One thing to look for in this video is the motion of the bushes in the background. The mic was only about 10 feet from them. Sometimes the bushes are moving about, yet there is little or no wind noise. With other setups, there's no motion at all, but the wind noise is high.

My mic has a low cut switch which I left off purposefully for the test. (I turned the filter on for the narrator bits and EQ'd to make up for the thinner sound.)

One thing I'm not sure of is if the Zoom's low-cut filter is digital or analog. If it's digital, it won't stop the wind from overloading the A/D converters. You really need the low-cut to be in the analog domain, such as on the mic.

Jon:

Thanks for another very informative test !

Chris,

You're welcome!

Tramm,

The visual representation is really telling. The Zoom sounds much quieter than the BeachTek into the ML firmware, but both show a purple background. Looking closely at the BeachTek, the signal is weaker at high frequencies, so the noise might be similar, but the signal to noise is lower.

The JL/ML plot looks amazing. I'd love to test it against a Sound Devices mixer/recorder.

John,

I've been testing the Sound Devices MixPre and Magic Lantern, along with the Juicedlink and the Beachtek. I think the biggest issue for S/N here is the amount of attenuation you put onto the XLR out of the mixer and what level you set in Magic Lantern. I'm currently using a -45db PAD and the level from the Sound Devices needs boosting a little in the camera, I'm assuming a lower value PAD will yield a better result but I have no way to test it.

Also has anyone noticed that the level meters in Magic Lantern pump when being fed a constant tone (1khz) from either a mixer or a tone generator into the Juicedlink. Not sure what's going on there??

Dan

It is likely due to bad sampling in my code leading to aliasing. My audio level task samples every few ms and does a very simple moving weighted average. I noticed that Canon appears to vary the time between samples to try to avoid this problem.

In regards to the first few seconds of audio being messed up, that is due to Canon's mvr_rec_start() function re-writing the audio registers to their parameters. The task in 0.1.4 re-writes them at 1 Hz, but I have figured out how to register a property handler to get notified when recording starts so that I can re-write them only when necessary.

I'm also working on disabling the power save functions so that the camera won't turn off when Magic Lantern is running. And a config file with gains and zebra levels/enable. Also hdmi output support, slow contrast edge detection and headphone volume control. Coming soon in 0.1.5!

Thanks for the update Tramm. The config files in particular will greatly improve work flow for shooters.

Its like Christmas, Tramm you are a star!

Dan

Tramm,

I just bought one of these FiiO E5 Headphone Amplifier in Black HiFi Headphones little headphone amps which works a treat with your hack giving good monitoring levels and a hard button volume control. Only problem is that I need to modify the mini jack at the camera end as it is the three strip type not a stereo one, does anyone know where I can get the right jack plug without butchering a Canon AV cable?

Dan

HOSA GRF-341 TWO FEMALE RCA TO 3.5MM ADAPTER

B&H has it but I got it from Dale Pro Audio because it was a little cheaper and free shipping.

If anyone has a wiring diagram for hacking the AV cable, though, please post. Even though I got the adapter, I'd rather opt for a shorter AV cable wired directly to a headphone jack.

Min Lee, thanks for the link. That adapter is precisely what I'm trying to avoid :) anyone else? I assume you can get the Canon cable as a spare part but I bet its not cheap.

Dan

I hacked up my A/V cable since I never use it. I've also successfully used an old iPhone earbud, but the enamel coated wires are hard to solder. The problem with the GRF-341 is that it doesn't split the video off correctly since it is only a stereo jack.

The part that you're looking for is a TRRS connector. 4 Conductor 3.5mm Plug - TRRS 3.5, 2.5mm Mini Connectors at Markertek.com is one source for just the connector, and this looks promising for an assembled cable (but I haven't tested it): 3.5mm. (TRRS) Right Angle Male Plug to (3) RCA Male (video & left/right audio) Camcorder Cable, 6ft

One problem, as I'm sure you have run into, is that the Canon firmware will try to use the composite video output when it boots, which requires the headset to be unplugged and plugged back in. Perhaps using the HDMI output will fix that in the next release of my firmware.

