Log — Music-Tech Wiki
Append-only history. Each entry starts with ## [YYYY-MM-DD] <op> | <title> where
<op> is ingest, query, lint, or split, so grep "^## \[" log.md | tail -5 works.
[2026-07-19] ingest | Analog vs digital guitar pedals (Guitar World, hub-routed, URL-only, T3)
Spoke's first guitar pedal / stompbox source and first treatment of the analog-vs-digital argument in effects (prior coverage was synths). Continuous circuit vs ADC→DSP→DAC; analog leans OD/fuzz (circuit nonlinearity = tone), digital leans delay/reverb/complex modulation (memory + math), hybrids blend both. Created source summary analog-vs-digital-pedals (Article, url-only, T3) + 3 Thing pages: guitar-pedal (DefinedTerm, effect-class — hardware cousin of audio-effect-plugin) and analog-vs-digital (DefinedTerm — the founding through-line promoted to its own node now that it spans instruments + effects), plus entity guitar-world (NewsMediaOrganization publisher; guitar-facing sibling of synth-anatomy/ synthtopia). Folded an "argument leaves the instrument, lands on the effect" extension into synthesis §2. No new runner-up (single clear match). Guitar World = introductory gear journalism, qualitative claims → T3.
[2026-07-09] ingest | Tasty Chips GR-MEGA — digital granular synth, Firmware 1.7 (hub-routed)
gr-mega (T2, gearnews) — hardware granular synth & sampler (€1,500+VAT): 4 layers, 20 voices/layer, 5,000 grains, 5 engines (sampling/granular/tape/granular-slice/spectral-phase-vocoder), 20→100 mod matrix. FW 1.7 reworks the granular engine (grains across all layers, pan/pitch/patch spray, pitch quantization w/ user scales, granular filter, multiband resonator). Value: the spoke's first clearly digital/DSP instrument — anchors the digital pole of the analog-vs-digital thesis with an instrument, not a plugin, and reframes it from "which is better" to "digital enables sounds analog can't" (new concept granular-synthesis). Also surfaced two cross-instrument motifs: "spray" randomness (also in beast-repeat) and user-scale microtuning (also in cyma-forma-alt). Maker paged tasty-chips. Updated synthesizer (analog↔digital
- synthesis-method axes) + synthesis thesis #2. +3 pages.
[2026-07-09] ingest | Arturia KeyLab mk3 "Ultra Orange" — MIDI controller (hub-routed, thin)
keylab-mk3 (T3, mixdownmag) — thin source: a colorway launch of an existing MIDI keyboard controller, no functional change. Ingested per the never-drop rule, kept mainly because it opens two new nodes: the midi-controller category (control surface, generates no sound — the input end of the chain, distinct from synth/effect) and arturia (French maker; V Collection/Analog Lab software + KeyLab hardware). Adds a fourth answer to synthesis "who moves the controls" (dedicated control surface) and a new bundle business model (hardware sold with software instruments) vs the free-plugin pattern. Thinness recorded; controller/DAW- integration flagged as a growth edge.
[2026-07-09] ingest | Beast Repeat — free Max for Live glitch/beat-repeat device (hub-routed)
beast-repeat (T2, synthanatomy) — dillon-bastan's free M4L rebuild/expansion of Ableton's stock Beat
Repeat: chance triggering, pitch-shift, granular/phase-vocoder stretch, per-repeat decay/width/fades, 5 built-in
effects, 8-slot snapshots. The spoke's first Max for Live source — a new format axis distinct from
standalone VST3/AU (shinronin, next-oxx): host-locked to Ableton, deeply integrated, extends a stock
device. Extended audio-effect-plugin (M4L format + 3rd free effect) and synthesis thesis #4. Maker paged as
dillon-bastan (Person — individual M4L author, vs studio/label). New ## Person group in index.
[2026-07-09] ingest | Cyma Forma ALT — sensor-driven analog soundscape synth (hub-routed)
cyma-forma-alt (T2, gearnews) — French all-analog synth for drones/soundscapes/generative melodies. Distinctive on three fresh axes: sensor control (mic = "ears", light sensors = "eyes"/theremin), microtuning (14 scales incl. just intonation across 5 oscillators), and a pin-matrix interface. Two MS-20-style multimode filters in series — the direct structural opposite of filtopia's parallel five-filter bank, which sharpened the "filter as identity" sub-thesis. Maker paged as cyma-forma. Wove into modular-routing (pin matrix = 4th instance), synthesizer (new mono/poly, filter, control, tuning axes), and synthesis (new thesis #3: the synth you don't play by keys — sensor/generative/microtuning). First source to push the spoke past note-in/sound-out toward ambient/generative + alt-control gear (Eurorack edge).
[2026-07-09] split | Spun out from hub _inbox cluster electronic-instruments (broadened → music-tech)
Founded from a same-session burst of five hub-routed sources. Two (polyUAnalog, Filtopia) had been parked in
_inbox as cluster electronic-instruments; three more arrived together (ShinRonin, MusicRadar
test-equipment, Next OXX), tripping the ≥3 spin-out trigger. The cluster was broadened from
"electronic-instruments" (too narrow — excludes software effects) to music-tech: sound creation &
processing across hardware synths, DIY/open-hardware, effect plugins, and technique.
Ingested (all URL-only, no raw docs):
- filtopia (T2) — Analogue Solutions all-analog mono synth, 5-filter parallel filterbank. Two write-ups (synthanatomy + gearnews) folded into one page.
- polyuanalog (T1) — University of Angers open-source (MIT) analog poly synth, conductor–voice, DCO.
- shinronin (T2) — Audio Damage free modular delay/filter/saturation plugin (VST3/AU/CLAP/AUv3).
- next-oxx (T2) — Nextlight free OTT-style multiband compressor, Windows-only VST3.
- music-from-test-equipment (T2) — MusicRadar feature on lab/test-gear music ("better sine waves").
Concept nodes: synthesizer, modular-routing, audio-effect-plugin. Maker/origin orgs:
analogue-solutions, audio-damage, physics-dept-angers. Synthesis seeded with three theses
(routing-over-components; the analog-vs-digital sound argument; free-plugins-as-audience-building) and two
open tensions. Boundaries drawn vs ../loudspeaker-design-wiki (reproduction) and ../speech-audio-wiki
(audio AI models). _inbox/polyuanalog-synth.md and _inbox/filtopia-synth.md removed (now owned here).
Eurorack flagged as the next growth edge. Registered in hub wikis.md; hub log.md split entry added.
[2026-07-09] ingest | KVR Developer Challenge 2026 free synths (BPB roundup) (hub-routed, Telegram)
Ingested the BPB roundup (James Nugent, bedroomproducersblog.com) of five free software synths from the KVR Developer Challenge 2026 (50+ freebies; winners 4 Aug 2026). New pages: source summary kvr-dev-challenge-2026-free-synths (BlogPosting, T2); Event kvr-developer-challenge; five synth SoftwareApplication pages — tetraop (Tilr, 4-op FM/wavetable hybrid), ouranos7 (John Crosby Acoustics, 6-op FM/DX7 homage), identity-synth (SocaLabs, formula-oscillator poly), double-freak (Bjango, audio-following pitch-tracker), the-hydra (bdEnergy, additive); two new synthesis-method DefinedTerms fm-synthesis + additive-synthesis (siblings of granular-synthesis); five maker Organizations (tilr/john-crosby-acoustics/socalabs/bjango/bdenergy); publisher bedroom-producers-blog (NewsMediaOrganization, also covers the earlier next-oxx). Synthesis: new thesis #5 (free-software synth pole; synthesis-method as the axis of difference; counter-pole to boutique hardware) + sources updated. All T2 — press roundup, feature lists from devs, subjective sound remarks, standings unresolved. avoid-ai-writing applied (clean; house-style em-dashes kept). +14 pages.
