It's raining here today, so this seemed like an appropriate time to catch up my rainy day D.I.Y. projects on the blog. This time out I am attempting at improving the Canon 5D Mark II audio so that it can produce usable production audio. ...just fair warning - this blog entry may go off the edge of the nerd-cliff just a bit.
Why can't it record good audio? Well, one of the reasons is certainly because of the aggressive high-pass wind filter Canon applies to all mic input audio. For some reason they don't let you defeat it in the user menus. When I say aggressive, think of Homer Simpson finding beer flavored donuts. Yep, that bad.
The first part of any engineering effort is data gathering, so I set about to measure the input frequency response of the 5D. My end goal was to use the Marantz PMD661 field recorder as a mic pre-amp and primary production audio recorder, however I also wanted to take the audio line output from the field recorder and feed it into the 5D to capture "good enough" production audio. To start off I created a stepped audio file in Adobe Audition (20-200Hz: 20Hz steps, 200-1KHz: 100Hz steps, 1kHz-20kHz: 1kHz steps) and upload it to the Marantz field recorder SD card. Then it was simply a matter of playing back the audio into the camera with a simple resistor divider circuit to compensate for the level difference between a line out and a mic input.
The stepped audio waveform you see below is the audio extracted from the 5D video file. The top (left channel) waveform is simple resistor divider that also represents the default frequency response of the 5D audio mic input. The bottom (right channel) waveform is the effect of the resistor divider, plus custom designed pre-emphasis circuit. I also plotted the normalized results in Excel to make the data a little easier to see. The bottom line: The 5D has a high pass filter at about 120Hz. The compensation circuit has a high pass around 60Hz. The default Canon filter makes people sound like Gilbert Gottfried with a cold.
Like any design there were some trade-offs. The first, and foremost, is that I had to use whatever components I had sitting around the test bench, sort of like the cook in the kitchen with a bunch of semi-random ingredients. No, my filter components aren't necessarily optimal but they're good enough to get the job done with respectable engineering design margin.
Second, I had to accept some input loss for the pre-emphasis circuit because the purpose of the circuit is to boost the lower frequencies. If you boost the lower (60-200Hz) frequencies without lowering the input signal you'll get audio clipping, which is even worse than a nasty/nasal-y high pass filter. Here I chose to trade off 5-6dB in signal because it put my high pass filter point (60Hz) well below the frequency response of human speech. I could have traded off more signal for more lower frequencies, but it seemed counter-productive to the signal-to-noise ratio. ...just my engineering judgement. To compensate, just set the 5D manual audio level to 7 (starting from 0). This sets the audio level between the Marantz field recorder and 5D to exactly match (within design margin). A different 5D manual audio level setting may allow you to use the circuit with the Zoom H4N, but I don't have one to test nor know much about it.
The resulting pre-emphasis filter circuit is shown below. It's pretty simply to build really. It wasn't worth it, monetary-wise, to build a PCB, so I didn't. Point to point soldering, as in the first photo of this blog entry, was sufficiently robust. Believe me, the circuit has already been tossed around, beaten down, and generally production abused like a car in a Micheal Bay film. I recently used it during a 48 hour film making competition.
What was the result? Well, I can say that I've tried it with both a shotgun mic and a lav and found the results were good enough for most small production needs. I doubt Rodney Charters nor Gale Tattersall are going to use this circuit, but for local commercials, indie films, and sit down interviews this is perfect. Since I don't have any specific production examples yet, you'll just have to take my word for it that speech sounded almost identical to the field recorder and the S/N ratio was very surprisingly good- enough to make most directors happy!
If you build and use this circuit, leave a comment and let me know how it goes.
HOW WE DID IT: Doc Style
2 years ago