Wow, I am going to check my set up out again. I didn't think I was having any issues. I just plugged a cord that came with my Boostaroo headphone amplifier into the AV out port, and I think I was getting both channels out. I will check it out tonight. It is just a standard miniplug male as far as I can recall.

Thanks Tramm, I'll check those out. I was going to ask about having to unplug the headphone out every time, glad to see your on top of that too!

Dan

Okay, I checked my rig out, ME66 to Juiced Link, to camera in, using JuicedLink cable, then Monitor out to my Boostaroo. I have the Juiced Link set for center on the channel mic is fed to. I ran Magic Lantern.

I get sound in both ears, though left is louder than right. I do note there is a difference in level reported on the level meters too, so this tells me that different levels are coming out of the Juiced Link for each channel. Or is it possible Tramm set up different levels on each channel in the camera ?

All plug ins in the set up out of the AV jack into the Boostaroo are the standard stereo miniplugs.

Hi Chris,

I haven't done any channel comparisons. I've been running the right channel on everything, simply because in stereo, the BeachTek puts the pilot tone into the left channel.

The key to measurement is to eliminate variables. Either send an identical signal to both channels by connecting the tip and ring on the mic plug into the camera, or record one channel at a time into a single recorder channel and compare.

I would recommend setting the gain of the unused signal to 0, low, and switching it to both channels as well. If the unused channel is switched to one side or the other, that might affect things.

Thanks Jon. I tried that, too, but still seems to be some difference. The difference isn't that great usually, but you would think that center panning would give a unified signal that I am not seeing at all times on the camera meter. This may be an incamera set up issue.

Again the difference is workable and not extreme, but just enough to ask questions about. Maybe Tramm can tell us if he had to set channels independently. It kind of reminds me of the way we set up sound with one channel higher gain, and one lower so if the higher went to high and distorted, the other channel was low enough to be usable still.

The channels are set independently, but to the same gain value. There may be minor variations in the hardware amps as well.

Does your headphone amp output noise when the Canon firmware is running? Everytime I've tried using a stereo headphone connection into the camera with the video output it produces lots of squealing noise on one of the ears. With ML running there is no video signal, but there is still some problems on the same ear due to it shorting a pin to the (inactive) video circuit.

Also, I misunderstood the way that the GRF-341 is being used. If you plug the RCA jacks into the Canon A/V cable, it will work just fine. For some reason I was thinking it would be connected to the camera on the 3.5mm cable side.

I suspect that we do have a variation in the in camera hardware. It would be interesting to have Robert Rozak's (JuicedLink designer) input on this issue.


I will check out noise with headphones plugged in regular Canon firmware this evening. As you may remember, early on I indicated that when we plugged in an outside mike or or an XLR adapter, and then tried to monitor using and LCD, I got screech too, though it wasn't really loud, but it recorded to the camera, causing havoc with the audio Capture.

Interestingly, I never even considered using the Canon AV cable out of the camera. I suppose this came with my experiences with the Canon HV20, where with the AV switched to headphone's in the menu system, I could plug a straight mini plug headphone into the camera to monitor with headphones. I did not have to adapt the existing cable, nor do I think it is necessary here, as I monitor headphones through the standard cable provided by Boostaroo.

Small ticks
 

When I use external microphones in the 5D with ML I get small ticks when there is sound. If I keep silent I don't get them, but when I talk I keep hearing them. They are subtle but enough to keep irritating me. If I feed the mics through the H4n and record at the same time I don't get ticks with the H4n but in the camera. I have tried with mics straight into camera, with H4n and another preamp.

I searched but didn't find anything in the forums about it. Anyone got the same problem?

Rode NTG-2 mics. I have attached a sample file. Yes, the tone I'm making is terrible, but you can hear the ticks.

I can clearly hear them in your sample. They sound like they are occuring at 1 Hz, which is almost certainly due to the firmware re-writing the AK4646 control registers once per second.

If that is the case, they will go away in a soon to be released verison of the firmware that only re-writes the registers at boot, when recording starts and when it ends. The quick hack of writing them every second was since I didn't have enough info to know when recording started.

Now that I have property event slaves working, my code will be notified when these events occur so that it will only need to re-write the registers when appropriate. This should also eliminate the first second of bad audio when recording begins.

Then I'm really looking forward to the next release. Thanks for all your work, you will receive a donation from me soon.

Part 3 loaded!!!
 

3. Canon 5D Mark II Audio Exposed - Wireless Lav on Vimeo

In Part 3 of Canon 5D Mark II Audio Exposed, I compare the Microtrack II, Zoom H4n, BeachTek DXA-5D, and the juicedLink CX 231 when recording from a lavalier microphone over a Sennheiser EW100 G2 wireless link.

The only processing of the comparison audio was changing gain to match levels. You can download the uncompressed comparison audio here: http://p3pictures.com/audio_5d2/Audi...irelessLav.zip

Don't miss...
Part 1: Boom Mic - 1. Canon 5D Mark II Audio Exposed - Boom Mic (juicedLink, Zoom H4n, Microtrack II, BeachTek) on Vimeo
Part 2: On Camera - 2. Canon 5D Mark II Audio Exposed - On Camera (juicedLink, Zoom H4n, Microtrack II, BeachTek) on Vimeo
...
Part 4: Foley (pending)
Part 5: Noise Tests and Final Conclusions (pending)

A couple of notes...
1) I screwed up on the BeachTek into the Magic Lantern firmware, which is a shame. It would have done well with the strong signal from the G2 system, and
2) You can hear the clicks into the Magic Lantern firmware when I speak. As mentioned above, that will be fixed in the next firmware release.

Enjoy!

Hi Jon: FYI: I shot some hard wired AT399 lavs into my 5D using the JuicedLink, using Magic Lantern. I used the default sound setting in Magic Lantern 0.1.4 I probably ran Juiced Link too low. It was set at high on both tracks , but I trimmed back to about the 11 oclock position on 1, and the one oclock position on the other. Pan for both was set at center. I actually used the mics as plant mics near the talent. I seemed to have plenty of headphones volume but I noted a bit of hiss. the levels came out low not exhibiting his, but I am going to have to boost, and will have to filter. ( By the way, do you have recommendations on how to filter the hiss in post.) From this shoot I learned it is important to not rely on boosted headphone volumes as an indication of the levels. In fact it would be wise to set levels based on meters, and then adjust headphones with volume controler attached to the headphone amp system. I should have bumped the level on the trims dials all the way up.

Doesn't magic lantern have audio peaking meters? I was counting on that.

Mark: Yes it does. However, I was shooting in a bright daylight situation, static on tripod, and made the mistake of relying on the phone levels I was hearing. I have a Boostaroo that raises volume level for monitoring. However, the level coming out of the Boostaroo is very hot, and I made the mistake of paying attention to levels from phones, and ignoring the level meters that Tramm had made just for this situation. Mistake, and through that mistake I learned you can actually rely on the meter levels, because if I had paid attention, I would have boosted the JuicedLink levels more.

Chris, I run everything at Analog +10dB ("5") and Digital at 0dB. I run the JL at "H" fully cranked, unless the signal is too hot, as it was from the Sennheiser G2 rack mount receiver. I trimmed to 12 o'clock for that test, but everything else has been full throttle.

That said, the JL/ML can sound a bit harsh if the levels are too high. The noise is quite low, so I prefer recording on the cool side, leaving lots of headroom and providing a clean sound.

For removing noise, I use Sound Forge 9. Look into Gold Wave for a cheaper alternative.

