How to Sync Audio and HDV video

[Gary Nattrass]

Originally Posted by Natan Pakman

Steve, just to clarify this point, it is correct to say that, if I am using the FX1, and I want to be able to sync up audio and video without physically connecting the camera and the audio recorder, then sending a constant blackburst signal with a composite video cable is not the way to go?

This is a huge shot in the dark, but is there any way to send a blackburst signal from the camera to the audio recorder in a wireless fashion? I figured I'd at least ask.

This is a semi-unrelated question, but how was audio synced to film on motion pictures because NLEs?

If you are looking at a wireless soloution then why not just send the audio from the sep recordist via a radio mic?

It is possible to do this and still record the sound separate if you wish, that way you have sep sound and a guide on the Fx1 that may be useable as well.
A clapper board will mark and keep sync checked.

[Natan Pakman]

Originally Posted by Gary Nattrass

If you are looking at a wireless soloution then why not just send the audio from the sep recordist via a radio mic?

It is possible to do this and still record the sound separate if you wish, that way you have sep sound and a guide on the Fx1 that may be useable as well.
A clapper board will mark and keep sync checked.

Gary, using a plug-in transmitter and recording onto the FX1 directly is fine, but in this setup, is it possible for the boom operator to listen to the recording via headphones? If the plug-in transmitter had a headphone out, that would be ideal, but it seems that the cameraman would have to monitor the sound. I haven't yet found a plug-in transmitter with this feature.

[Andy Tejral]

Well, you can run the boom mic through a mixer, have sound guy listen to mixer and send the main output wirelessly to the camera.

[Steve House]

Originally Posted by Natan Pakman

Steve, just to clarify this point, it is correct to say that, if I am using the FX1, and I want to be able to sync up audio and video without physically connecting the camera and the audio recorder, then sending a constant blackburst signal with a composite video cable is not the way to go?

This is a huge shot in the dark, but is there any way to send a blackburst signal from the camera to the audio recorder in a wireless fashion? I figured I'd at least ask.

This is a semi-unrelated question, but how was audio synced to film on motion pictures because NLEs?

First of all you have to get clear just what you mean by the word "sync." There's "sync" in the sense that you have an alignment point where a known instant in the sound file, a specific audio sample, can be aligned with the exact frame in the video file that was captured when that sound was recorded. And then there's "sync" in that when you align that known point in the sound with the video frame when it was recorded at the start of a clip and hit "play" the two files STAY in sync for the duration of the shot. A clapper or timecode does the former, video blackburst to slave the sample clocks does the latter.

I'm not aware of a simple system that would send video sync wirelessly. Wireless timecode is no problem but wireless sync is much trickier. That doesn't mean it doesn't exist - Andy mentions a product but I'm not familiar with it in order to say yeah or nay.

Back in the film and analog audio day, extending even into the DAT days, either a sync signal was sent from the camera, or coming from the power mains, or the camera speed was controlled by a crystal oscillator that also sent a signal to the audio tape, or in the case of DAT a linear timecode signal, was recorded on a track that ran parallel to the audio on tape. As the first step of postproduction, the analog or DAT tape was "resolved," ie, re-recorded, onto perforated magnetic filmstock. The tone recorded on the tape or the linear timecode recorded on the DAT controlled the speed of the resolver recording device so that 1 second of recorded audio went onto exactly 24 frames length of magnetic film as indicated by the edge perforations. Then as Andy said, the magnetic perf was physically aligned to the film workprint using a geared "sync block" that locked the two film strips together via their sprocket holes.

