How about good quality Audio?

[Bernard Racelis]

Originally Posted by Steve Mullen

What you hear is a very good job of preventing audio clipping. When the overall signal gets near clipping, the gain is reduced which naturally lowers the background noise. It's hardly objectionable. Clipping, which would ruin your audio would be far worse.

Of course, a human might think far enough ahead to set the gain based on the loudest noise. If they knew ahead of time what that would be at a race track. And, where exactly on such a tiny camera would you put the knob? Or, do you want someone to stop and use a menu. I really doubt anyone shooting in gun & run mode at a race track would have time to do this.

Most operators shooting with a $2500 camera would be very happy to get the shot for the nightly news or a webcast.

If and when Canon decides that a video from a DSLR can be used for purposes other than the web or quick gun & run video grabs, there are more than enough buttons on the camera for manual control.

The top dial is already being used to adjust the speaker volume during video playback so it makes sense to put the gain there. That dial is normally used for the shutter when shooting stills, but since you don't normally change the shutter during video recording, the shutter can be put in the menu. And you still have the rear dial for the aperture. The EF mount doesn't have servo-zoom so you don't need a button for that.

[Steve Mullen]

Not to beat a dead horse, but here is a paragrapg from a TI TSC2100/01 application note. These are the "rules" by which this AGC operates:

Without speech or signal, decrease the Maximum Gain Applicable variable to have an acceptable level of amplified noise.

Next, with the lowest speech or signal source, increase the Noise Threshold
variable to a level so that the weak signal can be detected and gained.

Then, increase the attack time if the speech initially sounds bad; decrease the decay time if the speech sounds too noisy for too long at the end.

"An AGC can automatically adjust the gain so that the signal is maintained at a certain nominal constant level (called target gain) over a changing range of real-world
conditions."

This simple IC has many parameters:

1. Target gain = –5.5 dB
2. Attack time = 8 ms and decay time = 500 ms
3. Noise threshold = –90 dB
4. Maximum gain applicable= 59.5 dB
5. Hysteresis = 2 dB
6. Debounce time from normal to silence = 0 ms
7. Debounce time from silence to normal = 0 ms

Please note I'm not claiming any of these still cameras use this type of chip.

[Dan Sharp]

Could anyone with experience using a 'double-system' for audio and video offer some pointers? Specifically, I'm wondering what the best way to carry a digital flash recorder with internal mics in an X-Y pattern (Sony PCM-D1) at the same time that I'm shooting with the 5D Mark ii alone. Do I just have to find a friend?

[Evan Donn]

Originally Posted by Steve Mullen

What you hear is a very good job of preventing audio clipping. When the overall signal gets near clipping, the gain is reduced which naturally lowers the background noise. It's hardly objectionable. Clipping, which would ruin your audio would be far worse.

Of course, a human might think far enough ahead to set the gain based on the loudest noise. If they knew ahead of time what that would be at a race track.

Preventing the audio clipping is fine, it's the slow creep up of ambient noise afterwards that's objectionable... you can argue all day about whether it's really 'AGC' or not by a variety of definitions but in the end I don't want ambient noise creeping up like that and then clamping down as soon as someone talks or there's a loud noise - it's the audio equivalent of autoexposure clamping down when someone walks into frame with a white shirt. It sounds unnatural, it sounds bad, and it sounds unprofessional....

You don't need to know exactly the loudest noise ahead of time... you pick a comfortable level for what matters to you and set one channel -10db to that level. If you get a sudden unexpected spike in noise you can recover it from the lower channel. Or you can add a limiter which only clamps on peaks that would clip and otherwise leaves it at the level you set it.

If you could set it.

At least they gave us exposure lock.

[Don Miller]

Is there a device that can record audio and simply write a time code to the 5DII audio track. The audio wouldn't be used from the 5DII, just the time code.
Someone half jokingly talked about recording the broadcast time signal to both. But we would want something that would sync automatically in a non-linear edit program.
A poor many gen lock. We don't need extreme accuracy.

[Don Miller]

Originally Posted by Dan Sharp

Could anyone with experience using a 'double-system' for audio and video offer some pointers? Specifically, I'm wondering what the best way to carry a digital flash recorder with internal mics in an X-Y pattern (Sony PCM-D1) at the same time that I'm shooting with the 5D Mark ii alone. Do I just have to find a friend?

You probably want to get it away from the camera. These mikes often hear what's directly behind too. Experiment.
Start the audio and video and then clap you hand in a location visible to the camera. This will allow you to easily sink audio and video. If your doing a lot of takes you can buy a clapboard.
I suggest leaving the audio on in the camera too as a backup and reference for syncing.

[Don Miller]

For music I use the Apogee Duet to a Macbook Pro (Mac only):

Apogee Electronics: Products: Duet

This has an A/D converter close in quality to their recording studio products.

[Larry Vaughn]

I wonder if you could feed the audio input with some kind of jam or reference signal, an ultrasonic one that only dogs can hear, at a constant or adjustable level, which would then set the auto gain accordingly. Then you could filter out the jam signal in post, and have your human audio where you want it.

[Mike Calla]

Can anyone share their wisdom on using a separate recorder / a slate and syncing in post? In the future I’ll be doing A LOT of interviews and maybe there’s a tidbit of priceless info that’ll save me time and headaches:)

[Don Miller]

Originally Posted by Mike Calla

Can anyone share their wisdom on using a separate recorder / a slate and syncing in post? In the future I’ll be doing A LOT of interviews and maybe there’s a tidbit of priceless info that’ll save me time and headaches:)

Plenty of info in the audio forum. The 5D specific suggestion I would make is to consider recording audio to both camera and external, if convent. Also, in an interview situation the 5D audio might work fine alone if the 12 minute record time is acceptable (with an off camera mic).

