Audio Capture and Magic Lantern at DVinfo.net

Bill Binder · September 29th, 2009, 01:42 PM

So, just wanted to provide some feedback/observations to the forum. First, I should mention that I do a lot of music events primarily from a stills and audio perspective. It's a little niche I've carved out for myself. <SMILE>

When it comes to audio for these events, I pretty much ALWAYS roll double-system. My typical setup would be taking a stereo soundboard feed plus a stereo pair of ambient (most often 2xAKG480+ck61/ck62/ck63) into a modded R44. So, it's not multitrack, it's very much a smallish field rig, but I can make great recordings with it, much better than soundboard alone or ambient alone.

That all said, with the recent purchase of a 5D2 earlier in the year, and some past dabbling in video (I know just enough to be dangerous, LOL), I've started to incorporate video into some of my work -- albeit via monopod with manual focus pulling off the LCD (LMAO). So, now it's audio first, stills seconds, and video third (usually one or two full songs), all at the same event as a solo recordist. If things work out, they get edited into a multimedia package of a single song caught on video with the audio synced in and stills laid on top for coverage and b-roll.

Anyway, prior to ML, I had some difficulties. For starters, manual control of ISO and aperture, but that's not an issue anymore. But more importantly, it was easy as hell to blow out spot lit musicians without completely realizing it. Now, with zebras, no problems. Second, at loud shows, the in-cam audio was so clipped that sometimes I couldn't even use it to do manual sync -- I had to go off video alone. Well, ML has fixed that too. In fact, with all the gain settings at zero, I'm actually quite AMAZED that the 5D2 produces audio that isn't really distorted at all in very high SPL environments. I would normally expect their mic to completely distort despite gain settings, but it doesn't. This has made finding sync much easier for me. I'm actually shocked at the SPLs the internal mic can handle at zero gain, it's pretty impressive actually -- and a quite shocking difference from what the AGC does on its own (it goes from completely useless and painful to your ears to entirely listenable, which is saying something).

Anyway, just wanted to give a HUGE shout out to Tramm and all his hard work. It really makes a difference for me. I think the only other thing that would be massively helpful for my personal work and workflow would be continuous recording (aka auto-restarting a new file once 4 gig is met), but can't have everything!

David W. Taylor · September 29th, 2009, 02:00 PM

Bill,
I use an R44 as well, and Tramm's ML has allowed me to feed back a pair of the R44 tracks back into the 5D2. To do this I have a pair of phonos plugged to the say the first 2 of the 4 separated R44 playback outputs, going to a stereo minijack into the 5D2 ext mic in.
With the ML letting me set the analogue gain to '0', this gives me the same audio onto the 5D2 has I have on a pair of the R44 tracks. Perfect for syncing and it's a backup of course.
I use NeoScene and Premiere Pro in editing and whilst there will be some slippage of sync between the recorder and camera, as neither are 'locked' in any timecode or video reference way , it's never a problem. I run the audio recorder at 48k 24bit, even though the camera audio is 44.1k of course.

The R44 has an 8gb card and make sure a new set of batts is in it, as if the R44 stopped I wouldn't have audio on the 5D2 either.
It works great for me but then I don't do long concerts.
DT

Jon Fairhurst · September 29th, 2009, 03:23 PM

Bill,

The problem with the ALC is that it is performed on the digital gain. Canon set the analog gain to +31dB, which explains why it overloads - and why the noise is so high when things get quiet.

David,

How do you compare the R44 audio to the R44-->5D2/ML recordings? In my tests with the juicedLink, the audio is really clean and crisp, and (only) slightly crunchy. My gut feel is that the anti-alias filter in the camera's audio circuit is weak.

Of course, you're probably taking the audio out of the headphone output, so that could add some noise. They're unlikely to spend as much on a headphone opamp as good manufacturers spend on their preamp signal path.

Chris Barcellos · September 29th, 2009, 03:44 PM

Quote:

Originally Posted by Jon Fairhurst

Bill,

and why the noise is so high when things get quiet.

Jon:

Yesterday, I got curious about a free program I had gotten off the internet for my IPhone... called Dog Whistle. I thought about times before ML, and the use of a track tone. So I loaded the camera, without using Magic Lantern and using Firmware 1.10, and ran a test with the tone on, and the tone off. I was surprised that there was very little change at all in the noise level. I took the camera into an a very quiet room, and didn't get any indication that the AGC was in search for a sound level, and the hiss we got with 1.07 was not showing up. So that raised a question. Did Canon put a limit on the auto gain with 1.10, or did we do something permanently with M.L. to the AGC.

Can you check and see if you have a similar result ?

Jon Fairhurst · September 29th, 2009, 04:22 PM

Quote:

Originally Posted by Chris Barcellos

Can you check and see if you have a similar result ?

Hi Chris,

My son has the camera up in Bellingham right now. I can't check nuttin'! :-(

Bill Binder · September 29th, 2009, 05:35 PM

Quote:

Originally Posted by Jon Fairhurst

The problem with the ALC is that it is performed on the digital gain. Canon set the analog gain to +31dB, which explains why it overloads - and why the noise is so high when things get quiet.

Ah, makes sense now. However, that said, as someone who deals with high SPLs regularly, I'm still quite AMAZED that the internal mic itself doesn't overload (brickwall) before it even hits the internal preamp. Makes me wonder if the internal mic is powered by more than your typical plug-in power (e.g., surely not phantom, but maybe 5v instead of 1.5v or something). Anyway, that's a pleasant surprise considering most cheesy little mics like that would brickwall for sure under the same SPLs.

Quote:

Originally Posted by Jon Fairhurst

Of course, you're probably taking the audio out of the headphone output, so that could add some noise. They're unlikely to spend as much on a headphone opamp as good manufacturers spend on their preamp signal path.

Nope, no need to use the headphone out (which is noisy) because there are four nice (and fairly) clean RCA line outs on the R44 from which to tap. Keeps the phone jack open for monitoring too.

Quote:

Originally Posted by David W. Taylor

Bill,
I use an R44 as well, and Tramm's ML has allowed me to feed back a pair of the R44 tracks back into the 5D2. To do this I have a pair of phonos plugged to the say the first 2 of the 4 separated R44 playback outputs, going to a stereo minijack into the 5D2 ext mic in.
With the ML letting me set the analogue gain to '0', this gives me the same audio onto the 5D2 has I have on a pair of the R44 tracks. Perfect for syncing and it's a backup of course.

Since for me, I'm primarily music-based and very much focused on stills, no need to patch into the 5D2 really. I don't have to deal with takes, LOL, just one long recording. For sit down interviews though, I patch in just like you do, less for sync and more as a backup. Also, I've hacked a DVD battery to power my unit (it's actually gaffed under the unit permanently), so it will literally record 4 channels of 24-bit with phantom on all four for something like 8 hours nonstop, heh. I basically never plug it in during recording, ever -- it's so nice not having to worry about finding an outlet. And since the R44 makes seamless splits, I can just let it roll practically forever (great for long shows and in festival scenarios).

Quote:

Originally Posted by David W. Taylor

I use NeoScene and Premiere Pro in editing and whilst there will be some slippage of sync between the recorder and camera, as neither are 'locked' in any timecode or video reference way , it's never a problem. I run the audio recorder at 48k 24bit, even though the camera audio is 44.1k of course.

Yeah, I've found that if I sync to the middle of the 5D2 clip (not the beginning), then with only 6 minutes on either side of the sync point (per clip), the audio never really drifts enough for it to matter in that short period of time. When I used to use MiniDV, then I'd sync the head, and stretch/shrink the tail to get sync over the whole clip (ah, the good ol days, only one manual sync per camera, although, my video quality was no where near the 5D2, so I guess it's worth it!).

Peer Landa · September 29th, 2009, 10:47 PM

Quote:

Originally Posted by Bill Binder

My typical setup would be taking a stereo soundboard feed plus a stereo pair of ambient (most often 2xAKG480+ck61/ck62/ck63) into a modded R44.

Just curious -- what modifications have you done to your R44..?

-- peer

Bill Binder · September 30th, 2009, 09:44 AM

Quote:

Originally Posted by Peer Landa

Just curious -- what modifications have you done to your R44..?

-- peer

See "Oade Brothers" on google. They've been modding recording devices for a long, long time. They are well respected in music recording circles, and they stand behind their mods. It depends on the specific model you'll dealing with as to what they can do (and what's worth doing for that matter as some devices are cleaner than others). Typically it's about lowering the noise floor, swapping out op amps, and/or taking things out of the chain that are adding noise, etc. To be honest, it's been a couple of years now, so I can't remember specifically what they did to my R44, but it runs very, very clean now (especially with that +6db on my AKGs, heh, noise floor is waaaay down there...)

David W. Taylor · September 30th, 2009, 12:28 PM

Jon,

You asked :
<<How do you compare the R44 audio to the R44-->5D2/ML recordings? In my tests with the juicedLink, the audio is really clean and crisp, and (only) slightly crunchy. My gut feel is that the anti-alias filter in the camera's audio circuit is weak.>>

I'm sorry to say I haven't used the 5D2 audio, as I only put it back in to give a good sync waveform. I only use that from the R44 as after all, with my workflow, the 5D2 audio gets recorded in camera at 44.1k but I am using 48k projects. I don't want too many sample rate conversions.

