What's a Few Perfs Between Friends?
by Andy Somers
It's a struggle to keep up with the progress of technology. It seems that if you miss one trade show or don't stay in touch with the emerging trends for a few months, you're completely out of date. Well, maybe not completely, but with the time demands placed on you while you're editing, it's a challenge to remain abreast of the new tools at your disposal.
Perhaps this is one of the reasons we tend to be relatively slow in adopting new technologies in feature film work. Specifically, I'm talking about sound post production, a field in which new technologies are being adopted only gradually. Digital audio workstations (DAWs), for example, have been around since the late '80s, but while other industries were quick to embrace them (the commercial market, in particular), the feature film industry has come around more slowly. And that adoption was an evolutionary integration as opposed to a revolutionary replacement of existing mag technologies.
Initially, DAWs were used to edit, with their output transferred to mag for playback on the stage. Later, that output was transferred to DA-88s or two-inch 24-track, with timecode. Now we're regularly playing back directly from the DAW; in the last year or two, we've also begun recording to DAWs as a matter of course (although feature stages still tend to record on mag as a backup to any DAW recorder).
As a result of this new advanced technology co-mingling with the technologies of the past, an issue has developed that has yet to be fully resolved. It manifests itself as a minor sync difference between mag and timecode-based units. On most dub stages, the difference is two perfs, or one half of a film frame. That is, the DAWs and other timecode-driven machines, like two-inch recorders and DA-88s, tend to end up two perfs late in relation to any mag units that happen to be running, including the mag recorders.
Timecode numbers always refer to the leading edge of a frame, but the stage controller lines up this leading-edge timecode number with mag that's parked on the middle of a frame. As a result, the timecode-based sources (and timecode-based digital recorders) are two perfs late compared to the mag-based units.
You may be asking, "What are two perfs between friends?" Well, first of all, there's potential for this error to become cumulative if left unchecked. More important, perhaps, is that when working with multiple source devices (i.e., mag recorders and DAWs), the sync pops from them don't line up. The result is a pop that can be longer than a frame in length on the record unit.
Two-pops that are one-and-a-half frames in length are quite common, and in some cases can be as much as two frames in length. You'll find these extra long pops on predubs, and even stems, in some cases. Since these pops occupy more than a frame, it makes sync ambiguous. Personally, I find ambiguous sync kind of scary - especially in light of the fact that when we're talking about film, the only real sync reference is the pop!
There are two other sync issues that should be mentioned in this context. First, if a videotape is providing the timecode reference for the stage, the video-tracking adjustment on the video machine also adjusts the relative phase of the linear timecode on the tape by plus or minus half a frame or so. Note that tracking does not affect the phase of the VITC, just the LTC. As a result, VITC should be used whenever possible.
Another sync issue to be aware of is digital propagation delay. Digital audio equipment introduces a delay in the signals that pass through it. This delay is different depending on the device, what it is doing to the signal, and if it compensates for the delay. In one room recently I noted a system delay of about five or six milliseconds through the digital console and into the digital recorder. That amounts to about half a perf (an eighth of a frame). While not a major sync difference, it too has the potential to be cumulative, resulting in increasing sync ambiguity.
So how do we deal with these problems? Some editorial departments address this issue by running a test on the dub stage in which a blank project - except for a two-pop and a tail pop - is played from the DAW. This project is locked to the stage, and after recording the pops, an operator takes the mag off the recorder, puts it on a sync block and measures where the pop is, as opposed to where it's supposed to be. All of the DAW projects on the stage are then adjusted to reflect the measured offset.
This method seems cumbersome, but it's a good way to ensure that projects are recorded in true sync. You might ask why an editorial change is made to a DAW session or project, as opposed to simply entering an offset in the DAW playback machine. One reason is that DAWs can crash and require rebooting, and when rebooted the offset is erased. With the pressures on a dub stage, it's very easy for an operator to forget to replace the offset of half a frame. And because it's small, it'll likely go unnoticed - that is, until you're back in editorial conforming the predubs and the scene changes don't line up.
But this is simply a response to a problem resulting from a clash of two different technologies. As we saw above, the problem really stems from the method used to lock the timecode and mag units together, and the method used to park a unit on the start mark. Since this is where the problem starts, it seems to me that this is where the problem should be fixed. I see two simple and practical solutions:
Possible Solution #1
Change the way the film/mag units are parked at the start mark. Instead of parking them on the center of the frame, park them on the leading edge of the frame. There's a big problem with this, however; it goes against the way mag has been set up for decades. Such a change now could lead to human error, and that's one thing we always try to avoid.
Possible Solution #2
Have the manufacturers of synchronizers and stage controllers provide a firmware upgrade that sets the timecode output of 1:00:00:00 to line up to half of a film frame before 0. For instance, when the footage counter is 0, the timecode is 1:00:00:00.625 (top of the hour plus 62.5 subframes, assuming a 100 subframe system). This is because .5 film frames equals .625 video frames, as video frames are 20% shorter than film frames. This method, to me, seems to be the most logical fix, since the synchronizers and stage controllers are the devices that are sending out the timecode that is not quite lined up to the true frame edges of the film. Of course, this solution will require cooperation on the part of the manufacturers of this equipment.
It would be nice to have an industry-standard method of dealing with this issue, some way of determining that wherever we go, sync is really sync. Perhaps it would be useful to start a committee to evaluate this problem. Now that we're all in the same union, we should be able to use our strength to find a solution that's best for all and can be instituted industry wide. Timecode has been around since 1968, yet after 31 years we're still having integration problems. I myself would like to see this problem put to rest once and for all.
Andrew Somers is a member of the Guild.
If you have any questions or comments and
you'd like to discuss this issue with him further,
contact him viaGeneral Titles & Visual Effects
The Motion Picture Editors Guild Newsletter
Vol. 20, No. 2 - March/April 1999