[Bill Binder]

Originally Posted by Mike Calla

Can anyone share their wisdom on using a separate recorder / a slate and syncing in post? In the future I’ll be doing A LOT of interviews and maybe there’s a tidbit of priceless info that’ll save me time and headaches:)

For interviews, I'd personally do something like:

1. Mic(s) into stand alone recorder, just let it roll for the entire interview. Recording sound separately like this might have another benefit in that if the 5D2 cuts off, at least you still have continuous sound that you could drop some B-roll over if need be.

2. Outputs of recorder fed into inputs of 5d (as mentioned in above post), probably will need adapters, and will also need to experiment with levels to make sure the 5D2 can handle what you're giving it (experiment before the interviews obviously since you can't actually monitor the 5d2).

3. Optional, but recommended. Somehow slate each take on the 5d2. The most simple form of this would be to have someone use a slate/clapper board, or to just have the talent clap once once the camera is rolling. This will make syncing later much easier, but as long as you have the same audio on the camera, you should be able to manually sync regardless (just WAY FASTER if the clap is there). Another low-budget take on slating that I use sometimes is to record the display of the recorder at the begging of the take. Although it won't be "true" timecode, it still can help locate the "approximate" location of the clip in post. In fact, with some recorders, like the one I use (an R44), both the record time AND the filename is show on the display, so if I record that on the video, I get both what audio file to sync to and an approximate timecode in that file. Although during an interview doing that might be disruptive.

4. Even more optional would be to slate the tail of the 12-minute video clip. This would only be needed if the video and your audio get significant drift over a 12-minute clip (e.g., you sync the head, but the tail is out of sync still). Sometimes you can sync the middle of the clip, say 6 minutes in, and then spread the drift out across the whole clip (slightly off at beginning, good in middle, slightly off by the end). But for 12 minutes, I bet drift isn't that bad. I record a lot of music, and over the course of an hour or two (obviously using a video cam without the 12-min limit) drift is a real issue that almost always needs to be addressed if you aren't completely synced in real-time using high end gear. But since this cam can't do more than 12-minutes, drift probably won't be an issue.

Personally, I think the 5D2 would be tough in a long interview situation if you are going solo mode (just yourself, no crew). It seems to me that there's a real chance of disrupting the flow of the interview with the start/stop of the camera (especially if swapping out CF cards, maybe not so much if just hitting record again). However, if rolling audio separately, maybe that's mitigated by the fact that if needed, you can miss some video but can depend on solid audio (think b-roll again). On the other hand, the interview footage will look so freakin sweet with the 5D2, and the ability to do it in low light settings, it might just be worth it...

[Salvador Garza]

The camera records 30fps, while NTSC is 29.97fps. How does this 0.1% difference affect syncing? How long in a take until it's noticeable?

[Jon Fairhurst]

Regarding latency, when playing a keyboard, 10ms is the threshold. When watching video, the threshold for audio delay is closer to 50ms. We're more sensitive if the audio is early, than if the audio is late.

With a 0.001 error, the delay will be noticeable (50ms) by about 50 seconds. The bottom line is that if you squeeze the video, you also need to squeeze the audio. The other option is to drop the occasional frame. The maximum error is then about 17ms.

[Mike Calla]

Thanks Bill !!– did a homemade clapper test run last night.
Suggestion # 1 is great; having additional audio might be a lifesaver in the future!

#2 When using a standalone audio recorder > running audio from the audio recorder into the camera is great because it made syncing the audio so much easier. Without the camera’s audio waveform I was doing a video(clapper) to sound sync which was a little bit tedious – When I had the camera’s audio waveform as well as the audio recorder’s waveform I could easily and quickly identify the sync positions.

#4 is good practice for longer takes with equipment that doesn’t have some form of time code sync. I come the audio world and having a master word(time)clock is imperative!

Actually, I’ll be doing one-man interviews – lots of them in fact! Each interview will last at least an hour; I’m a little worried about the stopping the interview 5 or so times! But I’m pretty good with interviewees and my interview topic is quite light hearted so hopefully it won’t be troublesome, and you’re right the visual outcome will more than make up for it!

[Mike Calla]

Originally Posted by Salvador Garza

The camera records 30fps, while NTSC is 29.97fps. How does this 0.1% difference affect syncing? How long in a take until it's noticeable?

This is confusing Salvador, and if my limited knowledge of video time code is correct the difference between 30fps and 29.97fps is in how the frames are counted and not actual frames

30fps is referred to as “non-drop frame”
And
29.97 is also referred to as “drop frame”, but in actuality it does not drop a frame but merely drops a frame number: Simple example: 1,2,3,4,6,7,8,9. The five is dropped, there are only 8 frames but the counter would display “9”

3000(or any amount of) frames in both 30fps and 29.97fps will last the same amount of time – is it only the time code that is calculated differently. These two time codes only count video frames and they count differently.

I don’t want to do the math but a quick a google search will give you the differences in time code running time display for real world duration/frame amount.

However, and more importantly – Broadcasters only accept Drop Frame - AND what makes it so confusing… and this seems crazy… is that 29.97 is in fact the actual frame rate for NTSC! You can edit your 5D footage in drop frame or non-drop frame – doesn’t matter – but for broadcast your final output should be in 29.97 drop frame!

Drop Frame vs Non Drop Frame for details or to confuse yourself more.

If i was wrong, someone PLEASE correct me:)