Part 2: Cleaning Up The Audio

I was recently asked to record two pieces of piano music played by my friend Jason Hardink (pianist with the Utah Symphony). In this article—the second of 5—I’m going to describe the process of cleanup. I’ll talk about the tools I used and what my intentions were.

Nowadays just about any classical engineer will tell you that she (or he) spends a lot more time in cleanup than they did actually getting the recording. Each track (and there may be dozens in an orchestral session) may require a different treatment. Over the years I’ve had to remove all sorts of sounds from recordings—coughs, early applause, speaking, page turns, air-conditioning, traffic, chair creaks, phones—even motorcycles. These noises can be removed with surgical precision, and the overall quality of a recording can be improved in breathtaking ways. You’re not going to hear wistful longing for tape and vinyl in this newsfeed!

The first place my tracks go when they come into my editing/mix system is iZotope RX (currently RX8). This is the tool that makes people think I’m a magician—which I’m not. The first thing I do in RX is remove the worst background noises. That includes heat, A/C possible traffic rumble and so on. It’s that bed of schmutz that you hear all the way through a recording. The hall at Westminster is very nice acoustically for chamber music. But the HVAC system is really noisy—especially on a hot day like the day of the session. I’d recorded there before so I knew this in advance. I asked Jason to give me 5-6 extra seconds of silence after each take, purely so that I could have plenty of environment to sample over the course of the day. I’m not going to give a tutorial on RX here, but one of its many strong areas is in reducing environmental noise. Expose it to as much of the noise as you can, and it does a good job of getting most of it out. I’ve been able to get as much as 15dB of background noise out of recordings, pushing the noise floor to inaudibility. There are always some surprises, so let’s take a top-level peek at what I found in this session:

Here’s an RX view of the entire Eckardt session over the full frequency range of about 90K (my sample rate was 192K). You’ll see a bit of ‘haze’ overall. That’s the broadband part of the air conditioner, along with room tone. There’s a bit more of that noise down around 100Hz. You’ll also notice 3 areas of constant whistle at 16K, 46K and 65K. Sounds like this are in a lot of modern spaces. I think the noise at 16K is probably a fan bearing. I have no idea what the other two are. They’e quite faint to begin with (I doubt any human can hear the top two) and the RX spectral denoiser can do a good job getting rid of if.

RX will allow you to aggregate all channels (I had 8) for cleanup, but I prefer to do it a pair at a time. Each mic or pair of mics has a different perspective on the room and it’s worth a little extra time to do them separately. I kept each noise profile so that—if I needed to—I could do an extra pass on the occasional short section. Each mic position will bring its own particular type of noise. For example, my close-in M/S pair will get a lot of mechanical noise from the piano and only a little of the HVAC. My high-flying room pair gets almost none of the piano noise but a lot of overall environmental stuff. My A/B omnis will pick up more in the way of bench squeaks and so on.

After removing the overall environmental noise, it’s time to lessen some of the noises of performance. Those noises might include deep breaths by the performer, damper pedal noises, bench squeaks and perhaps page turns by the composer in attendance. This is a place where you must be careful—doubly-so when the recording is for video. You can, with sufficient time, get it all out. Every shoe squeak, every sigh, every thump. But doing so can detach it from the fact that a human is playing—you don’t want to turn a Steinway D into a MIDI piano! So the trick here is to diminish the most egregious stuff but not to be afraid of every single noise. It is essential to have a subwoofer that takes you down to 20Hz or so. Sometimes there’s a lot of thumping down there and it’s always distracting. If you can’t hear it in your studio you’ll never know it’s there (although a listener may hear it). RX is very helpful in that you can see the entire recording, spread out over the frequency spectrum. You can zoom in or out as needed. A thump might be a big blob on the bottom of the display. A snap might appear as a narrow line from top to bottom. A light hand is required. It might be sufficient to lower just the high frequency component of a noise—the part that draws your attention. You may still hear some residual stuff while you’re editing, but in the context of the entire recording you might not even notice that sound a week later. Let’s take a look at a couple of these undesirable noises in this same session. These will be zoomed in as necessary.

Don’t worry if you can’t quite see those things in the overall image. You do get the hang of it, given time. Any session has its own particular noises, and they usually leave a pretty obvious graphic signature.

When I’m through with RX (I’m not really through—I just think I am), I export everything as a set of stereo wav files and ship them off to ProTools. Everything I export remains at 192K sampling rate and 32-bit floating-point resolution. I’ll now be waiting for a rough-cut video based on the edit list from the performer.

In the next segment I’ll talk about editing the performance from all of the takes. Look for it in a few days. In the meantime, Jason Hardink has made the first video public. You can see it here. The video of Steven Roens’ piece should be released very shortly.