MEMLNaut Firmware: NautLinear

Overview

NautLinear is a nonlinear tone and harmonics processor. It shapes sound with a range of waveshaping processors, creating tonal changes and adding warmth and complexity ranging from subtle to extreme. A preamp drives eight waveshapers side by side, a damper tames the result, and feedback sends the output back into the shapers. The neural network controls 47 parameters, and you shape them with the NISPS controls.

Signal path

flowchart TD
  IN["Line in<br/>L + R summed to mono"] --> AL["Auto-level"]
  AL --> PRE["Preamp<br/>drive, asymmetry, sag"]
  PRE --> SH
  FBIN(("Feedback")) -. "master × each shaper" .-> SH

  subgraph SH["Shapers · master drive, bias, 4× oversampled"]
    direction LR
    S1["Hard"] ~~~ S2["Tanh"] ~~~ S3["Knee"] ~~~ S4["Diode"]
    S5["Oct"] ~~~ S6["Asym"] ~~~ S7["Sine"] ~~~ S8["Cheby"]
  end

  SH -- "mix level of each shaper" --> DMP["Damper<br/>slew limit, saturation"]
  DMP --> MK["Makeup gain<br/>matched to the dry level"]
  MK --> WD{"Wet / dry<br/>knob Y"}
  IN -- "dry" --> WD
  WD --> OUT["Line out and phones<br/>mono on both channels"]

  DMP -. "FB Pre-Make" .-> TAP(("Feedback tap<br/>back to the top"))
  MK -. "FB Post-Make" .-> TAP

Dotted lines are feedback: the tap at the bottom feeds the Feedback node at the top. The Options screen chooses where the tap is taken from.

The preamp

The input is levelled, then driven into a soft saturator with an adjustable bias. The bias makes the curve asymmetric, which adds even harmonics. With sag, the bias also shifts with the input level, the way a driven valve stage does: loud attacks get more even harmonics, and the tone relaxes as the note decays.

The shapers

Shaper	Sound
Hard	Hard clipping: flat-topped, buzzy
Tanh	Smooth saturation
Knee	Clipping that goes from soft to hard
Diode	Diode-style clipping, softer on one side than the other
Oct	Octave-up fuzz: a rectifier that goes from half-wave to full-wave
Asym	Bends the positive and negative halves of the wave differently
Sine	Wavefolding
Cheby	A harmonic mixer that adds the 2nd to 7th harmonics

The first five share a master drive, each with its own trim. Hard, Tanh and Knee also share a bias, which adds even harmonics. Sine and Cheby have their own input levels instead. Each shaper has its own mix level and its own feedback amount.

The damper

At the end of the shaped signal, and so inside the feedback loop:

• a slew-rate limiter, which only acts on fast, loud movements, so it tames screaming high-frequency feedback without dulling quiet playing;
• a soft saturator, which bounds the level going round the loop.

Level

NautLinear matches the loudness of the shaped sound to the dry input, so changing the parameters changes the tone more than the volume. A soft limiter keeps the output from clipping.

Inputs and outputs

The processor takes a stereo line input. Left and right are summed to mono before processing. Mono processed sound is output to both channels of the line out and headphone sockets. Set the input level on the Levels screen.

Quick start

1. Flash the firmware, connect a sound source to the line input and turn up G.
2. Set the input level on the Levels screen.
3. Set the wet/dry balance with knob Y.
4. Move the joystick, then like, dislike, drag and jolt as described on the NISPS page.
5. Turn shapers, feedback, the preamp and the damper on or off on the Enable screen.

Controls

Click the panel to enlarge it. Controls ringed in gold work differently from the standard NISPS controls.

The controls are the standard NISPS controls, with this change:

Control	In NautLinear
Knob Y	Wet/dry mix (instead of reward scale)

The network’s input defaults to the joystick. You can change it on the Input Source screen, including to machine listening of the line input.

Screens

NautLinear adds three screens after the Neural Net screen. The Levels screen, with the input gain, comes at the end as in every firmware (see the NISPS page).

Neural Net: the 47 outputs are labelled by section: Pre, Shape, Mix, Damp and FB.

Focus: choose which parts the network changes. The parts are Hard, Tanh, Knee, Diode, Oct, Asym, Sine, Cheby, FB, Pre and Damp. Each shaper’s group includes its mix level and feedback amount. Pre covers the input levelling and the preamp. The focus is not saved.

Enable: turn each shaper, the feedback (FB), the preamp (Pre) and the damper (Damp) on or off. Highlighted parts are on. Turning a shaper off takes it out of the blend, and the others rebalance to fill the level.

Options: choose where the feedback is taken from.

• FB Pre-Make (the default) takes the feedback before the level matching. The feedback can keep sounding, and even howl, after the input stops.
• FB Post-Make takes it after the level matching. The feedback mostly changes the tone, and dies away with the input.

Saving

The Enable and Options settings are saved with each model, and come back when you load it.

Parameters

Outputs	Section	What they control
0–6	Pre	Input levelling, preamp drive, asymmetry, sag amount and sag recovery, master drive, bias
7–27	Shape	Drive trims for Hard, Tanh, Knee, Diode and Oct; knee hardness; diode softness; octave rectifier; Asym curves; Sine input level, fold depth and asymmetry; Cheby input level and the 2nd to 7th harmonics
28–35	Mix	Level of each shaper
36–37	Damp	Slew limit and saturation
38–46	FB	Master feedback, and feedback into each shaper

MIDI

This firmware doesn’t send the neural network outputs as MIDI. It does listen for MIDI CCs, so you can control NISPS from a footcontroller or similar:

CC	Action
1	Positive reward
2	Negative reward
3	Scramble neural network
4	Clear replay memory
5	Exploration noise
6	Training speed

If you choose a MIDI controller on the Input Source screen, CCs 1-8 become inputs to the neural network instead.

Files

Code

https://github.com/MusicallyEmbodiedML/MEMLNaut-NISPS. To build this firmware, define MODE_NAUTLINEAR in MEMLNaut-NISPS.ino.

Firmware Download

MEMLNaut-NISPS.NautLinear.1.2.1.uf2

See the firmware guide for installation instructions.