Lessons from Aegispolis
2025-12-28
I made a cover of Aphex Twin's song Aegispolis this month, for fun and learning. In this post I reflect on what I've learned.
Learning goals
I make my own music and I'm always looking to get better at it. Something I still find difficult is how to go from a musical idea to a 3-5 minute song. There are "formulas" for this in pop music (verse - chorus - verse - chorus - bridge - chorus) but it doesn't feel right when I apply these to my own music. How does Aphex do it?
Another goal in making this cover was to get to know my equipment better. I've been experimenting with limited, simplified setups to learn new things, but learning new tools while writing music can be a distraction. By making a cover version of someone else's song I can focus on the mechanics of how to use the equipment without worrying about what to write.
Equipment
Ever since I bought a gadget to put my own sounds in it, I have been having a lot of fun with the Yamaha RX5 drum machine, programming rhythms into its built-in sequencer. This is a nice change from my usual routine of writing everything on the Cirklon.
The RX5 is great for drums and percussion but melodies don't come natural on this machine. I've been trying to find a way to keep the experience and feeling of the beats I can write on the RX5 but add some sequenced melodies next to it.
For this cover I settled on using the 1984 Roland MSQ-700 to generate the MIDI notes of the melodies, and the Yamaha TX802 to turn those notes into sound.
Roland MSQ-700, Yamaha RX5 and Yamaha TX802.
I don't believe Aphex used any of these particular machines in the song but that does not bother me. My goal was to learn, not to make an exact recreation.
Analyzing Aegispolis
Aegispolis is the fourth track on the Selected Ambient Works 85-92 LP. I analyzed the structure of this song by drawing little colored blocks in Logic Pro.
I made this diagram by first loading in the original audio and adjusting the project tempo so that the bar changes in Logic match those in the song. Then I started listening carefully when different instruments drop in and out of the song, and putting empty MIDI blocks on the timeline when an instrument was playing. I didn't worry about the notes or how to make the sounds: I only wanted to map out the structure.
Structure of the song
The song contains three groups of sounds: drums and percussion, melodic instruments, and ambient sounds. Each group has its own structure in the song.
The drums and percussion are organized into 2-bar patterns. There are 5 drum patterns which repeat several times before changing. You can see the alternation of the drum patterns in the purple "pattern" row in the diagram above.
In the melodic group there are three synthesizer sounds: a lead melody, a voice that plays slow block chords (labeled "strings" in the diagram), and a bass melody. The bass and lead sounds play distinct 8-bar melodies. The block chords, Fm and B♭, are held for 2 bars at a time, forming a 4-bar repeating pattern. The melodic patterns repeat throughout the song without ever changing. They are sometimes muted on the mixer but when they come back in they are exactly at the point in their respective notes where you would have expected them.
There are occasional far away percussive noises in the track that feel like ambient sounds but they are actually part of the rhythm track. The only proper ambient sound is the sound of rushing water that comes in around 2:00. The water shows no obvious signs of looping or repeating so it is probably one long field recording that is brought in and out with the mixer.
This organization into three groups turned out to fit well with my choice to use two different sequencers: I could let the RX5 sequencer control the drums, the MSQ-700 could control the melodies, and for the ambient sounds I don't even need a sequencer. I just played a long recording of nature sounds out of the headphone socket of my computer.
What did I learn
This selection of equipment is enough to make cool songs
My way of breaking down the song works. I had to do a minimal amount of sequencing (3 8-bar melody patterns on the MSQ-700, and 5 2-bar drum patterns on the RX5) and the rest could be done by chaining patterns on the RX5 and fading melodies in and out on the mixer.
As I write this blog post I'm wondering if maybe it would have been even simpler to not chain the RX5 drum patterns and change them manually instead. It could be that this is actually what Aphex did, because it would explain the unusual timing of the drum pattern changes: these could be live performance "mistakes".
You don't need a lot of different musical ideas to create a 5-minute song
As I mentioned just now, there are just 3*8 + 5*2 bars of MIDI data in the song but it's not boring.
In the original, Aphex patiently reveals the different elements of the song one at a time, creating a sense of development. While rehearsing the performance I found I kept rushing the beginning. The bass only comes in in bar 33 but I would often be ready to introduce it at bar 25.
At the same time he does something subtle where he occasionally briefly brings in percussion sounds which help to make the different sections feel different from each other.
Combining static loops with linear structure
A lot of my own songs fall in one of two categories: either there is a development over time with a "linear" arrangement, with lots of new rhythms and melodies emerging over the length of the song, or the song is a loop-based performances where none of the notes in the loop changes: the only thing that changes is whether you can hear them.
