Zoom H2 for video?

[Sam Kanter]

I know a H4N is better, but I have a Zoom H2 and want to use it for a small two-cam 60D plus 5DII music/dance project and later sync with Pluraleyes.

What are best setting for H2 to give best chance of good sync? I assune 48K is better than 44.1?

Thanks in advance...

[Matt Davis]

And 24 bit 96 KHz is better than 16/48 - you can downsample later.

Despite what I was told by a famous trainer, I do get better results recording at 24/96, as there's more info in the audio to play with, the noise floor is lower, I can adjust levels without getting in trouble, and things just sound better IMHO.

Think of it as 4:4:4:4 audio rather than 4:2:0 audio. Although it's all going to end up compressed every which way, at least you've got plenty to work with at the edit stage.

As for sync, the original H4 was a bit iffy, but I think they've got the whole sync thing licked now.

Pluraleyes is definitely made of magic smoke. Quick, efficient, accurate and spooky.

[Sam Kanter]

I was more concerned about sync than audio quality (which is probably fine at 44.1 16-bit). Are there specific rates that are better for sync?

[Matt Davis]

You're fine with sync - the reason I mention bit rate was that it's better to go higher at first and come down, rather than start low and bump up, and 'not quite' bit rate was purported to be the issue with the original Zoom 4's drift.

'Probably fine at 44.1k'

Most edit codecs expect at least 48KHz, and upsampling isn't ideal.

Try 24/96 some time. You may be converted too.

[Sam Kanter]

Thanks, Matt. I will go for 48K.

Since I have no audio person, do you think AGC is conceivable, or hope manual will work? Sound is a live, improvised bass that will be bowed. I can test levels beforehand...

[Matt Davis]

Okay - nobody else has jumped in, so apologies for banging what appears to be the same drum. But...

AGC is like Autofocus - when things are hairy and wild and there's a possibility for human error because there's just too much going on, AGC and Autofocus will get you at least something. And if the content isn't pretty taxing, it will get you something pretty usable. But there's a cost.

In music, if there's any dynamism - loud bits, soft bits - and this seems to happen a lot in live performances, AGC will duck and dive with your levels and it's all a bit too heavy handed to make comfortable listening for your audience. But fixing the levels may risk being taken by surprise with peaks in the levels, sudden dips - so you either get a sound engineer to ride the levels, or suffer noise (boosting the quiet bits), distortion (the loud bits) or the rubber sheet of a limiter (which is roughly the top half of AGC) with fixed audio levels.

Or, and I apologise in advance for sounding like a stuck record over this...

24/96 olds a HUGE dynamic range which you can pull up and down to suit, so get your meters wiggling in the middle and you get the peaks *and* the troughs. Then you get to ride the levels (and do anything else) in post.

[Jon Fairhurst]

Regarding 24/96...24 bits gives a huge dynamic range. 96 kHz doesn't. I'm a 24/48 fan as this allows us to avoid the additional downsample from 96 to 48 kHz.

We don't shoot video at 48 fps to downsample to 24 fps, but we want as many clean bits in each frame as we can get.

[Matt Davis]

Originally Posted by Jon Fairhurst

We don't shoot video at 48 fps to downsample to 24 fps, but we want as many clean bits in each frame as we can get.

Fair point - being awash with 32GB cards, necessity hasn't driven me to cook up any variations on standard settings, just turn it up to 11 and enjoy the difference.

Is it all down to dynamic range, though? I'm sure I can hear the difference between 48 KHz recordings and 96 KHz recordings.

[Jon Fairhurst]

It depends on the overall signal path. Are you listening in 96 kHz? If not, what software are you using to downconvert from 96 to 48? It also depends on the recorder. It's possible that a given piece of gear has better analog anti-aliasing filters for 48 kHz recording than 96 kHz recording, or vice versa.

Most professionals record film dialog at 48 kHz. For sound design, where the sound may be re-pitched, or for music, where you might deliver 24/192, higher sample rates make sense.

Anyway, 24-bits is a good thing, though if the preamps are cheap, you won't get much our of those additional eight bits. Higher sample rates may or may not be an advantage, depending on what you are doing and what equipment you're doing it with.

If recording in 96 and downconverting sounds better on your equipment than recording in 48, then keep doing it. Especially with 32GB cards! :)

[Matt Davis]

Very illuminating. Thank you.

And yes, it was music where I heard the difference, rather than dialog - though the halo effect may also play a factor.

