Berghain’s Säule

For some years, I have been interested in the possibility of composing not only for a space, but with it. We tend to think of a venue as the place where music is played, a more or less neutral environment in which a composition unfolds. Yet every room has its own acoustic character. It reinforces certain frequencies, absorbs others, and allows some sounds to linger long after their source has gone silent.

I wanted to know whether these properties could become part of the composition itself. Not something to correct or work around, but something to play.

When I was invited to perform at Berghain’s Säule, I thought it was the perfect opportunity to pursue this idea. I would explore the room’s acoustic character and use that understanding to compose a techno live set specifically for the space. Alongside this, I wanted to bring the Neurorack, a neural synthesiser I had developed during my doctorate at IRCAM and had recently begun enhancing.

These were two quite different experiments, but they came from a similar curiosity: what happens when we approach the conditions through which sound is produced not as fixed constraints, but as material for musical invention?

An instrument made of architecture

An instrument is never simply a means of producing sound. Its interface, construction, limitations and particular behaviour influence what can be played and, inevitably, what we imagine playing on it.

I have spent much of my research trying to understand this relationship, particularly through the design of new digital and hardware instruments. With Säule, I wanted to extend the question beyond the instrument itself.

A room is not passive at all. Its dimensions, materials and geometry influence the way sound propagates and is perceived. Anyone who has spent time on a dancefloor knows this physically: a bass note that seems to fill your entire body in one position can feel very different elsewhere. Some sounds linger, others disappear almost as soon as they are produced.

In sound reinforcement, these effects are usually regarded as problems. Considerable effort goes into minimising them, making sure that the audience hears something as consistent as possible across the room. However, I was interested in the opposite approach.

If a room naturally responds to certain sounds, why not compose with that response? If a sound continues to resonate after the synthesiser has stopped producing it, could that decay become part of the rhythm? Could the architecture introduce something into the music that I had not entirely determined myself?

I did not want to turn Säule into an acoustic curiosity or make a demonstration of its resonances. I wanted to find a way for these properties to participate in my techno set, where sound is experienced as much through the body as through the ear.

Learning to listen to the room

Before the performance, I was fortunate enough to have access to the empty venue and its sound system for an acoustic exploration.

I spent time listening to the space, carrying out measurements and observing how sound behaved in different parts of the room. I was interested in understanding its particular acoustic character, not as an abstract architectural object, but as the space in which the music would eventually be heard.

This process required a different kind of listening from the one I usually practise in the studio. Rather than focusing exclusively on the sound I was producing, I was paying attention to what happened after it left the speakers. How did the room respond? What remained of a sound as it travelled through the space? How did its character change depending on where I stood?

The observations offered a way of approaching the room as an active participant in the music, with its own tendencies and limitations.

I was not looking for a perfect acoustic model or a way to make every part of the space behave identically. What interested me was precisely its singularity: the possibility that a composition could develop a different character through its encounter with a particular architecture.

I returned to the studio with a new way of thinking about the material I wanted to write.

A room that is not a box

Most of what makes the room interesting is also what makes it awkward. Heavy pillars stand inside the room rather than at its edges, so the low-frequency field is interrupted in the middle of the space instead of only at the boundaries. The ceiling is not at a constant height, which smears the vertical behaviour across a range rather than concentrating it where a rectangular room would put it. And the back of the room is a wall of glass, a surface that behaves nothing like the concrete around it.

Two further variables sat outside my control, and they behave very differently from one another.

The first is the audience, and I assumed it would matter less than one might expect. At these frequencies a wavelength is many metres long and a human body is a small fraction of it, so a crowd scatters and absorbs far less here than it does higher up the spectrum. I should be honest that this is a working assumption rather than a settled fact: published absorption figures for an audience are mostly measured in seated halls, where much of what is being measured is the seating rather than the people, and the data thins out badly below the low-mid bands. A standing crowd on a concrete floor at sub frequencies is not a well-documented case.

The second is temperature, and that one is not an assumption at all. The speed of sound in air rises by roughly 0.6 metres per second for every degree Celsius, about 0.17 per cent, and the frequency of every room mode is directly proportional to that speed. An empty room measured in the afternoon and the same room full and warm hours later are therefore not acoustically the same room: a ten degree rise moves every resonance up by nearly two per cent. For a sharp, high-Q mode that is a meaningful fraction of its own width. I measured a cold, empty room and then played a warm, full one, and that difference is the single largest source of error in the whole approach.

Writing for something I could not fully control

The question was now how to turn that experience into music.

I did not want to compose an entire piece around an acoustic analysis, or allow the measurements to determine every musical decision. I wanted them to inform the composition while leaving room for intuition, experimentation and the unexpected.

I began thinking more carefully about how sound occupies a space, and how the relationship between presence and absence can shape our perception of it.

Sometimes I wanted the low end to feel immediate, dry and tightly controlled. At other moments, I wanted to allow the room to participate more actively, letting sounds accumulate, resonate or extend beyond the gestures that produced them.

The interesting part was not simply finding sounds that would interact with the architecture, but deciding when to make that interaction perceptible.

A resonance that is constantly excited quickly becomes part of the background. For the room’s contribution to emerge, the music also needs moments of restraint. Silence, contrast and the distribution of energy became as important as the sounds themselves.

There was something appealing about composing for an effect I could never entirely reproduce in my studio.

I could anticipate certain behaviours, but the final result would depend on the actual room and on the sound system. The composition would only fully exist in that particular situation.

Bringing the Neurorack back to the stage

At the same time, I was returning to another project that had occupied several years of my life.

