Skip to content
IkonaBIT Instruments written by musicians, for musicians. Sign in

2026 · W36 2 min

What a normalised note hides

A common way to check a model against the instrument it copies: play one note, hold it, bring both recordings to the same loudness and compare the harmonics, the extra tones that give a note its colour. It is a good check, and we used it too. But it looks at one note at one level. It cannot tell you what happens when you play more.

Colour that follows the level

A plain volume control only makes a sound louder: the colour stays the same. An amplifier stage built from transistors does more. Push it harder and it adds more colour, so the sound gets brighter, not just louder. Bring everything to the same loudness and that difference disappears.

Arco49, our string ensemble, models the two transistor stages that the original’s schematic shows on the way from the keys to the output. How much colour they add depends on the signal level inside the machine, which the schematic does not give, so that level is a setting we chose. Hold more keys and more signal reaches these stages: a chord colours the sound more than a single note, and the colour follows each note as it swells and fades. The VOLUME knob sits after them, so it changes how loud, not how bright.

The amount of colour is a design choice. We aim for a gentle rise in the upper harmonics as the sound gets louder, a target we took from public recordings of the original’s crescendo. We have not measured the machine itself, so it is our target, not a measurement of the original.

What it changes in practice

Arco49 answers to how much you play, not only to the volume knob. A single note is gentler; stack a chord and the sound opens up.

It was not always so. At first Arco49 had a plain volume control where the original has those two stages, and the colour did not change with the level. We found the gap and added the stages.