NIVIEM EQP-1

NIVIEM EQP-1 is a component-level digital emulation of the Pultec EQP-1A Program Equalizer — the 1961-designed, 1962-production passive tube programme equaliser built by Pulse Techniques, Inc. of Englewood Cliffs, New Jersey, and the single most recorded equaliser in the history of professional audio. The plugin models the EQP-1A's entire signal chain: the passive LC/RC network that gives the unit its musical, shelf-only voicing; the parallel low-frequency boost and cut topology that produces the famous "Pultec trick" as an emergent property rather than a scripted curve; the series-resonant LC tank that turns the high-frequency boost into a bandwidth-dependent peak; the ECC83 → ECC82 push-pull make-up amplifier that recovers the network's ~15 dB of passive insertion loss with the tube character that defines the chassis; the interstage and output transformers that load the amplifier and close the global negative-feedback loop; and the supply-sag envelope that droops the shared B+ rail as the unit is driven.

Around that schematic-faithful core, the plugin layers a set of Niviem-authored extensions — continuous frequency mode per section, opt-in HPF and LPF helpers, Mid/Side routing, channel-unlink, swappable input valves (ECC83 / 5751 / 12AY7), variable mains voltage, tube wear, transformer headroom, a Character (Drive) control that partially opens the feedback loop, Smart Auto Gain with per-side strength, A/B chassis slots, a click-free preset system with a 52-entry factory bank, the Nivipedia in-plugin encyclopedia (65 entries), and a premium WebView UI rendered as a brushed, anodised chassis. A single Hardware Accurate master switch returns every extension to its schematic value the instant it is engaged, so the chassis behaves either as a strict 1961 EQP-1A or as a modern programme equaliser that grows out of one — never as a confused hybrid of the two.

What Makes This Plugin Special

• Component-level passive EQ: the LF boost and LF cut networks are summed in parallel (out = in + (boost − in) + (cut − in)), not cascaded, so the simultaneous boost-and-cut "Pultec trick" emerges from the turnover asymmetry of two independent shelves rather than from a scripted dip — same way the 1962 cap-bank produces it.
• Hardware Accurate per-band turnover ratios: every dial frequency on the LOW FREQUENCY and ATTEN SELECT switches carries its own boost-corner and cut-corner ratio (not a single global factor); ratios are derived from EQP-1A3 and DIYRE EQP5 cap-bank measurements published by the GroupDIY and DIYRE communities.
• Maillet six-term polynomial for the ECC83 input valve: a measured RCA 12AX7 plate-current curve fitted by Pascal Maillet, evaluated along the AC load line per oversampled sample, with the operating point pinned to the schematic's ≈140 V / 1 mA quiescent.
• Koren 1996 space-charge model for the ECC82 output valve and the 5751 / 12AY7 input options: the de-facto standard tube simulation form, fitted from Philips datasheet curves with self-bias load-table cross-check, ≈4 % mean error against measured points.
• ADAA-1 anti-aliasing on every nonlinear stage: Parker / Esqueda 2016 antiderivative method — for each baked tube and transformer transfer curve, the precomputed antiderivative is evaluated as (F(x_now) − F(x_prev)) / (x_now − x_prev), giving the band-limited mean transfer across the input span rather than the aliasing-prone instantaneous read.
• Polyphase halfband oversampling cascade targeting ≈352 kHz internal rate: a chain of three identical 2× halfband stages (taps 127 / 61 / 25, Kaiser β = 12, ~120 dB stopband); the factor is auto-selected from host rate so 48 kHz → 8×, 96 kHz → 4×, 192 kHz → 2× — the nonlinear cascade sees the same high rate regardless of session.
• CON (continuous) frequency mode per section — the LOW FREQUENCY, HF BOOST FREQ and ATTEN SELECT switches each carry a continuous-sweep position that breaks free of the engraved stops; Hardware Accurate forces stepped mode the way the 1961 unit was built.
• Opt-in HPF + LPF helpers: surgical first-order high-pass and low-pass that sit outside the passive EQ chain, intended for the cleanups the EQP-1A's musical shelves can't do; disabled by default and disabled under Hardware Accurate.
• Full M-S routing: process Mid and Sides through independent instances of the chassis (separate tube types, voltage, age, transformer headroom and EQ settings per side) or run as a conventional stereo pair.
• LINK control: with LINK on, both channels (or both M-S sides) carry the same parameter values; with LINK off, every knob splits into two independent controls — the conventional dual-mono workflow of a hardware EQP-1S.
• Smart Auto Gain: a 64-point pink-weighted spectrum scan derived from the live EQ + character settings produces a per-channel make-up trim so that bypass A/B is honest — not a LUFS post-processor but a parameter-derived prediction, with a Strength knob that scales the compensation from 0 to 100 %.
• Click-free transitions on every parameter change: passive-EQ coefficient ramps (~8 ms), bypass crossfade, valve-swap crossfade, CON↔stepped crossfade, HPF/LPF engagement crossfade, M-S routing crossfade — every move that could click is bridged by a sample-aligned smoother.
• A/B chassis slots: two complete chassis snapshots living side-by-side, switchable instantly, paste either direction, save either as a user preset.
• Nivipedia in-plugin encyclopedia: 65 entries across nine categories (EQP-1, Sections, Controls, Innovations, Theory, Tubes, History, Workflow, Presets), with wiki-link cross-references between articles; reachable from a "What is this?" item on every right-click context menu.
• 52 factory presets across 9 categories: Init, Vocal, Drums, Bass, Guitar, Mix Bus, Mid/Side, Creative, Hardware Accurate — every preset a complete chassis snapshot, hand-tuned for level consistency across the bank.
• Premium WebView UI: a brushed-aluminium chassis with engraved silkscreen labels, photoreal milled-aluminium knobs, real-time spectrum analyser, value editor on every control with keyboard entry, gesture-aware host-automation bracketing on every drag, full keyboard navigation and ARIA scaffolding for screen-reader users.