[2026-07-10] ingest | OXI Instruments E16 performance MIDI controller (FW 1.0) (hub-routed, Telegram)
Ingested the synthanatomy launch write-up of the OXI E16: a 16-endless-encoder performance MIDI controller (192 mappable CCs across 16 scenes × 12 pages, encoder grouping, snapshot morphing, knob recording; dual USB-C, TRS MIDI, Bluetooth MIDI; €449). Firmware 1.0 adds an 8-min/16-track MIDI parameter looper (3 sync modes) + Lua scripting API. New oxi-e16 (Product, T2) + maker oxi-instruments (Organization; OXI One sequencer house). Updated midi-controller concept to a keys-vs-knobs split (keyboard/bundled-software keylab-mk3 vs standalone-premium knob-box E16 that records and replays control gestures → sequencer-adjacent). Synthesis thesis #3 ("who moves the controls") extended. T2 vendor specs, no hands-on. avoid-ai-writing applied (clean). +2 pages, 2 updated.
[2026-07-12] ingest | Akai MPC One G2 standalone sampler (RA news, hub-routed, Telegram)
Ingested the RA (ra.co) announcement of the Akai MPC One G2, the 2nd-gen standalone MPC sampler/groovebox (8-core "4× perf", 4GB RAM / 64GB storage, MPC 3.9 OS with a DAW-like linear arranger, 32 plugin instruments + 16 stereo audio tracks, 24-in/24-out over USB-C, built-in synth engine, Ableton Live Control, 20GB+ bundled sounds incl. NI Analog Dreams MPC Edition; blue MPC1000/4000-nod finish). $799 / MPC Key 37 G2 $999; released 12 Jul 2026. New pages: Product/source summary akai-mpc-one-g2 (T2); maker Organization akai (first workstation vendor here); new instrument-class DefinedTerm sampler (sampler↔groovebox, the sampler↔synth boundary via granular-synthesis). Synthesis: new thesis #6 — the workstation: integration/workflow as the value, not a single sound-shaping trick (host not just sound source; standalone hardware migrating toward the computer-DAW arranger workflow); resolved the "sampling/DAW cluster" open question — union scope holds, MPC is the widest single instrument yet. T2 vendor-announcement specs, no independent measurement. avoid-ai-writing applied. +3 pages, 3 spine files updated.
[2026-07-13] ingest | Xnullx: five free sound-design plugins (Synth Anatomy, hub-routed, Telegram)
Ingested Synth Anatomy's coverage of Xnullx, a brand-new dev that shipped five free sound-design plugins at once (VST3+AU, Mac+Win): Grainbrain (multiband granular), O2 (three-band saturation+width), Bunka (beat-slicer sampler, MPC-style), Foldspace (phase-mod + wavefolding), Halide (granular freeze/spectral-smear). The load-bearing angle isn't a single plugin — it's that the reviewer suspects the plugins are AI-generated (generic site/UIs, all-at-once drop; reader reports of install bugs + CPU-pegging site). New pages: source summary xnullx-five-free-plugins (BlogPosting, T2); maker xnullx (Organization, watch node); publisher synth-anatomy (NewsMediaOrganization — was an unpaged recurring source: Filtopia/ShinRonin/OXI E16); new method DefinedTerm wavefolding (5th node on the synthesis-method axis, West-Coast waveshaping, ships paired with phase modulation). Decision: did NOT page the five plugins individually — held for evidence given the AI-shovelware doubt; techniques cross-linked to their method pages instead. Updated audio-effect-plugin (+batch-release/AI-generation wrinkle), granular-synthesis (granular-as-effect: Grainbrain per-band, Halide spectral smear). Synthesis: extended thesis #4 with the free-plugin shadow (machine-generated shovelware; the meta-irony that the hub itself is LLM-run), added wavefolding to the thesis #5 method fan-out. avoid-ai-writing applied (clean). +4 pages, 3 updated.
[2026-07-14] ingest | M-VAVE FM-1 pocket FM synthesizer (Synthtopia, hub-routed, Telegram)
Ingested Synthtopia's announcement of the M-VAVE FM-1, a ~$75–100 pocket FM synth: six-operator, 32-algorithm engine (= the Yamaha DX7 spec), 12-voice, 27-key silicone keyboard, 7 knobs + TFT, built-in arp + step sequencer, multi-FX, presets; triple simultaneous MIDI (3.5mm TRS + USB-C + Bluetooth), built-in speaker + 3.5mm out, 2000mAh / ~12h battery, Win/mac/iOS/Android. New pages: Product/source summary m-vave-fm-1 (T2); maker m-vave (Organization, budget-gear pole); publisher synthtopia (NewsMediaOrganization — recurring but previously unpaged). Updated fm-synthesis (first hardware FM instance; the DX7 6-op/32-algo config as a commodity target hit at both free-software and budget-hardware ends). Synthesis: added a third economic point to thesis #5 — cheap hardware between boutique-expensive and free-software, splitting price from format; spoke's first hardware FM + first pocket/portable instrument; triple-MIDI blurs it toward the controller role. avoid-ai-writing applied. +3 pages, 2 updated. T2 vendor announcement, no hands-on.
[2026-07-14] ingest | Korg NTS-4 DIY performance mixer (AudioTechnology, hub-routed, Telegram)
Ingested AudioTechnology's announcement of the Korg NTS-4, a DIY self-assembly six-channel analog performance mixer kit (4th in the Nu:Tekt line): 4 stereo + 2 mono 3.5mm inputs (mono pair Eurorack-level with switchable attenuator), per-channel CUE/MUTE, per-channel SEND FX (7 options) + master TOTAL FX (10 options), stereo SEND/RETURN, MAIN L/R on ¼", USB-C audio (48k/24-bit) + bus power, 3.5mm TRS MIDI OUT, steel chassis, 195×125×38mm. (Fetch: WebFetch 403-blocked → firecrawl recovered full text per HUB edge rule; not a transient error, no stub needed.) New pages: Product/source korg-nts-4 (T2); maker korg (Organization, Nu:Tekt DIY line); new instrument-class DefinedTerm mixer (routing/summing with no sound source); publisher audiotechnology (NewsMediaOrganization, thin). Updated modular-routing (+mixer as routing-as-the-whole-product limit case; Eurorack edge touched). Synthesis: extended thesis #1 (mixer = purest routing instance; performance CUE/MUTE ties to thesis #3; 2nd DIY device after polyuanalog; analog path + digital FX hybrid); updated the Eurorack open question (touched from the edge, not crossed) and added a new scope question (is a pure-infrastructure mixer in-domain, or studio-signal-chain drift?). avoid-ai-writing applied. +4 pages, 2 updated. T2, no hands-on.
