Take apart the sound you love

First, thank you. To everyone who’s been along for this since the early days, who replied to these emails, sent a bug, or told a friend. I read all of it, and some of what I’m building right now started in one of your replies.

Right now I’m in the lab most of the day. The big update for the other four plugins is coming together, and the way I work on it is a bit odd, so I thought I’d show you.

The frame

When richard feynman died in 1988, his blackboard at caltech still had this on it: “what I cannot create, I do not understand.” just below it: “know how to solve every problem that has been solved.”

That’s most of how I learn sound. I take things I love apart, rebuild them, and find out where my rebuild is wrong.

It’s an old trick. Benjamin Franklin taught himself to write this way. He took essays he admired, jotted short hints of each sentence, put them away for a few days, and rewrote the essays from the hints alone. “then I compared my spectator with the original, discovered some of my faults, and corrected them.”

The research mostly backs him up. In a 1985 study, novices who studied worked examples went on to solve similar problems faster than the ones who only practised solving. A finished mix you love is a worked example. And the classic deliberate practice paper (ericsson and colleagues, 1993) describes practice that is structured, aimed at getting better, with feedback on the result. Honest footnote: a 2014 meta-analysis found deliberate practice explained 21% of the differences in music performance. It matters. It isn’t everything.

How I do it with plugins

There are plugins I really love, FabFilter’s among them, and I study how they behave. I don’t open anyone’s code. I send them test signals (short tone bursts, a steady tone buried in noise, a hit on one side only) and measure what comes out. How fast does it react? How narrow is the cut? What does it leave alone? Then I run the same signals through mine and compare, and the answer usually lives in the small stuff: a few milliseconds of attack, a cut that spills a little into its neighbours.

Then I put them side by side in blind listening, levels matched, and vote without knowing which is which.

The devil is in the details, and you still have to let it flow. The measuring tells me where to look. The blind vote decides.

What to take into your next session

Rebuild a reference, franklin style. Pick a mix you love. Write one line per section (how loud the kick sits, how bright the vocal is, how wide the chorus opens). Close it. A day later, rebuild those moves in your own session from the notes only. Then compare and write down where you were wrong. That last list is the lesson.

Match levels before you compare. In noise, a loudness difference under half a dB can already be heard (Zotter and Frank, ambisonics, §4.9.8). So an unmatched a/b is partly a level test. Trim the louder one until they feel equal, then judge.

Listen to the side alone. Side is (left minus right) divided by two. Flip the polarity of one channel of the reference and sum to mono, and all that’s left is what they put wide. You’ll often find there’s less out there than you thought.

Null test your own moves. Duplicate a track, flip the polarity on the copy, and put your plugin on one of them. What you hear is exactly what the plugin did, nothing else. (your daw has to line up the plugin’s latency, most do it automatically.)

Do the delay maths. 60,000 divided by the bpm gives you a quarter note in milliseconds. At 120 bpm that’s 500 ms, an eighth is 250. And an echo arriving within about 35 ms of the first sound doesn’t move the sound, you still hear it where it started (Loy, musimathics vol. 1, §6.13.3). At 100 ms it becomes its own slap.

The polymath bit

One more study, because it’s my favourite. Root-bernstein and colleagues (2008) found nobel laureates were almost three times as likely as other scientists to have an arts or crafts hobby. It’s a correlation, so I won’t claim the painting caused the prize. But I’ve always been a polymath by heart, and I’ve gotten used to the idea that nothing you take apart is wasted. A film, a recipe, a song. It all transfers.

From the studio

It’s been quiet on YouTube, sorry. I really want to get back to the films, and I will, even though the channel is small. I just want these updates out the door first. There’s a Danish saying, øvelse gør mester (practice makes the master), and right now that’s my whole week.

I’m also building SYNC for real now. Two producers on the same Ableton project, each on their own machine, both keep working, and nothing gets lost. More on that soon.

What’s a sound you’ve tried to take apart? Reply and tell me. I collect these.

-Jonas

kernaudio.io