Who Is This Plugin For?

The EQP-1A is at home anywhere musical, broad-stroke tone shaping is wanted — its shelf-only voicing makes it the sweetener of choice for vocals, mix-bus, mastering chains, and any source that benefits from gentle character rather than surgical correction. The plugin reproduces that voice and adds the modern niceties (M-S, CON sweep, tube swap, Auto Gain) that a programme equaliser earns its place in a 2026 mix with.

Pulse Techniques, Inc.

Pulse Techniques, Inc. was founded by Gene Shenk in 1959 in Teaneck, New Jersey (later Englewood Cliffs), as a small precision-audio shop serving the FM-broadcast and recording-studio markets of the East Coast. The first product was the EQP-1 (no "A") in 1960 — a passive programme equaliser pitched at radio stations and mastering rooms, with a slightly different capacitor bank and a simpler make-up amplifier than the unit that followed. The EQP-1A went into production in 1962, with the cap bank revised for smoother integration with a redesigned ECC83 → ECC82 push-pull cascade closed by a global negative-feedback loop — and that 1962 revision is the version that became the gold-standard.

Shenk's design priority across every Pulse Techniques unit was musicality over surgical precision: every band a shelf or a tank, no notch filters, no parametric peaks, every control engraved with a small number of broad stops. The unit was built to sweeten — to make broadcast voices and recorded music sound better — not to fix problems. That focus is why the EQP-1A remained in continuous production through 1980, and why Pulse Techniques is still in business in 2026, hand-building new EQP-1A3 units (a modern revision) at the original facility to original specifications.

What Came After

The Chassis

The original EQP-1A is a 3U rack unit. Across the front panel, left to right, the hardware exposes:

• LOW FREQUENCY — 4-position rotary selecting the LF section's corner (20 / 30 / 60 / 100 Hz).
• LF BOOST — continuous knob, the LF section's parallel-shelf boost (0 to ≈+13.5 dB).
• LF ATTEN — continuous knob, the LF section's parallel-shelf cut (0 to ≈−17.5 dB).
• HF BOOST FREQ — 7-position rotary selecting the HF tank's centre (3 / 4 / 5 / 8 / 10 / 12 / 16 kHz).
• HF BANDWIDTH — continuous knob, the HF tank's Q (broad ↔ sharp; gain and Q are coupled by design).
• HF BOOST — continuous knob, the HF tank's drive (0 to ≈+18 dB at sharp, 0 to ≈+9 dB equivalent at broad).
• ATTEN SELECT — 3-position rotary selecting the high-cut shelf's dial frequency (5 / 10 / 20 kHz).
• HF ATTEN — continuous knob, the high-cut shelf's floor (0 to ≈−16 dB).
• IN / OUT — bypass toggle (passive EQ + make-up amp in circuit, or relay-bypassed).

The Niviem chassis carries all nine of those exactly, in the same order, with the same engraved stops; on top it adds — every one opt-in, every one pinned to schematic value by Hardware Accurate — a NIVI button (Nivipedia), a STEREO / M-S routing pill, a LINK toggle, HPF and LPF opt-in cleanup filters, a TUBE swap rotary (ECC83 / 5751 / 12AY7), A and B chassis-slot chips, a SAVE button with a preset navigator at the chassis name, and a POWER master bypass at the right edge.

Original Hardware Specifications

The EQP-1A has been on so many records since 1962 that a complete user list would read as a list of professional recording engineers; the unit is in nearly every classical mastering chain, nearly every Hollywood scoring stage, nearly every New York and Los Angeles mix-bus rack. What follows is a representative slice, not an exhaustive enumeration.

Studios

Engineers

Famous Records

• The Beatles — Sgt. Pepper's Lonely Hearts Club Band (1967) and Abbey Road (1969): Geoff Emerick used the studio's EQP-1A on bass, lead vocal and string sweetening throughout both sessions.
• Pink Floyd — The Dark Side of the Moon (1973): Alan Parsons mastered through Pultec at Abbey Road; the EQP-1A shaped the broadcast-ready clarity that made the record sit so well on cheap consumer systems.
• Led Zeppelin — IV (1971) and Houses of the Holy (1973): Andy Johns ran Pultec on the bass and the bus across both records; the low-end heft is partly its work.
• Stevie Wonder — Innervisions (1973): Robert Margouleff and Malcolm Cecil used the EQP-1A on the Rhodes and the Moog basses; the warm, broad low-end signature is a Pultec move.
• Quincy Jones — Thriller (1982): Bruce Swedien's mix chain on the Westlake console included EQP-1A on the bass, the kick and the lead vocal across the record.
• Nirvana — Nevermind (1991): Andy Wallace's mix bus carried Pultec EQ on the master; Bob Ludwig's mastering chain at Gateway carried it too.
• Daft Punk — Random Access Memories (2013): a deliberately analogue-chain record — Mick Guzauski's mix bus and Bob Ludwig's mastering both ran through EQP-1A units; the album's lush bass and air are Pultec territory.

When you reach for an EQP-1A you are reaching for a sound that is in the listening canon — listeners know what a Pultec sounds like without ever having heard one in isolation.

Methodology

NIVIEM EQP-1 was built from the EQP-1A schematic, not from a sample bank or a reverse-engineered impulse response. The passive section is modelled by folding the schematic's LC ladder into the analytical s-domain transfer function for each band — a low shelf for LF boost, a low shelf for LF atten with a different turnover ratio, an RBJ peaking biquad whose pole/zero pair is derived from the series LC tank's resonant parameters for HF boost, a first-order high shelf at corner = dial ÷ 7.5 for the ATTEN SELECT section — and bilinear-transforming each to a biquad at the host's sample rate. The three biquads run in series, each with its own ~8 ms coefficient-crossfade smoother so any knob move stays click-free; the LF boost and LF cut paths are summed in parallel inside the LF biquad's coefficient bake so the Pultec trick emerges from the schematic, not from a scripted shape on top.

The make-up amplifier follows the same first-principles approach. The ECC83 input stage uses Pascal Maillet's six-term polynomial fit to a measured RCA 12AX7, pinned to the schematic's ≈140 V plate / 1 mA quiescent operating point and solved along the AC load line (reflected interstage primary impedance ≈ 100 kΩ per triode) per oversampled sample by a 3-iteration Newton step warm-started from the previous solution. The ECC82 output stage uses Norman Koren's 1996 space-charge equation fitted to Philips datasheet curves over 21 measured points with ≈4.1 % mean error, pinned to ≈250 V plate / 8–10 mA / Rac = 12.5 kΩ ÷ 4 (one quarter of the Peerless S-217-D plate-to-plate primary), self-bias load-table cross-checked. The interstage transformer is modelled as a Thévenin source feeding a nonlinear iron core — flux integration with a tanh-shaped saturation knee, asymmetric for the two flux polarities at saturation onset, ADAA-1-read so the magnetising current band-limits cleanly. The global negative-feedback loop from the output transformer's tertiary winding back to the ECC83 cathodes is closed by the same Newton iteration. Hardware Accurate pins the operating points, the tube types, the iron loading, the mains voltage and the supply-sag depth back to schematic values, and the chassis behaves as a strict 1961 EQP-1A from that moment on.

Key DSP Implementations

1. Passive EQ network. Three series biquads — one combined LF section (parallel boost + cut shelves with independent corner ratios baked into the coefficient design), one HF tank (peaking biquad), one HF cut shelf — each with an ~8 ms equal-power coefficient crossfade between old and new coefficients on every parameter change. The LF parallel sum is computed once per coefficient bake so the run-time path is a single biquad per band per channel; the trick emerges from the bake, not from runtime summing.

2. Make-up amplifier. Two-stage Newton load-line solve baked at startup into a 2048-point LUT per supply-sag curve, with 7 sag curves spanning the slow B+ droop the unit shows under sustained load. The runtime path is a LUT read with ADAA-1 antiderivative evaluation: each sample reads (F(x_now) − F(x_prev)) / (x_now − x_prev) from the precomputed antiderivative table, giving the band-limited mean transfer across the input span. Iron stages (interstage + output transformer) run in the same loop at the same oversampled rate with their own ADAA-1 tables.

3. Anti-aliasing. A polyphase halfband FIR cascade of three 2× stages with tap counts 127 / 61 / 25 and Kaiser β = 12, designed for ~120 dB stopband rejection at the boundaries. The internal rate target is ≈352 kHz, with the factor auto-selected per host rate (48 kHz → 8×, 96 kHz → 4×, 192 kHz → 2×, 384 kHz → 1×). The total round-trip latency is an exact integer of base-rate samples and is reported to the host so the bypass A/B, the dry-path parallel chain and any sibling-plugin delay compensation stay phase-aligned.

Hardware-Accurate per-band turnover

The LF section's boost-corner and cut-corner ratios are not a single global constant. Each engraved dial frequency (20, 30, 60, 100 Hz) has its own boost ratio and its own cut ratio under Hardware Accurate, because the original cap-bank that selects between the four LF dial positions does not just swap a frequency-determining capacitor — it swaps the entire two-pole network that drives the shelf, and each network has slightly different component values. The ratios below are what the plugin uses when Hardware Accurate is on; toggle HA off and the chassis collapses to a single nominal ratio per direction (the "creative" simplification that the original chassis cannot do).

These ratios are derived from EQP-1A3 and DIYRE EQP5 cap-bank measurements published by the DIY tube-electronics community on the GroupDIY and DIYRE forums; they are the best-available numeric reference for a Pulse Techniques cap-bank short of measuring an original 1962 unit on the bench, and they match the Ian Bell measurement set within the published tolerances.

What We Added Beyond the Original

What We Did NOT Add

Some omissions are deliberate and should stay that way; calling them out explicitly helps users understand the chassis's design intent.