The mistaken belief that timecode can both establish sync and maintain sync dates to those days. Timecode on a DAT was recorded as linear timecode, laid down continuously alongside the audio tracks. When the tape was played in telecine, the recorded timecode actually controlled the playback speed during the ingest process. But that's not what happens with modern recordings. A file-based recording, if it has timecode at all, only grabs it as a single point timestamp, the reading of the timecode counter when the first audio sample was recorded. Any later running code is calculated on-the-fly by adding the number of samples played to the starting timestamp. As a result, timecode DOES NOT control playback speed - whether you are copying the audio file as a BWF directly into your NLE or going through a telecine process, the playback speed is 100% dependent on the degree to which the recorder's sample clock at the time the recording was made matches the speed of the playback device's sample clock. A $5000 Sound Devices 744t's sample clock is going to be running at pretty close to spot-on the 48kHz standard. A $400 hand-held Zoom? It's anybody's guess. With some recorder/camera combinations you might get 15 minutes before it becomes noticable. With others, it can be much shorter. Using various compressed recording formats like MP3 or WMA can be a total disaster.

[Mike Demmers]

I'm not aware of a simple system that would send video sync wirelessly. Wireless timecode is no problem but wireless sync is much trickier. That doesn't mean it doesn't exist - Andy mentions a product but I'm not familiar with it in order to say yeah or nay.

Isn't this sad? It seems to me that manufacturers are missing a market to solve sync problems in this market. Almost all of these prosumer cameras have no way to sync to each other, or anything else, video or audio.

I can't see any way to solve the video sync problem without genlock inputs or something equivalent, but the audio should be much easier, provided you just need one camera synced. I think that is reasonable, since you can use that camera as your master shot and edit other cameras into that timeline.

I'm not sure how useful sending video sync pulses would be, these cameras have no genlock inputs and it is not the best for audio sync (wordclock or linear timecode would be better). But it would be technically trivial to send vertical sync wirelessly - a $1 sync separator chip would derive the 30Hz pulses which could be sent to any wireless transmitter.

But what I would really like is a box that is small enough to mount on a camera that does just about the opposite of what a lockit does. A box that accepts NTSC video from a camera and outputs linear timecode locked to the video. It should cost no more than $300 or so.

It could be very simple - just input, output, and a reset button to restart the timecode at 01:00:00:00

Any wireless mic transmitter could be plugged into this. Heck, even a $15 hobby FM transmitter module might work, anything that will transmit audio with a 12-15K bandwidth would work.

On the receiving end, anything that locks to linear timecode could be used. Could be the Tascam unit, or something like the Motu UltraLite-mk3 or Traveler-mk3. This would allow multiple audio tracks to be phase locked to the camera, eliminating drift.

Maybe I have just missed it, but I have not been able to find such a box.

-Mike

[Andy Tejral]

Originally Posted by Mike Demmers

Almost all of these prosumer cameras have no way to sync to each other, or anything else, video or audio.

Well, you said it right there: prosumer. This is not a function that most prosumer users need.

But what I would really like is a box that is small enough to mount on a camera that does just about the opposite of what a lockit does. A box that accepts NTSC video from a camera and outputs linear timecode locked to the video. It should cost no more than $300 or so.

Ya know, it wouldn't be too hard and I wouldn't be surprised if a company like Horita couldn't slap one together. You could do exactly what you describe with their existing products but it would take several boxes and cost much more than your asking for.

But volume of sales would be a problem: this would be a very small niche. You need the features of a professional but don't want to pay the price. Bad combination.

Oh, another thought: I'm not sure how much R&D companies are going to do for an NTSC based product.

[Mike Demmers]

Say, good call on Horita!

I just looked at their site and realized there is pretty close to a bingo with their TG-50 Generator/Inserter. Horita Co., Inc. Affordable Time Code and Video Equipment

This is a timecode generator that does window dubs. To do that it MUST lock to the incoming video, so it is synced to the video. And it has a timecode output - if this is, as I assume, linear timecode, and active while in generate/window dub mode - then there it is.

Runs off 9V, so easily portable.

List $399, probably street price a bit lower, so it is pretty close to the $300 mark I suggested. As a bonus, it will work as a window dub generator. ;-)

Now just feed that timecode into any really cheap wireless setup and you are good to go.

