Skip to content

About

Guitar, bass and acoustic amp simulator with a drag-to-reorder pedalboard. AU + VST3 + Standalone for macOS, built with JUCE.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Latest commit

 

History

9 Commits

Folders and files

Repository files navigation

Friendlitronics Amp

An amp simulator, pedalboard and colour box in one plugin — AU + VST3 + Standalone, macOS, built with JUCE and CMake.

Guitar, bass and acoustic amps with freely mix-and-match amp / cab / mic and continuous mic placement; eight pedals on a chain you can drag into any order, with the amp itself as one of the draggable items; and a line path with tape and lofi colour for using the whole thing on a bus or a master.

Everything is a parametric model — analytical filters and waveshapers, no impulse responses and no sample libraries — so the plugin is self-contained and every control is continuous.

The amp page

The pedalboard page

Install

Download the zip from Releases, drag Friendlitronics Amp.vst3 into ~/Library/Audio/Plug-Ins/VST3 (or the .component into Components for the Audio Unit), and rescan in your host.

Release builds are signed with a Developer ID and notarised by Apple, so there are no security prompts and nothing to approve.

Requires macOS 11 or newer; universal for Apple Silicon and Intel.

What it models

Everything is a parametric / analytical model (no impulse responses), so the mic position is a continuous control and the plugin is fully self-contained.

Amps (Source/dsp/AmpVoicing.h)

  • Twin Twin — blackface clean, high headroom, scooped mids, spring reverb

  • Ack Thirty — top-boost chime, class-A EL84 bloom, "Cut" control, present mids

  • Stacked Plexi — Plexi/JCM cascaded-gain crunch, tight lows, strong presence

  • Champion — single-ended class-A, early asymmetric breakup, lots of sag

  • Acoustic (piezo DI) — AER/Loudbox-style clean full-range amp with a piezo "imaging" front end (Source/dsp/AcousticImager.h) for DI'd acoustic guitars:

    • Body — body resonators (air ~98 Hz, top/back ~196/235 Hz, 520 Hz plate mode) + warmth bell + short darkened early reflections: the air/wood a mic hears that an under-saddle piezo doesn't
    • De-Quack — cuts boxy 400 Hz, nasal 1.2 kHz, brittle 6.5 kHz shelf, and a 2.7 kHz dynamic cut that deepens on pick attacks (onset-keyed, level-independent)
    • Comp — stereo-linked RMS leveling compressor (evens out strums)
    • Colour — AER-style contour: bass/treble lift, mid dip
  • Bass: Soft Freezer — Ampeg SVT-style tube head: big headroom, 40 Hz / 800 Hz / 4 kHz tone stack (Ampeg's published centres), with the SVT's deliberately lopsided mid (+10 dB boost, −20 dB cut)

  • Bass: Bee 51 — 25 W Portaflex: warm, dark, early breakup and lots of sag. Baxandall ±10 dB @ 40 Hz / ±18 dB @ 5 kHz, per the Heritage reissue spec

  • Bass: Funky Circle — Acoustic 360: clean and very deep. Its Variamp is a series-LC band whose cut is a near-total null while its boost is modest, so the Mid knob is asymmetric (+6 / −20 dB); Bass/Treble act as scoop depth

  • Bass: Modern DI — hi-fi/Aguilar school: near-flat with huge headroom, the indie & pop pick-bass starting point

All four bass amps add a bi-amp drive front end (Source/dsp/BassStage.h):

  • Split — crossover, 90-400 Hz. The low band is never distorted; the high band is derived by subtraction, so at Blend 0 the stage is bit-exact unity

  • Grit — fuzz drive with octave-up rectification, auto-levelled to the clean band (so Grit changes character, not loudness) and high-passed at the crossover so its intermodulation can't muddy the lows

  • Blend — how much of the high band arrives driven rather than clean

  • Sub — analogue octave divider built the way an OC-2 is: the flip-flop only inverts alternate cycles of the rectified dry signal rather than synthesising a square, so the sub inherits the note's own envelope and stays phase-coherent. Monophonic, like the hardware; tracks best from E2 (82 Hz) up (on a low E the octave lands near 20 Hz and is filtered away)

  • Line / Colour (no amp) — no preamp, no power stage: just a gentle mix EQ plus whatever LoFi / Tape you dial in, for colouring a bus or a master. Set cab and mic to "None (line out)" and it is flat within 0.03 dB from 40 Hz to 16 kHz with every colour control at 0.

