Carmilla

2024-09-26 MSK 008 MSK 014 music composition

I posted my latest composition project on the audio server a few days ago. It's called the Carmilla suite, and it's a neo-Baroque piece for piano four hands.

There are two kinds of instrumentation that might be called a piano "duet": a piece might be written for two people each playing a piano of their own, or it might be written for two people sitting side by side and playing the same keyboard at the same time. This piece is that second kind. In earlier times when it was more common for many people to have pianos in their homes and to know how to play them, "piano four hands" was a popular style of music because of its social aspect.

Even if it doesn't have much obviously to do with modular synthesis, I spent enough time on this project (more than a year) that I need to get some value for it, and it does raise some interesting points for modular music-making too, so I'll discuss them here. Even though the techniques and conventions of piano music are different, and a lot of things that are easy to do on the piano are difficult on the modular, we can still learn from those things in ways that may enrich modular music.

The entire project with notes about the musical aspects is posted on the North Coast Synthesis audio server. I'm trying not to duplicate those notes in this posting, so if you want to read about the story the music is intended to tell, or to download MP3 or FLAC versions, go visit the audio server. Single-track streams and downloads are there as well. Here on this page, you can watch a video version (which goes through the sheet music, reformatted to fit on a screen) or click the player to play the whole suite as a single audio track. Either way, it is about 28 minutes in length.

This music is based on the book Carmilla by Joseph Sheridan Le Fanu, and you can go read the book via Project Gutenberg at that link. It's a vampire story, and notably it's an early vampire story, published in 1872, which was 25 years before Dracula. Many of the things we take for granted in vampire books today were new at the time of Carmilla, and on some important points, actually originated in Carmilla. On the other hand, many other things we expect in vampire books, which actually originated in Dracula, simply were not yet part of the vampire lore and don't show up in Carmilla.

Separating pitches and durations

In the Baroque period, both composers and performers tended to see a composer's written score as being more a sort of guideline or suggestion than specific instructions on what to play. There was a lot left to the performers' imaginations. They might play extra notes, use different timing from what the written music might seem on its surface to indicate, and supply extra details like dynamics and tempo that would not be indicated or might be only sort of indicated. That was just the way it was: both composers and performers expected that some parts of the music were deliberately unspecific.

For a while in the 17th Century in particular, some composers were writing so-called unmeasured preludes, where the written music says what pitches to play but it's up the the performer to decide the durations of the notes, and therefore the rhythm. Two performances of the "same" unmeasured prelude might sound quite different while both being correct in the sense of being what the composer wrote.

I decided to include an unmeasured prelude in the Carmilla suite partly because I thought it was a fun concept, and partly as a clue to performers that actually the entire suite is intended to be played in something like the Baroque tradition, with improvisation and some deviations from the written text considered okay. Getting my notation software (Lilypond) to produce the special notation for it was an interesting challenge, too.

For some time after the Baroque period there was a trend toward specifying more and more detail in written music. The expectation fell on composers that they would tell the performers exactly what to do, and on performers that they would actually do what they were told, not add their own ideas. It was only in the 20th Century that there came to be a strong move back in the other direction, with "aleatoric" and what modular users now tend to call "generative" music, where the composer might specify a way to choose what to play, but the notes actually played would be somehow decided at performance time.

How could we bring the "unmeasured prelude" concept to modular? I see two patching ideas that stem from this concept. Both amount to applying the timing from one source to the pitches from another source.

First, we could apply human-performed timing to a predetermined sequence. It's easy to do that by just plugging the gate output of a MIDI interface like the MSK 014 Gracious Host into the clock input of a sequencer. The patch below is just a basic subtractive synthesis patch; I've used North Coast Synthesis modules for that part too because I want to plug them, but the really key part is the Gracious Host. Note that its pitch CV output is not connected to anything; instead, the gate CV (as well as controlling the envelopes) is clocking the sequencer. Play notes on the MIDI controller plugged into this patch and you'll be getting the preselected pitches from the sequencer, but with human-controlled timing. The human player can create a wide variety of different effects all from the same predetermined sequence - much like playing an unmeasured prelude from the 17th Century.

