DSLR best practices from the perspective of a location sound recordist

[ryanpeds]

I have done a lot of DSLR shoots all with different workflows. A common workflow that is basic and works is just syncing with a dumb slate. I have done a couple of large projects shot with DSLRs and we sent timecode to camera. We built special cables that sent timecode from a lockit box to channel 1 and then mounted a small shotgun on the top for reference audio to channel 2. The timecode channel has a built in pad. They would then sync everything in Avid since it reads timecode from audio tracks. We would also sync with a dumb or smart slate mainly for logging the info of the takes. One main note is to remember that the Canon DSLRs have a 2 frame audio delay so everything has to be shifted to 2 frames to line up. Not sure what the delay is on other DSLRs.

[dominiquegreffard]

One main note is to remember that the Canon DSLRs have a 2 frame audio delay so everything has to be shifted to 2 frames to line up. Not sure what the delay is on other DSLRs.

So just to make sure that i understand typically on a canon dslr audio is behind video from 2 frames?

Therefore i cannot compensate onset by adding audio delay because it can only make it worst.

[Constantin]

The advantage of the scratch track is the improved quality. This helps wirh instant playback, for one. But then it also helps with PluralEyes in those situations where the camera and the sound guy are some distance from each other. And it's even more helpful when there are two cameras and the second is perhaps even further away. <br />I've seen a few producers who have regretted not going with a scratch track. In many cases the on-board mic will do the job just fine, but in many cases it won't. It's not more hassle for the camera op to mount a small receiver on the camera than it would be to have a mic on there. And with an additional TC box it's even worse. <br />The 2-frane offset can easily be fixed in post.

[ryanpeds]

So just to make sure that i understand typically on a canon dslr audio is behind video from 2 frames?

Therefore i cannot compensate onset by adding audio delay because it can only make it worst.

[ryanpeds]

Correct. The audio on the Canon DSLRs is 2 frames behind. I would imagine this is from processing inside the camera.

As far as mounting goes, I use a small set of rails with Velcro on the back to hold the TC box just behind the camera. This usually doesn't get in the way especially when doing interviews.

[dominiquegreffard]

Correct. The audio on the Canon DSLRs is 2 frames behind. I would imagine this is from processing inside the camera.

As far as mounting goes, I use a small set of rails with Velcro on the back to hold the TC box just behind the camera. This usually doesn't get in the way especially when doing interviews.

Ok thanks for clarifying this!

[Boomboom]

It's not the only camera type/brand/model that does that. And the TC offsets are different between them.

[Boomboom]

Tests made here by a post audio mixer/editor lead to these results between 3 cameras:

Arri Alexa : correct TC / audio slightly preceding video

RED (don't have the model) : TC preceding 2 frames / audio & video synced

Sony EX-3 : TC preceding 1 frame / audio & video synced

Bottom line : knowing that, I'll leave you to your own conclusions and receipes concerning your shooting practices.

[ryanpeds]

The Sony F3 is really bad. It's usually 4 frames offset but I've seen it change to 3 or 5 or 7 even while being physically connected to a lockit box.

[Marc Wielage]

Correct. The audio on the Canon DSLRs is 2 frames behind. I would imagine this is from processing inside the camera.

 

Audio is typically ahead, because of processing on video that delays the picture slightly (which is the usual problem with video signal paths in general). I haven't seen a sync problem with Canons, but anything is possible. But I'm the guy who keeps saying that audio on DSLRs should be regarded as a scratch track at best.

[Stewart]

Just want to take a moment to thank Constantin, Alexander Burstein, Ryan, Dominique and Pierre for their comments. And Marc for his additional post.

 

There's not only some very helpful raw information in this thread, but really interesting and helpful differences in perspective/workflow.

 

Cheers.

[JDirckze]

I've only done one show with a 5D, and for that particular show the workflow we used was to send a scratch track to channel one and TC (via Lockit) to channel 2. The camera op had rails and a V-Lock system powering the camera so mounting was easy and in fact helped with balancing the camera. We never had any complaints from post regarding sync.<br /><br />Understandably without rails and a shoulder mount this option would be difficult.<br /><br />Seems most camera ops and shows here in Australia that want the 5D look have now moved to using the C-300 or C-100.

[Nate C]

Doing a doco at the moment shot on two 5Ds. Actually since moving back to Sydney been doing a more stuff on the 5Ds than I would prefer. That said at least they are quiet. 