Cabs — 2×12 open-back, 2×12 alnico, 4×12 closed-back, 1×10 small combo, Acoustic Full-Range (8" twin-cone, near-flat to 15 kHz), Bass 8×10 sealed, Bass 1×15 vintage, Bass 1×18 folded horn, Bass 4×10 + tweeter, None (line out).

Bass cab shapes come from Ampeg's published ±3 dB figures where they exist plus the only third-party measurements available (gated CLIO curves of comparable cabs) — so the 8×10 has its real −8 dB dip at 1.2 kHz and the 2.2 kHz paper-cone breakup lobe that is the audible "top" of a sealed 8×10, rather than a flat band to 5 kHz. No anechoic data exists for any bass guitar cab; these are informed approximations, not measurements of the real thing.

Mics — bright dynamic, dark/fat dynamic, large-diaphragm condenser (U87-style), small-diaphragm condenser (KM84-style), None (line out)

Amp, cab and mic are selected independently.

Pedalboard

A second tab (PEDALS) holds eight pedals, each with its own enclosure, stomp switch and LED, plus a chain strip along the top.

The strip shows the whole signal path as draggable tiles — the eight pedals and the amp itself. Drag to reorder. Anything left of the AMP tile feeds the amp input, inside the oversampled region, which is where dirt belongs (its harmonics get band-limited before they fold back down). Anything right of it runs after the cab and mic, like an effects loop — right for modulation and delay, though a fuzz placed there runs at host rate and can alias. Clicking a tile switches that pedal on or off.

The order is saved with the session and with user presets.

Pedal Knobs Notes
Wah Pedal / Q / Auto Cry Baby-ish resonant peak, 400 Hz - 2.2 kHz. Automate Pedal for a real sweep, or turn up Auto for an envelope filter
Dive Pitch / Mix / Double Delay-line pitch shift, ±1 octave (5 = unison), with pitch-synchronous grains. Double adds a detuned ADT voice
Sustainer Sustain / Attack / Level Dyna Comp-style 5:1 squash with auto makeup
Screamer Drive / Tone / Level TS-808: the 720 Hz gain-leg split keeps bass out of the clipper, which is the mid-hump
Fuzz Sustain / Tone / Level Big Muff: three cascaded soft clips inside a 90 Hz - 1.2 kHz band, with the 1 kHz scoop
Chorus Rate / Depth / Mix CE-2-style; the wet path is band-limited like a BBD, and the right channel's LFO lags 90°
Phaser Rate / Depth / Fdbk Four allpass stages swept together, feedback sharpens the notches
Delay Time / Fdbk / Mix / Tone Analogue echo: the repeats darken and saturate inside the feedback loop; Time bends rather than jumps

Every pedal switched off is a bit-exact bypass.

Why the pitch shifter is built the way it is. Reading a delay line back at the wrong rate transposes it, but the read head eventually has to jump, and that jump is what people hear as choppiness. Two details fix it:

  • the grain length is snapped to a whole number of periods of the incoming note, so the two ends of a splice sit at the same point in the cycle and the join is phase-continuous. The period comes from a two-stage autocorrelation tracker: a coarse pass on a decimated copy finds the right octave, then a narrow full-rate pass gets the precision. (Zero-crossing detection, the obvious approach, locks onto a harmonic on a plucked low E — it read 140 samples where the true period was 582 — which left every splice misaligned.)
  • because phase-locked grains are correlated, the splice uses an equal-gain crossfade, not equal-power — equal-power sums correlated signals to +3 dB and puts a bump on every splice.