[2026-07-17] ingest | Reason 14 review (MusicTech, hub-routed, Telegram)
Ingested MusicTech's 8/10 review of Reason 14 by Hollin Jones — an experienced Cubase user reviewing Reason Studios' 25-year-old DAW from the outside. The spoke's first proper software DAW — the exact case the synthesis open question named ("whether Eurorack or a DAW proper lands"). Reason's identity is the virtual rack you patch with cables (Tab to flip it) + the Combinator; ships in two forms — a standalone DAW and a rack plugin (Reason Rack) that runs inside a rival DAW, sans sequencer/mixer; 68 devices, 50k+ patches, VST host; £299 Reason+ / £199 Rack (perpetual) + monthly tiers the reviewer finds "redundant." New pages: source summary reason-14-review (Review, T2); Product/SoftwareApplication reason; new DefinedTerm daw (software-class); maker reason-studios (Organization); publisher musictech (NewsMediaOrganization); author hollin-jones (Person); comparison node cubase (SoftwareApplication — the review's yardstick, Cubase Artist £273). Updated modular-routing (Reason = routing at production scale, biggest instance) and index. Synthesis: added thesis #7 (the DAW: integration in software + routing thesis at production scale + the dual-form feature-pressure valve); answered the DAW half of the scope open question (Eurorack remains the open edge). avoid-ai-writing applied. +7 pages, 2 updated. T2 — single named practitioner's hands-on review; counts/prices are the vendor's.
[2026-07-17] ingest | HiChord vs Orchid ORC-1 vs Minichord — portable chord synths (Piano & Synth Magazine, hub-routed, Telegram)
Ingested a Piano & Synth Magazine three-way comparison that opens a new instrument class, chord-synthesizer (chord buttons + strum surface, Omnichord lineage; organized by how you address harmony, not how sound is made — the sharpest "who moves the controls" answer yet: a non-pianist plays because the instrument encodes the chords). New pages: source chord-synths-compared (Review, T2, url); instrument-class chord-synthesizer; three Products — hichord ($320, 7-button + looper/drums/sampler/ sequencer, "mini groovebox" = workstation corner), orchid-orc-1 ($549, voice-leading engine, 16-voice, VA/FM/reed, Kevin Parker design input = harmonic-intelligence corner), minichord ($169, open-source via Seeed Studio, USB-MIDI out = DIY corner); publisher piano-and-synth-magazine (NewsMediaOrganization); kevin-parker (Person, Tame Impala). Updated midi-controller (nearest relative — interface-as-point but makes its own sound; Minichord straddles via MIDI-out). Synthesis: added thesis #8 — the chord synth as an interface-to-harmony class, triangulated across the spoke's existing axes (workstation #6 / DIY / synthesis-method + artist-endorsement), evidence the category is real not a gimmick. avoid-ai-writing applied. +7 pages, 2 updated. T2 editorial comparison; specs/prices are makers'.
[2026-07-22] ingest | RackDroid — VCV Rack on Android (GitHub, hub-routed, Telegram)
Ingested rackdroid (SoftwareSourceCode, T2, GPLv3): a touch-optimized Android port of
VCV Rack 2 — 66 built-in modules, finger-dragged patch cables, pinch zoom, long-press numeric
knob entry, Oboe/AAudio full-duplex low-latency audio, USB + BLE MIDI, WAV recording, sideloadable
.rdmod packages (Bogaudio, Valley, Befaco). Android 10+/arm64/GLES3, ~40 MB APK, 30 tutorials. Adoption
caveat recorded honestly: 9★ / 1 fork, nobody in the sources has run it.
This crosses the spoke's longest-standing open edge — Eurorack — but not from the expected direction. The corpus reached the modular thread three times without crossing it: modular-routing named Eurorack as the growth edge, korg-nts-4 touched it from outside as an integration point, reason scaled the rack metaphor to a whole DAW. VCV Rack is the thing itself (any-developer modules, CV and audio as the same cable, no fixed path, the patch as the instrument) — and it entered as free software on a phone rather than as hardware. New Thing page vcv-rack holds the platform; RackDroid is the source that reveals it.
Analytical payoff recorded in two places. modular-routing gained an "unbounded case" section: every prior form exposed routing inside something bounded (one instrument / effect / mixer / DAW catalogue), Rack removes the boundary, making it the strongest instance of the page's thesis. Synthesis thesis #1 took the same point plus the split it forces — modularity is architectural, analog is a signal-path property; the corpus had only ever met them together, and a virtual rack keeps the first while dropping the second, so it works as a control on thesis #2 (whatever survives in software wasn't coming from the circuits). Thesis #5's free-software pole extended from single instruments to a whole free platform (free infrastructure hosting other people's instruments), widening the boutique gap from "one synth vs zero" to "a Eurorack case vs a free APK", and pairing with m-vave-fm-1 on the phone-as-instrument axis.
Two port-engineering details worth keeping: platform code isolated in native/port/ over unmodified Rack
v2.6.4 so the port tracks upstream (its value depends on desktop modules continuing to work), and the
graphics rebuilt under GPLv3 to make distribution possible — a modular platform's modules are their
panels, so faceplate artwork is a separately-licensed asset that can bind redistribution even when the DSP is
free. First time licensing-of-visual-design shows up in this spoke.
Open question updated: Eurorack half-answered — architecture in the corpus, hardware/economics/panels still out; Eurorack hardware is now the sharper open edge because there's finally a software control to compare it against. Entity: nowheel (thin — repo carries no biographical detail). avoid-ai-writing self-pass. +3 pages, 1 page updated, 3 spine files.
[2026-07-23] ingest | Behringer CS Mini — 3-voice CS-80-inspired analog mini synth (Synth Anatomy)
Routed from the hub (route in ../log.md). URL-only, T2 — synth-anatomy recurring editorial source,
but relaying a pre-release leak (spotted in Behringer's Synthtribe app v3.2.2; no official launch, ~$99
estimate). Specs volatile/unconfirmed.
New pages: behringer-cs-mini (Product, source) and behringer (Organization). CS Mini = 3-voice
paraphonic analog synth inspired by the Yamaha CS-80: 3 osc (saw/sq/sine) + sub w/PWM + noise +
ring-mod, resonant LPF + resonant/modulatable HPF (the CS-80 twin-filter nod), ADSR + AR-as-LFO, touch
keys, MIDI, USB-C, motion sequencer.
Dedup: no Behringer/CS-80 page existed. Did not page the Yamaha CS-80 — it's the inspiration, one passing
mention; kept inline (thin-node caution), page it if a second CS-80 reference lands.
Gap-relevance: adds a new pole to the analog-vs-digital thread — the budget mass-market analog clone,
where every prior analog instrument was boutique (filtopia £1,799, cyma-forma-alt €1,279) or DIY
(korg/polyuanalog). Synthesis: added a paragraph — analog moves to the cheap end, complicating the
"better sine waves" purism's economics; a third maker-type (mass-market reissue) beside boutique/DIY; and the
paraphonic≠polyphonic asterisk on the synthesizer voice model (shared filter/amp path = cheap "3
voices," unlike the fully-poly CS-80).
Entities: behringer created (Music Tribe budget-clone maker; Synthtribe pipeline). Author "Tom" (Synth
Anatomy) not paged (byline; the outlet is the paged node).
Verify deferred per hub policy (content-only). avoid-ai-writing run.
[2026-07-25] ingest | Miltone 4Exp — Oberheim Four Voice System recreation (DJ Mag)
Routed from the hub (route entry in ../log.md; runner-up: none — an analog hardware synth is squarely
this spoke). URL-only, T3: announcement relay from a DJ/club outlet, no specs, no hands-on, delivery a
year out.
New pages: miltone-4exp (Product, source: true), vintage-recreation (DefinedTerm — new axis),
plus entities miltone, oberheim, manfred-veber, dj-mag.
Dedup: grep for oberheim|miltone|djmag|reissue|recreation|clone across wiki/ → only
behringer-cs-mini / behringer. entity-index for Oberheim / Miltone / DJ Mag → NO MATCH, created new.
