to it. Thanks for the assistance and apologies for startling up a hare.

johnmeyer wrote on 11/16/2009, 7:58 AM

final conclusion is that the "telecine" wasn't constant in speed and I should have looked more closely at my linear plot of the difference between the audio and picture. But weren't you able to play back the film and sound on your film editing machine and they were in sync?

If this was the problem, it is another good reason to always try to get a Workprinter, Cintel, or similar transfer where you have one frame of video for each frame of film. Then, at least that part of the equation is pristine and will not vary.

RalphM wrote on 11/16/2009, 8:07 AM

No apologies needed. I couldn't contribute much but I did pick up a great expression: "startling up a hare".

RalphM

Serena wrote on 11/16/2009, 2:12 PM

John, yes the picture and sound were in sync and fine, but after transfer to video and audio they were not. Absolutely agree about the transfer process, at least for this type of project where sync sound has to be married to the video. Obvious once learned!

WayneM wrote on 4/29/2011, 4:59 PM

I was searching for info and came across this thread which I am going keep a copy of since there is some great info in here. I am working on transfer/recovery of a 40 year 16mm project where I have the final cut film on 16mm and the matching audio on separate 16mm sprocketed mag tape. I won't face the challenges I hope that occurred here.

My question relates to another project.

I have about 2000+ feet of 16mm silent film from the 1920s. This was shot with a spring wound home movie camera and the majority is sharp as a tack and from a tripod. Includes a good deal of world travel including ship board and a variety of great scenes.

The issue is, I don't have a 16mm projector that will run at 18fps AND that has a telecine shutter. I'm going to do a simple aerial image transfer to digital via video camera from at 24fps projector with five bladed shutter.

Am I likely to be able to do a decent job of bringing this back to speed by transfer at 24fps and then in Vegas Pro (I have 8.0C) reducing the clip playback speed to bring it to that of the camera. I’ll probably go with resampling off also.

Note that I didn’t mention the likely frame rate. I believe the camera was a Kodak 16mm camera (like I have in my collection) and it did not have a variable, at least intentionally, frame rate. I think I will just try a couple speeds and see what looks best. Precision isn’t a big deal and there are some interesting areas where I may even want to slow it more.

Do you see any issues with where I’m headed with this?

Thanks much.

SuperG wrote on 4/29/2011, 7:14 PM

I didn't see any mention in the thread - so I'll just mention that if you right click on media properties from the time line, you have three choices for audio time stretch:

None
Classic
Elastique

Supposedly, Elastique is the better method and, as I understand, a lot of companies (including SCS) license the technology. I don't know if this will help, but just in case...