Together those took the inharmonic junk at -1 octave from -22 dB to -77 dB, and on real plucked notes the pedal now adds only 0-8% envelope ripple over what the note already has. PITCH is also smoothed per sample, so sweeping it bends rather than steps. On input with no detectable period (chords, noise) it falls back to fixed grains, which is the classic warble.

Presets

The menu holds four kinds of thing:

  • FACTORY RIGS — the 23 built-in whole-rig presets (also exposed to the host as programs).
  • FACTORY PEDALBOARDS — 10 boards (see below). These set the pedals and the chain order and leave the amp alone, so any board can be tried through any amp.
  • MY RIGS / MY AMPS / MY PEDALBOARDS — whatever you save.

SAVE asks for a name and a scope: Everything, Amp only, or Pedalboard only. That split is the useful part — a rig is usually "this board through that amp", and the two halves get swapped independently. An amp preset never moves your pedals; a board preset never touches your amp. DEL removes the selected user preset. They live as ordinary files in ~/Library/Application Support/Friendlitronics Amp/Presets/.

Loading a factory rig first resets every parameter to its default, so presets are self-contained and can never inherit a stray setting from the one before.

Factory pedalboards

Board What it is
Shoegaze: Glide Fuzz into slow deep chorus, pitch pedal detuning underneath
Shoegaze: Shimmer No dirt: compression, chorus, long phased delay
Shoegaze: Wash Edge-of-breakup drive, enormous delay after the cab
Indie: Wobble Very slow, very deep chorus on a clean amp
Indie: Jangle Doubler Light compression, a hint of chorus, short slap
Funk: Auto-Wah Clav Envelope filter, squashed hard so every note opens it equally
Blues: Screamer + Slap Mid-humped overdrive and one repeat
Lead: Octave Up Octave above, blended under the dry note, then overdriven
Ambient: Infinite Wash Sustainer, delay near self-oscillation, phaser stirring it
Grunge: Cocked Wah Fuzz Wah parked as a fixed filter in front of a fuzz

How the distortion is built

The things that make an amp sim sound like an ice pick are mostly avoidable, and each has a specific cause:

  • Every curve is monotonic. A transfer curve whose slope goes negative folds the waveform back on itself, which sounds like ring modulation. The old push-pull curve tanh(x - 0.15x³) did exactly that above x = 1.49, i.e. whenever Master was up. It is now (1-m)·tanh(x) + m·tanh³(x), which has the same flat-middle/round-rails shape and cannot fold.
  • Every stage has a Miller low-pass. A real triode rolls its own top off through the grid-to-plate capacitance, so each stage in a cascade feeds the next a darker signal. Without it, harmonic order climbs stage over stage and the result is fizz. Later stages roll off progressively further (x0.78 each).
  • The output transformer has a bandwidth. A real one does not pass 20 kHz; rolling the top off after the clip is what stops power-stage distortion from reading as grit.
  • The shapers are anti-aliased (first-order ADAA: y = (F(x[n]) - F(x[n-1])) / (x[n] - x[n-1]), using each curve's closed-form antiderivative). Measured in isolation this improves alias rejection by 33 dB at 1x. Inside the amp's 4x oversampled core it adds little on top of the oversampling, but the drive pedals can run at 1x when placed AFTER CAB, and there it matters.