Gap-relevance: the payoff is the pairing. behringer-cs-mini (2026-07-23) and this (2026-07-25) both
recreate a five-figure 1970s analog polysynth and disagree about what recreation owes the original —
Behringer keeps the idea and deletes the price (paying in paraphonic architecture), Miltone keeps the
circuit and the price and sells fidelity. That's a distinct axis from analog-vs-digital, so it got its own
node, vintage-recreation, which also absorbs the DX7-spec homages (ouranos7, m-vave-fm-1) and
minichord's Omnichord lift that had been sitting unconnected. Synthesis §2 extended with both the
pairing and a third reading of "analog" — archival, where restoration expertise is the manufacturing
qualification.
Weakness recorded on the page in its own section: the source gives no voice count, oscillators, filters,
envelopes, programmer, MIDI or CV — only "analog." Flagged for a first-party or hands-on refresh.
No contradictions. Verify deferred per hub policy (content-only, no page moves). avoid-ai-writing run.
[2026-07-26] ingest | 5 of the Best Software and Hardware Trackers (Attack Magazine)
Routed from the hub (Telegram; URL arrived with Facebook fbclid/sfnsn tracking params — canonical URL
stored). T2 — Adam Douglas, 2026-07-22; a buyer's guide, so specs and prices are the makers' and there
is no hands-on. Prices dated as a 2026-07 snapshot.
New pages: attack-mag-5-trackers (Review source), tracker (DefinedTerm, interface-class),
renoise + sunvox (comparison nodes), attack-magazine + adam-douglas (entities).
The five, inline on the source page: SunVox (free / $5 iOS), TRX-12 ($29.99, 35 step-FX commands), Renoise
($88; Redux $56, v3.5, market leader since 2002), Dirtywave M8 Model:02 ($699, 8 mono tracks, 255 patterns),
Polyend Tracker+ ($999; Mini $799 with battery + field mic). History per the source: Ultimate Soundtracker
(1987, Amiga) first, OctaMED most popular, jungle/breakcore the adopting audience.
Only two of the five got their own page, and for reasons outside the roundup: renoise because daw
was explicitly waiting for DAW-vs-DAW comparisons, sunvox because it's a modular-routing instrument
that arrived through its sequencer. TRX-12 / M8 / Tracker+ stay in the source table — one line each, no
node-level role — per the xnullx-five-free-plugins precedent for roundups.
Updated daw: new section — three DAWs, three layouts (tracks-and-mixer / virtual rack / typed grid), so
the category splits on layout, an axis the Cubase depth-yardstick can't see. Updated sampler: a
tracker corner of the groovebox class — M8 and Tracker+ pass the integration test with a typed grid instead
of pads-and-steps, which shows the MPC layout is a choice inside the category, not its
definition.
Synthesis: new §9 — the tracker as a class defined by how you enter time, the second interface-defined
class after chord-synthesizer; the source's causal claim (per-line editing is cheap → jungle/IDM
gestures became idiomatic) reverses the usual instrument-serves-sound direction and gives thesis #3 a fourth
answer to "who moves the controls": the person typing. Two open questions added — whether the spoke's real
taxonomy is sound-source × control-surface (testable against the next arrivals), and whether tracker
hardware carries MIDI/CV (the source is silent, and modular-routing cares).
Dedup: no tracker/Renoise/SunVox/Attack Magazine page existed anywhere in the hub. grep across all spokes
for Attack Magazine / Adam Douglas / Alexander Zolotov → no match, created new.
Weakness recorded on the page: no hands-on, no losers, nothing on MIDI/CV, file interchange, or the cost of
memorizing hex effect commands.
No contradictions. Verify deferred per hub policy (content-only, no page moves). avoid-ai-writing run.
[2026-07-26] ingest | remodevicocomposer.eu — Remo De Vico, composer and sound designer
Routed from the hub (Telegram; Facebook fbclid stripped). T1 — first-party site, authoritative for
what he publishes and who he is. Took four fetches: the landing page reads as a portfolio, and the routable
substance only appears once you follow the nav (/maxforlive/, /pure-data/, /musical-mirror/).
New pages: remo-de-vico-site (WebSite source — a new @type section in the index),
remo-de-vico (Person), patching-environment (DefinedTerm, software-class).
Core facts: 13 M4L devices named on page 1 of 8 (Tetraligeti, Videodrums, Loop Paradox, Collapse,
Vectordrone, Oblique Strategies, 4Samples, Sweetspace, Generative, Glitch Loop, Iceberg, Broken Sphere,
Irrational Delay); two free Pure Data patches (Granular Philosophy + LFO for Granular) distributed to run
as DAW plugins via plugdata; sample packs; a "Musical Mirror" essay series (Raymond Scott, Ligeti, Moog,
Daphne Oram, Tomita, vector synthesis, generative music, rights organizations). Training: La Sapienza
(thesis on Luciano Berio), Cosenza Conservatory, Santa Cecilia under Giorgio Nottoli, masterclasses
with Francis Dhomont and Mario Bertoncini — an acousmatic/electroacoustic lineage, a first here.
One concept node rather than two: patching-environment covers Max/MSP and Pure Data together (same
lineage, same author) with M4L vs plugdata as the two DAW-wrapper routes inside it. Splitting them would
have duplicated the argument, and the argument is what's new — a Max or Pd patch wires the primitives a
vcv-rack module is made of, so the routing surface and the programming surface are the same thing.
Updated modular-routing (new section: routing as the programming language — the thesis below the
instrument layer), audio-effect-plugin (plugdata as the host-portable counterpart to host-locked M4L).
Synthesis §4 gains a third read of "free": not a strategy at all. No branding, no pricing, no docs — the
devices are byproducts of scoring, published because a patch costs nothing to copy. That completes a
five-point spectrum (studio reviving IP / label seeding producers / developer courting attention / suspected
machine-generated batch / composer emptying his tools folder), and it complicates the xnullx suspicion:
a large undocumented all-at-once drop read as the tell for machine generation, and this is a human one with
the same shape.
Dedup: no Remo De Vico, Max-for-Live, Pure Data or plugdata page existed; M4L had only been described inline
on beast-repeat / dillon-bastan, which is why a second M4L source earned the concept node.
Weakness recorded in its own section: no device is documented — names and download links only, no
descriptions, prices, or Live/Max version requirements. Everything about what an individual device does is
inference from its name, and flagged as such. Whether the M4L devices are free is not stated; only the
two Pd patches are explicitly free.
Entities: remo-de-vico only. His teachers (Nottoli, Dhomont, Bertoncini, Berio, Rostagno) are evidenced
but tangential under the ENTITIES relevance gate — biography, not spoke domain — so named in prose, not paged.
No contradictions. Verify deferred per hub policy (content-only, no page moves). avoid-ai-writing run.
[2026-07-28] ingest | Hermetic Modular Alchemy Lab — open-source DSP module platform (synthanatomy.com)
Routed here by the hub (runner-up: ../embedded-iot-wiki — it's an STM32H750 board, but that spoke's own
rule routes audio hardware by purpose, and a Eurorack module's purpose is music). T2 — Synth
Anatomy's usual editorial news piece, not a hands-on; every spec is the maker's and nothing is measured.
New: hermetic-alchemy-lab (source), daisy (platform), hermetic-modular (Organization).
Updated: modular-routing, analog-vs-digital, synthesis (Eurorack question closed + new
thesis note), index. Reused synth-anatomy — already paged, not duplicated.
This closes the spoke's standing Eurorack open question. VCV Rack / RackDroid had crossed the format in software on 2026-07-22, leaving hardware, panels and economics explicitly open. All three land: 12HP, skiff-friendly, $299 introductory / $389 MSRP, 6 pots + 3 buttons + 10 jacks + 102 addressable RGB LEDs on a physical panel.