Aegispolis shows how you can have both. The drums form a linear arrangement and the melodic parts form a constant loop with live show/reveal performance. This is an important lesson for me because I felt stuck doing one or the other but not both!
There is a tension between the two approaches that you can see happening in Aegispolis. The bass, which is a melodic loop-forever part, has a companion drum part which reinforces the bass melody. You can hear this when the bass comes in for the first time at 1:17. The moment I start live-performing by turning parts of a loop on and off I loose my sense of time. But the drum sequence is linear and it will start playing the bass-complementing drum pattern at a fixed time, whether I am dropping in the bass melody or not!
I had to take notes and practice to make sure I could do the live bits with the melodies in time with the drum sequence. This is new for me.
An alternative approach would be to change the drum patterns manually. If you do this, lining up the bass melody with the bass-complementing drums becomes easier. Because of the pattern changes in the drums you still get more development than you would with only muting/unmuting on a single loop.
Mixer fades are harder than MIDI mutes but they are not that hard
I have been using MIDI mutes to do live performance of turning parts on/off in my music for some time now. This is similar to the thing I did with the mixer in my Aegispolis cover, but MIDI mutes (on the Cirklon at least) make it easier to stay in time. You won't get notes that cut off or drop in in the middle. With the mixer on the other hand, if you move a fader in the middle of a note rather than in between notes, it can sound sloppy. I think I have been avoiding the faders because of this.
But this time, I was performing a song with sequencers that don't have MIDI mutes, so I had to use the mixer. And besides that, you can hear that the original uses the mixer, so I was really out of excuses not to do it.
Play with tuning knobs more
The tuning of the three melody parts in Aegispolis is unusual. They are all tuned to a different pitch and none of them is tuned to A440.
I never noticed this as a listener but if you try to play along with the original you run into this.
Especially with digital synthesizers, which have perfectly stable pitch, it's easy to fall into the habit of never playing with detuning like this, but it's worthwhile because it sounds good in this song.
Using multiple sequencers is interesting
For a long time now I have tended to write my own songs entirely on the Cirklon. This is nice but it makes it hard to have different parts of the song develop independently. In Aegispolis the melodies stay in one loop while the drums play a series of different loops. The Cirklon wouldn't let you do that because the melody loop is 8 bars and the drum patterns are 2 bars long. You can do it if you really want to but it would be awkward.
If you use two sequencers it becomes very simple to have them behave more independently.
Covers don't have to be perfect if your goal is to learn
I used to get frustrated when making covers because I could never make them sound quite like the original. But now I don't feel like that is the point. The point is to learn and experiment.
On the other hand, I had several moments when working on this cover where I thought, "I'm not going down the rabbithole of getting that detail right", and then an hour later I did it anyway, and I learned even more.
There are people on Youtube who make perfect recreations of recorded songs and for some time I fell in the trap of thinking that has to be the goal. It doesn't.
Make more covers
Covering Aegispolis was surprisingly fun, challenging and insightful. And as I write this blog post I am getting even more ideas I want to try out. So I think I will make more covers!
The MSQ-700 is fun in spite of its limitations
I had read online that the MSQ-700 is fun to work with in spite of being limited. From my own initial impressions I could clearly see the limitations but I was less sure about the fun. This cover project helped me understand where the fun is.
On the Cirklon I can record myself playing in real time and then use quantization and editing to achieve a precise sequenced feel. Editing on the MSQ-700 on the other hand is very crude so if you want that sequenced sound you either need to be able to play like that, or you have to use the step entry mode.
Step entry means spelling out to the sequencer what you want it to do. Play a C for this long, then a B♭ for this long, etc. With real time recording, you just play what you want so you don't have to verbalize it. With the step sequencing mode, in order to not loose track of what I was entering, I wrote down the melodies in classical music notation.
These notes contain many mistakes. The lead melody is Fm pentatonic, not Cm. The chords are wrong. The eighth notes in the lead are really 16th notes followed by 16th rests. The bass part is written in Em but it should be Fm.
Of course you can use a step sequencer without a strict goal and embrace the unpredictable results. This is the technique behind TB-303 "acid" bass lines. But if you have a melody in your head and you want to get that exact thing into a step sequencer, I find that it helps to write the melody down on paper first.
Conclusion
I used to feel that making covers was frustrating and un-creative, but now I think that it's fun and I'm learning a lot. I'm glad I tried this.
Tags: music