Back to blog
December 12, 2019

Compression on piano | Greg Wells




In this excerpt from Episode 2 of the "Start to Finish Series, with Greg Wells," we find Greg working on the piano part that he recorded for the song "Lucky Number," which he's producing for singer/songwriter Bryce Drew.

Put the Squeeze On

Greg starts by listening to his piano track and doing a few minor edits to clean it up. But the main task we see him doing is reducing the dynamic range of the piano. He says the part he played was quite dynamic, and the piano he was playing was on the punchy side. As a result, there are a lot of loud peaks.

Greg's settings for the Focusrite d3 compressor/limiter

As an aside, he makes a thought-provoking observation about the difference between hearing an instrument or vocal un-amplified in a room versus hearing it coming back through a speaker. Wells says what sounds great in person often needs work before it sounds good through a speaker.

"This is all so unnatural; speakers, microphones, it's all so weird," he says. He gives an example of a dynamic vocal that sounds great in the room (without microphones), but when you hear it through speakers, it seems "amateurish." What he means is that it suddenly sounds too loud or too quiet.

He inserts a Focusrite d3 compressor/limiter plug-in on the track. He says he really likes that plug-in, and, not surprisingly, the hardware version, too. He mentions that both Chris and Tom Lord Alge use the hardware unit, known as the "Red Compressor," on the mix bus. The Focusrite plug-ins were unavailable for a while, he says, but were rewritten for AAX and now are back. According to Wells, they're easy to set.

Another processor Greg uses on the piano track is the UAD Neve 33609 emulation.

He doesn't want to get rid of all the dynamics, but he wants to control them, so they sound good in speakers. He uses the limiter to shave off the peaks of the loud, punchy chords. It will still sound like he played aggressively, but the peaks won't be jumping out.

Next, he inserts the UAD Neve 33609 compressor plug-in. He says it's incredible on any source. He tells the story of many years ago when he was working with producer Joe Ciccarelli, where Joe showed him how good the 33609 sounds on piano. He's never forgotten that.

On the Attack

When you're setting a limiter to reduce peaks, as Greg did in the excerpt, you need to be careful how you set the attack time. If you're dealing with an instrument with a lot of hard transients, like a piano, and you're trying to keep it sounding natural, you don't want to make the attack too fast. Wells used a super slow attack of 50ms on the Focusrite. That allowed those transients to get through but still reduced the peaks.

The following examples (1a-1c) demonstrate what can happen if you set too fast an attack when processing a piano track or other transient-heavy recording.

Example 1a: Here is a short excerpt from a piano recording with no compression or limiting.

Example 1b: This time, a PSP FetPressor plug-in (an 1176 emulation) was inserted with a ratio of 16:1, which is high enough to be considered limiting. The attack time is at its slowest setting (10ms), which allows the transients to sound natural.

Example 1c: The same example, but this time the attack was set to 0.1ms, its fastest setting. The threshold was set higher to -6 to reduce the amount of compression (it was -12 on Example 1b) because with the faster attack, it sounded way over-compressed at the -12 threshold setting from example 1b. Even with the threshold raised, the piano loses its punch and seems too obviously compressed as a result of the fast attack time.

If you look at the waveforms for examples 1a-1c, you'll see that 1b reduced the peaks but kept plenty of dynamic range, whereas 1c got a lot more squashed.

A Bite Out of Gain

Another option for taming peaks is to use volume automation or, if your DAW supports it, "clip gain," aka "bite gain." The difference between them is that clip gain is pre-fader and volume automation post-fader. So, if you turn the clip gain up too high, you'll overload the channel input. If you set the volume automation too high, you won't see the increase reflected in the channel fader, but it will show up in the level on the master bus or other busses you're sending signal to.

Sometimes, there might be a couple of spots in a track where the peaks are really high, perhaps too high for your compressor or limiter to reduce sufficiently without sounding overly compressed. For those kind of situations, you could use clip gain or volume automation to reduce only those trouble spots and then compress it afterward.

Vocal tracks are great for automated fixes like that, because they have a lot of spaces between words and phrases, where you can draw in your automation without the level changes sounding artificial. That's often not the case on a piano track, where there are a lot of sustained notes and chords. If you change level as something is sustaining, it can sound unnatural.

The next example demonstrates using volume automation to smooth out loud peaks on a vocal track.

Ex. 2a: Here's a segment of the original vocal. The words "pain," "gain" and "bet" are all pretty loud.

Ex. 2b: The same track, but this time the loud words were reduced using volume automation.

Here you see the dynamic adjustments using volume automation in Example 2b

You'll probably find it a lot easier to draw in the automation rather than trying to use a fader to write it during playback. For brief boosts or cuts, such as those in the previous example, it's hard to be accurate enough with a fader.