#4 in a series of posts about the tracks on my latest album, Small Things.
Self-remix! Trying a new structure
The main feature of ‘Trellis’ is that I basically wrote a whole track and then chopped it up, in a really systematic way, to see if it would make it better. I really liked the results and did some edits to maximise the good times, to my taste.
I often find it really fruitful to establish a system or process, follow that, and then bail on the rules I set myself when I’ve gone far enough that I know where I want to go next. This is an example of that.
Here’s the pre-cut-up version. On some level it might sound really familiar already, because the material is most of what you hear in ‘Trellis’.
The parts are basically:
- A fairly constant, shimmering bed of sound, an echoing arpeggio based on one chord, with some looping and other stretched sections layered in.
- Some long, single synth notes on top, coming in loudly near the start.
- The piano-like layers that come in later in this version and slowly evolve. They’re shorter and punchier than the above: an electric piano, an acoustic piano, a couple of synth sounds.
- Incredibly mooshy drums.
I exported something that I could have called finished (yeah, the above audio), chopped off the end and then sliced the rest into two-bar phrases. Say I took 96 bars, that would give me 48 slices.
Then I swapped all those slices around, moving the last slice so it came between the first and the second, the second-to-last slice so it came between the second and the third, and so on.
Here’s an image to try and demonstrate, using 24 slices so it’s a bit less hectic than with 48. The top row represents the order of the original audio and the bottom row is how I rearranged it. (Click to enlarge if it’s really tiny.)
I made some edits afterwards, from memory I chopped out 8 bars here and 8 bars there to tighten things up: obviously the final track isn’t almost 8 minutes long like the above audio!
On top of all of this I repeated the process with some piano parts, but I reversed them.
One chord, but picking notes out of it
This track really sits on one chord throughout, with the individual notes picked out as an arpeggio. There’s a lot of echoes so notes often overlap and once they pile up it becomes a bit of a wash, the “shimmering bed of sound” I mention above.
The most traditional voicing of the notes being played would give you a D#m7 chord with a flat 9 (E) on top. We’re in D# phrygian throughout, which is like a minor scale with a flat second (that E note).
But I’ve got those notes really spaced out, so much so that I reckon it feels incredibly open and ambiguous, and not like a minor 7 chord much at all. The notes span over three octaves. Here’s what they look like on a piano roll.
More than one time signature at once
I’m not actually letting that washed out arpeggio play out, even in the pre-chopped track. Instead I sample and loop 1 beat in every 10. The track is just in 4 4 time, so this effectively creates a polymeter of 10 over 4, I guess. Another way to think about it is the loops are 2 1/2 bars long, given the track is in 4 4.
I think I love polymeter and use it so much because it’s kind of disarming: after decades of getting deep into techniques of music making, I can honestly never listen to music like a normie does, but making patterns I find it near impossible to follow is one way to try to get past that technical brain. (This is called a post hoc justification, gentle reader.)
Banging on metal, fed into a vocoder
There’s only a couple of sounds that I added on top of the sliced up track I’ve been describing so far. I’ll just write about one.
When you’re brushing your teeth, do you ever change the shape of your mouth and notice how the sound changes?
So, that’s the guts of how a vocoder works. (I mean it’s even closer to a talkbox or what Peter Frampton did with his talking guitar, but shush.) You take a fairly constant sound (a synth playing a melody or chords) and you shape how much of that synth reaches people’s ears with a second source, very commonly a person’s voice. ‘Around The World‘ is a classic vocoder-led tune, but you can go back to a zillion ’80s electro classics.
The robot voice stereotype is entirely legit, but you’ve probably heard more subtle uses of a vocoder without noticing it before. You’ve probably heard drums sent through a vocoder, like Royksopp’s ‘Remind Me’ or the later part of Stereolab’s ‘Refractions in the Plastic Pulse’. (That Stereolab album has been a big influence, btw – connect the “dots” to Mouse On Mars and Tortoise members on production duties.)
So, let’s break down a vocoder part that comes in around 2m15 in ‘Trellis’.
This is the absolutely awful sounding synth part that I’m playing.
Wow. I’d like to apologise for bringing shame upon my entire family.
Anyway, then I’m shaping that synth sound with this recording.
That’s me banging and tapping on a telecom tower, on the top of Te Ahumairangi Hill in my hometown. I used contact mics taped to the tower, so I’m recording the sound as it travels through metal rather than through air. Most places in Wellington are very windy and there are many birds, so contact mics are handy when you don’t want either of those things in the recording.
So here’s the vocoder magic. I really like this combination of the above sounds already!
Finally, I send that through a big ol’ reverb. Here is what you hear in the final version of ‘Trellis’, as a kind of added sprinkle on top of the cut-up sections.