Three findings. (1) It splits the routing thesis. modular-routing says character lives in how blocks are wired. The Alchemy Lab keeps the format and moves the wiring inside: the panel is fixed, the signal path is firmware, and even which CV a jack carries is an SDK decision — the same 12HP face becomes ECHOA (four delay lines per channel, 1 ms–16 s, digital/BBD/tape/vinyl degradation, Schroeder reverb via four-stage all-pass) or SPAGYROS (three-slot performance chain). Set against vcv-rack it's the exact mirror: VCV is unbounded patchable topology with no panel, this is a real panel with a reloadable topology. The physical patch and the reconfigurable architecture come apart — Eurorack had always shipped them together, and each turns out to be available without the other. (2) One board under several brands. The article's comparison set is the finding: Noise Engineering's Versio/Alia/Legio and Bastl's Kastle 2 "are also based on the Daisy." Four product lines, one commodity Electrosmith board. That's analog-vs-digital read at the business layer — in analog the circuit is the product (filtopia's filter bank can't be dropped into someone else's panel), in digital the compute is fungible and identity moves up into panel, jacks and DSP code. Paged daisy as its own node because it's a hub with four evidenced product lines hanging off it, not a passing spec. (3) The open-source claim is narrower than the headline. Quoted: "Every hardware abstraction layer of the Alchemy Lab module is open-source" — that's the HAL. The article confirms an SDK exists but states no license, no language, nothing about whether hardware design files are published, and nothing about how firmware gets loaded. Recorded as a gap, not smoothed into "open-source platform," because this spoke's existing open-source corner is specific (polyuanalog MIT with board files, minichord publishing hardware) and HAL-only would be materially weaker than either. Needs the project's repo. Residual gaps noted in synthesis: the economics arrived as one price point, not a market, and a purely analog Eurorack module — the case where circuit and character can't be separated — is still unsourced. Verify deferred per hub policy (content-only). avoid-ai-writing run.
[2026-08-04] lint | re-tiered on what pages rest on, not who published them
The curator's resolution of the standing question this spoke had carried since 2026-08-03 and the hub
had re-deferred twice: 23 pages graded T2 on the strength of enthusiast/trade mastheads, where a
literal reading of ../QUALITY.md put them at T4 (trade press).
Neither answer was taken. The ladder gained a rule instead — tier the page by what it rests on: someone tested, measured or compared it → T2; the maker's specs and price relayed → T3, because a press release passed through a masthead is the vendor speaking, which is what T3 already describes; filler → T4; a project's own repo stays T1 whoever links to it.
Result across the 23: 16 → T3, 6 stay T2, 1 → T1. Plus akai-mpc-one-g2 T4 → T3 (an announcement is first-party information, not filler).
Spoke tier mix: T1 2 / T2 23 / T3 2 / T4 2 → T1 3 / T2 6 / T3 19 / T4 1.
The spoke was already inconsistent with itself. keylab-mk3 (Mixdown) and miltone-4exp (DJ Mag) were sitting at T3 for the same kind of coverage a dozen pages had at T2. This pass is mostly the spoke applying its own two exceptions to the rest.
Kept at T2 because somebody evaluated something: reason-14-review (scored 8/10 by a named practitioner), chord-synths-compared and attack-mag-5-trackers (comparisons across three and five products), kvr-dev-challenge-2026-free-synths (five chosen from 50+), xnullx-five-free-plugins, music-from-test-equipment (reporting a practice, not a product).
rackdroid → T1: never press at all, a GPLv3 project's own repository, the grade polyuanalog already carried.
Publisher pages now say the grade follows the page. synth-anatomy, synthtopia, audiotechnology, attack-magazine and bedroom-producers-blog each carried a sentence asserting a default tier for the outlet; all rewritten. BPB is the cleanest illustration — its roundup stays T2 (selecting five standouts from 50+ is a judgement the developers did not make) while the five per-instrument pages it seeded moved to T3 (feature descriptions relayed). The evaluative act sits in the roundup.
What this costs, stated plainly: the floor reading goes from T2-heavy to T3-heavy. No page's content changed and no claim was withdrawn. What changed is that the tier no longer implies a press release was independently verified. Growth edge 5 closed; edge 1 (analog purism, measured) is unchanged and is now the only thing that would actually raise the floor.
One page the rule does not settle: analog-vs-digital-pedals stays the lone T4 — a Guitar World explainer resting on neither testing nor first-party information. Recorded as a genuine edge case rather than forced into a bucket.
[2026-08-05] ingest | MD-7 Pocket Synthesizer (Mach Devices store page)
Routed from the hub (Telegram, Google share link → machdevices.com). The storefront lists exactly one product, so the "Products" page and the product page are the same source. Fetched both plus the marketing sections via Firecrawl after WebFetch hit the shop's SPA shell.
New pages: md-7-pocket-synthesizer (Product, T3, first-party store copy for an unshipped device) and mach-devices (Organization).
What's new for the spoke: an instrument whose pitch is haptic feedback — a FOC motor disc under the encoder rendering detents, virtual walls and per-parameter textures. Every control surface the corpus held until now was one-way, so this opens a new synthesis question: does the control surface push back, and is that a second direction on the existing control axis rather than a new instrument class? No answer available — nothing here is tested.
What the source does not contain: any synthesis-engine description. Voices, oscillators, filters, polyphony — none of it is stated, so the page cannot say what kind of synth this is. The three ports are never identified either. Recorded as a hole, not smoothed over.
Also added to Contradictions: "open source" now appears at three different strengths in the corpus — polyuanalog (boards + code, MIT), hermetic-alchemy-lab (HAL only), and the MD-7 (promised, repository not published).
Price HK$3,130 and the late-September-2026 pre-sale date are volatile; re-check at release.
[2026-08-06] ingest | Csound (Wikipedia)
Routed here by the hub. It arrived inside a batch of ~14 programming-language sources heading for
../programming-languages-wiki, and it is the one that did not belong there: Csound is a language, but
its substance is sound synthesis, so the dominant-substance rule puts it in this spoke.
New page: csound (T2). What it adds that nothing here had: the MUSIC-N lineage back to Max Mathews at Bell Labs, and the orchestra/score split — a patch written as text and a performance written as data. Set against vcv-rack and the Eurorack pages, it is the same build-a-signal-path idea a few decades earlier and without cables. Its ~1,700 opcodes also make fm-synthesis, additive-synthesis and granular-synthesis into components rather than product identities, which is the opposite of how every instrument page in this wiki is organized.
Held back: version 6.18.1 dates from November 2022, so the "actively maintained" claim rests on the GitHub pointer rather than on a recent release. Language-design angle noted as cross-spoke adjacency, not duplicated.
[2026-08-06] ingest | Pure Data (pure-data/pure-data)
Routed here by the hub in the same batch as csound — both are programming languages for sound, and both land in this spoke because sound is the substance.
New page: pure-data (T1, project repo). The pairing is what makes it worth having. Pd draws the instrument and plays it live; csound writes the instrument as text and the performance as a score. Between them the spoke now covers both ways a synth can be written down, and both descend from the Bell Labs / MIT computer-music line. Pd's canvas is also modular-routing taken literally — the cable carries the program's control flow, not just audio — which puts it next to vcv-rack in form and a long way from it in purpose.
Puckette's authorship of Max connects this to beast-repeat (a Max for Live device) and so to mainstream production tooling. Noted as context: the repository page does not state it.
Held back: no version number and no release cadence on the page, so currency rests on the commit count (7,866) rather than on a dated release. Licence is "custom, see LICENSE.txt" — not read.