 

My method is I send a scratch track to camera wirelessly - Lectro IFB - either my mono mix or if I am doing split track (L/R) my left track. This way the camera has a track that is exactly the same as one of my tracks to help plural eyes out.

 

Then I set the time on the camera to within a second of my TC. Often the Op has no idea how to do this.

 

Lastly I get the cam op to shoot the TC display on the Nomad, whilst I do a verbal ID and tone at the beginning of the day. This way the editor can see that TC and camera time are similar and they have information on the camera of what folder to find the days audio, what track I am sending them, to - do not use camera audio as it is a guide track only and not of broadcast quality - and what I am recording at Bit/sample/tone date etc. Kinda like the old days of sending bars and tone.   

[Marc Wielage]

Doing a doco at the moment shot on two 5Ds. Actually since moving back to Sydney been doing a more stuff on the 5Ds than I would prefer. That said at least they are quiet. 

 

One big issue, though, which I encountered on long interview shoots: the 5D and/or 7D will just shut down when they get too hot. This depends on the ambient room temperature as well as the length of time of the specific shot. Contemporary directors -- even for documentaries -- don't seem to get that interview subjects need to have a break every 20-30 minutes, and I think they push the subjects too hard when they go to 45 minutes, 50 minutes, even an hour under the lights. Not only do the cameras heat up, the people on camera get really stressed out and tired over time. It's not the same as doing a dozen 3-minute takes in a row, which is much less stressful (for people or gear, in my opinion). 

[S Harber]

We're now on season 3 of a show called Wilfred where we have been using DSLR cams for the run.

It started with 5/7D Cannons and then last season moved to the Nikon as they leapfrogged ahead of Cannon in the chip dept for a while there. 3 cameras in play and it won the ASC award last year (not that awards are real arbiters of quality or success).

Our methodology was to jam padded lockit boxes into the audio input and mark the shots with a timecode slate. They tried various workarounds as tests-Pix 240,Black Magic Shuttle drives etc. but nothing seemed to work like simply recording on the camera as there were delays in outputs that would have made folks crazy.

That's how we do it on Wilfred.

 

Scott Harber

[Christian Spaeth]

Thanks for sharing your workflow, Scott. A little off topic, but I watched the first two seasons of Wilfred and was always amazed at how clean the dialogue is. Obviously with three cams in play I suppose a lot of lav in the mix, but all very clean and crisp, I enjoyed listening! I always wondered how Wilfred is lav'ed. My guess is in the hood, middle of forehead?

[S Harber]

Thanks for sharing your workflow, Scott. A little off topic, but I watched the first two seasons of Wilfred and was always amazed at how clean the dialogue is. Obviously with three cams in play I suppose a lot of lav in the mix, but all very clean and crisp, I enjoyed listening! I always wondered how Wilfred is lav'ed. My guess is in the hood, middle of forehead?

Hi Christian and thanks.

We give the director the first pass in it's entirety staying out of all the cameras then tend to break the C/wide camera after 3 seconds or so after the 1st take and he can either push in or hose the rest of his shot. Thereafter, we push in and try to force the cameras in to singles and then start shooting the actual footage that gets used. The show gets about 3 passes of a scene on a take (18 mins or so) as the director is averse to cutting.  We basically do about 2 series of takes per setup.  A pain at times but it is working well as I'm testing out the Pix260 so the longer takes bode well as the metadata aspect of the recorder is still being sorted out at this point.

We put a lav in Wilfred's hood and have loops sewn in to get the cable to his back. Nothing very special there. Jason/Wilfred has a fantastic voice and Elijah is a pro who is a dream to work with. They both have the chops to be able to deliver a quiet voice in a loud fashion as well as a thorough understanding of what works on the technical side. Being fairly egoless helps immensely as well.

 

Sorry for the sidebar there everyone, but... I'm not a fan of the DSLRs and it's not my call (just to bring it back to the topic).

 

S

[Marc Wielage]

Jason/Wilfred has a fantastic voice and Elijah is a pro who is a dream to work with. They both have the chops to be able to deliver a quiet voice in a loud fashion as well as a thorough understanding of what works on the technical side. Being fairly egoless helps immensely as well.

 

It's always terrific to deal with actors who understand the art of the stage whisper. Mr. Deichen and I were on a set a few months back where we had five actors WHO ALL SPOKE UP HERE and then another actress who spoke way down here, within the same scene. I was stunned that the director didn't reign her in. One "assumes" it was all sorted out in the final mix, but that was a very, very tough project for sound. 