I might just remove the circuit board from that box and put it, along with the board from a wireless setup, into a more suitable one.

-Mike

[Steve House]

Originally Posted by Mike Demmers

Say, good call on Horita!

I just looked at their site and realized there is pretty close to a bingo with their TG-50 Generator/Inserter. Horita Co., Inc. Affordable Time Code and Video Equipment

This is a timecode generator that does window dubs. To do that it MUST lock to the incoming video, so it is synced to the video. And it has a timecode output - if this is, as I assume, linear timecode, and active while in generate/window dub mode - then there it is.

Runs off 9V, so easily portable.

List $399, probably street price a bit lower, so it is pretty close to the $300 mark I suggested. As a bonus, it will work as a window dub generator. ;-)

Now just feed that timecode into any really cheap wireless setup and you are good to go.

I might just remove the circuit board from that box and put it, along with the board from a wireless setup, into a more suitable one.

-Mike

I think you're making the common error of confusing timecode with sample clock sync. Sending timecode to a file-based audio recorder does not prevent sync 'drift.' File-based recorders do not record LTC nor do they sync their sample clocks to incoming timecode. The only thing such a recorder uses any timecode that is fed to it for is to timestamp the timecode reading associated with the first audio sample into the file header. That's it - timecode does NOT control the speed of either recording or playback. Wordclock does, The key is you need some means of generating wordclock that is in sync with the video sync burst. There are two ways - either a common clock feeds video sync to the camera through genlock AND also feeds wordlock to the audio recorder; OR the camera acts as the master and its video output goes through a box that outputs wordclock slaved to the camera's video sync. The Tascam (and the Sound Devices 788T (but not the 702T or 744T) and the Nagra VI) does that in the recorder, don't know off the top of my head about Deva or Cantar. The only way to prevent 'drift' is that the camera video sync and the audio sample clock must derive from the same master timebase, regardless of how you arrange it. Tiemcode's clock rate is just not high enough to accomplish that.

[Mike Demmers]

I am not confused about word clock vs smpte.

The Motu units sync their word clock to SMPTE:

MOTU.com - Clock, sync and MIDI

"Direct Digital Synthesis™ clock

The Traveler-mk3 is equipped with Direct Digital Synthesis™ (DDS), a DSP-driven phase lock engine and very high frequency digital clock source that produces imperceptibly low jitter characteristics (well below the overall noise floor), even when the Traveler-mk3 is resolved to an external clock source via either word clock or SMPTE time code. "

These are not recorders, 'resolved and 'phase lock' mean just that.

The Tascam HD-P2

TASCAM

HD-P2 Synchronization features:

SMPTE/LTC timecode input on locking XLR balanced jack

Timestamps Broadcast WAVE recordings from SMPTE input

Chase locks to incoming SMPTE timecode

Video clock input resolves to house clock

Tri-level sync support for HDTV applications

Includes Frame Lock, Lock and Release and flexible Freewheel settings for unpredictable timecode sources

Pull-up and Pull-down sample rates included for video format compatibility

I am less sure about this one, but 'Video clock input resolves to house clock' sure suggests this.

And 'chase lock' (not just 'chase' ) has a pretty specific meaning to me. 'Lock' being shorthand for 'phase lock'. You can chase without lock, you can lock without chasing.

The sound Devices units (except Sound Devices 788T), and others, make no such claims.

If I am wrong, there is some pretty deceptive advertising going on here...

Tiemcode's clock rate is just not high enough to accomplish that.

That is why phase locked loops were invented.

Linear time code has an included, derivable 4800Hz clock included. Hook that to a phase locked loop with a divide by ten counter in the loop, and you get 48000Hz.

Wordclock...

All that is required is that the vertical sync pulse have less than 1/48000 of jitter for this to work. Since any modern camera is operating off a much higher frequency clock, from which this is derived, that is pretty well guaranteed.

If the audio is recorded phase locked to the camera, there will be no drift. Both the camera and a digital audio recorder are really frame oriented devices.