[2026-08-08] ingest | Two AES papers on digital transparency — the oldest edge, closed and contested
Research pass, hub ## Most wanted edge #9 ("analog purism, measured" — the spoke's oldest open
question). Two T1 peer-reviewed AES sources, and they disagree.
meyer-moran-adda-loop — Meyer & Moran, JAES 55(9), Sept 2007. A CD-standard 16-bit/44.1 kHz A/D/A loop switched in and out of high-resolution playback, double-blind, over a year, on professional monitors and electrostatic systems, with engineers, students and audiophiles listening. Undetectable at normal-to-loud levels by any subject on any system.
digital-filter-audibility — Stuart & Craven, AES convention paper. Isolates the anti-alias and mastering filters and finds listeners are sensitive to them; concludes some audible signals cannot be encoded transparently by a standard CD.
Both kept, written into ## Contradictions / tensions rather than resolved — and they are
reconcilable: a system can be transparent in ordinary use while one component is discriminable under
conditions built to expose it. Recorded position: transparent in practice, not provably transparent
in principle. Interest declared on the page — Stuart & Craven's side is the argument behind
high-resolution and MQA advocacy, which is a reason to want a third party, not to lower the tier.
The substantive result for the spoke: the analog-character claim does not come from sampling. analog-vs-digital had reasoned its way there; it now has measurement. Successor edge written for the untouched half — nobody here has measured an analog circuit against its own digital emulation.
Floor: T1 4 → 6 on a spoke whose floor has been its standing weakness.
[2026-08-10] ingest | textbeat — plaintext music sequencer and MIDI shell
Routed here from the hub (Telegram). Runner-up ../foss-applications-wiki, declined: an MIT tool a user
runs locally does touch that spoke, but the substance is sequencing and music theory, not
self-hosting. Noted on the page as a cross-spoke adjacency rather than split.
textbeat written (T1, the project's own README and repo metadata; nothing run, no audio heard).
It extends tracker in a direction the spoke had not held. Every tracker here is a program with a grid inside it — renoise, SunVox, the M8. textbeat keeps the vertical column convention and hands editing to vim, so the tracker gesture arrives inside a general text editor with macros, registers and version control applying to the score.
The finding worth keeping is the cell contents. Trackers spell notes absolutely; textbeat's cells
hold scale degrees and chord functions (1maj, 2m, b7), resolved against the current scale at
playback. A phrase is stored as a relationship, so transposition is a scale change rather than a
rewrite. That puts it beside chord-synthesizer — hichord and minichord make theory
playable by putting it under buttons; this makes it editable by writing it as text. Same idea, opposite
ergonomics.
Dormancy recorded, not smoothed over. The README calls the project "still very new" and the repo has not been pushed since 24 March 2023 — a claim that stopped being true without anyone editing it. Planned features (VST, recording) have three quiet years behind them and are not read as capabilities.
[2026-08-12] ingest | Where the warmth actually is (research pass, via hub quality cycle)
Hunted for growth edge 1's successor — the circuit half of the analog argument, which meyer-moran-adda-loop and digital-filter-audibility leave alone because both test conversion.
moog-ladder-nonlinear-digital (T1, Antti Huovilainen, Helsinki University of Technology, Proc. DAFx-04, 4 pp, read in full). The paper opens on this spoke's own argument — existing digital Moog filters "are linear and some people feel that they sound 'digital' and lack the 'warmth' that is charasteristic of analog filters" — and answers it by deriving the filter from Moog's 1965 transistor ladder rather than approximating its response.
The finding, and it is a mechanism rather than a verdict. The nonlinearity is not a flavour added to a digital filter; it falls out of the circuit. Each differential pair passes a tanh of the base-emitter voltage difference over the thermal voltage, and because the four stages are buffered the ladder discretizes as a cascade of one-pole IIR sections with the nonlinearity embedded. Hence: "there are no extra coefficients that would need to be tuned by ear — the 'warmth' is determined by the input amplitude."
Why it lands on both sides. It grants the purists' premise — the character is in the circuit's
nonlinearity, the same place analog-vs-digital-pedals puts overdrive — and denies the conclusion that
the nonlinearity is unmodellable. What is left of the digital shortfall is specific: oversampling is
mandatory for the embedded nonlinearity, and the feedback path's unit delay adds
4·pstage + 180·fc/Fs, so resonance drifts off cutoff and the 3 dB-per-stage attenuation no longer
holds, meaning resonance needs per-frequency compensation. Engineering costs with known fixes.
Edge 1 re-specified, not closed, and the reason is on the page. There is no bench measurement, no hardware in the loop and no listening test here — the validation is analytical. What the edge wants now is narrower than before: a discrimination result, an ABX or bench comparison of a ladder circuit against a model of that same circuit. Another modelling paper will not do it.
Folded into synthesis thread 2, which had been conducted entirely in adjectives until now.
[2026-08-17] refresh | textbeat (re-sent source, already paged 2026-08-10)
The same GitHub URL arrived via Telegram a second time, seven days after the first ingest. Per the edge rule this refreshes textbeat in place; no new page, and it is logged as a refresh rather than an ingest because most of the substance was already here.
Nothing moved upstream. Re-checked against the GitHub API: 443 stars, 12 forks, 1 open issue, last push 2023-03-24, not archived, MIT — identical to the first read. The snapshot stands.
Two things the first pass missed, both now on the page. The claim that "the same text transposes by
changing the scale" had no mechanism attached; key and scale are global variables (%k=3,
%k+6, %s=dorian), so a re-key is one character and the score is untouched — with the caveat that
letter names are still accepted, so the notation permits the absolute style it argues against. And
the theory library has a size: 147 chord definitions plus a fuzzy-matched General MIDI patch
list, which is the concrete backing for the "theory in the cell" argument the page makes against
ordinary trackers. Also recorded the full roadmap, including chord analysis from MIDI input —
the one item that would reverse the tool's direction, and the least likely to ship after three
dormant years.
[2026-08-20] lint | Quality cycle — two orphans connected
Both pages were reachable only from index.md, which is not a connection.
moog-ladder-nonlinear-digital now hangs off analog-vs-digital, where it belongs: Huovilainen's derivation puts a tanh per transistor differential pair at the centre of the ladder's behaviour, so the "specific nonlinear character" that page credits to analog circuitry is a describable property rather than a mystery, and the warmth tracks input amplitude. The page's standing caveat is kept and sharpened — the mechanism is now sourced and the analog-versus-its-own-model comparison still is not, because the paper publishes no bench figures and no listening test. That is growth edge 1, unmoved.
textbeat now hangs off tracker. It shares the vertical column grid and discards the application around it, and its cells hold scale degrees rather than note names, which is the one notation choice none of the trackers on that page makes.
[2026-08-22] quality | Cycle: loudness, the delivery spec the spoke never held
Coverage edge 8 closed with both halves. ebu-r128 (T1) read from the EBU PDF in its November 2023 V5 revision: Programme Loudness normalised to −23.0 LUFS, true peak ≤ −1 dBTP, gated measurement per ITU-R BS.1770 equation 7, plus the revision history (relative gate −8 → −10 LU in 2011). spotify-loudness-normalization (T3, first-party policy) for the streaming half: −14 LUFS, negative gain on hot masters with no added distortion, limiter only on the −11 "Loud" preset. loudness-normalization written as the concept page joining them.
The finding worth keeping is that the loudness war was a metering failure — the EBU says so in its own reasoning clauses — and it ended when the meter changed, not when the argument was won. That runs opposite to this spoke's usual analog-versus-digital verdict.
index.md gains a TechArticle section and three entries; synthesis.md gains a thesis section and
a successor edge asking for a measurement over real releases rather than another published target.