Another idea would be to go in the opposite direction, with a human partially choosing the notes, but the modular patch making more specific choices from among the human-chosen set, and controlling the timing. Here's a patch for that.

In this patch, the MSK 014 Gracious Host is used as a quantizer. Three sine waves from the Fixed Sine Bank are mixed to generate input; the top knob (positive offset) is adjusted to keep the total voltage (which would otherwise be bipolar) centred within the quantizer range. Then the output of the Gracious Host goes to a sample-and-hold module. Meanwhile the timing of the sample and hold comes from a Mutable Grids module: this will produce a regular rhythm but one more interesting than just a straight repetitive beat. You could use some other trigger sequencer, like perhaps a so-called "Euclidean" sequencer, to taste. The point is that it's something where the machine, not the performer, chooses the timing. Downstream of these modules, I've shown in the sample patch the use of two Transistor ADSRs to first lengthen the trigger pulse into a gate, and then give it a nicer shape, before feeding the typical subtractive patch.

This second patch reverses the concept: whereas the first allowed the performer to choose the timing with machine-chosen notes, here the performer chooses the notes, or a selection of notes, by playing on MIDI channel 3 or 4; then the machine chooses the timing, and if the performer holds more than one note simultaneously, the machine chooses among them. A variation of this patch might use one of the "arpeggiator" modes of the Gracious Host instead, with a trigger sequencer driving the 1 PPQN clock input, instead of a sample-and-hold module. Another interesting possibility would be to use the "gate" output on the Gracious Host, here unused, to clock the Grids. That output gives a pulse each time the pitch CV input crosses a quantization boundary, so it would be a way for the melodic pattern to interact with the rhythm while still not being directly under the performer's control.

Ornaments

Ornamentation was one of the typical features of Baroque music. This was sometimes another thing composers left up to the performers, but it was also sometimes included in the written music. The general idea was that quite often there would be extra notes, or special ways of playing the notes intended to embellish the basic tune or make it prettier without actually becoming part of the basic tune.

Sometimes an ornament would be relatively simple and standardized - an added note or two and a change to timing, like these downward mordents. A composer might indicate this kind of ornament with a special symbol in the written music.

In the "Phantomia" movement of Carmilla, there's a segment which is supposed to represent the feeling of pushing through icy water, which is how Laura describes what it's like to surrender to the predation of the vampire. I specified a lot of upward and downward mordents in this passage which help make the whole thing sound more watery.

In that same clip you can see and hear a less standardized ornament right at the end. This kind of thing might be written in advance by the composer (there are several examples in Carmilla) but it would also be expected that performers might add their own.

Sometimes the line between what is melody and what is ornament can be a bit fuzzy. Consider the third movement of Mozart's Piano Sonata no. 11, "Alla turca." (You know it, even if you don't recognize the name; watch this YouTube video if necessary to refresh your memory.)

How did WAM think up all those little notes? Well, his point of view would probably have been that he's just that good. But it also could be that he took a much simpler sequence of notes, added standard, traditional "turn" ornaments, and then wrote them out as separate notes instead of using turn symbols:

Instant music! Compare that with what Shinkichi Mitsumune did in the "Sunlit Garden" prelude from Revolutionary Girl Utena (and here's a YouTube link for the full piece):

Here, very similarly, it's just a few notes ornamented with a repeating pattern - a "reverse turn" and something like what in drumming is called a "flam" - except not really. The first reverse turn is played straight, but in the second one, the last couple of notes are lower than the standard reverse-turn pattern would call for. And the third one is even more significantly modified, either by lowering the first note or raising the last three, depending on your point of view, so there's a big jump in the pattern instead of the smooth steps usually found in a reverse turn. These bits of melody have the general shape of reverse turns, but shifting the notes up and down increases the amount of tension. Those modified notes catch the listener's attention and demand the release of the perfect fifth at the end of the quote.