48000 = 1 audio frame, one second.

1 second of video = 30 frames, or 48000 audio clocks, 48000/30 = 1600 audio clocks per frame.

Feed a video file into your nle, it counts 30 frames and says, that is one second.

Feed an audio file in, it counts 48000 samples and calls it one second.

So there should be no drift, provided both were originally synced to the same clock - the cameras, in this case.

Most file-based audio recorders just time stamp. The specfic devices I mentioned at least claim to also phase lock to incoming linear time code.

... I just looked at the Sound Devices 788T manual. It actually does NOT claim to lock to external linear time code. It only locks to word clock or video (which of course would also solve this problem). This surprises me, though.

-Mike

[Gary Nattrass]

Originally Posted by Andy Tejral

Well, you can run the boom mic through a mixer, have sound guy listen to mixer and send the main output wirelessly to the camera.

Thanks I was just about to say that as this is the most common application of sep sound.

I have a sign 44 mixer that can record direct to a sony D50 but I can also send a radio link back to the camera as well if required.

[Steve House]

Originally Posted by Mike Demmers

I...
... I just looked at the Sound Devices 788T manual. It actually does NOT claim to lock to external linear time code. It only locks to word clock or video (which of course would also solve this problem). This surprises me, though.

-Mike

Yep - perhaps I was unclear - I was trying to say that the SD788T and the Nagra VI share the Tascam's ability to lock their sample clock to incoming video sync but the other SD recorders do not. As for the sample clock chasing LTC, I exchanged emails with the good folks at Sound Devices a couple of years ago about that exact issue. I asked whether their "T-series" recorder's sample clocks would lock to incoming LTC and their response was that they did not. They specifically said the internal timecode counter chases external LTC but the audio sample clock does not, the reason being that it was impossible to derive a sufficiently stable clock from the LTC. It's my understanding that when external timecode is present, the system jams its value into the timecode count register in lieu of the internal clock's count but it never actually influences the internal clock oscillator frequency. It's not actually controlling the clock but instead it merely sets its value by continuously jamming into the register. In pocket watch terms, it's like continually twisting the crown to manually keep the hands showing the same time as the clock on the wall without opening up the watch to adjust the rate it ticks.

I think you are correct regarding the Tascam - page 21 of the manual claims the audio clock chases external timecode's embedded clock when present. Of course with consumer cameras like the OP's FX1, getting the darned LTC out of the camera and over to the recorder in the first place is a major stumbling block.

[Mike Demmers]

Originally Posted by Steve House

They specifically said the internal timecode counter chases external LTC but the audio sample clock does not, the reason being that it was impossible to derive a sufficiently stable clock from the LTC.

I think it is a shame they left this out. I can uderstand their reasoning in terms of following something like an old analog deck, but that is not the only scenario anymore.

It looks to me like we have fairly inexpensive solution to the lock/drift problem in the in the TG-50.

The video output of the camera is locked to its master clock, and hence to all video and audio it deals with directly.

The Horita. because it is also doing a window dub, MUST be locked to roughly twice the horizontal frequency accuracy in order to do that. And so will be the timecode it outputs.

Now that timecode clock is 4800 Hz but it's actually locked to a much higher freqency. But timecode is limited to about 11Khz bandwidth (by its spec). That is still less than the audio bandwidth it may be transmitted over, so that is not limiting its accuracy any more than the inherent bandwidth of the timecode clock (which must be at least 9600 to pass thatclock accurately per Nyquest).

So, worst case, we are looking at a lock inaccuracy of about 1/9600 or about 0.104 milliseconds. For frame accurate lock to video,you only need about 1/60 or 16 ms.

I call that good enough. The only thing you may not want to do is split a stereo track with one track on the camera and one on the remote recorder. But no one with any sense is going to do that!

Of course there is no way to get the absolute code, there will be an offset. But it will be a consistent offset - set once and everything else will fall into place and remain in sync.

