• Reality at 25/24 Frames per Second

    By Peter Becker

    Here’s a Criterion discussion that won’t die. It has to do with Berlin Alexanderplatz, and it came up again this week, thanks to a couple of customers writing in. We were standing there in a clump outside our production manager’s door—the disc producer, the head of audio, and a few more of us—running through the same arguments one more time and ending up, once again, at the same conclusion. It all starts when Rainer Werner Fassbinder chooses to shoot Berlin Alexanderplatz , his epic masterpiece, at 25 frames per second (fps). It makes sense, since in Europe television runs at 25 fps, and the film was being shot for European television. But what happens when you need to make a 24 fps HD master? Or a print that will be projected in theaters at 24 fps? Either you do what we’ve done, let the film run naturally at 24 fps—which means the running time will be 4 percent longer and the pitch of the audio will drop down slightly—or you could try to solve the problem with digital processing and pitch correction, crushing 25 frames worth of information into 24 frames.

    If the film weren’t fifteen hours long, we probably wouldn’t even be having this discussion. The actual differences in timing and pitch are mostly perceptible in a side-by-side comparison, and on an episode by episode basis we’re really only talking about a couple of minutes—an hour-long episode would run a little over sixty-two minutes. By sacrificing actual clock time we preserve the integrity of the picture, ensure natural movement between frames, and avoid introducing digital artifacts. Subjecting the master to massive signal processing based on the ugly math it takes to display 25 frames worth of data each second, but using only 24 frames, would result in a huge amount of interpolated picture information that doesn’t actually exist on celluloid. Instead of a frame-accurate picture of the film, you get an image of what the frames would look like if we start from the assumption that frames 1, 2, 3, and 25 are actually .96, 1.92, 2.88, and 24. Clearly there is no 2.88th frame, so the signal processor has to derive one. Here’s an example of what that would look like:




    Changing the running time also causes a 4 percent change in the pitch of the audio. Just as with a record player, when the sound gets played back more slowly, the pitch drops. The only way to correct the pitch would be to change the actual sample values, which would mean introducing a whole new series of interpolations as we replaced the entire soundtrack with a derivative soundtrack based on math. In order to avoid creating artifacts and distortion, we chose to present a frame-accurate rendering of the image and sound and put up with the 4 percent time and pitch difference. This is the same compromise (in reverse) as the one that is made when a 24 fps theatrical film is mastered at 25 fps for a PAL broadcast or DVD release.

    Still, because Berlin Alexanderplatz is almost fifteen hours long, the 4 percent difference adds about thirty-five minutes. Surely that must change things. That’s a half hour more of my life dedicated to this already epic movie. If I start watching this film at the same moment as my colleague Robert Fischer in Germany, he will meet Barbara Sukowa’s character, Mieze, something like fifteen minutes before I will. It means that what took a minute of real time on Fassbinder’s set will take 1.04 minutes on a theatrical print or on our DVD. I know it’s the right way to handle it, but I’m still having a hard time accepting it. Isn’t that a form of distortion too? I know that the alternative, the processed image, looks terrible—jumpy, distracting, and unacceptable on every level. But theoretically speaking, does audiovisual fidelity necessarily outweigh the obligation to replicate the experience the director intended us to be having over a certain interval in clock time? Would Fassbinder have cut the film differently if he looked at the total running time as fifteen and a half hours instead of fifteen? These are not merely technical questions; they are artistic ones, and unfortunately there is no good answer, just a best one, and that’s why we keep having the conversation. Yesterday I promised that we’d be having the conversation for the last time, so I just thought I’d take a moment and get the whole thing out of my system once and for all. Isn’t that what blogs are for?

3 comments

  • By AustinPrior
    September 13, 2010
    08:19 PM

    All of this seams oddly irrelevant to the finished product: An NTSC DVD box set running at 29.97 fps. Why bother with a 24fps stage at all? Once you've got every frame stored on the computer you could double every fifth frame ending up with something that's running at 30fps and is the right speed. Then from there you can run those frames slightly slower out to NTSC at 29.97 giving you a much smaller distortion in sound and timing. I'm certainly not claiming to be anywhere near as knowledgeable on such matters as the fine folks at Criterion but I don't see what is impossible about my suggestion.
    Reply
  • By Sean Pak
    March 29, 2012
    05:07 PM

    But why single out this release? Wouldn't Scenes From a Marriage and Fanny and Alexander have the same issue, as they were also shot originally for European television? You should at least mention those two in the article.
    Reply
  • By Franco Ferrer-San Miguel
    January 21, 2014
    10:38 AM

    Don't forget "The Magic Flute", which, in its PAL DVDs, has a more faithful speed (Even listening to the music in its original pitch). (Also, discussion needs to be made of its subtitles....)
    Reply