<<Of course, you're probably taking the audio out of the headphone output, so that could add some noise. They're unlikely to spend as much on a headphone opamp as good manufacturers spend on their preamp signal path.>>

The R44 has 4 phono analogue audio direct ops from the 4 tracks, which is what I use., usually the first pair. They are at the 'domestic' -10dB but they feed into the 5D2/ML ext input if you set the analogue gain to it's '0' lowest level in the ML config file.
DT

September 29th, 2009, 01:42 PM	#1
Bill Binder Major Player Join Date: Jun 2005 Location: San Francisco Posts: 542	Audio Capture and Magic Lantern So, just wanted to provide some feedback/observations to the forum. First, I should mention that I do a lot of music events primarily from a stills and audio perspective. It's a little niche I've carved out for myself. <SMILE> When it comes to audio for these events, I pretty much ALWAYS roll double-system. My typical setup would be taking a stereo soundboard feed plus a stereo pair of ambient (most often 2xAKG480+ck61/ck62/ck63) into a modded R44. So, it's not multitrack, it's very much a smallish field rig, but I can make great recordings with it, much better than soundboard alone or ambient alone. That all said, with the recent purchase of a 5D2 earlier in the year, and some past dabbling in video (I know just enough to be dangerous, LOL), I've started to incorporate video into some of my work -- albeit via monopod with manual focus pulling off the LCD (LMAO). So, now it's audio first, stills seconds, and video third (usually one or two full songs), all at the same event as a solo recordist. If things work out, they get edited into a multimedia package of a single song caught on video with the audio synced in and stills laid on top for coverage and b-roll. Anyway, prior to ML, I had some difficulties. For starters, manual control of ISO and aperture, but that's not an issue anymore. But more importantly, it was easy as hell to blow out spot lit musicians without completely realizing it. Now, with zebras, no problems. Second, at loud shows, the in-cam audio was so clipped that sometimes I couldn't even use it to do manual sync -- I had to go off video alone. Well, ML has fixed that too. In fact, with all the gain settings at zero, I'm actually quite AMAZED that the 5D2 produces audio that isn't really distorted at all in very high SPL environments. I would normally expect their mic to completely distort despite gain settings, but it doesn't. This has made finding sync much easier for me. I'm actually shocked at the SPLs the internal mic can handle at zero gain, it's pretty impressive actually -- and a quite shocking difference from what the AGC does on its own (it goes from completely useless and painful to your ears to entirely listenable, which is saying something). Anyway, just wanted to give a HUGE shout out to Tramm and all his hard work. It really makes a difference for me. I think the only other thing that would be massively helpful for my personal work and workflow would be continuous recording (aka auto-restarting a new file once 4 gig is met), but can't have everything!

September 29th, 2009, 03:23 PM	#3
Jon Fairhurst Inner Circle Join Date: May 2006 Location: Camas, WA, USA Posts: 5,513	Bill, The problem with the ALC is that it is performed on the digital gain. Canon set the analog gain to +31dB, which explains why it overloads - and why the noise is so high when things get quiet. David, How do you compare the R44 audio to the R44-->5D2/ML recordings? In my tests with the juicedLink, the audio is really clean and crisp, and (only) slightly crunchy. My gut feel is that the anti-alias filter in the camera's audio circuit is weak. Of course, you're probably taking the audio out of the headphone output, so that could add some noise. They're unlikely to spend as much on a headphone opamp as good manufacturers spend on their preamp signal path. __________________ Jon Fairhurst

September 29th, 2009, 02:00 PM	#2
David W. Taylor Regular Crew Join Date: May 2009 Location: Pinewood Film Studios, Bucks, United Kingdom Posts: 80	Bill, I use an R44 as well, and Tramm's ML has allowed me to feed back a pair of the R44 tracks back into the 5D2. To do this I have a pair of phonos plugged to the say the first 2 of the 4 separated R44 playback outputs, going to a stereo minijack into the 5D2 ext mic in. With the ML letting me set the analogue gain to '0', this gives me the same audio onto the 5D2 has I have on a pair of the R44 tracks. Perfect for syncing and it's a backup of course. I use NeoScene and Premiere Pro in editing and whilst there will be some slippage of sync between the recorder and camera, as neither are 'locked' in any timecode or video reference way , it's never a problem. I run the audio recorder at 48k 24bit, even though the camera audio is 44.1k of course. The R44 has an 8gb card and make sure a new set of batts is in it, as if the R44 stopped I wouldn't have audio on the 5D2 either. It works great for me but then I don't do long concerts. DT

September 30th, 2009, 12:28 PM	#9
David W. Taylor Regular Crew Join Date: May 2009 Location: Pinewood Film Studios, Bucks, United Kingdom Posts: 80	Jon, You asked : <<How do you compare the R44 audio to the R44-->5D2/ML recordings? In my tests with the juicedLink, the audio is really clean and crisp, and (only) slightly crunchy. My gut feel is that the anti-alias filter in the camera's audio circuit is weak.>> I'm sorry to say I haven't used the 5D2 audio, as I only put it back in to give a good sync waveform. I only use that from the R44 as after all, with my workflow, the 5D2 audio gets recorded in camera at 44.1k but I am using 48k projects. I don't want too many sample rate conversions. <<Of course, you're probably taking the audio out of the headphone output, so that could add some noise. They're unlikely to spend as much on a headphone opamp as good manufacturers spend on their preamp signal path.>> The R44 has 4 phono analogue audio direct ops from the 4 tracks, which is what I use., usually the first pair. They are at the 'domestic' -10dB but they feed into the 5D2/ML ext input if you set the analogue gain to it's '0' lowest level in the ML config file. DT