Now consider those two passages together. Are these melodies really just a few big notes with some ornamentation, or are they lots of small but distinct notes? Mozart's sounds more like "a few big notes" to me; by sticking strictly to the pattern, which was a pattern his listeners in the 18th Century would have heard many times before, he made that pattern fade more into the background so that listeners would hear the big notes. Mitsumune, in his piece written in the 1990s but intended to sound older, is bending the pattern. The individual small notes are brought into the foreground, even though the pattern remains recognizable.

I think that for the most part ornamentation as the Baroque composers imagined it would be more something you would incorporate into a pre-planned sequence, or perform on a keyboard controlling a modular synth, rather than trying to build it into a patch as such. But it's fun to think about how to achieve it just by patching, and I think the key may be to use the "ratcheting" effect familiar to many modular users. Here's a patch making use of the Doepfer A-160-5 "Ratcheting Controller," along with the North Coast Synthesis MSK 008 Octave Switch. To simplify the graphics, I've only shown the control voltage generation here; it's to be assumed that this would take a clock from somewhere, and would feed a "voice" patch to make use of the resulting pitch control voltage.

The Doepfer module is a voltage-controlled clock multiplier; depending on the input CV, it either passes its input clock through unchanged, or at a multiplied rate, generating the classic "ratchet" effect. (In order for this multiplication to work well, the input clock must be at a consistent, stable rate.) The TipTop Z8000 module is several step sequencers in a single module, sharing a grid of knobs with each sequencer using the knobs in a different pattern; it's not important to use this particular module but its relatively small size makes it convenient.

The input clock (1x) drives the first row 4-step sequencer and one of the boustrophedon 16-step sequencers. The output of that 4-step sequencer goes to the Octave Switch to be quantized. In the image you can see an interesting trick: I've patched the two unquantized inputs of the MSK 008 to each other in order to defeat the normalling across from the other channel. Both unquantized inputs of that channel will read as zero volts rather than being copies of what's patched into the other channel, so the channel output will only be the whole-volt-quantized sequencer output. Patching input to input does not necessarily have predictable or useful effects on every module, but it works in this case.

The quantized output of the first row sequencer (normally 0, 1, or 2 volts) gets fed back to control the ratcheting of the Doepfer ratcheting controller; then output from there goes to the second row sequencer clock. Meanwhile the 1x clock also goes to reset the second row sequencer. The idea is that normally, all but one of the knobs in the first row will be set to voltages near zero which will all quantize to zero. One knob will be set higher. Then on every fourth beat, a nonzero voltage will come out of the quantizer and cause the ratcheting controller to produce a burst of pulses, running through the knobs in the second row; but at the start of each beat it'll always reset to the first of those knobs. The performer would set up the second-row knobs to describe a pattern like a "turn," and then use the knobs of the third and fourth rows to make up an interesting melody for the 16-step sequencer, working between the notes fixed by the knobs already set.

The outputs of the second row and boustrophedon sequencers both go to the Octave Switch again to be precision added in the left channel. The result of that is the pitch control output. So we've got a 16-step sequence, but every fourth step (or more often, depending on the knob settings) the regular "step" will be replaced by a fixed pattern something like the "turns" above, transposed to that step's pitch: an ornament. Adding ornaments makes the overall melody more complex and interesting than just a straight repeating sequence with all notes the same length.

It's easy to think of variations on this patch that would make it more flexible or complex. One might use separate sequencer modules for the "ornament" and "basic melody" patches so that they can be controlled with more independence. If the two sequences are of relatively prime length (such as 5 and 13 steps, instead of 4 and 16 as shown here) then the overall pattern becomes much longer, although such rhythms tend to be less danceable.

Pianos are lousy modules

A piano is a terrible Eurorack module.