Measured with 3.7 kHz in (any energy below it can only be foldover), and as a 2-6 kHz vs 200-800 Hz tilt on a plucked-string burst:

brightness tilt before after
Marshall Crunch +8.9 dB +4.3 dB
Marshall Lead +7.1 dB +2.5 dB
AC30 Chime +7.1 dB +5.9 dB
Twin Clean +8.8 dB +7.1 dB

Signal chain

input
 -> InputStage (coupling HPF + bright cap)
 -> AcousticImager (Acoustic amp only: de-quack -> comp -> body -> colour)
 -> [oversample 4x]  Pedals set to INTO AMP
                  -> BassStage (bass amps only: split -> fuzz/blend -> sub)
                  -> Preamp (cascaded asym tube stages)
                  -> ToneStack (FMV / Vox / Champ / acoustic / bass)
                  -> PowerAmp (sag + class-A/AB clip + presence + resonance)
    [downsample]
 -> SpringReverb (Twin / Champ only)
 -> Cabinet (resonance bump + cone breakup + steep HF roll-off)
 -> Mic (per-mic response curve)
 -> MicPosition (distance / axis / angle)
 -> RoomSim (ambient room blend)
 -> Pedals set to AFTER CAB
 -> LoFiStage (bit crush + sample-rate hold + hiss/crackle + narrowing)
 -> TapeColour (any amp: tape saturation + head bump + HF loss + flutter)
 -> output trim + wet/dry mix

Oversampling wraps the whole nonlinear core in a single up/down bracket. Latency is reported to the host and the dry path is delay-aligned for the Mix control.

Mic placement

  • Distance — proximity bass up close, HF air-loss + level drop as you pull back
  • Axis — cone centre (bright/aggressive) → edge (dark/smooth), with a subtle comb
  • Angle — on-axis → off-axis high-shelf cut (the classic "angle the 57")
  • Room — a darkened small-room reverb mixed in (distant-mic feel)

Tape colour (vintage)

Source/dsp/TapeColour.h, adapted from the sibling tape_emulator project, because half of why old bass tones sound like that is the tape they were cut to. It keeps that project's principles:

  • derivative-normalised saturation — tanh(k(x+bias))/k is unity at low level and compresses as it is pushed, instead of adding hidden makeup gain
  • pre/de-emphasis around the shaper (exact algebraic inverse), so highs saturate harder than lows while the linear response stays flat
  • programme-dependent drive — a slow envelope adds breathing compression
  • head bump as a resonance, not a shelf — a bell at 70 Hz
  • gap-loss HF rolloff — a gentle 4 kHz shelf plus a low-pass, not a brickwall

Flutter is transport instability (summed incommensurate wow sines + low-passed noise + an ~8 Hz scrape) read from a fractional delay line, shared by both channels so the wobble stays mono-safe. With Flutter above 0 the delay line adds ~1.5 ms that is not reported to the host, so keep Mix at 100% when fluttering.

Using it as a colour box (bus / master)

Pick Line / Colour (no amp) with cab and mic both None (line out), or just load one of the Master: presets. That path has no preamp, power stage, speaker or mic in it: Bass/Mid/Treble become a gentle 100 Hz / 1 kHz / 10 kHz mix EQ, and the colour comes from LoFi and Tape. Gain, Presence, Resonance, Sweetness, Master and Reverb are greyed out, because there is no amp for them to act on.

  • Tape — saturation + head bump + HF loss, the glue setting
  • Flutter — transport wobble (keep Mix at 100%; see the latency note above)
  • Crush — 16-bit down to 4-bit quantisation, no dither
  • SR — sample-rate hold, host rate down to ~3 kHz, aliasing included
  • Noise — hiss plus sparse vinyl crackle, constant rather than gated
  • Narrow — mid/side collapse toward mono

Every one of those at 0 is a bit-exact bypass, and the Master: presets are level-matched to within about 0.6 dB of the input so bypass A/B is honest.

Building

Requirements: CMake 3.22+, a C++17 compiler, and Xcode command line tools on macOS. JUCE is fetched automatically — you do not need to install it.

cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j

That produces AU, VST3 and Standalone builds as a universal binary (arm64 + x86_64), and installs the AU and VST3 into ~/Library/Audio/Plug-Ins unless the CI environment variable is set.