• No oversampler off-switch. The oversampling cascade is not user-defeatable. The nonlinear stages cannot run cleanly at base host rate; turning the cascade off would alias and the chassis would no longer be honest. The factor is auto-selected from host rate so the CPU cost is bounded.
• No HF Boost Q free-knob. Bandwidth and peak gain are coupled at the schematic — the LC tank's Q rises and falls with the peak amplitude by physics, not by choice. A free Q knob would break the EQP-1A's voicing; the BANDWIDTH control honours the coupling.
• No third LF stop between 60 and 100 Hz. The original cap bank has four stops, not five; CON mode is provided for everything between, and an in-between engraved stop would be a Niviem invention pretending to be a schematic value.
• No MEQ-5 mid-band. The MEQ-5 is the EQP-1A's sister product and a different chassis; building the MEQ-5's 200 Hz–7 kHz mid-band into the EQP-1 would be a confusion of two distinct units. The MEQ-5 deserves its own emulation, not a sub-mode.
• No solid-state make-up amplifier path. The EQP-1A's voice is the tube + iron cascade; a solid-state make-up amp option would be a different unit, not an extension of the EQP-1A. Character (Drive) and Transformer Headroom give the chassis its variable-colour range without changing what the chain is.
• No phase-linear toggle. The passive EQ's minimum-phase response is part of its sound — the phase shifts the LF parallel shelves produce contribute to the Pultec trick's perceived depth. A phase-linear mode would flatten that out, and the chassis would sound less like itself.
• No A/B persistence across project save. A/B is a session-scoped comparison tool; the saved project carries the currently-active slot only. User presets are the persistent comparison store.

The EQP-1 ships as a single signed installer per platform — pkg on macOS, zip on Windows. Both installers carry the same plugin code; only the host wrapper formats differ. Throughout this manual the product is written EQP-1 with the hyphen for the brand, and EQP1 without the hyphen in filenames and bundle paths (host scanners and Windows filesystems tolerate the hyphen but several DAWs surface the no-hyphen form in their plugin lists, so the on-disk binary uses it for cross-host consistency).

System Requirements

macOS

Windows

macOS Installation

1. Sign in at niviem.net and open your account page. The EQP-1 download appears under My Plugins once the purchase clears.

2. Download Niviem_EQP1_v1.0.0_macOS.pkg. The installer is signed and notarized by NIVIEM AUDIO LTD — Gatekeeper will not prompt on a current macOS.

3. Double-click the pkg and follow the installer prompts. The default destinations are:

   Format	Path
   AU	~/Library/Audio/Plug-Ins/Components/Niviem EQP1.component
   VST3	~/Library/Audio/Plug-Ins/VST3/Niviem EQP1.vst3

4. Launch your DAW and let it rescan the plugin folders. In Logic Pro, the AU validator runs automatically on first sight; in Studio One / Cubase / Live, trigger a manual rescan from the plugin-manager dialog if the EQP-1 does not appear immediately.

Windows Installation

1. Sign in at niviem.net, open your account page, and download Niviem_EQP1_v1.0.0_Windows.zip from My Plugins.

2. Unzip the archive. Inside is a single bundle: Niviem EQP1.vst3.

3. Copy Niviem EQP1.vst3 into the system-wide VST3 folder:

   C:\Program Files\Common Files\VST3\
   

   Per-user installs into %LOCALAPPDATA%\Programs\Common\VST3\ also work; the system path is preferred so every DAW account sees the plugin.

4. Launch your DAW. If the chassis loads as a blank white panel rather than the brushed-metal faceplate, the host is missing the Microsoft WebView2 Runtime — install it from the official Microsoft download page (https://developer.microsoft.com/microsoft-edge/webview2/), restart the DAW, and rescan. Windows 11 ships WebView2 by default, and Windows 10 picks it up through Windows Update on most current installations.

User Preset Location

The factory bank is read out of the bundle resources and is never written to disk. Your own saved presets land in a per-user directory you can back up, sync, or browse outside the plugin:

Each .preset is a JSON document carrying the parameter snapshot, the category, an optional description, the author tag, and a UTC ISO timestamp. They are stable across versions — a preset saved in 1.0.0 loads cleanly in any future build, and the directory survives plugin reinstalls.

License Cache

The activated license is cached as a signed JWT on the same machine that activated it, so subsequent launches are offline-clean and instant:

Deleting the cache file forces the plugin back into demo mode on the next launch. Use this as the manual "deactivate this machine" path if you cannot reach the niviem.net dashboard — the activation pool entry is released the next time the plugin checks in.

First-launch flow

On every instantiation, the plugin checks three entitlement sources in parallel. The first one that returns a positive answer wins; the others stop being consulted:

1. Cached Niviem JWT. The plugin reads eqp1.jwt from the License Cache path above and verifies the signature locally against the embedded Ed25519 public key. No network round-trip — a licensed machine is offline-clean on launch.
2. MuseHub probe. If MuseHub Desktop is installed on the machine and reports an entitlement for eqp1, the plugin treats it as an active license for the duration of the session. The MuseHub bridge is asynchronous and never blocks audio.
3. Demo mode. If neither of the above resolves, the plugin enters demo mode. The splash screen dismisses, and an activation dialog floats over the chassis until the user enters a key or chooses Continue in Demo.

Demo mode

Demo mode passes the full DSP chain — including every Niviem extension — at full quality. The only restriction is a 60-second-on / 5-second-mute repeating cycle that ensures the chassis cannot be used unattended in production. The cycle counter resets on every prepareToPlay (transport start, sample-rate change, host buffer-size change), so quick A/B passes against bypass land cleanly inside one 60-second window.