Consider: the piano can play over a frequency range of a little over seven octaves, which is pretty good by Eurorack standards. But it doesn't actually track over anywhere near that range. Octaves on a piano are typically tuned wide by 10 cents per octave or more, so by the standards usually applied to a synthesizer oscillator, it'll be unacceptably out of tune beyond just a couple of octaves. And there's mandatory quantization on the pitch control, so you can't compensate for the mistuning with Gracious Host frequency calibration or similar. Apart from the poorly tracking oscillators, the high notes on a piano end up with significantly different timbre from the low notes. Someone listening to a piano and believing it was a synthesizer, would wonder whether the filters were tracking at all.

There's potentially a lot of polyphony on a piano. Unlike with most modular synthesizers, it's routine to play six or seven notes at once on a piano. But even at that level, and becoming a greater issue as the polyphony increases further, there start to be limitations related to the performer's anatomy and skill: people usually only have two hands, of limited size, and with at most five fingers per hand, so it's not possible to play multiple notes that are too far apart, or too many simultaneous notes in total; and the more simultaneous notes are being played at once the harder it is to control them precisely. You can get more polyphony, and more flexible polyphony, by composing for "four hands" as I did in Carmilla, and you can even go beyond that with alternate technique like the tone clusters for which Henry Cowell is famous. But it just keeps getting more complicated.

If you said on a synth forum that a piano was polyphonic, you'd find yourself at the bottom of a dogpile of well-actually kids explaining to you that because it doesn't have per-note timbre control, a piano is really only "paraphonic." (They sincerely believe that is a word.)

Also, I got into a big argument with the flight attendants last time I tried to bring a piano as carry-on luggage on a plane. What idiot designed this thing to consume so many HP?

And yet pianos remain popular.

I'm being somewhat facetious here, but that's one of the lessons to learn: it is silly to apply the standards of synth modules to a piano. Although people sometimes say that limitations can boost creativity, it's also possible to say that some of the limitations of a piano are not even limitations in the first place, when properly considered.

For instance, the stretched-octave thing comes about because (for reasons of Science!) the "harmonics" in the sound of a piano string are not exact integer multiples of the fundamental frequency. They tend to be a little higher than integer multiples, and for that reason are more properly called overtones than harmonics. Stretching the octave tuning compensates for the non-integer overtone frequences, with the result that two C keys an octave apart on the piano keyboard sound better in tune with each other when tuned more like 1210 cents apart, than if they were tuned 1200 cents apart (a perfectly just octave). It's not a problem, it's actually a good thing, as long as the piano is only playing by itself. But this is one reason really big pianos (concert grands instead of little uprights) are preferred for playing in ensembles with other instruments: the physical effects that lead to the stretched tuning become less significant with longer strings, so a big piano can be tuned closer to the tuning used by other instruments and still sound good.

It also isn't possible to get rid of the slightly inharmonic overtones and keep the piano sound. There are some tricks you can attempt with phasing and frequency shift, but a simple resynthesis based on integer harmonics, such as we might be likely to attempt with analog electronics, comes out noticeably lacking. That's one reason piano synthesis is usually based on samples rather than modelling. If you do insist on physically modelling a piano, you end up using digital techniques to do modal and waveguide synthesis as you might for a drum: treating the piano as a percussion instrument, which, ultimately, is what a piano really is.

Compositions for piano often aim to positively use features like the varying timbre at different pitches. I did that a lot in Carmilla, for instance in the "dogfight strain" near the end of the Phantomia, which is structured as call-and-response. The player on the right (traditionally called "Primo") plays a couple of bars that are meant to sound like a question, and then the player on the left ("Secondo") plays a response to it, so it sounds like they're having a discussion or argument, and because they're playing in different pitch ranges, the call and response are in different timbres. The higher register has more of a plinky sound with some "air" in it, whereas the lower register is more full-voiced. That is not just a difference in pitch. The shapes of the spectra are different, because of the physics of how the piano works.

Marching band music will often include a dogfight strain with back-and-forth between two different groups of instruments; but a piano, specifically because of its seeming "limitation" of different timbre in different registers, can do it all on one instrument.

