Compressing Snowballs and Other Dark Arts

So, I spent a little time this weekend trying again to get a decent vocal recording with the Snowball. There are a couple of problems remaining that I have little confidence that I can resolve.

Even with a different computer, a different USB cable (yet again), and a completely new installation of MacOS X 10.4 (Tiger), I am still getting occasional, random dropouts in the recording. This results in chopped-up waveforms, which means nasty glitches on playback when the chopped ends don't meet smoothly. The glitches do not seem to be strictly periodic, and they do not seem to be always 512 samples in length. This happened before under some conditions using my PowerBook G4. In some very brief testing recording to my PC, it did not happen. I can't record to the PC, though, because it is much too loud. So even if I could get a good voice level, and do a perfect reading, it's got random noise in it. I'm tolerating that for now, while I have queries into BLUE's support forum.

The levels of the resulting recording are, in general, too low, even after setting the only parameter I have control over, the microphone level, to its maximum. I'm sitting very close to the microphone, and I'm speaking in my best radio voice. (I've been a college radio deejay, and never had any trouble getting a decent vocal level when on the air). The gain structure of this thing may work better for recording a singer, or a live band when the cut is engaged, but it is just not right for spoken word. In other words, definitely not ideal for podcasting.

I just read a review of the Samson USB microphone in Sound on Sound magazine -- apparently, Samson provides a little control panel that lets you alter the gain structure directly. The reviewer found this very helpful for working around the limited dynamic range available with a 16-bit A/D.

The Snowball doesn't provide this control. It makes me think I should have purchased the Samson mic -- although the reviewer also mentioned problems with bleed-through of digital noise, when connected to a G5, and was less than enthusiastic about the 16-bit range, or the level of self-noise. So I probably should have have never gone down the path of attempting to use a USB microphone at all. That money is spent, though, so for now I am still trying my best to make do with what I have.

Since I don't have control over the microphone's built-in preamplification at the analog stage, I have to normalize the levels in the audio-editing program afterwards. This results in an an unacceptably high level of background noise.

Where is that noise coming from? Well, it is a very sensitive mic, and I'm using it in a real room, not a vocal booth. (I don't seem to have a spare vocal booth handy.) Recently, I've been recording in a bedroom of my apartment very late at night. I can tell you what it _isn't_, because the mic will detect all these things: it isn't air-moving equipment. It isn't fluorescent lighting. It isn't fan noise from the computer. It isn't the apartment's air conditioner. It isn't traffic noise. It isn't birds outside. The only sound I can hear in the room is faint noise from the hard drive of the computer I'm using for recording, and an air conditioning unit somewhere else in the building, or possibly from a building across the parking lot. It's quite a quiet room, and I ought to be able to get a decent, if not quite broadcast-quality, voice recording, but I can't, and I'm becoming increasingly convinced that it is the weird, compromised gain structure of this mic, and the glitchy USB data transfers, rather than something I'm doing wrong.

I've done some experimentation with using a compressor and noise gate. The plug-ins I'm using are a set called the "fish fillets" -- the Blockfish compressor, Floorfish expander/gate, and Spitfish de-esser. The results aren't promising. I like the plugins -- they seem very flexible, but to get anything resembling a loud output I've got to crank the output level of the compressor, which results in the plug-in applying a lot of gain, which brings up the background noise to really nasty levels.

The gate/expander can help with that -- it can effectively cut the background noise (when I'm not talking) to zero. but then you wind up with the background noise swelling up and down along with my voice. But it's clear that the presets are expecting a much hotter signal, because it isn't kicking in at a low enough level. This means it is muffling the beginnings of words. I have to crank the sensitivity way up to get it into the right range, and tweak the frequency range it is working with as well, and it still doesn't sound very good. I don't think the problem is the plugin -- that seems to sound quite good, but I'm just not giving it a decent signal to work with.

I've done a little playing around with EQ or notch filtering on the fly to reduce the noise, but the noise does not seem to be in a narrow frequency range, so this doesn't help much. There's another approach I could try -- there are noise-removal plugins, such as the one that comes with Audacity. These work, but at least on a mono voice, seem to leave a weirdly phased signal behind, which is hardly a better solution.

Maybe I'll put up some sound files and screen grabs so I can at least document the problem. I also want to try recording again on the PC. If I can get the audio glitching problem to appear, even once, I have a pretty strong case that this thing is completely unusable.