While in demo mode the chassis carries two visual reminders in the upper-left header: a DEMO badge sitting between the EQP-1 logo and the NIVI button, and an ACTIVATE button immediately to its right. The badge dims and the activate button hides as soon as a valid entitlement arrives — no DAW restart required.

Activating

Click ACTIVATE in the header to open the activation modal. Paste the license key from your niviem.net account or order confirmation email and press Activate. The plugin POSTs the key plus a machine fingerprint (hashed hostname + boot-volume UUID) to https://admin.niviem.net/api/v1/activate. The dashboard issues a signed JWT bound to that fingerprint and returns it; the plugin verifies the signature locally against the embedded Ed25519 public key and writes the JWT into the License Cache. No further network calls are made by the plugin after that point.

Every Niviem license carries a 5-machine activation pool. You can move the EQP-1 between studios, laptops, render nodes and rehearsal rigs freely within that pool — the activation list is editable from your niviem.net account page, where individual machines can be deactivated to free up a slot.

MuseHub customers

If you purchased the EQP-1 through MuseHub, no Niviem license key is required. The entitlement is carried by MuseHub Desktop on every launch — the plugin probes the MuseHub bridge during the parallel entitlement check above and silently activates when the probe returns. The two paths coexist on the same install: a Niviem-key activation will activate the plugin even if MuseHub is not installed, and a MuseHub session will activate even if the Niviem cache file is missing. Whichever responds first wins the current session.

Quick Start (90 seconds)

1. Insert the EQP-1 on a stereo bus, lead vocal, or full mix. The chassis loads at its Default preset — every knob flat, both helper filters off, TUBE engaged, ROUTE in STEREO with LINK on. The audio path is bit-identical to bypass at this state apart from the input/output transformer pair and the make-up amplifier's small-signal pass — the chassis is harmless until you reach for a knob.
2. Click the preset name plate in the centre of the header to open the preset browser. Walk through the Vocal — Air & Body category. The categories collapse and expand; clicking a preset loads it immediately, and the preset name plate updates.
3. Click the A chip in the header A/B group to claim slot A for your current settings, then click A→B to copy them into slot B, then click B to flip to the slot. The two slots are now identical — start dialling slot B and use the A/B chip to compare without losing either snapshot.
4. Right-click any knob and pick What is this?. The Nivipedia overlay opens directly at the entry that describes that control — boost shape, frequency mapping, hardware lineage, Niviem extensions, the lot. Use it as the in-context reference.
5. Click any cream chassis area outside the controls to dismiss any open overlay (preset browser, Nivipedia, activation dialog). Escape works too, on every modal.

Signal Flow

INPUT
  │
  ▼
[HPF helper]*       ← opt-in 12 dB/oct Butterworth, off-state bypassed
  │
  ▼
[LPF helper]*       ← opt-in 12 dB/oct Butterworth, off-state bypassed
  │
  ▼
[Input transformer] ← iron-saturation low-end colour, level-dependent
  │
  ▼
[Passive EQ network]
  ├── LF Boost  (first-order low shelf, corner ≈ 6 × dial)
  ├── LF Atten  (first-order low shelf, corner ≈ 13.5 × dial)   parallel → Pultec trick
  ├── HF Boost  (RBJ peaking, gain-Q coupling via Bandwidth)
  └── HF Atten  (first-order high shelf cut, corner = dial ÷ 7.5)
  │
  ▼
[ECC83 input pair: Maillet polynomial · Newton load-line · ADAA-1]
  │
  ▼
[Interstage transformer]
  │
  ▼
[ECC82 output pair: Koren space-charge model · ADAA-1]
  │
  ▼
[Output transformer]
  │
  ▼
[Global negative feedback loop — tertiary winding → ECC83 cathodes]
  │
  ▼
[Auto Gain]*        ← opt-in param-derived loudness trim, unity-bypassed
  │
  ▼
[Output gain]       ← ±12 dB per-side trim
  │
  ▼
[BYPASS mixer]      ← equal-power crossfade, delay-compensated dry path
  │
  ▼
HOST OUTPUT

* = opt-in Niviem extensions, off-state bit-identically bypassed (no DSP cost when idle).

Under ROUTE = M-S, the entire chain above is wrapped in an M = (L+R)/2, S = (L-R)/2 encode at the input and a matching L = M+S, R = M-S decode at the output. The bare-ID parameters drive the Mid path; the _R siblings drive the Side path. Under STEREO with LINK off, the routing stays in L/R and the two channels diverge cleanly.

First Preset Tour

The factory bank is the fastest way to get a feel for the chassis. Walk these five in order with a test source playing:

After the tour, Init returns you to a clean starting state without un-installing the chain.

The EQP-1 chassis is rendered as a brushed-metal slab carrying engraved silkscreen, photoreal milled-aluminium knobs and rotary selectors, a recessed scope window with live spectrum and EQ curve, and three large section panels — LOW, HIGH, OUTPUT. Every interactive surface carries WAI-ARIA scaffolding so the plugin is operable from a screen-reader or keyboard-only workflow as well as the mouse.