[ryanpeds]

Audio is typically ahead, because of processing on video that delays the picture slightly (which is the usual problem with video signal paths in general). I haven't seen a sync problem with Canons, but anything is possible. But I'm the guy who keeps saying that audio on DSLRs should be regarded as a scratch track at best.

Mark,

It might be the other way around. I can't remember as it was over 2 years ago when we did the test. What we did was shot time code and slate and then ran time code into the camera and then checked the delay in the Avid. It was always a constant 2 frame delay from the visual slate time code and the time code being read on the audio track. If I have a chance I'll do it again and pull a screen grab.

[ryanpeds]

We should start a post with all the camera delays and time code issues so that they can be quickly referenced. I know on the new Ambient ACL204s you can offset the time code but I think it's only increments of 5 or 10 frames. I wish it were single frames to get it dialed exactly in.

[Christian Spaeth]

Hi Christian and thanks.

We give the director the first pass in it's entirety staying out of all the cameras then tend to break the C/wide camera after 3 seconds or so after the 1st take and he can either push in or hose the rest of his shot. Thereafter, we push in and try to force the cameras in to singles and then start shooting the actual footage that gets used. The show gets about 3 passes of a scene on a take (18 mins or so) as the director is averse to cutting.  We basically do about 2 series of takes per setup.  A pain at times but it is working well as I'm testing out the Pix260 so the longer takes bode well as the metadata aspect of the recorder is still being sorted out at this point.

We put a lav in Wilfred's hood and have loops sewn in to get the cable to his back. Nothing very special there. Jason/Wilfred has a fantastic voice and Elijah is a pro who is a dream to work with. They both have the chops to be able to deliver a quiet voice in a loud fashion as well as a thorough understanding of what works on the technical side. Being fairly egoless helps immensely as well.

 

Sorry for the sidebar there everyone, but... I'm not a fan of the DSLRs and it's not my call (just to bring it back to the topic).

 

S

Wow, really insightful post. Thanks Scott for sharing. I will try that approach on my next multi cam shoot.

[stevegrider]

One thing to be wary of when using PluralEyes:

If you're shooting a :30 spot when the dialog is wall-to-wall, it will confuse PluralEyes. If the talent is consistent, the waveforms will be virtually identical. The work around for this situation is have the AC announce the full scene/take information at the head of the take. If they simply say "mark" it's not enough to differentiate one take from another.

[pvanstry]

For me two workflow work and I offer both to the producer and post.

1- feed an audio scratch track mono to the camera. The idea is for post to use plural eyes to resync. But I record only a mix and no Isos track. The reason is that whatever the sound file contains as to be pretty much the same that I am sending to the camera ( basically the camera gets a summed mono feed of my stereo mix). If I was to do Isos then every time I fade down an input ( but it is still being recorded in Isos ) pluraleyes will be confused and have issues in resincing. I send a 1k tone at every take kind of like a beep/slate.

2- feed audio LTC ( timecode ) into a track on the camera. Then post uses AUX FCP TC reader from Videotoolshed and is able to convert the image file by reading the audio Timecode and converting it to a timecode image file. Then they use BWF QT merge and merge the audio file with the image file using timecode. Import into editing software and voila. All good.

Works good as long as everyone is in the loop.

[dominiquegreffard]

But I record only a mix and no Isos track. The reason is that whatever the sound file contains as to be pretty much the same that I am sending to the camera ( basically the camera gets a summed mono feed of my stereo mix). If I was to do Isos then every time I fade down an input ( but it is still being recorded in Isos ) pluraleyes will be confused and have issues in resincing.

I ve looked into the pluraleyes user manual and i have not seen anything regarding the use of multitrack but i doubt that having let s say a mono mix on track 1 and iso lav's on let s say 2,3,4 will mess up plural eyes.

I think that as long as one channel is identical to what you scratch feed to camera then pluraleyes is fine. The other iso channels will simply follow. I m not 100% sure about this one but i tend to think that. I ll contact the develloper to get that info confirmed.

http://www.redgiant.com/docs/pluraleyes/

[TRCS]

I was just dealing with this today- plural eyes can sync it all fine (2 stereo files), but its what to do after that's not clear... We've found it easiest to sync pre video editing by replacing the camera audio with recorder audio in PE (problem is you can only replace it with two tracks (LR)).