I'd recommend recording the same setup with the mic(s) in the quietest room in your house surrounded by (but not touching) blankets. Record at least 30 seconds. Use that as your noiseprint. That way, you'll be removing the hiss, but not the room.

And Mark, yes, ML includes meters. However, it does take some faith to run the juicedLink full bore. We're so used to noise when the volume is at 10 that I understand why Chris backed off the levels. The juicedLink breaks the rules - crank it all the way when you need the signal!

Tramm, I reported these same ticks a some time back demonstrating the tone method of shutting down the AGC. It occurred when I raised the volume level on the tone to the higher levels. I thought at that time that the camera was cycling something to do with AGC and that was causing the ticks.

They appear as black vertical lines in the last section of the spectragraph. Otherwise it looks lots like the original, with the exception that the noise in the Beachtek test appears to be ramping back up during the silent section.

We don't have another "black dots" issue do we? If you heard the ticks with no magic lantern then we may have a problem. Of course Tramm can fix anything, right?

Mark, I am not so sure it is a problem... The ticks were showing up when I was pumping tone in at high levels from my mp3 player. I remember I was trying to find the level that disabled AGC completely. This turned out to be around 25 on my volume meter on my MP3 players, and if I pushed up over that, I started hearing the ticks. I remember thinking at the time that it may be activating some sort of limiter in the camera. My tone was going in at 15,400 hrz.

I have managed to generate the click in my set up, using a 15,400 inaudible tone, lined into camera using my juiced link. Clicking occurs as volume is applied to the track. It occurs the same in both ML and stock. I have forwarded track to Tramm to help him sort it out.

I don't think it is directly my problem anymore. The stock firmware results in the first graph fairly conclusively show a problem in either the audio device or something in the chain leading to it. The stock firmware also looks like it was put through a limiter -- there is clipping far below full scale and your voice sounds very clipped. The left channel of the sample with Magic Lantern sounds clean without any clicks, other than one word where you clipped the full scale.

Both of these two spectragraphs show the interruption on the right channel very clearly and the first also shows that the pilot tone is not doing a very good job of defeating the AGC with the stock firmware -- note how the bottom channel "swells" during the silent sections.

One oddity, however, is that the ML firmware is showing the clicking problem at a much, much lower volume than the stock one. This may be an artifact of the different gain and we might be able to fix it with different MGAIN and DGAIN parameters.

My hearing isn't that great after too many hours in airplanes, but I could hear the pilot tone pretty clearly in my right ear. The analyzer shows it at 15.3-15.4 kHz, which I thought was above my hearing.

Tramm

With the stock firmware track, it I had no way, of course, to monitor the input, and I had the hot ME 66 right at my. The clipping cleaned up as I set the mic under my monitor. Thats really what Magic Lantern is doing for us.


Interestingly, when I attempted to use the pilot tone initially with my ENG44 I could not feed the tone through the ENG44.. it caused a very prounounced cycling of tone levels. When I brought that up back then I suggested it might have something to do with the Eng44 preamps, or an onboard limiter..

I had not noticed that before, but you are right, it was there on the lowest volume level, 25 level on the volume input. I whatched a levels meter on it, thought it is inaudible to meit runs a lot higher than any other sound being recorded.

I can't hear it. Older ears.

This is all an interesting exercise, but from my viewpoint, the only issue arises with abnormally high sound levels. I have not heard any of the tick in my recordings, except when I am putting a tone in the camera. So when you get your version out that allows us to lock in settings, we should be able to easily test various set ups to arrive at combinations that eliminate the issue completely.

Thanks again for your hard work..

Part 4 - foley - loaded!!!
 

In Part 4 of Canon 5D Mark II Audio Exposed, I compare the Microtrack II, Zoom H4n, BeachTek DXA-5D, and the juicedLink CX 231 when recording foley sounds with a Rode NT1-A large condenser microphone.