Of course you also need a recorder that locks its clock to the timecode clock like the Tascam.

But the basic problem I was looking at - that few of these cameras have actual timecode outputs or sync of any kind - seems to have a solution. If I want to record a 2 hour concert to a multitrack and have a master shot that is in sync, here is a way to do it without having to use a camera with genlock. Before this thread I was pretty much convinced I would have to build this myself, or use a (big, not portable) Motu Midi Timepiece.

-Mike

[Steve House]

I'm looking at the Horita info and don't see how they can help the issue. Neither the WG-50 nor TG-50 output wordclock. You can feed them composite video and they'll read and output the embedded timecode but unless you're using a recorder like the Tascam that can lock its sample clock to incoming timecode we haven't gained anything. And with a recorder like the Tascam, the Horita becomes superfluous since the recorder will clock directly to the video anyway. True, it solves the issue of getting timecode out of a camera that does not have a timecode output but with most file-based recorders timecode is only a glorified slate anyway.

[Mike Demmers]

Originally Posted by Steve House

I'm looking at the Horita info and don't see how they can help the issue. Neither the WG-50 nor TG-50 output wordclock. You can feed them composite video and they'll read and output the embedded timecode but unless you're using a recorder like the Tascam that can lock its sample clock to incoming timecode we haven't gained anything. And with a recorder like the Tascam, the Horita becomes superfluous since the recorder will clock directly to the video anyway. True, it solves the issue of getting timecode out of a camera that does not have a timecode output but with most file-based recorders timecode is only a glorified slate anyway.

What is gained is that the Horita sends timecode, which is specifically designed to be sent over any audio path. The Horita is small, and a small cheap wireless transmitter could be attached to it. So it can be pretty portable.

I suppose you could wirelessly transmit the video output from the camera, but that would require a video transmitter - more expensive than something to transmit audio, and not as likey to be in an average kit. Lots of people probably have less-than-great old wireless systems around that they used before they got something good...but would work fine for this purpose.

Work clock is not, like time code, really designed to be sent over an audio path. So while I suppose you might find some device to translate video to word clock, and maybe manage to send it over an audio link, it woud be a hack at best. Also, timecode readers are usually designed to handle transmission errors gracefully. Word clock inputs - not so much.

With timecode, even though the exact timecode will have offset, at least there *some* there and the Tascam can use it to stamp the file.

Also, the device on the remote end may be something else than a Tascam, it could be a computer, perhaps a laptop. With my tiny JLCooper’s PPS-2 timecode to midi timecode converter, I could (loosely, but sometimes good enough) lock anything that will lock to midi timecode (almost any audio/midi production program), with this small box instead of a larger piece of rackmount gear. Timecode/midi timecode is very flexible stuff, you can use it to control lighting, smoke machines, etc. I have a tendency to push the technical edges a bit on productions...

So it's about smallest, most portable, most flexible, easily wireless, cheapest. The cheapest way to do something I don't expect to have to do very often, but which could be very useful from time to time.

Does this make better sense now? Its really about the wireless option. Cheaper than a whole lockit system.

Even the window dub feature might be useful for cueing purposes in certain situations. Sent to any cheap video monitor you get a timecode display almost for free.

-Mike

[Mario Vermunt]

I am too looking for a small and inexpensive device that can convert Blackburst to wordclock. My SD702 accepts only wordclock. I do not want to invest in an expensive (and mostly complex) device like the Aardvark since most of the times the audio stays in sync anyway. If Tascam can build such a device in a recorder that is far cheaper than my SD702 it should be possible to have one for say $500.

The PluralEyes software that automatically syncs audio sounds great but I think it will only work to synchronize startpoints and not hold sync for a prolonged time. Some sort of automatical slate therefore (am I right?).

Anyway not to complain though. I used to work with a Uher 4200 taperecorder with a built in sync module that could sync perfotape to the flashoutput of my 8mm filmcamera!