Entities: 0 created.
[2026-08-30] ingest | Headphonesty — "192 kHz Is Worse Than 44.1 kHz for Most Music"
Routed here by the hub router from Telegram (runner-up: ../loudspeaker-design-wiki).
Model: claude-opus-5 · effort: high. The URL arrived with Facebook fbclid and utm_*
parameters; stripped before recording. WebFetch returned 403, so the page was read with
firecrawl per ../HUB.md.
headphonesty-192khz-vs-441khz (T3 — it relays other people's studies and quotes and runs no
test of its own; the spoke's rule is to tier a page by what it rests on). No published: — the
URL path gives /2025/05/ and the page carries no byline or date, so a day would be invented.
The dedup was the interesting part: this spoke already owns the argument at primary level. meyer-moran-adda-loop (AES, T1) and digital-filter-audibility (AES, T1) were ingested 2026-08-08 as a kept contradiction, and analog-vs-digital carries the resulting position — transparent in practice, not provably transparent. So this source could not add evidence. What it adds is the popular version of the debate, six named practitioners, and one distinction the wiki had not drawn.
New concept page sample-rate, splitting the sampling question into the two halves the sources keep merging: playback, where no audible gain above 44.1 kHz has been demonstrated and the corpus's own primary is a null; and production, where oversampling headroom is genuinely needed by pitch shifting, time stretching, extreme EQ and any nonlinear processor that generates harmonics above Nyquist — most of what this spoke's plugins do. High rates are a studio tool and a store specification, and those are different questions.
Two problems with the source, recorded on its page rather than smoothed:
- It reads a null as a direction. Meyer & Moran found the conversion loop undetectable, which supports "192 kHz is pointless," not the headline's "worse." The "worse" claim rests on Montgomery's ultrasonic-intermodulation mechanism and Lavry's converter-accuracy argument — both plausible, neither measured here on real program material.
- It omits digital-filter-audibility, the AES paper pointing the other way. A reader arriving through this article would not learn the literature is split.
Also noted: Bob Katz's quoted line is about 24-bit depth, a different axis, filed under a sample-rate argument without marking the switch; "experts agree" counts six people chosen for agreeing; and the page is affiliate-monetized (disclosed) while advising readers to spend on headphones instead of file formats.
New pages: headphonesty-192khz-vs-441khz, sample-rate, monty-montgomery (same-as
Wikipedia + Wikidata Q1086856, checked), headphonesty.
Updated: analog-vs-digital (the production/distribution split), meyer-moran-adda-loop (a note
on how it is being cited downstream), digital-filter-audibility (a note on being left out of the
popular account), index.md, synthesis.md (new tension entry).
Entities deferred under the recurrence rule — named once each, in a quote list: Dan Lavry / Lavry Engineering, Bob Katz, Dan Worrall, Paul Maunder, Mark Waldrep / AIX Records. Lavry and Katz are the strongest candidates if a second source touches either.
Cross-spoke lead recorded on headphonesty: ../loudspeaker-design-wiki already names this
publisher's February 2026 gear-depreciation dataset as a source it wants, and that would be
original data rather than a roundup.
[2026-09-01] ingest | MIDI 2.0, from the standard (quality cycle)
Coverage edge 7 asked for the MIDI Association specification and the Association publishes it free.
Downloaded M2-100-U v1.1 (15 June 2023) and M2-104-UM v1.1.2 (27 October 2023) and read both with
pypdf: midi-2-0-specification (T1) for the document set and midi-2-0 for the protocol.
Two findings beyond coverage. The spoke's stepping complaint is about 7 bits, not about digital — MIDI 1.0 gives a controller 128 steps and the MIDI 2.0 Protocol gives it 32 bits with over 32,000 controllers, so the zipper artifact that gets filed under analog-vs-digital is a control-data limit with a protocol fix, and it is not evidence in the converter argument (meyer-moran-adda-loop, digital-filter-audibility). And the badge is weak: a device claims MIDI 2.0 by implementing MIDI-CI or UMP plus one feature of that branch, so two compatible devices may share nothing above MIDI 1.0.
Edge 7 closed; successor written (nothing here records which half any product implements).
Updated synthesis.md (new section), index.md. No new entity pages — AMEI and the MIDI
Association have no verified identity URL checked here, so both stay plain mentions. Budget: 0 of 8.
[2026-09-01] ingest | Orphea, a 13-key pocket electronic piano (designboom)
Routed from the hub. orphea-designboom (T3, designboom, 1 Sep 2026) — a design-press showcase of Orphea, a one-octave pocket piano from Pianni.
Fetched the primaries rather than trusting the write-up, and it changed the page. The Kickstarter is
pre-launch — "Launching soon", 39 followers, creator "First created · 0 backed", San Francisco, no price,
no goal, no dates (checked today via Firecrawl; WebFetch gets 403 there). And pianni.co sells
children's music-learning books and keyboards, listing Orphea as its "Signature Instrument … screen-free
and child-safe". So the three accounts of this object disagree about who it is for: a sketchbook for
travelling musicians (designboom), "creativity anywhere" hardware (Kickstarter), or a child's first
instrument (the maker). Only the maker names a buyer. Filed as a tension in synthesis.md, not resolved.
Two more findings. The spoke has its first digital-piano — every keyboard here so far was either a synth or a midi-controller — and it is the clean inversion of thesis 3 (the interesting portable instruments are the ones you don't play by keys): Orphea keeps the keyboard and shrinks everything else. Thesis 3 gained a paragraph saying what that sharpens. And no source anywhere names the sound engine — not the maker's site, not the Kickstarter, not designboom. That is the second pocket instrument in this spoke with no engine published, after md-7-pocket-synthesizer, so it became coverage edge 10 rather than a note on one page.
New: orphea, orphea-designboom, digital-piano, pianni, jordan-daniels,
designboom (Wikipedia + Wikidata Q5264412, checked). Updated: index.md, synthesis.md
(thesis 3, a new contradiction, coverage edge 10). Entity budget: 3 of 8. A Los Angeles pianist of the same
name appears in search results and is not recorded on jordan-daniels — nothing checked here links him
to the San Francisco Kickstarter listing.
[2026-09-04] ingest | New Discoveries in Digital Audio — Harley interviews Wadax (TAS)
Routed here by the hub from Telegram. https://www.theabsolutesound.com/articles/new-discoveries-in-digital-audio/ — the-absolute-sound, 2026-09-01, robert-harley interviewing wadax CEO javier-guadalajara. WebFetch got a 403; Firecrawl read it.
Why here and not loudspeaker-design-wiki. The first read said "audio hardware, therefore the electroacoustics spoke," and that was wrong. This source is about converter and interface transparency, which is the argument this spoke already runs on three sources (analog-vs-digital, meyer-moran-adda-loop, digital-filter-audibility) and that spoke runs on none. Pages were drafted there and deleted before commit; nothing was left behind.
Paged: wadax-digital-audio-interview (Article, T3, published: 2026-09-01) plus four
entity nodes — wadax (same-as wadax.eu), javier-guadalajara, robert-harley,
the-absolute-sound (same-as Wikipedia + Wikidata Q3519694). Harley's author page 403s to
automated requests, so he carries no same-as rather than a guessed one.