4. Canon 5D Mark II Audio Exposed - Foley on Vimeo

The only processing of the comparison audio was changing gain to match levels. You can download the uncompressed comparison audio here: http://p3pictures.com/audio_5d2/Audi...art4_Foley.zip

Don't miss...
Part 1: Boom Mic - 1. Canon 5D Mark II Audio Exposed - Boom Mic (juicedLink, Zoom H4n, Microtrack II, BeachTek) on Vimeo
Part 2: On Camera - 2. Canon 5D Mark II Audio Exposed - On Camera (juicedLink, Zoom H4n, Microtrack II, BeachTek) on Vimeo
Part 3: Wireless Lav - 3. Canon 5D Mark II Audio Exposed - Wireless Lav on Vimeo
...
Part 5: Noise Tests and Final Conclusions (pending)

Enjoy!

Thanks Jon, another interesting test.


I am going out to a party in a while, where some mariachi's will be playing. I am going to shoot the performance, using ML in its current state, and see what we come up with. Problem is I am probably going to record at default settings cause I know I will not be able to change again after camera shut downs, or time outs, etc.,

Set the timeout longer. You definitely want the +10dB analog gain.

Also, record with the levels on the low side to avoid clicks.

"Ai, yai, yai, yai..."

Have fun!

Here's the spectragram. Not much to note on it, other than the gain on the Microtrack might be a little high after equalizing them and that the DXA-5D without the pilot tone is producing a sample that is not symmetric about 0. You can see a slight DC offset in the second data set if you view the image in full.

Yeah, the Microtrack seems too high. I'll double check my calibration and re-do it, if needed. I normalized on the pin drops. Between the higher noise and elevated bass, I expect the latter parts to be higher, but not that much higher.

I'm not sure where the DC offset came from on the BT/ML. I ran the pilot tone only into the left channel and the signal only into the right. I then selected the right channel only. The only adjustment I made aside from selecting the active channel was gain.

I've now posted Part 5: Noise. During that work I triple checked my calibrations. Tonight I'll go back and make any needed adjustments to Part 4...

http://www.vimeo.com/5474562

In Part 5 of Canon 5D Mark II Audio Exposed, I compare the noise of the Microtrack II, Zoom H4n, BeachTek DXA-5D, and the juicedLink CX 231 using a now noise, Rode NT1-A large condenser microphone.

I first measured the sensitivity of the equipment with a 500Hz tone sequence. I then measured the noise with the microphone in a bathroom in a basement in the center of the floor plan. The mic was suspended in a space surrounded by thick blankets. When you hear my voice, I am outside of the bathroom, speaking through the door.

In post, I normalized all of the mics by adjusting the gains, based on the sensitivity measurements. On the second pass, I added another 18 dB of gain to all the samples to make the noise easy to hear on any playback system.

The uncompressed comparison audio is available here: http://p3pictures.com/audio_5d2/Audi...art5_Noise.zip

Don't miss...
Part 1: Camera Mounted Mic - 1. Canon 5D Mark II Audio Exposed - Boom Mic (juicedLink, Zoom H4n, Microtrack II, BeachTek) on Vimeo
Part 2: Camera Mounted Mic - 2. Canon 5D Mark II Audio Exposed - On Camera (juicedLink, Zoom H4n, Microtrack II, BeachTek) on Vimeo
Part 3: Wireless Lavalier - 3. Canon 5D Mark II Audio Exposed - Wireless Lav on Vimeo
Part 4: Foley - 4. Canon 5D Mark II Audio Exposed - Foley on Vimeo
...
Part 6: Conclusions (pending)

wireless direc to camera
 

Jon, your perseverance and the depth of your testing is impressive - well done, thanks for sharing so much effort! I'm looking for a down and hopefully not too dirty method to use right away with the gear I have. I'd like to run a lectrosonics SR two channel receiver into the camera, and need to have a cable made. Is there power on the mic jack for the canon?
I'm assuming I can adjust the gain off the receiver to a high mic/low line kind of setting, and avoid buying an additional preamp.
Does this sound reasonable based on your experience?