Notes:

  • JUCE version. Pinned to 7.0.12 via FetchContent. JUCE 7.0.12 predates the macOS 15 SDK, which removed an API it calls, so patches/juce-7.0.12-macos15.patch is applied automatically at fetch time. Without it the build fails on Xcode 16+.
  • Using your own JUCE. Drop a checkout (or symlink) at ./JUCE and it will be used instead of fetching. Override the fetched version with -DAMPSIM_JUCE_TAG=<tag>.
  • Xcode project: cmake -B build -G Xcode.

Validate the Audio Unit:

auval -v aufx Amp1 Frnd

Releasing

scripts/sign-and-notarise.sh signs with a Developer ID, submits to Apple for notarisation, staples the ticket to both bundles and repackages. CI runs it automatically on a v* tag when the signing secrets are present, and falls back to an unsigned zip when they are not, so forks still build.

A build you make yourself is not quarantined and needs no signing to use locally — this only matters for builds other people download.

Parameters

Knob Notes
Amp / Cab / Mic independent model selectors
Gain preamp drive
Bass / Mid / Treble tone stack (topology per amp)
Cut AC30 top-cut (greyed for other amps)
Presence / Resonance power-amp HF shelf / LF bump
Sweetness macro for harmonic character: morphs even-harmonic asymmetry, cab bell-peak lift, and class-A bloom together (5 = neutral, higher = more chime/shimmer)
Master power-amp drive into sag
Reverb onboard amp reverb (Twin / Champ / Acoustic)
Body / De-Quack / Comp / Colour Acoustic amp's piezo imager (greyed for other amps)
Blend / Split / Grit / Sub bass amps' drive front end (shares the panel row with the acoustic controls)
Tape / Flutter tape colour, works with any amp; 0 = bypassed bit-exactly
Crush / SR / Noise / Narrow lofi degradation, any amp; 0 = bypassed bit-exactly
Distance / Axis / Angle / Room continuous mic placement
Output / Mix / Bypass level, wet/dry, bypass
Pedals (8) on / placement / 3-4 knobs each, on the PEDALS tab

Known v1 limitations

  • Tone stack is a voiced-biquad approximation, not a discretized passive RC network.
  • No true tube hysteresis / state-space circuit model.
  • Spring/room reverbs are JUCE-reverb approximations, not physical models.
  • Tone/placement filters recompute per block (no per-sample smoothing yet), so very fast automation of EQ-type knobs can zipper.

Licence

GPL-3.0-or-later. See LICENSE.

This project links against JUCE, which is dual-licensed. JUCE 7's own terms name the GPLv3 as the alternative to a paid tier, which is why this project is GPLv3: it means the plugin can be distributed freely, with source, and without the JUCE splash screen. Anything you build on top of it has to be GPLv3 too. The VST3 interfaces are Steinberg's, under their GPLv3-compatible terms.

Credits and references

The tape stage is adapted from a sibling tape_emulator project by the same author: derivative-normalised saturation, pre/de-emphasis around the shaper, programme-dependent drive, head bump as a resonance rather than a shelf, and gap-loss high-frequency rolloff.

The voicings are informed by manufacturer documentation and service manuals (Ampeg SVT and B-15 manuals, the Acoustic 360/361 service manual, AER and Fishman documentation), by published measurements where they exist, and by the schematics of the pedals being modelled. Where no measurement exists — nobody has ever published a frequency response for a B-15, and there is no anechoic data for any bass guitar cabinet — the values are informed approximations, and the source comments say so.

A note on names

The amps, cabs, mics and presets carry invented names. Each voicing is modelled on a classic design, and the source comments name those designs openly, because that is what the numbers in them refer to and hiding it would make the code harder to follow. Nothing here is affiliated with, endorsed by or connected to any manufacturer, and all trademarks belong to their respective owners.

About

Guitar, bass and acoustic amp simulator with a drag-to-reorder pedalboard. AU + VST3 + Standalone for macOS, built with JUCE.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages