Monday, July 09, 2007

Timecode capture / line up in Final Cut Pro

Have just investigated lining up/ capturing separately recorded audio and picture via timecode in FCP.

In the previous post, the idea of jam-synching was touched upon. Basically, the Field Recorder is jam synched (or forced) to adjust its internally generated timecode to that of the camera (or vice versa). Once this has been done, the physical connection between to the two can be removed for at least a couple of hours without drift. This method has the added benefit that the sound recordist is not restricted to being in close proximity to the camera, as no physical link is required (although it can be useful if the sound op wants to monitor sound coming off camera too, for backup purposes etc.). This method only works in FREE RUN (F RUN) mode which is where the timecode progresses regardless of whether the tape is moving or not and means that the two devices do not have to be started and stopped precisely at the same time.

Any way, a Fostex FR-2 was set to generate FREE RUN timecode and then its timecode output was connected to the Timecode input on a Canon XH-G1 camera. Immediately, the camera picked up the timecode and matched it. The cable was then disconnected and both devices carried on generating their own separate timecode, but precisely in time with each other. A small amount of footage was recorded to both camera and field recorder and it was discovered that it helps if the camera is started slightly before and stopped slightly after the field recorder (more on why this is shortly).

Firstly, the wave file recorded by the FR-2 was imported, via USB, to FCP and because of its BWF (Broadcast Wave File) system, the timecode start and stop times immediately appeared in the "Media Start" and "Media End" columns within FCP. Secondly, the "Log and Capture" window within FCP was opened so that the matching footage from the camera could be grabbed. By Option/Dragging the timecode value in the Media Start column into the IN point within the Logging window (and by doing the same for the Media End column too) an exact capture could be carried out by using "Capture Clip". Once complete, the footage and wave file matched perfectly. By letting the camera start before and end after the field recorder, it is easy to get FCP to capture footage between the start and stop times of the wave file from the field recorder. If this wasn't carried out and the camera started after the field recorder, then FCP would not be able to find the required timecode on the tape.

***Addition (Feb 29th 2008) - the last paragraph discussing lining up the separately recorded audio with the picture from the camera, via timecode is not the best method. A much neater method is to use the MAKE MULTICLIP option within FCP and when asked how to align to audio with the video, simply choose MATCH BY TIMECODE. Once this is carried out, FCP will create a multi-clip with the separate audio perfectly aligned to the picture, regardless of whether it was started before or after the camera (the audio attached to the video file from the camera can then easily be removed from the multiclip). This method is also really neat when aligning footage from a multi-camera shoot to audio recorded on a separate device. If a clapperboard was used to aligned the cameras and audio recorder, IN points could be marked on each of the video files (i.e. at the frame where the clapper closes) and then the matching IN point could be marked on the audio file. By SHIFT+clicking on the same audio file with each of the camera angles one at a time and then creating a MULTI-CLIP and selecting ALIGN BY IN POINT, each of the camera angles will now have identical audio attached to them.

Levels and line up



Had the chance this morning to investigate audio levels when recording to a different devices. A Canon XH-G1 camcorder has just been purchased together with a Fostex FR2 with optional timecode board and a Sound Devices 302 field mixer.

The Canon XH-G1 was chosen because, at present, it appears to be the cheapest camcorder available which will both generate and accept timecode synchronisation signals via a standard BNC connection. The idea is to get students to record their audio to both camera and field recorder and to maintain sync between the two via timecode. In some situations, when it is favoured to have the sound operator to be a long way from the camera, the useful feature of jam synching can be used. This feature allows the slave device to force its internally generated timecode to that of the master and then free run without the need for a permanent cable connection (although the cable should be reconnected every few hours to maintain stability). This system only works in FREE RUN mode, which is where the timecode continues progressing whether the tape is moving or not and thus means that the field recorder and camera do not have to be started and stopped exactly at the same time (a useful Flash demonstration of this principle can be found here: http://www.usa.canon.com/app/html/HDV/XHG1/inputs_outputs.shtml (click on the SMPTE TIMECODE section within the large black box). Regardless of physical start and stop times of both the camera and field recorder, the footage required from both will have identical timecode meaning lining up the files in an NLE like Final Cut Pro becomes very easy (see later).