Layout

The default chassis is 1400 × 778 pixels at 1.0 host DPI; the WebView scales linearly with the host's editor scale factor. The faceplate splits into three vertical bands — Header, Scope + Control Deck, Footer:

┌───────────────────────────────────────────────────────────────────────────────┐
│ NIVIEM EQP-1   [DEMO] [ACTIVATE]  [NIVI] [STEREO] [LINK] [HPF] [LPF] [TUBE]   │
│                              [A] [A→B] [B→A]  [SAVE] [◀ Default ▶]  [⏻]         │
├───────────────────────────────────────────────────────────────────────────────┤
│                                                                               │
│  ┌─────────────────────────────────────────────────────────────────────────┐  │
│  │       SPECTRUM + EQ CURVE   (live FFT bars · gold transfer line)        │  │
│  └─────────────────────────────────────────────────────────────────────────┘  │
│                                                                               │
│  ┌──────────────────┐  ┌────────────────────────────┐  ┌────────────────────┐ │
│  │  LOW FREQUENCY   │  │   HIGH FREQUENCY · KCS     │  │                    │ │
│  │   · CPS rotary   │  │   rotary (3/4/5/8/10/12/   │  │                    │ │
│  │  20·30·60·100·CON│  │   16 · CON)                │  │                    │ │
│  │                  │  │                            │  │                    │ │
│  │ HPF  Boost  Atten│  │ LPF Boost  BW  Atten       │  │ Character Strength │ │
│  │              LF  │  │              HF  Atten     │  │              Gain  │ │
│  │              CON │  │              CON  CON      │  │ [Auto Gain · ON]   │ │
│  │                  │  │                            │  │                    │ │
│  │  [Hardware       │  │   ATTEN SEL · KCS rotary   │  │                    │ │
│  │   Accurate]      │  │   5 · 10 · 20 · CON        │  │                    │ │
│  │                  │  │                            │  │                    │ │
│  │       LOW        │  │           HIGH             │  │      OUTPUT        │ │
│  └──────────────────┘  └────────────────────────────┘  └────────────────────┘ │
├───────────────────────────────────────────────────────────────────────────────┤
│                                                                       v1.0.0 │
└───────────────────────────────────────────────────────────────────────────────┘

(The actual rendered chassis is a continuous milled faceplate with photoreal knobs and recessed LCD-pillbox readouts — the diagram above shows logical grouping only.)

Header

Reading left to right, the header carries:

• NIVIEM EQP-1 logo — brand mark plus product type (Program Equalizer). Non-interactive.
• DEMO badge — visible only when unlicensed. Reminder that the chassis is in the 60 s / 5 s mute cycle.
• ACTIVATE button — visible only when unlicensed. Opens the activation modal.
• NIVI — opens Nivipedia (the built-in encyclopedia, full-screen overlay).
• STEREO / M-S — Route flip. Toggle between independent L/R processing and a Mid/Side encode-process-decode wrap.
• LINK — couples every L/R control pair so the two channels stay equal. Forced on under Hardware Accurate.
• HPF — engages the opt-in 12 dB/oct Butterworth high-pass helper filter (frequency knob lives in the LOW section).
• LPF — engages the opt-in 12 dB/oct Butterworth low-pass helper filter (frequency knob lives in the HIGH section).
• TUBE — opens the Make-Up Amplifier panel. The chip text reflects the current mode — TUBE when the make-up amp is engaged, CLEAN when it is bypassed.
• A / A→B / B→A — the A/B comparison group. The leading chip shows the active slot and toggles between them; the two copy chips push the active slot's state into the other slot.
• SAVE — opens the Save Preset dialog (name + category + description fields).
• Preset Navigator — left arrow, preset name plate (click to open the preset browser), right arrow. A small dot lights on the name plate when the current state diverges from the loaded preset.
• Power (⏻) — chassis bypass. Lit = effect engaged; dim = bypassed (equal-power 25 ms crossfade, dry path delay-compensated).

Main Chassis

The chassis splits into three sections side by side. Each section carries a top selector (where applicable), a row of knobs at a single horizontal baseline, a bottom toggle or selector, and a section label. Knob row alignment is preserved across the three sections so the eye reads a single horizontal line of controls across the whole chassis.

Footer

A thin engraved strip below the control deck carries the build version on the right-hand side (v1.0.0). The strip is non-interactive — it exists purely as a visual rest line that grounds the cream faceplate against the chassis edge.

Value Readouts

Every continuous knob carries a recessed engraved-LCD pillbox immediately below its skirt, showing the current value in engineering units (dB, Hz, %, V). The readout refreshes live during a drag and during DAW automation — there is no settling delay. Rotary selectors show their selected stop label in the same recessed pillbox style above the stops; the LOW FREQUENCY, HIGH FREQUENCY and ATTEN SEL rotaries each render their position as a milled aluminium pointer over the engraved stop ring.

Interaction Patterns

Right-Click Context Menu

Every interactive control exposes a custom Niviem context menu in place of the browser default:

• Reset to Default — knobs only. Restores the parameter to its factory default value through the gesture-wrapped automation pipe (single recorded edit, never a smear).
• Copy Value — knobs only. Copies the current formatted value (e.g. 4.5 dB, 60 Hz) to the system clipboard.
• Paste Value — knobs only. Parses the clipboard and writes the value if it matches the parameter's expected format. Out-of-range values clamp silently to the parameter's min/max.
• What is this? — every control on the chassis, including chips, rotaries, A/B group, preset chip and Power. Opens Nivipedia at the entry that describes the clicked control.