I tried to make a virtue of something similar in the design of the MSK 009 Coiler VCF: it was an unavoidable consequence of the electronic design that it wouldn't be able to track accurately and would have somewhat different spectral effects depending on the pitch range, so I've tried to present that as a positive feature. Like a piano, the Coiler can create a variety of different sounds in a single patch, just by giving it different pitch ranges. That would be much harder to achieve with a module like the MSK 007 Leapfrog VCF, which has advantages of its own but tracks precisely.

I think sometimes the criticisms brought forward of synthesizer modules, which would be unfair if applied to a piano, are actually unfair for synthesizer modules too. Before complaining, think: am I metaphorically trying to stuff a concert grand into the overhead bin? Typical limitations of modular synthesizers such as lack of polyphony may be better understood as just what the synthesizer is rather than something wrong with it.

Sustain

One of the most characteristic features of the piano is the "damper" or "sustain" pedal. Understanding it calls for knowing a bit about how a piano works. You press a key, and that triggers a hammer to hit the string, but the key also holds a felt pad called the damper away from the string. You get a loud transient when the hammer hits (the "attack" and "decay" phases in synthesizer terms), and then the string rings for a while (the "sustain" phase) while you hold the key down. When you release the key, the felt damper pad is allowed to touch the string again, and that brings the note to an end (the "release" phase of the envelope).

Stepping on the pedal holds all the dampers away from the strings, regardless of the operation of the keys. So notes begin with the usual hammer strokes and then, at least in theory, the notes don't end. The ADSR envelope becomes more like "ADSS... S... S..." Of course the vibrations will eventually run out of energy and fade away, but on a good piano, notes played with the pedal may last many seconds after the key is released. There are also some secondary effects resulting in particular from sympathetic vibrations between sets of strings. A loud note played with the sustain pedal will cause higher-pitched strings at harmonics of that note to vibrate too, making the sound even fatter.

It's possible to increase the interest in piano music by selectively using the pedal to change the sound at different points; something like a reverb effect. You can hear the difference in the first minute or so of the "Vampire Waltz" movement in Carmilla. This is a longer example and may take careful listening. The player widget links to the entire track rather than a clip. There is a strain repeated twice at the start, and I called for "pedal only on the second repeat," which is the translation of the Italian text highlighted in the screenshot. Parts of the second repeat, with pedal, sound bigger and "fatter" even though the are just the same notes as the first time around.

This would be a very difficult effect to achieve with a modular synth. It isn't even easy for the software MIDI synthesizer I used to actually generate these tracks.

The first thought might be to use an ordinary reverb with a wet/dry control. Pedal down switches to moderately wet, pedal up switches back to fully dry. That would be easy to patch.

The piano damper pedal is kind of like a reverb, but not exactly. A frequent maneuver for piano performers, visible in the score image above at the end of measure 6, is to release the pedal momentarily to terminate all the then-in-progress decay tails, and then immediately re-apply the pedal to catch the next few notes, wiping out the background texture that had been building in order to start in a different direction. You can't do that with a simple wet/dry control; you need some kind of "reset to silence" feature.

Not many hardware reverb modules can simulate the piano sustain pedal. I think the software synth implements it by selectively ignoring the MIDI "note end" messages, allowing each note to continue until the "pedal up" signal is received or a new note at that pitch is played. But that means the software synth needs to support a very large amount of polyphony, with potentially as many as 88 notes playing at once. It's a challenge, especially if someone wants it to work real time. And the secondary effect of unstruck strings vibrating in sympathy, if we want to simulate that too, raises further implementation issues.

Many pianos go even further by implementing a "sostenuto" pedal, which is a kind of selective damper pedal. Step on the sostenuto pedal and it holds the dampers up, but only for those notes which were being played with keyboard keys at the moment the sostenuto pedal was activated. Dampers continue to move normally for all the other notes. So it's possible with sostenuto to lock in a chord and have it play for a while in the background, while other notes play normally (without the sustain effect) on top of that chord. It's even possible to prepare some notes by playing them so gently they make almost no sound, activate sostenuto to hold up those dampers, and then have the prepared notes available later to play with sustain or to use as sympathetic resonating strings.