What it adds. A third position on the spoke's oldest question, and one that does not play by
the same rules: audible and unmeasurable. Bit-identical data said to sound different through a
different one-foot USB cable; rise time, droop and receiver sensitivity sold as listener-adjustable
controls. Unblinded, unrepeated, and reported by the vendor selling the remedy. Filed as a position
held in the field, not as evidence — analog-vs-digital and synthesis.md both say so
explicitly, because folding it in beside two double-blind AES papers would flatter it.
Two things in it are worth more than that framing suggests: the intra-process / inter-process split (errors inside the converter vs errors arising between devices — analog path, grounding and power, the digital interface) names a layer the spoke had no page for; and the feed-forward IC is falsifiable at the output even though no measurement is published. New growth edge recorded: a third-party bench or blind test of any digital-interface claim.
Dedup: rg found no Wadax, Harley or Absolute Sound page in any spoke.
[2026-09-05] research | the ladder filter, measured → music-tech-wiki (1 source, T1, edge half closed)
Growth edge 1 has failed two hunts asking for a bench comparison of an analog circuit against its
digital model. The absence-claim drain found it instead: moog-vcf-measured (Paschou, Esqueda,
Välimäki & Mourjopoulos, Aalto / Patras, APSIPA ASC 2017, T1), read in full from the Aalto
post-print with pypdf after the browser fetch returned binary.
A real Moog Mother-32 — current production, not a vintage unit and not SPICE — measured against D'Angelo & Välimäki's model. Linear response by 6-second exponential sweep, 10 Hz–22 kHz, deconvolved and windowed; nonlinear response by 12 Vpp sine and sawtooth at 500 Hz, cutoff ≈3.7 kHz. Result: "accurate within a few dB in the linear regime" with very similar nonlinear behaviour, odd harmonics matching, and antialiasing needed to suppress the components the trivial implementation aliases in.
Half close, not a close. The edge asked for a bench comparison and a discrimination result. This is the bench half; the paper never asks a listener anything, which confirms from the inside the diagnosis the 2026-08-28 failed hunt reached — the field evaluates models against signals, not against hardware. Written as edge 1a, with the note that a few dB around a resonant peak is exactly where the character claim lives and no paper here draws a threshold.
The unexpected finding got its own edge (1b). Measuring a current product turned up three things its maker does not publish: the cutoff knob labels do not match the measured cutoff, the resonance knob maps to no scale at all — the authors report it in % of travel and fit their own curve, with nonlinear behaviour from 70% and self-oscillation around 80% — and an undocumented input attenuator had to be matched by ear. This spoke treats instrument specs as first-party fact; here a peer-reviewed measurement contradicts a front panel.
Retracted in place on moog-ladder-nonlinear-digital: "there is no measurement of a real ladder
filter" (dated, original kept struck). Dedup: rg found no Paschou, Mother-32 or APSIPA page anywhere.
[2026-09-09] ingest | Vintage Emulator Studio (VES) — antoniograzioli/VintageEmulatorStudio
Arrived via Telegram as a bare GitHub link; routed here by the hub (runner-up: foss-applications-wiki).
This spoke's registry note settles it: "software emulations of gear stay here." Read the repository
API metadata, README, LICENSING.md and docs/MAME_PATCHSET.md directly rather than through the
product page, so the ingest describes the project rather than the sales copy.
Source: vintage-emulator-studio, T1 — a first-party project repository, the strongest
evidence for what the software is and how it's built, and no evidence at all for how it sounds.
freshness: 60d: created 2026-08-30, pushed 2026-09-06, version 0.9.289, 69★. Pre-1.0 and moving.
Two new concept/entity nodes and one substrate page. firmware-emulation (the mechanism),
mame (the emulator it embeds, SoftwareApplication, same-as Wikipedia + Wikidata Q1054055 +
mamedev.org, all checked), autodafe (the maker; GitHub antoniograzioli, display name "Autodafe",
Milan). A separate Person node for Antonio Grazioli is deferred — brand and individual are one node
and the only evidence is a profile page. The seven unpaged instrument makers in VES's machine list
(Yamaha, Casio, Ensoniq, Roland, Sequential, Linn, PAiA) are deliberately not paged: a machine list is
not evidence about a company. akai and oberheim already existed and are linked.
Why it earned three pages. It fills the bottom rung of vintage-recreation — software emulation, which had been a one-line stub with no source — and then breaks the rung. VES does not model an instrument's sound. It boots the instrument's own ROM on a MAME-simulated CPU, so for a digital-voiced machine the output is the original computation rather than an approximation of it. The spectrum's organising trade (fidelity against price) stops applying, and analog-vs-digital gains a case where the fidelity question dissolves instead of being answered. Both pages updated; synthesis gains a section.
Three costs recorded, none about sound quality. ROMs, firmware and NVRAM are excluded from source control on purpose, so the free software does not boot until the user solves the copyright problem themselves. Accuracy is inherited from upstream MAME drivers, which the maker says outright and cannot answer for. And the technique picks its era: ~45 machines, nearly all FM/PD/PCM/sampler instruments from 1983–1995, against the 1975–1983 analog polysynths every other recreation rung rebuilds.
A licensing and business shape the spoke had not seen. AGPL-3.0 source with a five-licence component
map, MAME vendored at tag mame0289 commit f34f02505e32c1993c6a782b6814232cbfc74e36 with an ordered
patch series and a check-ves-mame-patches.sh that verifies the reproduction — then the compiled builds
are sold at autodafe.net. Strongest provenance discipline of any software source here.
Recorded as unmeasured: nothing in this corpus compares VES's output to a real CZ-101 or DX100, and the Prophet-5 and Six-Trak (analog voices under digital control) are the machines where the technique's reach is least clear. Neither the repo nor this ingest can say how much of those voice paths is modelled.
[2026-09-14] ingest | AudioNodes — free modular synthesizer and DAW (BPB)
Routed in from the hub (Telegram).
Source: audionodes, T2 — Tomislav Zlatic on bedroom-producers-blog, a recurring
source; vendor feature list and prices quoted. published: 2026-09-10. freshness: volatile — alpha
software (0.14.3) with subscription pricing.
Refreshed: daw gains a fourth layout (node graph as the DAW); vcv-rack linked. Spine: synthesis thesis #7 extended; index under Product / SoftwareApplication. AudioNodes LLC deferred (one source, no identity beyond the product site).
[2026-09-14] ingest | MusicRadar on Orphea — via quality cycle (absence-claim drain)
orphea said no source stated its sound engine; MusicRadar (2026-09-02) does. Source:
orphea-musicradar, T2, published: 2026-09-02. Retracted the engine sentence in place with the
date; the rest of the "nobody has published" list (voices, battery, weight, transport, price) stands.
[2026-09-14] lint | freshness pass — 15 dated announcements regraded stable, polyUAnalog re-verified
Quality cycle. The spoke held 16 of the corpus's 38 stale volatile pages; 15 were press announcements
(synthanatomy, BPB, gearnews, RA, mixdown, synthtopia, audiotechnology) whose text does not change —
regraded stable under ../QUALITY.md's 2026-08-04 rule that a dated publication is a record, not a
live quantity. polyuanalog is a repository and stays volatile: re-read from the GitHub API (87★,
last push 2026-04-07 — dormant), page unchanged.
[2026-10-07] ingest | Thicket (zc devices) — Bedroom Producers Blog
Routed from the hub (no runner-up spoke; clean fit). New page thicket (T3): a free, polyphonic node-based synth plugin in open beta until 2026-11-08, from a new one-person studio. Cross-linked with the closest sibling, audionodes (also node-based, also free, but Thicket builds one fixed instrument rather than an open DAW-and-rack). Self-generating macros from a patch is a feature not seen here before. No pricing after beta, no repo, one source only.