The Neurorack began during my doctorate at IRCAM, as an attempt to bring neural audio synthesis into a physical, playable instrument. I was interested in what happens when a neural network is no longer something we interact with through a computer interface, but a sound-producing system that can be played, patched and explored in real time.

I have always been interested in designing instruments as a way of expanding musical possibilities. An instrument proposes a particular relationship with sound. It gives access to certain gestures, resists others, and occasionally leads us somewhere we would not have thought to go.

With the Neurorack, I wanted to explore what a neural synthesiser could offer within this relationship. In the months leading up to Säule, I spent a considerable amount of time rebuilding and extending it, exploring new ways of interacting with its sonic material.

Two ways of making a sound

The module does not synthesise from an oscillator. The model is conditioned on a small set of spectral descriptors, loudness, brightness, noisiness, bandwidth and pitch among them, and reconstructs a waveform from that description rather than building one up from scratch. Three such descriptions are loaded at once, and a good part of playing the instrument consists of moving through the space between them. A morph control traverses that space continuously.

Rendered straight from the instrument rather than taken from the night, this is that movement on its own: a phrase that begins deep and dark and arrives somewhere brighter and more metallic, moving one step at a time.

A second pair of controls lets one description lend only its tail to another, so a strike can begin as one material and end as something else, which morphing alone cannot do because it moves every characteristic at the same time.

Alongside the neural engine there is now a granular one, reading the same recordings directly rather than through the model. A single control crossfades between the two, and the positions in between are genuinely both engines at once. A texture mode holds a sound open past its natural length, so an instrument built around impacts can also sustain.

Inside the set rather than beside it

The part that mattered most for a live set is that it behaves like any other module in the rack. There are four control-voltage inputs and two gate inputs on the panel. The first CV is pitch, one volt per octave, and can be quantised to semitones. The other three can be assigned to any parameter in the instrument, each with its own attenuverter, offset, response curve and slew, so a single modulation source can be gentle on one destination and extreme on another. The gates do more than fire a note: they can freeze the current sound, regenerate it as a fresh take, step through the loaded descriptions, act as a clock, or hold a texture open for as long as the gate stays high.

This is what allowed the Neurorack to sit inside the set rather than beside it. The sequencers, envelopes and modulation sources I was already using could address it directly, and its sonic character became something to play and automate rather than something to configure beforehand.

One behaviour took some getting used to. Because the model is conditioned per note, some controls respond instantly while others can only take effect on the next trigger. A filter opens under your hand; a change of material waits for the next hit. I stopped treating that delay as a limitation to work around and began composing with it, thinking in terms of what could move continuously and what had to be placed.

What it is like to play

What interests me most is not the technical complexity of the synthesis itself, but the way the instrument can be played. I wanted to interact with the character of a sound, its weight, brightness, noisiness or texture, rather than exclusively with the technical processes responsible for producing it.

These qualities become something that can be shaped through gestures, modulation and experimentation. The musician does not simply select a result but develops a relationship with a system whose possibilities gradually reveal themselves through playing.

The aim is not to make the technology disappear. It is to develop an interaction in which the technical processes become part of a musical practice rather than an obstacle to it.

This balance between discovery and control has been central to the project from the beginning. I want an instrument that can surprise me, but that I can also learn, practise and develop an intuition for.

Bringing the Neurorack to Säule was an opportunity to explore this relationship in a context very different from the laboratory or the studio, where musical decisions have to happen in the moment and the instrument becomes part of a larger performance.

Between two instruments

Looking back, I think the room and the Neurorack were two expressions of the same artistic preoccupation.

One was a neural instrument I had spent years designing, gradually refining how it could be controlled and played. The other was an existing architectural space whose acoustic behaviour I was only beginning to understand.

In both cases, I was interested in making music through something that has its own particular way of responding.

I did not want the Neurorack to function as a technological demonstration, any more than I wanted the room measurements to become an exercise in acoustics. I wanted both to become part of a musical language that could stand on its own.

Building the performance around techno was never really in question. It is the music I love making most, and its relationship with repetition, physical intensity and low frequencies made it particularly suited to exploring the relationship between music, space and dancers. It gives the room something to respond to, while allowing me to move between more direct rhythmic material and passages in which textures, resonances and less stable forms of sound could take over.

The challenge was to bring these elements together without making the music feel like a sequence of experiments.

The night itself

There is a particular uncertainty in preparing a performance for a room that cannot be reproduced anywhere else.

I had spent weeks composing, testing and adjusting material in anticipation of how Säule would respond. But from the booth, I could not hear what was happening across the dancefloor. The acoustic interactions I had composed for were taking place somewhere beyond my own listening position.

I had to trust the preparation and, ultimately, the music.

Only afterwards, speaking with people who had been standing in different parts of the room, could I begin to understand how some of those decisions had translated into the experience of listening.

I find this uncertainty interesting in retrospect. A performance is never entirely under the control of the person producing it, but this project made that condition unusually explicit. I had composed with the expectation that the space would contribute something of its own, while knowing that I would not be able to hear or direct that contribution in its entirety.

Säule was one of the most demanding projects I have undertaken, both technically and musically. It brought together years of research, months of instrument development, and a composition that could only really be tested on the night itself: even if it can be played on any dancefloor, it would be felt differently in Säule.

More than anything, it reinforced my interest in making music through systems whose behaviour can be explored, understood and shaped, but never completely reduced to intention.

Acknowledgements

I am particularly grateful to the Säule team, and especially Krishan and Martin, for giving me access to the empty venue and its sound system, and for their professionalism and help throughout the process. Being able to investigate the room under the conditions of the eventual performance was essential to the project.

Massive thank you also to Egregore and Terenor Records for the invitation and the help, for trusting an unusual proposal, and for giving me the freedom to pursue it.