Dynamic Tube: shape colour around the input envelope

Use the accents as your reference

The original eight-second synthesized bass below contains a strong note followed by quieter notes. Each note also fades after its onset. Those two contrasts—between notes and within a note—give you a way to judge input-dependent processing. This is generated audio, not a recorded bass performance or a captured tube device.

Begin with type A, Drive +8 dB, Bias 55%, Envelope 1.20, Attack 30%, Release 55%, Output −6 dB and Dry/Wet 100%. These settings intentionally expose colour rather than aim for invisible processing. Start at a quiet playback level and follow one strong note into the next softer note before deciding whether more drive is useful.

Remove movement before studying it

Set Envelope to 0 and leave Bias alone. Compare Drive at +2 dB and +8 dB. Listen for a change in the upper edge of the accented note and for how differently the softer notes respond. A nonlinear process can create harmonics whose balance depends on signal level; the same Drive value does not produce an identical result on every recording or every note.

Reduce Output after increasing Drive until the comparison feels similarly loud. The two controls are not linked. Output is a gain stage after the colour-producing core, so turning it down makes the processed sound quieter without undoing that processing. If the accent loses its identity even after level matching, lower Drive rather than trying to restore dynamics with Output.

Bias is a character decision

Keep Drive at +8 dB and compare Bias 35%, 50% and 65%. Listen for how the balance of roughness and note body changes. Bias conditions the nonlinear model; it is not simply a treble boost or a second volume control. Avoid assigning a particular harmonic recipe to every position. The response also depends on type, drive and the incoming waveform.

Choose a position that leaves the quiet notes intelligible. If only the accented note benefits while the rest becomes scratchy, return Bias closer to the centre and reduce Drive. A useful setting should serve the phrase, not just its loudest moment.

Now let the envelope participate

Restore Envelope to +1.20, then compare it with −1.20. Its range extends from −3 to +3, so this is a signed amount, not a wet percentage. Listen for a different relationship between the onset and fading body. Do not assume positive means louder or negative means cleaner throughout the note; judge the movement against the same accented pattern.

With a nonzero Envelope, compare Attack at 20% and 80%, then restore it and compare Release at 20% and 80%. Change one at a time. Ask whether the colour is concentrated near the start or remains apparent farther into the decay, and whether the next quiet note retains its own identity. Back off the Envelope amount if the movement distracts from the rhythm. These controls are displayed as percentages; they are not documented millisecond times or a substitute for a compressor’s attack and release.

What the curve picture cannot tell you

Types A, B and C select separate learned processing models. Public Dynamic Tube feeds them through a short, original smoothing filter, not the captured input-filter response retained for private reference. The controls remain available, but this changes the timbre. No fixed alignment delay is added by this input filter.

The panel curve is an idealized illustration, not a measurement of the live waveform or an exact transfer plot of those models. Compare types with the same phrase and recheck Output level rather than choosing by the drawing. The current implementation does not expose a working Tone control; do not build a recipe around one.

Finish by trying Dry/Wet around 60% if the fully processed version is too insistent. This adds the original signal back rather than making the model itself more transparent. Keep a fully wet comparison available so you can tell whether the blend improves articulation or merely conceals excessive drive.