[3DRoundabout]

…It was our goal to make a film which uses the technology to create something unique, something that wouldn’t be possible without it. We wanted to convey the impression of flight and weightlessness in space – an immersive, purely visual and poetical experience for audiences. With that in mind, RISE was shaped more like a dream than a short film, where a young woman suddenly begins to fly as if gravity were suspended.  …

Production

During production, the few seconds captured in realtime at up to 2000fps ended up running for a few minutes on playout. Using AJA frame syncs to synchronise the signals for on-set playback in 3D on the JVC display, gave me, the director, a chance to review the result as it was being recorded. I could immediately reject a recording the moment I noticed that a take didn’t yield the desired result in order to save time on set and to save drive space (we ended up recording some 4TB in two days). The challenge as a director on a high-speed 3D production is to see that desired moment during realtime capture, and call ‘cut’ before it is eaten up in the ring buffer of the HS-2. Otherwise there’s nothing useful to record during 25p playback.

Synchronising

Since the Weisscam RAW codec is not 3D-muxed (meaning left and right streams are recorded into one file as is the case with an SI-2K 3D setup for instance), but recorded individually into separate files, we anticipated that syncing the 3D footage in post could turn out to be a challenge. Imagine visually identifying a sync problem between left and right images when the difference is 1/1000 of a second! As it turns out this looks more like a geometric offset at first.

A traditional head slate is not an option with the maximum recording time of less than 10 seconds at the frame rates we were shooting with. By the time the slate were moved out of the frame and the action had taken place, the slate would have been eaten up in the ring buffer, with no decent sync mark in place anymore. The same problem arises when using a tail slate. The clapper would have to move at the speed of light in order to position the slate and clap it before the valuable footage is recorded over. But the speed of light was precisely our solution!

We used professional photo strobes with a flash duration of about 1/1500s. These were positioned to illuminate some darker areas of the frame and were fired wirelessly by an assistant as soon as I called ‘cut’. A perfect lightning-speed 3D sync marker! At the end of the day, the genlock did the major work, and the sync was consistently off by only one frame anyway – although it would have taken a while to ascertain that fact without the strobe. A helpful side effect of this method was that it also gave the DIT department a heads up when the realtime transfer from the cameras to the DigiMags could be stopped, saving additional time because we didn’t wait to record useless minutes of footage at the end of a take.

…

Read the full story here: http://3droundabout.com/2011/09/4881/case-study-rise-a-high-speed-film-in-stereo-3d-at-2000fps.html