Non-Standard Optical Sound Offset On 1937 "Conquest" Trailer

[Damien Taylor]

Hi all, I have been scanning in 4 reels of nitrate before I deposit it with the national archives. It's mostly fragments a projectionist in the past seems to have cobbled together.
Everything has been normal except for one piece of film, a clip from the trailer for Conquest (1937), it must be an early trailer as it still carries the original title of Marie Walewska.
I'm 99% sure I have done the extraction using AEO-Light correctly as I have done hundreds of times before but for some reason to sync the audio I had to move the offset to approximately 41 frames!
I don't know for sure but don't think this was some kind of alternate standard I am assuming maybe a lab error, has anyone seen anything like this before?
There is a possibility that I have done something silly but my workflow is fairly well standardised. I even went back and extracted the soundtrack again making sure the AEO set-in frame was 0 and that the project was set to 24fps, the frame count of the optical audio and picture elements are identical.

Here's the piece of film, you can hear how long before the sound kicks in at the start.
https://www.youtube.com/watch?v=GkuSwaBzgEc

[Phillip Grace]

It is probably the result of a mistake made 86 years ago, either in preparing the negatives for printing, or when the printing machine was being threaded, particularly if the picture neg and sound neg. were printed to the positive in separate passes. Normal sync for 35mm optical sound is 20 frames ahead of the picture. I looks like someone inadvertently retarded when they should have advanced.
As a matter of interest, what equipment are you using to scan the 35mm Nitrate? I cant find anybody here who will touch it.

[Damien Taylor]

Thanks Phillip, a lab mistake was my assumption but glad to hear you agree. The offset is in the correct direction it just seems to be twice the amount required, appreciate your reply.

I'm using a DIY scanner made from mostly 3D printed parts, the design is freely available online as the "Gugusse Roller" At roughly 3-4 seconds capture time per frame it is an automatic but somewhat tedious process however I was able to build it at home for not a huge amount of money. The slow speed and cold light source make it reasonably safe if casually monitored. I don't really like having nitrate around the house so I'm scanning it so I can get rid of it.

[Kyle Mikolajczyk]

Very cool you got your hands on some Nitrate to scan, it looks and sounds great! I also just realized I have been subscribed to you for a while, love your scans keep up the great work.

[Damien Taylor]

Thanks Kyle, there is about about 75 minutes worth of material which I am very keen to upload publicly. I am just waiting to get the digital copies back to the film's custodian first.

I have found a new method of tracking scans in DaVinci Resolve Fusion (planar tracker) that has drastically improved stability which is something I was never fully happy with. They are now rock solid. The perforation tracking is now so good it is highlighting how poor some contact print registration is.

[Phillip Grace]

Hi Damian

I looked at the Gugusse Roller material on the internet. Amazing device, and thanks for pointing it out. It took a while to get my head around the way it works. I especially like the electronic perforation registration.and the film frame being held stationary for the exposure. The outputs as seen on you-tube look pretty good. I noticed a tiny bit of weave in the scans, but the vertical registration was extremely good.
Also interested in the Aeolight solution for reconstructing the sound track from still frames. Do you know how high a frequency it can resolve from the scans?

Cheers.

[Jay Schwartz]

Normal sync for 35mm optical sound is 20 frames ahead of the picture.

I believe it is actually 21 frames advanced (not 20).

[Damien Taylor]

Hi Damian

I looked at the Gugusse Roller material on the internet. Amazing device, and thanks for pointing it out. It took a while to get my head around the way it works. I especially like the electronic perforation registration.and the film frame being held stationary for the exposure. The outputs as seen on you-tube look pretty good. I noticed a tiny bit of weave in the scans, but the vertical registration was extremely good.
Also interested in the Aeolight solution for reconstructing the sound track from still frames. Do you know how high a frequency it can resolve from the scans?

Cheers.

The scanner creator is in the middle of a full overhaul to a unified design so lots of improvements and refinements have been made over the original concept. Registration while scanning is approximate as there is only so much precision that can be offered with home 3D printed parts at this point in time, full stabilisation is carried out digitally in post processing.
The original method of digital stabilisation that was found in DaVinci Resolve did not stabilise the image across the entire frame and did suffer from unacceptable weave, bob and skewing, especially as it did not correct movement caused by the film slightly twisting lengthways through the gate, something the scanner design does not account for. A fellow builder enlightened the community to a better method a couple of months ago which produces vastly superior results in less time, possibly on par with mechanical registration.

AEO-Light is indeed incredible software and the fact that it is open source is a credit to the organisations behind it and I am very appreciative that hobbyists such as myself have access to it. The documentation doesn't really give too much away regarding exact frequency response but I would imagine it is only limited by the quality of the original printing and the resolution of the scan. Variable area tracks in my experience produce a much more consistent high quality result. Density tracks require a lot more finessing to produce the best output.

[Damien Taylor]

I believe it is actually 21 frames advanced (not 20).

SMPTE specification is 21 which is what I use. In projection applications sometimes operators will thread for 20 to allow a slight delay in the time taken for the sound to reach the center of an auditorium.