At the level of sampled audio, the "freeze" effect of the Mutable Clouds (or its present-day successors) is a little like a sostenuto pedal. I think that feature is often ignored by users of Clouds-derived modules, but it's reasonable to use a sequencer gate output to turn "freeze" on and off, setting up and clearing away different harmonic backgrounds for a foreground melody. Some Eurorack reverb modules may offer something similar, whether it's called "freeze" or by some other name, and it could also be simulated on any reverb with a long reverb time, aggressive feedback, and a VCA on the input to control which notes it will or won't pick up to add to the texture.

At the control level, when playing a Gracious Host with a typing keyboard (meaning a "QWERTY"-style keyboard rather than a piano-style keyboard), the Caps Lock key functions a lot like a sostenuto pedal, locking some notes while allowing others to play normally. However, because the Gracious Host only has at most two simultaneous pitch outputs, the usefulness of the keyboard lock feature is basically limited to arpeggio and quantizer modes.

Even without hardware and software features in the instrument, sostenuto can be simulated at the composition level just by writing some very long notes and shorter notes over them. I actually ended up doing that when generating the audio for Carmilla because my software synth did not implement sostenuto, or at least, didn't implement it the way I wanted. So I had to explicitly override the notation software to generate the notes I wanted in the MIDI to simulate sostenuto, separately from the notes that would be displayed in the written sheet music.

Notation

That brings us to one other important aspect of the Carmilla project, the sheet music in conventional music notation with staves, clefs, crotchets, slurs, and so on. I used that for the examples in this article, and I spent a lot of time writing the parts and score with Lilypond, not so much because I can actually read this kind of written music fluently - I can't! - but because I would like to learn. I also think that this kind of notation looks cool, whether I can fluently read it or not. The nicely engraved "sheet music" for Carmilla was part of the art I wanted to create in this project even if (as is quite likely) nobody is ever really going to print off the PDF and sit down at a physical piano with a partner to play it.

I get the idea that conventional music notation is unpopular in the modular world, and much as people say you lose half your audience with every math equation, I may have lost half the audience for this article with each screenshot from the sheet music.

But I think there's room for some sort of notation in modular music-making and I wish the community would spend more time thinking about this. How can we write down the important parts of a patch, or a modular track, in a way that will be useful for sharing our works or for our own compositional purposes?

When it comes to patches: I've been using screenshots from ModularGrid, and there are a few of those in this article. Those are convenient but they may not be easy to read - in order to look at one of those screenshots and know what it's doing, you must know all the modules involved and where their jacks are, and then untangle the wiring in your head to see which functions are controlling which other functions and the overall signal flow. The same skills are needed to patch a modular in real life; but it remains that for written communication, maybe a block diagram would be better.

Even if you think you'll never need to communicate your ideas with other people, having some kind of graphical notation can be useful just for organizing your own thoughts. For example, when planning chord sequences I've gotten some use out of "SeeChord charts," which plot chords using graphical symbols on an axis organized in circle-of-fifths order.

The SeeChord Web site unfortunately seems to be succumbing to bit rot: I don't think it has been touched since 2015 at the latest (despite bearing a 2024 copyright date, presumably automatically updating), and it depends heavily on some kind of WordPress plugin for displaying the charts, which no longer works. So you can only see the charts that the maintainer happened to burn into ordinary image files, not the dynamic charts that the plugin would produce. Nonetheless, it's possible to get a fairly good idea of how the technique works just by reading the text.

I could write a whole article here explaining SeeChord charts in more detail, but the author of the SeeChord Web site is so enthusiastic about asserting intellectual property claims over them, I'm not eager to put serious work into a project like that when I'm not sure it would be appreciated. But given that copyright covers expression and not ideas, I feel okay about at least posting the following chart, which covers measures 91-100 of the Phantomia from Carmilla.

