Ever since Canon released the 2.0 firmware update for the 5D Mark II, I’ve been suspicious of the camera’s “manual” audio function. Using Magic Lantern http://magiclantern.wikia.com with the original firmware, I had been able to capture broadcast-quality audio (I’m speaking literally here: I used in-camera audio for, among other things, two regional TV spots that I shot and edited). However, on several shoots I’ve done with the new firmware, I’ve been noticing a hiss in the audio recorded in camera. On the shoots in question, I had the benefit of an audio operator running a top-of-the-line Sound Devices mixer/recorder. He recorded the sound on his hard drive, while feeding me the audio as a reference and/or work track. While I had planned to use the Sound Devices files for the final edit anyway, I was concerned that the poor audio quality I was hearing would be an issue for one-man-band shoots when I was flying solo.
Jon Fairhurst did some excellent tests comparing the CX231 to the Zoom (among other devices) while running Magic Lantern, but I haven’t seen any formal tests of the new firmware. So, I did a simple one.
I set up an Audio Technics AT897 shotgun mic, and recorded some test dialogue in a couple of ways:
1) Through a JuicedLink CX231, into the 5D2.
2) Into a Zoom H4N recorder.
3) Looped through the Zoom H4N (through the headphone jack on the Zoom) and into the 5D2.
You can download the audio from here: http://5dfilmmaking.com/audio_test.aif
Based on this test, I drew a few conclusions:
1) The hiss I heard in my field work doesn’t seem to have been caused by the 5D2. I suspect the wireless feed I was receiving from the audio operator, but will need to run separate tests using the Sound Devices to verify that.
2) The Zoom and the 5D2 both generate a lot of faint noise (in this image, red is signal and blue is silence. The purple haze around the vocal signal is noise.
3) The Zoom clamps the signal at around 20khz (theoretically not an issue, since that’s the the limit of “normal” human hearing, while the 5D2 does not.
4) The CX231 generates some sort of pilot tone close to 20khz. I can’t hear it, and I can’t explain it. I have a DN101 headphone amp/AGC disabler installed, but the tone was constant whether I had the DN101 on or off, and whether I took the output from the DN101 or directly from the CX231. I wrote to JuicedLink, and was pleasantly surprised to receive a very prompt and thorough reply: “The noise blip you see in the spectrograph around 20KHz is -50dB down, and is from the preamp switching power supply. There is no need to remove it in post. It’s not like the AGC disable tone which is injected at a high level to force the AGC lower and contains harmonic components. With an A-weighting filter (the human ear), it is attenuated by another 50dB. It is of no concern.”
5) The Zoom generates slightly less noise than the 5D2 (for this test, I had the CX231 set to maximum gain, and the internal gain in the 5D2 set to the 1/4 mark to get a nice hot signal), but looping the unbalanced headphone output of the Zoom into the 5D2 is the worst of both worlds. Clearly, this would only be a backup track (in case somebody forgot to press the “Record” button on the Zoom, which of course somebody would never do), or a nice solid track for “pluraleyes” to sync with.
Overall, I was disappointed by how noisy the signals looked, but I was relieved to find that they all sounded pretty good. I’ll test the Sound Devices mixer to see if I can identify the source of the hiss, and update accordingly.