The whine started as background noise
One afternoon the charger for an old laptop began emitting a steady high tone. It sat on the desk while I patched a simple delay line. The tone slipped into the input without any microphone. Within ten minutes the frequency had risen thirty cents.
I decided to keep it. The first loop was ten seconds long. Each pass through the delay added a little pitch wobble that matched the rising heat in the transformer.
Tracking the drift in real time
After twenty minutes the whine had climbed another semitone. The loop no longer matched the original timing reference. I reached for the delay time knob and nudged it forward by four milliseconds. That single adjustment kept the new pitch center from clashing with the older layers already cycling.
The charger cooled during a short pause to answer the door. When I returned the tone had dropped. The loop now contained two conflicting centers of gravity.
Instead of erasing the colder section I routed a second delay with a slightly longer repeat time. The two versions ran in parallel. One warmed, one cooled. Their difference created a slow beating that replaced any need for an LFO.
Three passes later the charger warmed again. The beating narrowed. I left it alone. The narrowing itself became the movement.
Why the artifact refused to be replaced
Software oscillators stay locked. This whine did not. Every time the room temperature shifted by even a degree the pitch moved again. I tried to mimic the drift with an analog VCO and a slow envelope. The result sounded too regular. The charger never repeated its exact path twice.
That inconsistency forced small performance choices every thirty seconds. A filter cut here. A feedback trim there. The piece grew from those corrections rather than from any pre-planned structure.
Similar accidental pitches have appeared in earlier experiments with contact mics on power strips. The charger version simply lasted longer because the heat cycle was slower.
Live patching without a safety net
During one take the whine jumped upward fast enough to break the loop into two separate rhythmic layers. I added a gate triggered by the same signal. The gate opened only on the louder peaks. Suddenly the piece had a kick pattern that had not existed five minutes earlier.
Nothing in the patch was stored. Each adjustment happened while the loop continued. Stopping to recall a preset would have let the charger cool and erased the current state.
The final recording runs fourteen minutes. Roughly half the material came from the charger itself. The rest came from reactions to its changes.
One listener asked whether the same sound could be recreated on tour. The answer is no. Different chargers, different rooms, different ambient temperatures all produce different drift rates. That is the limitation and the point.
Further notes on unstable sources appear at this technical log. A separate thread on heat-affected circuits lives at the maintenance archive. Additional examples of single-take hardware pieces sit at the performance diary.