Visual Feedback

The chassis surfaces several real-time visualisations the eye reads passively while working:

• Knob LED arc — a thin gold conic ring from the 7 o'clock minimum stop around to the current value, painted into the knob's outer bezel.
• Spectrum analyser — gold FFT bars rendered in the scope window from an off-thread FFT (10-order, 1024 samples per frame, log-frequency bin mapping). Never runs on the audio thread, never runs while the editor is closed.
• EQ curve overlay — a gold line through the scope window showing the current transfer function (the sum of helper filters + passive EQ + tube-stage equivalent magnitude response).
• Sister curve — when LINK is off, the right / side channel's curve paints under the bare curve in a contrasting hue (cyan under STEREO, magenta under M-S) so the two paths read independently.
• Knob dim states — controls that are temporarily inert dim to ~40 % opacity. Examples: LF CON dims while the LOW FREQUENCY rotary is on a stepped stop, HPF frequency dims while the HPF header chip is off, Auto Gain Strength dims while Auto Gain is off, the entire make-up amp modal dims under Hardware Accurate.

The EQP-1 carries 14 continuous knobs, 3 multi-position rotaries (LOW FREQUENCY · CPS, HIGH FREQUENCY · KCS, ATTEN SEL · KCS), 8 header chips, and the license-state chip pair (DEMO badge + ACTIVATE button) that surfaces only in unlicensed sessions. Every continuous knob has matching _R sibling parameters under LINK off / M-S routing, so the effective parameter count doubles when running the chassis on two independent channels.

Header

NIVI

Opens the Nivipedia encyclopedia overlay. Every control entry, theory page, and section walkthrough is rendered here — search, table of contents, cross-links between entries. Escape or click outside dismisses the overlay; the chassis state underneath is untouched.

STEREO / M-S (Route)

Options: Stereo / M-S Default: Stereo Parameter: route

Flips the chain between independent left/right processing and a Mid/Side encode-process-decode wrap. Under Stereo, the bare and _R parameter sets drive L and R respectively. Under M-S, the bare set drives the Mid path and _R drives the Side path. Hardware Accurate forces the route to Stereo.

LINK

Default: On Parameter: link

Couples every L/R parameter pair so the two channels stay equal. Turning LINK off does not erase the _R snapshot — the two sets retain their last state, the LINK listener simply stops mirroring. Forced on under Hardware Accurate.

HPF

Default: Off Parameter: hpfOn / hpfOn_R

Engages the opt-in 12 dB/oct Butterworth high-pass helper filter. The frequency knob lives in the LOW section. Forced off under Hardware Accurate.

LPF

Default: Off Parameter: lpfOn / lpfOn_R

Engages the opt-in 12 dB/oct Butterworth low-pass helper filter. The frequency knob lives in the HIGH section. Forced off under Hardware Accurate.

TUBE

Default: On Parameter: tube / tube_R

Opens the Make-Up Amplifier panel and controls the colour-path engage state. The chip text reflects the current mode — TUBE when the make-up amplifier (input/output transformers + ECC83/ECC82 stages + global feedback loop) is in the path, CLEAN when it is bypassed and the chassis delivers the pure passive-EQ curve. (Make-up amplifier details — Character, Strength, Valve, Wear, Amp Voltage, Transformer — are documented in their own section.)

A / A→B / B→A

Default: A active, B uninitialised Parameter: held in stateA_ / stateB_ snapshots (not host-automatable)

The A/B comparison group. The leading chip shows the active slot — click it to flip between A and B. A→B copies the active slot's state into the inactive slot without flipping the active marker; B→A does the same in reverse. (A/B workflow tips and gotchas — automation-thread implications, preset-load behaviour — are documented in their own section.)

SAVE

Opens the Save Preset dialog with name, category, and description fields. Saves to the User Preset Location under the entered name; the new preset appears in the preset browser's user category on the next open.

Preset Navigator

Layout: ◀ prev | preset name plate | ▶ next

The left and right arrows step through the factory + user preset bank, wrapping at the ends. The centre plate shows the current preset name; clicking it opens the categorised preset browser. A small dot lights on the plate when the live state diverges from the loaded preset (the "modified" indicator).

Power (⏻)

Default: On Parameter: bypass

Chassis bypass. Lit = effect engaged. Dim = bypassed; the signal passes through unchanged after a 25 ms equal-power crossfade with the dry path delay-compensated for any helper-filter / oversampler latency the wet path carries.

LOW Section

LOW FREQUENCY · CPS

Options: 20 / 30 / 60 / 100 / CON Default: 60 Hz Parameter: lowFreq / lowFreq_R (stepped) + lowFreqMode + lowFreqCont (continuous mode)

Selects the LF shelf corner frequency in cycles-per-second (CPS — the EQP-1A schematic notation, identical to Hz). The four stepped positions reproduce the original Pultec rotary; the CON stop engages Continuous mode and exposes the LF CON knob, sweeping the corner smoothly from 20 to 100 Hz on a log curve. Hardware Accurate forces CON off and snaps the selector back to its nearest stepped neighbour.

Boost (LF)

Range: 0 dB to 13.5 dB Default: 0 dB Parameter: lowBoost / lowBoost_R

First-order low shelf, corner at approximately 6 × dial (so 60 Hz dial → ~360 Hz shelf knee). Always a shelf, never a peak. Combine with Atten on the same frequency for the famous Pultec trick — the two parallel networks land at slightly different corners and produce a low-mid scoop with a fattened low-end shelf above and below.

