Synchronization of Audio and Video from Different Sources

Dale Paterson · April 22nd, 2006, 12:32 AM

Hi Jeff,

Definitely 48/16 but have also tried 48/24.

It is definitely this clock issue.

Anyway - I'm sure that somewhere down the line there is a solution which I WILL find!

Regards,

Dale.

Seth Bloombaum · April 22nd, 2006, 09:47 AM

Jeff,

Dale is experiencing two different synchronization errors that are under discussion here:

1) Offset error, produced by the time it takes sound to travel through the air to Dale's camera mic. vs. on-stage mics through wires to mixer and pc. This would be perfectly addressed by the wireless link you're suggesting.

2) Timebase error, produced by the clocks of the digital mixer and the camcorder running at slightly different speeds. This is not addressed by taking mixer audio to camera. This can be fixed in two ways:
a) Shrink or expand the duration of the mixer audio track in post until it matches the duration of the camera audio track. Made easier by beeps or other references that are visible in the waveform display. Inexpensive, and not very difficult once you've done it a couple times.
b) Upgrade camcorder and mixer (or digital recorder) to equipment that accepts timecode input and generates timecode output. Then, slave the audio recorder to the camcorder. Costs well over $10,000 but less than $20,000 for camcorder and audio recorder, depending on audio recorder selected.

To repeat, timebase error is not fixed by taking mixer audio to the camera, wireless link to the camera fixes only one of the two problems Dale is experiencing.

Why? Imagine two film cameras. Film a scene (with motion!) with one camera at 12 frames per second, and the other at 24fps. Take the processed film reels, and put them up on two projectors side-by-side and show them both at 24fps. The original 12fps-reel will show in half the duration at twice the (apparent) speed. This is timebase error, the two clocks were far from synched, it affects speed of action and duration when the content gets synch locked in projection, or on the timeline.

In this example, we had a 50% speed/duration difference. Dale's errors are much smaller, on the order of less than .1%, but that's enough to throw lipsync visibly off over an hour of content when the two audio tracks are locked to the soundcard's clock.

I've done a lot of synching in my time, with no beeps. Not so hard to fix offset error or timebase error in post. At least that's my experience, your mileage may vary.

Dale Paterson · April 23rd, 2006, 02:01 AM

Good Morning.

Thanks Seth for that explanation - explained as only a master could.

There is one other thing that is bugging me and would just like clarification (just for interest sake):

Correct me if I am wrong but even with cameras that have 'free run' timecode like the Sony Z1 you will still experience the same problem. Would I be correct in making this statement?

The reason I ask is that during my search for a solution to my problem I saw a thread somewhere (it was referring to the Canon GL2 but I think that you can follow the same procedure with the Sony Z1) that detailed how to start the 'free run' timecode with the remote on multiple cameras and then when it came to post things were just a lot easier as each tape (scene) had the correct corresponding timecode tape for tape / scene for scene.

After demonstrating that not one of three of my cameras run in perfect synch to each other I make the assumption that even with three or four Sony Z1's there would STILL be a difference between each cameras timecode i.e. timebase errors (learning all the time) over a long period. Correct? The 'free run' timecode would just make it easier to synch multiple camera shots IF AND ONLY IF the multiple shots were not of a lengthy duration.

Also - just for the record - I think I understand what a 'free run' timecode is but what is a 'record run' timecode?

Something else that has just come to mind: Does the difference between my cameras (for example) mean that they are not shooting at EXACTLY 25fps (PAL)? During capture when Vegas reports that my average fps for the capture of an entire tape is 24.98fps or maybe 25.03fps is it this timebase error that is being reported and should resampling each cameras video / audio track not solve the problem (although I did try this with no change in synch between the video /audio tracks from each camera)?

I know that none of the above has anything to do with the problem that I am trying to resolve - just want to know for interest sake.

Regards,

Dale.

Seth Bloombaum · April 23rd, 2006, 11:09 AM

Dale,

Free-run is handy, but, as you suspected, doesn't help with the offset and timebase errors you're experiencing.

