Video with the Canon 5D Mark II for still photographers part 2 Sound

Read part 1 form factor here 

Sound: OK here is the Canon 5D Mark II's Achilles Heel, sound. There are a bunch of interesting videos floating around on the web shot on the MK II. Notice what they have in common? NO ONE IS TALKING. They are silent films, we're talking Buster Keaton and Fatty Arbuckle.

live sound for this video is recorded on the Sony PCM D50 with an Azden sgmx1 shotgun mike watch it on vimeo here

Yes the MK II has an onboard mike -unlike the nikon D90- but the onboard sound quality blows. The onboard mike picks up tons of handling noise and if you are using an image stabilized lens, the servos sound like a lawn mower running in the next room.

Plugging an external mike into the camera's sound port only makes matters worse. There is so much line hiss that the sound is unusable. With any microphone I've tried or any that I’ve heard about it’s the same. Maybe a beachtek or a Field Mixer would take care of the problem but I don't know anyone who has actually tried one (if you have PLEASE write in).

So are we stuck with Buster Keaton? No the solution is to record sound separately and synch it in edit. I use the Sony PCM D50 Laforet likes the Fostex FR-2LE . To do this you'll need a shotgun mike. My AZDEN SGMX1 produces great sound and doesn't need phantom power, but there are many, many others. I secure the recorder and mike to my shoulder support to make things a little less cumbersome.

Synching is a matter of making a sharp noise at the start of the clip then lining the sound spike up in editing software. In this respect and others I’ll get to in later posts the MK II behaves much more like a movie camera than like a video eng. In fact it is a terrible eng. If you are gathering news, turn around times are going to be tight, there is not time to put sound clips and video together, and you sure can’t run around a fire scene or a car wreck slapping a slate together and yelling “take 2, rolling.”

Now a word –or more- about sound drift. Your sound recorder and the Mark II are going to record time differently. They shouldn’t I know. Why one minute to my Canon is different than one minute to my Sony recorder is beyond me. But I found out the hard way that they do.

It is not just with the Sony, people have had the same problem with other recorders. After 12 minutes of a continuous take they are almost a second out of synch!

On a short clip, less than a minute, you are never going to know the difference. But on a long one? It is going to be an issue. If the sound is even a frame off you will notice when people talk and suddenly it looks like you are shooting a badly dubbed Hong Kong fight flick.

Maybe someone can explain why this happens? My assistant thinks it is the compression, I think it’s because we are shooting video with a still camera.

Not to worry, there is a hack for this problem. In FCP I set the sound to play back at 99.98% (command J). Then I line up the sound spike at the fist of the clip and everything is cool.

I suppose that you could set the video to play back at 100.02% and get the same result but that requires rendering and who has time?

We’ll talk about editing next. Get ready to spend real money…

 -Stephen Alvarez

Read part 1 form factor here

 

 

Comments

[gary]

here is a link to a tv news weekend package using 5d2 with rode at -10db and camera mounted for all interviews and the reporter standup. i do simple one man band news with it, this is simply the raw audio no processing as it aired, altho the web site adds some distortion the way they post it.
only the voiceover was done without 5d2, at studio.

copy and click on
http://www.wlfi.com/dpp/news/local/local_wlfi_westlafayette_feastprep_20090927

[gary]

hi!
i have been using single system sound since december.
experimented like crazy with mics and levels and technique.
news and commercials and corporate video, with interview sound from 5d2.
for news bites and broll, use a rode shotgun mono with -10 db switch kicked in.
works on air no problem, no noticeable hiss.
for wireless lav, using a lectrosonics unit with receiver output adjusted to -28 db. ALWAYS have talent do a countdown from 5, and the autogain is fine.
THEN i run a simple noise reduction program contained in cheap adobe premiere 7 on all footage. the primary sound is untouched, the hiss is GONE. seriously.
sure beats syncing stuff up. sounds fine.
-gary

[mdoane]

Stephen-

Yes, it is. I was the AC on MissHIV, a doc about AIDS in Africa. Spent a couple of weeks in Uganda, but never ventured too far from Kampala. We heard many stories of the conflict in Northern Uganda. Your work from there is stunning. Great job.

[Stephen Alvarez]

Peter,
thanks for the info. It will be interesting to see how 2 5d mark IIs synch, I'm doing a 2 camera shoot with them next week and will post results...

[Stephen Alvarez]

responding to Matt:
Matt thanks that makes sense to me -in that sort of way that string theory makes sense- it seems like something is actually running at 29.97.

An unrelated question, at the opening of your reel is that downtown Kampala?

[peterg]

I think the issue is in oscillator built in devices. Every digital device has it. I usually notice this problem when I shoot 3d stereo movie with two, exactly the same camcorders (Sony via LANC controller). At the end of shooting, let's say 2 min clip, both clips can differ in one or two frames in length, which is of course unacceptable, because frames are not the same on left and right clip. To solve this problem I use special synchronizer. If you try to shoot video with two 5DMKII you will notice the same problem, both clips will be tiny different in length (if they are started at the same time). With different devices you will get bigger differences...

[mdoane]

The audio drift is likely because your recording at 30fps on the MkII and at something else(probably a sample rate of 48kHz) with the field recorder. The field recorder may not even have timecode, but FCP is probably reading it as something different than 30fps. Audio timecode is arbitrary,read easy to change, but it is necessary for sync. Since slowing it down to .02% worked, FCP is probably reading it at 29.97fps. A minute is a minute on both devices. What really matters is how many audio samples FCP is playing back in a minute, read speed, in your 30fps timeline.

I think I just confused myself. Does that make sense?