[Jon Fairhurst]

I definitely hear a difference between CDs and DVD-A discs. However, it's subtle. If the music is playing while I'm up and about the house, I don't notice the difference. However, if I'm sitting in the sweet spot, the difference is clear.

That said, it's not that the instruments sound "better" or even "crisper". It's that the sound stage is more clear, larger, and better defined. I think it's due to the improved phase accuracy, rather than the additional frequencies above 20 kHz.

If you put in a disc and asked me if it was a CD or DVD-A, I might not be able to guess correctly much more than half the time. That said, if you were to play the same song on CD or DVD-A, back to back, I'd expect to get near 100%. But only if listening from the sweet spot in a quiet environment.

To me, the value of SACD and DVD-A over CDs depends on how the listener consumes music. If it's casual listening, CDs are good enough. But if you really like to experience music, SACD and DVD-A are worth every penny.

Bringing this back to audio for video, it's much the same thing. If you are recording a great talent in a nice space with a good mic on a quiet recorder, the 48 vs. 96 kHz thing could be important. If the talent/mic/room/recorder are sub-par, changing the sample rate won't improve a thing.

But 24-bits? If your recorder can capture more that 16-bits of dynamic range, 24 bits rock!

[Sam Kanter]

The sound being recorded is this bass player who will pluck, bow, and make other sounds with his bass while interacting with the dancer. This was recorded with a Rode video shotgun. This is a church near Times Square, and I will have lots of problems with car street noise which will possibly be edited out.

This is not a documentation, but a creative piece that will be shot in many takes at different perspectives with two cameras, then edited together creatively. I'll have to find the best position to place the recorder, without it being seen on camera.

YouTube - In the Shadows

Any helpful ideas would be appreciated. Thanks in advance...

[Sam Kanter]

Originally Posted by Matt Daviss

Okay - nobody else has jumped in, so apologies for banging what appears to be the same drum. But...

AGC is like Autofocus - when things are hairy and wild and there's a possibility for human error because there's just too much going on, AGC and Autofocus will get you at least something. And if the content isn't pretty taxing, it will get you something pretty usable. But there's a cost.

In music, if there's any dynamism - loud bits, soft bits - and this seems to happen a lot in live performances, AGC will duck and dive with your levels and it's all a bit too heavy handed to make comfortable listening for your audience. But fixing the levels may risk being taken by surprise with peaks in the levels, sudden dips - so you either get a sound engineer to ride the levels, or suffer noise (boosting the quiet bits), distortion (the loud bits) or the rubber sheet of a limiter (which is roughly the top half of AGC) with fixed audio levels.

Or, and I apologise in advance for sounding like a stuck record over this...

24/96 olds a HUGE dynamic range which you can pull up and down to suit, so get your meters wiggling in the middle and you get the peaks *and* the troughs. Then you get to ride the levels (and do anything else) in post.

Yes - I pretty much know all this. There is no sound person. I'll probably keep on manual, set levels low and hope for the best...

[James Donnelly]

Issues around syncing have been covered, but I have a H2 and 24/48 is best from a fidelity point of view. 96k is actually worse as it introduces audible artifacts.

As for 24 bit it is definitely the way to go, but you won't see the full benefit in dynamic range increase that has been mentioned. You will see a marginal improvement if any.

The ADC in the H2 is the TI TLV320AIC32 which has a 92dBA signal to noise ratio, so 24 bits might seem a bit pointless given that this is below the theoretical SNR of 16bit audio.

At least you will see slightly less fidelity loss through processing a 24 bit signal in post, so there's no harm in it.

PS. I sometimes feed the line out from the H2 into the 550d with the analog and digital gain at 0 in the camera, but then you don't have ML yet on the 60d so you can't do this. It's not quite as good as going double system with the H2's built in mics, but it's close and avoids the whole sync issue.

Hopefully you'll get ML soon, looks promisiing that you will.

[Sam Kanter]

I think 16-bit 48K should be good enough. I'm more concerned with best mic placement, good levels, and outside noise than I am with minor differences in bit rates. It's just me directing sound, lights, another camera person, etc. I only have a 4G card for my Zoom - would need a 16GB to record at 24-bit. I just bought two 32GB cards for video (plus two 16GB I already have).

So far no news on ML for 60D - hopefully in the near future. 60D will record audio manually, but one needs a high-output mic.

EDIT: I think 24-bit 48K has no down side, and would give me enough time even with my 4GB card. Again, finding good mic placement and levels for H2 seems a bigger issue than bit rate.