The default for most prosumer camcorders is record-run, which runs more like a tape counter. Typically it starts at 00:00:00;00, what's recorded at the end of a one-hour tape will be about 01:00:00;00 or a couple minutes more, as tapes usually are 1-3 minutes longer than their stated capacity. So, the timecode generator lays down nearly continuous timecode on tape. (***edit - regardless of whether you've started and stopped the camera***)

So, the timecode generator only rolls when the tape is recording.

Then there is preset, which interacts both with record-run and free-run. Back in ancient times (as few as 10 years ago!), we were doing lots of linear editing with tape machines and editing controllers - they did NOT like seeing the same timecode twice in a project! Usual strategy was to "preset" starting timecode at hour 1 for tape 1, hour 2 for tape 2, etc, and use record-run.

Free-run means that the timecode generator rolls continuously, regardless of whether the camcorder is recording or not.

Free-run has all sorts of applications, but the one we're mostly interested in is so-called time-of-day timecode. Let's say it's 8:30am on the shoot day, and we've had our coffee and doughnuts. We'll preset all devices to 8:35am, switch them to free-run, then, whoever has the good wristwatch or the loudest voice will count down the last 10 seconds to 8:35, at which point everyone starts their timecode generator rolling.

Voila, all the timecode recorded now shows the time-of-day. With prosumer gear, that provides only rough sync - if done carefully, we're better than 10 frames, but that will still require a tweak in post, and, code will drift through the day. We might sync a few times if we care, but usually if this approach gets us within a couple seconds the editor can find the syncing points.

With pro-gear and a master timecode generator that is continuously wired to all devices, this is always frame-accurate. Usually in a multi-camera studio it's cameras, not camcorders, and all the tape decks are in a control room where they are frame-locked always (with master TC or house black, but that's another subject...)

Quote:

The 'free run' timecode would just make it easier to synch multiple camera shots IF AND ONLY IF the multiple shots were not of a lengthy duration.

I do this all the time with music recording, very similar to what you have done with your alesis digital mixer/laptop. Yes, I'll roll continuously over an hour tape, or even a 3-hour tape in a studio deck. I always have to correct for offset error, and sometimes have to correct for timebase error, depending on the accuracy of the individual clocks.

My attitude about this is that we now have $3-10,000 USD cameras doing most of what we used to spend $40-60,000 on, only with better pictures. I'm so amazed by the fact that I can personally own equipment and an NLE with these capabilities (instead of being tied to a multi-million dollar facility) that I don't worry too much about having to do these sorts of corrections.

You're concerned with how elegant the solution is - I'm very impressed that the beast is walking at all.

Quote:

Does the difference between my cameras (for example) mean that they are not shooting at EXACTLY 25fps (PAL)? During capture when Vegas reports that my average fps for the capture of an entire tape is 24.98fps or maybe 25.03fps is it this timebase error that is being reported...?

A frame is a frame, that is, it has its own begin and end information. So when you put it up on the timeline it's duration becomes (PAL) 1/25th of a second. If you lay it back to the Z1 after editing maybe it will be plus or minus some fraction of a percent different.

When does this matter? When it goes into large distribution systems like broadcast, cable or satellite. When there is a video switcher in use, and you want it to cut between sources glitch-free, which means cutting during the vertical interval between frames. And so, there are timebase correctors and/or frame synchronizers that retime frames during playback.

But it doesn't matter much for DV in general, because we usually either distribute direct to the consumer on DVD or VHS, or, we hand it off to a broadcaster who routinely fixes timebase errors during playback, because all tape machines have these errors - they are mechanical devices. I suppose things will be somewhat different when there are no tape players in use.

Shawn Redford · April 23rd, 2006, 01:07 PM

Seth - Thanks for your great explanation of various time-codes. So is there anything on the prosumer level that has time-of-day time code?

