Frequency based ducking.

I was driving back from the orchestra rehearsal on Thursday night and started thinking about a comment Ken Walker made about my latest musical composition. He didn’t like the fact that I had used “Voice Ahs” as the lead sound. The thing is though, I wasn’t using that sound. I was actually using something quite a bit more subtle/interesting. Unfortunately, all the subtleties were lost in the overall sound of the mix.

One of the things that producers do to fix this kind of problem is to use a parametric EQ to carve out a space in the frequency spectrum for particular sounds, thereby making them more prominent. The problem with this is that it is a manual process.

Suddenly, it hit me: Build an “inverted vocoder”.

Vocoders work by splitting the frequency spectrum of a signal to be processed into many (say 128 or 256) small bands and then setting the level of each band based on the amount of energy present in that part of the frequency spectrum of a separate modulation signal. When you feed something rich in harmonics, like a ramp wave in as the signal to be processed and use the sound of your voice as the modulation signal, you get that traditional “singing robots” vocoder sound.

But what would happen if you inverted the signal, so that by default all of the frequency bands were “full on” (i.e. not cutting out the sound) and as the energy went up in the modulation signal, it was automatically lowered in signal to be processed? (You would want to be able to control the overall amount of reduction that occurs. Using too much would probably cause the result to be “unnatural”.)

Now imagine, feeding the backing track for your mix in as the signal to be processed and the “lead” line (or whatever it is that you want to make more prominent) as the modulation signal. The result would be to automatically reduce in the backing track the frequencies that contained the most energy in the lead line. Effectively, you would be automatically EQ’ing the track to make the lead line more prominent. (Of course, you still have to mix the original lead signal and the processed backing track to get the final result.)

This is similar to another process used by producers called “ducking”, which lowers the overall level of one signal when another signal is present, but with this mechanism you only remove certain frequencies.

What’s really weird about all this is that, I have never heard of anyone doing it before. If anyone else knows of software (or hardware) that implements this algorithm, please let me know, I’d like to find out if it works.

As unlikely as it is, if it does happen to be a novel idea, remember that I thought of it first. 😉 I’m probably not going to have time to do anything about this myself, but if you decided to try it, I’d appreciate it if you let me know. (And if turns out to be the “next big thing” in studio technology and you start making millions, it would be cool if you sent me one. LOL.)

Software music studios, Logic and UI Design

Back in the good ol’ days when I used a PC to make music, Cubase was my tool of choice. When I got my G5, I tried running the Mac version, but for me at least, Cubase on the Macintosh is just too frustrating. Crashes, incompatibilities, the need to have a USB dongle, bah! The last time I rebuilt my hard drive, I didn’t re-install Cubase, and I don’t miss it.

In fact, I have been using Reason 3.0 for most of the music I’ve made lately. As an all-in-one software studio, it’s almost everything you need. Even though it doesn’t support software plug-ins, what’s there really is quite inspiring. The Combinator, in particular, has made a huge difference in the scope of things you can do with it. Unfortunately, if you want to mix in recorded sound (like a cello ;-)) Reason just won’t cut it. There’s just no way to reasonably (sic) get digital audio tracks into it.

So the obvious thing to do on a Mac is to try Garage Band. As I found, GB actually works quite well. It’s surprisingly powerful, for an essentially free piece of software, with features like commercial-grade software instruments, multi-track recording, the ability to freeze tracks — they call it “locking” –, and pitch correction. It also has an easy to use GUI, with “Mac-like” simplicity at first clance, but lots of flexibility/capability when you need it. I’m sure I could use Garage Band as my main DAW, but it’s missing a few features (like, score creation) that I occasionally need.

In the Macintosh world, the next step up from Garage Band is Logic Express. I had some cash to blow on tech toys (since I just sold my old PowerBook) so I picked up a copy. After playing with it for a day, I can say that it seems like it is what I was looking for, but man what an arcane GUI.

Now, I’m sure there are Logic-heads out there who will tell me that it’s the “one GUI to rule them all”, but it’s obvious that the Apple GUI designers lost the battle when they tried to bring that we-bought-the-company app into the fold. I’m heading out for dinner, so I don’t have time to do a full rant about it, but the capsule summary is “This one makes Cubase look easy to use.” ’nuff said.

GarageBand actually works.

It will happily host all of the AU plug-ins you can throw at it. Don’t believe me? Try this on:

Believe it or not, it has:

  • 1 ARP 2600 V
  • 1 Moog Modular V
  • 1 Boesendorfer 290
  • 1 Crystal
  • 1 Reaktor 5, and
  • 2 Absynths

plus a random selection of GarageBand drum loops.

Even with many of the synths playing more than one note, I was still only running at around 70% CPU according to Activity Monitor.