By way of quick orientation: the horizontal axis here is time, split up into measures with the measure numbers at the bottom. The vertical axis shows the root notes of chords, but it's not organized into higher and lower pitch; instead it's organized in circle-of-fifths (theoretically spiral of fifths) order. From each note the fifth or dominant of that note is the next one up (so, G above C) and the fourth or subdominant is the next one down (so, F below C). The two horizontal lines around C are to emphasize that this passage is in the key of C minor.

Chords on the chart are shown with different symbols corresponding to their quality: squares for major, diamonds for minor, a filled-in diamond for C minor in particular (the tonic chord of the key), other symbols for other qualities. There's a specific pattern, depending on the scale in use and easy to memorize, that indicates which chord qualities will fit into the scale, although of course it's possible to play other chords that don't. The symbol for a sus2 chord is actually one I made up (and the same symbol I used for suspended chords in my fractal sequencer), because the SeeChord Web site doesn't seem to specify a symbol for those. And I've written conventional chord symbols along the bottom to further aid in orientation.

This same passage is on pages 23-24 of the reference score. In conventional music notation it requires very careful attention to figure out what's going on with the harmony in measure 97. What's the significance of using a natural note (sharpened relative to the scale) here?

But you can hear the change in harmony there (about 37 seconds into the audio clip), and you can see it on the SeeChord chart. The G-minor chord changes into G-augmented (a chord I use a lot, to symbolize the vampire herself, in this piece), and then it takes a drastic turn into Ab-sus2.

The Ab chord is a big jump on the chart, and it stands out in the music more than the modified G chord does; but it also feels kind of like it makes sense because that Ab is really the tritone substitution of a D-sus2 chord (shown in brackets), which would be out of key for this passage in the key of C-minor but still a comfortable step away from G.

Planning out these kinds of chord sequences, and doing things like transposing them into other keys, is easier (at least for me) in the more abstract graphical chart than when working with raw notes and the complexities of staff notation.

The composer Vangelis, best known for his film scores, for many years used an elaborate MIDI setup which you can see in this YouTube video clip. The big white boxes with rotating button-legend displays seem to be custom-made, and it appears that their main function is basically that of the pistons on an organ console: he can press a button to recall a pre-designed configuration ("registration" in organ terms) that will specify which instrument sounds his keyboards, or different parts of the keyboards, will produce. In the middle of a performance he can press another button to switch to another registration and get a different selection of instruments, without interrupting his playing.

Apparently there are also triggered-sequence things going on here, too: some of the "registrations" include setting up preplanned sequences that will be triggered by single notes on the keyboard at transpositions controlled by the notes he plays directly. This aspect is a little like the auto-ornamentation patch discussed above, but with more complicated conditions for triggering the sequences. This was in the early 1990s and it's interesting to think how such things could be taken further with present-day machine learning techniques, to make the machine "play along" with human improvisation.

That is all cool from a performance perspective, but what really interests me in the Vangelis clip is what he's doing with his felt pen. Vangelis famously never learned to read and write conventional music notation, and he would often compose by just recording his improvisations directly in one take; but at the start of the clip it's clear that he's making some kind of written plan in advance for what he intends to play, using some kind of notation that is meaningful at least to him even if to nobody else. And then he's also got a system of symbols that he seems to use both on the registration-selection boxes and on the "score" he's drawing, apparently a symbol for each registration of the console.

If someone's making EDM with a modular, can they use graphical notation too? Probably! I'm not sure exactly what kind of notation makes sense for that, and it would no doubt depend on the individual anyway. But the "tracker" interface of some sequencing software seems like it's just a hairsbreadth away from a graphical grid already, and the displays of some drum machines all the more so. I'd say it's a question worth thinking about, with the question maybe more important than the answer: assuming you aren't doing so already, how would you write down your compositions, in whatever genre you're composing? And if you are already using some kind of written notation, what effect would it have on your process and your music if you tried to use a totally different kind of notation?

◀ PREV Currency options in the Web storefront