Seth Bloombaum · April 23rd, 2006, 06:06 PM

Quote:

Originally Posted by Shawn Redford

...So is there anything on the prosumer level that has time-of-day time code?

The manufacturer doesn't call it time-of-day, they'll call it free-run and preset.

Most of the higher end prosumer cameras, such as PD150/170, XL1, XL2, Z1, DVX100 have this capability. I think PDX10 does. If I remember right, GL1/2 does.

Not sure about the lower cost cousins of the Sony cams such as VX2000/2100, FX1. I'm not that familiar with Panasonic's DV line other than the DVX100.

Dale Paterson · April 23rd, 2006, 11:51 PM

Good Morning,

FYI - I can confirm that neither the VX2100E nor the FX1E have either free run or record run timecode settings / options / presets.

I found this link on my travels that refers to the PD150 timecode settings:

http://www.urbanfox.tv/workbooks/son...0timecode.html

Regards,

Dale.

Dale Paterson · April 29th, 2006, 02:45 AM

An update:

I received my Control-L (LANC) to MIDI Time Code Cable (Generator) by Avit Research.

The cable / generator works perfectly - the cameras time code (FX1E / VX2100E / TRV27E) is sent to Vegas / Acid via the LANC port - no problem there (clearly visible when opting to view MIDI time code 'in').

Problem:

Although you can trigger Vegas / Acid recording and playback using the camera the time code from the camera is not recorded or should I say that when recording, Vegas / Acid do not 'chase' the camera's time code. So even this device makes no difference when trying to record, and keep in synch, audio recorded externally - the same consistent drift is evident.

If anybody knows how to 'instruct' Vegas / Acid to 'chase' the external MIDI time code 'in' then please let me know although I do not think that it is possible as, after reading the Vegas manual, Vegas can generate a MIDI clock and MIDI time code so this would work if the prosumer cameras had time code 'in' but they don't i.e. if they did the camera's would 'chase' the Vegas / Acid clock or time code but not the other way around.

Regards,

Dale.

April 22nd, 2006, 09:47 AM	#47
Seth Bloombaum Inner Circle Join Date: Sep 2003 Location: Portland, Oregon Posts: 3,420	Jeff, Dale is experiencing two different synchronization errors that are under discussion here: 1) Offset error, produced by the time it takes sound to travel through the air to Dale's camera mic. vs. on-stage mics through wires to mixer and pc. This would be perfectly addressed by the wireless link you're suggesting. 2) Timebase error, produced by the clocks of the digital mixer and the camcorder running at slightly different speeds. This is not addressed by taking mixer audio to camera. This can be fixed in two ways: a) Shrink or expand the duration of the mixer audio track in post until it matches the duration of the camera audio track. Made easier by beeps or other references that are visible in the waveform display. Inexpensive, and not very difficult once you've done it a couple times. b) Upgrade camcorder and mixer (or digital recorder) to equipment that accepts timecode input and generates timecode output. Then, slave the audio recorder to the camcorder. Costs well over $10,000 but less than $20,000 for camcorder and audio recorder, depending on audio recorder selected. To repeat, timebase error is not fixed by taking mixer audio to the camera, wireless link to the camera fixes only one of the two problems Dale is experiencing. Why? Imagine two film cameras. Film a scene (with motion!) with one camera at 12 frames per second, and the other at 24fps. Take the processed film reels, and put them up on two projectors side-by-side and show them both at 24fps. The original 12fps-reel will show in half the duration at twice the (apparent) speed. This is timebase error, the two clocks were far from synched, it affects speed of action and duration when the content gets synch locked in projection, or on the timeline. In this example, we had a 50% speed/duration difference. Dale's errors are much smaller, on the order of less than .1%, but that's enough to throw lipsync visibly off over an hour of content when the two audio tracks are locked to the soundcard's clock. I've done a lot of synching in my time, with no beeps. Not so hard to fix offset error or timebase error in post. At least that's my experience, your mileage may vary. Last edited by Seth Bloombaum; April 22nd, 2006 at 04:42 PM.