Calibration / Line Up
Some tests were carried out to establish exactly at what level both the camera and field recorder were recording at (as the digital peak level meters can be somewhat inaccurate). The first test was to connect the Sound Devices 302 field mixer directly to the Canon XH-G1 via two standard XLR cables. The 302's attenuation was set to 0, meaning that the 1kHz sine wave it generates is calibrated so that 0 on its bar graph display equates to a level of 0dBu and also provides 20dB headroom beyond this point. The recommendation is to set the input level on the recording device so that 0 on the 302 corresponds to -20dBFS (dB FULL SCALE) on the camera, meaning that +20dB on the 302 corresponds to 0dBFS on the camera, i.e. absolute maximum level.

With the camera's XLR inputs set to LINE and the input gains set to maximum, the 1kHz sine wave input from the 302 was recorded to tape (in DV mode, not HDV). The meter on the camera is very difficult to accurately read as no numbers are provided, however it showed a level of approx. 1 increment below the third division (I said it wasn't that useful). To discover exactly what level was recorded to tape, the footage was captured, via firewire to FCP and then exported to Apple's Soundtrack as its meters are slightly better (although the software itself is pretty useless, but that's another story...). The meters for both channels showed a level of -20.6dBFS which is an extremely useful and useable level. This therefore means that at 0 on the 302's bar graph display, the signal going down onto DV tape was -20.6dBFS, meaning that even if the 302 hit its +20dB segment (maximum) the camera would hit just under 0dBFS, i.e approx. -0.6dBFS.

A second test was carried out, this time using the both the Fostex FR-2 field recorder and camera to record the test signal from the 302. The signal chain was as follows: XLR line outputs on the 302 into the XLR inputs on the Fostex FR-2 (with input set to LINE). The Phono RCA line outputs from the FR-2 were connected into the XLR inputs on the camera and then finally, to monitor the signal level from the camera, its headphone output was input to the monitor return path on the 302. Providing this monitor line from the camera enables the signal to be monitored "post tape / disc". With the same camera settings as previously discussed, and with the input level on the FR-2 carefully adjusted so that it showed the required level of -20dBFS, both the camera and field recorder were set to record. Upon capturing the footage (and copying the wave file from the field recorder). the levels were as follows:

FR-2 LEFT: -18.3dBFS
FR-2 RIGHT: -19.3dBFS

Camera LEFT and RIGHT: approx. -28dBFS

It can therefore be seen that the line outputs on the FR-2 are attenuating the signal slightly and also, its digital meter isn't particularly accurate, giving a reading of -20dBFS for an actual recording level of -18.3dBFS and -19.3dBFS for the left and right channels respectively. It was also difficult to accurately adjust both the left and right input gains to give the same level (as can be seen from the results here). The outputs from the FR-2 are unbalanced, so this might explain the drop in signal level.

As the input gain on the camera was at its maximum setting, there was no way to increase the signal level any further without changing the input sensitivity to the MIC setting. To avoid overdriving the inputs to the camera in this state, the input gains were reduced to minimum and also the camera's ATT (attenutation) function was switched on. Without the ATT button switched in, and even with the input gain at a minimum, the input stage would distort due to overdriving. Gradually the input gains were turned up to give a higher level than what was possible in LINE mode and after a few attempts (which won't be discussed here) it was found that with the camera's meter showing a level of one increment under point number 3, a recording level of around -21dBFS could be achieved; thus allowing 1dB of headroom above the 302's maximum output level.

An alternative method to using the ATT function on the camera, would be to use the 302's built in attenuation mode, however the 302 has a somewhat awkward menu system, so it is probably easier to keep the 302's output level at its default of 0 = 0dBu.