April 22nd, 2006, 12:32 AM	#46
Dale Paterson Major Player Join Date: Jun 2005 Location: Johannesburg, South Africa Posts: 200	Hi Jeff, Definitely 48/16 but have also tried 48/24. It is definitely this clock issue. Anyway - I'm sure that somewhere down the line there is a solution which I WILL find! Regards, Dale.

April 23rd, 2006, 02:01 AM	#48
Dale Paterson Major Player Join Date: Jun 2005 Location: Johannesburg, South Africa Posts: 200	Good Morning. Thanks Seth for that explanation - explained as only a master could. There is one other thing that is bugging me and would just like clarification (just for interest sake): Correct me if I am wrong but even with cameras that have 'free run' timecode like the Sony Z1 you will still experience the same problem. Would I be correct in making this statement? The reason I ask is that during my search for a solution to my problem I saw a thread somewhere (it was referring to the Canon GL2 but I think that you can follow the same procedure with the Sony Z1) that detailed how to start the 'free run' timecode with the remote on multiple cameras and then when it came to post things were just a lot easier as each tape (scene) had the correct corresponding timecode tape for tape / scene for scene. After demonstrating that not one of three of my cameras run in perfect synch to each other I make the assumption that even with three or four Sony Z1's there would STILL be a difference between each cameras timecode i.e. timebase errors (learning all the time) over a long period. Correct? The 'free run' timecode would just make it easier to synch multiple camera shots IF AND ONLY IF the multiple shots were not of a lengthy duration. Also - just for the record - I think I understand what a 'free run' timecode is but what is a 'record run' timecode? Something else that has just come to mind: Does the difference between my cameras (for example) mean that they are not shooting at EXACTLY 25fps (PAL)? During capture when Vegas reports that my average fps for the capture of an entire tape is 24.98fps or maybe 25.03fps is it this timebase error that is being reported and should resampling each cameras video / audio track not solve the problem (although I did try this with no change in synch between the video /audio tracks from each camera)? I know that none of the above has anything to do with the problem that I am trying to resolve - just want to know for interest sake. Regards, Dale.

April 23rd, 2006, 01:07 PM	#50
Shawn Redford Regular Crew Join Date: Jun 2005 Posts: 158	Seth - Thanks for your great explanation of various time-codes. So is there anything on the prosumer level that has time-of-day time code?

April 23rd, 2006, 11:51 PM	#52
Dale Paterson Major Player Join Date: Jun 2005 Location: Johannesburg, South Africa Posts: 200	Good Morning, FYI - I can confirm that neither the VX2100E nor the FX1E have either free run or record run timecode settings / options / presets. I found this link on my travels that refers to the PD150 timecode settings: http://www.urbanfox.tv/workbooks/son...0timecode.html Regards, Dale.

April 29th, 2006, 02:45 AM	#53
Dale Paterson Major Player Join Date: Jun 2005 Location: Johannesburg, South Africa Posts: 200	An update: I received my Control-L (LANC) to MIDI Time Code Cable (Generator) by Avit Research. The cable / generator works perfectly - the cameras time code (FX1E / VX2100E / TRV27E) is sent to Vegas / Acid via the LANC port - no problem there (clearly visible when opting to view MIDI time code 'in'). Problem: Although you can trigger Vegas / Acid recording and playback using the camera the time code from the camera is not recorded or should I say that when recording, Vegas / Acid do not 'chase' the camera's time code. So even this device makes no difference when trying to record, and keep in synch, audio recorded externally - the same consistent drift is evident. If anybody knows how to 'instruct' Vegas / Acid to 'chase' the external MIDI time code 'in' then please let me know although I do not think that it is possible as, after reading the Vegas manual, Vegas can generate a MIDI clock and MIDI time code so this would work if the prosumer cameras had time code 'in' but they don't i.e. if they did the camera's would 'chase' the Vegas / Acid clock or time code but not the other way around. Regards, Dale.