Atten (LF Cut)

Range: 0 dB to 17.5 dB Default: 0 dB Parameter: lowCut / lowCut_R

First-order low shelf cut, corner at approximately 13.5 × dial (so 60 Hz dial → ~810 Hz shelf knee). The corner sits noticeably above the Boost corner — that 2.25× ratio between the two shelves on the same dial is the structural reason the Pultec trick works. Stack Boost and Atten on the same band and the network leaves a low-mid scoop between the boost shelf and the cut shelf rather than cancelling.

LF CON

Range: 20 Hz to 100 Hz (log) Default: 60 Hz Parameter: lowFreqCont / lowFreqCont_R

The continuous LF corner frequency, active only when the LOW FREQUENCY rotary is on the CON stop. Dimmed and inert when the rotary is on a stepped position. Niviem extension — Hardware Accurate forces it off the path.

HPF Frequency

Range: 20 Hz to 800 Hz (log skew 0.3) Default: 80 Hz Parameter: hpfFreq / hpfFreq_R

Cutoff for the opt-in 12 dB/oct Butterworth high-pass helper filter. Engaged via the HPF header chip; the knob dims while the chip is off. No Q control — the classic Pultec-adjacent helper filter character is a clean Butterworth roll-off without a resonant knee.

Hardware Accurate

Default: Off Parameter: hwAccurate

Master toggle for the chassis's authentic-EQP-1A mode. Pins every Niviem extension to its neutral (off / unity) position so the unit behaves like a strictly authentic EQP-1A. (Hardware Accurate is documented in its own section — the per-parameter pin-list lives there.)

HIGH Section

HIGH FREQUENCY · KCS

Options: 3 / 4 / 5 / 8 / 10 / 12 / 16 / CON (kc / kHz) Default: 8 kHz Parameter: highFreq / highFreq_R (stepped) + highFreqMode + highFreqCont (continuous)

Selects the HF peaking-boost centre frequency in kilocycles (KCS — schematic notation, identical to kHz). Seven stepped positions reproduce the original EQP-1A rotary; the CON stop engages Continuous mode and exposes the HF CON knob, sweeping the centre smoothly from 3 kHz to 16 kHz on a log curve. Hardware Accurate forces CON off.

Boost (HF)

Range: 0 dB to 18 dB Default: 0 dB Parameter: highBoost / highBoost_R

The HF peaking boost, voiced as an RBJ peaking biquad with gain-Q coupling — pushing the boost makes the peak narrower the way the original LC tank does on real silicon. This is the EQP-1A's most musical control: sharp narrow boosts read as presence, broad lower-Q boosts read as air or sheen, and the boost stays clean up to the 18 dB ceiling without crossing into harsh territory.

Bandwidth

Range: 0 (broad) to 100 % (sharp) Default: 0 (broad) Parameter: highBoostBW / highBoostBW_R

Controls the Q of the HF peaking-boost — broad ≈ Q 0.6 at 0 %, sharp ≈ Q 2.5 at 100 %. Maps the dial directly to the biquad's Q on a continuous curve so the audible character moves smoothly from a sheen-style wide bell to a tight presence peak. The Boost knob's gain-Q coupling rides on top of this — a sharp Bandwidth at 18 dB Boost produces a narrower, more pronounced peak than a broad Bandwidth at the same Boost setting.

Atten (HF Cut)

Range: 0 dB to 16 dB Default: 0 dB Parameter: highCut / highCut_R

First-order high shelf cut sitting at the ATTEN SEL corner divided by 7.5. This is not a low-pass filter — above the corner the response drops to the shelf floor and stays there rather than rolling off forever. The shelf floor is set by this Atten knob; the corner is set by the ATTEN SEL rotary below.

HF CON

Range: 3 kHz to 16 kHz (log) Default: 8 kHz Parameter: highFreqCont / highFreqCont_R

The continuous HF boost-centre frequency, active only when the HIGH FREQUENCY rotary is on the CON stop. Dimmed and inert when the rotary is on a stepped position. Niviem extension — Hardware Accurate forces it off the path.

Atten CON

Range: 5 kHz to 20 kHz (log) Default: 20 kHz Parameter: highCutFreqCont / highCutFreqCont_R

The continuous ATTEN SEL dial frequency, active only when the ATTEN SEL rotary is on its CON stop. The actual shelf corner is the dial value divided by 7.5 (the EQP-1A's first-order shelf characteristic). The 5 kHz dial floor keeps the corner inside Pultec-authentic territory — dropping below would put the shelf knee into the upper mids and break the chassis's musical voicing.

LPF Frequency

Range: 1 kHz to 20 kHz (log skew 0.35) Default: 12 kHz Parameter: lpfFreq / lpfFreq_R

Cutoff for the opt-in 12 dB/oct Butterworth low-pass helper filter. Engaged via the LPF header chip; the knob dims while the chip is off. No Q control — same clean Butterworth character as the HPF helper.

ATTEN SEL · KCS

Options: 5 / 10 / 20 / CON (kc / kHz) Default: 20 kHz Parameter: highCutFreq / highCutFreq_R (stepped) + highCutFreqMode + highCutFreqCont (continuous)

Selects the dial frequency for the HF Atten shelf. The actual shelf corner is dial ÷ 7.5 (so 20 kHz dial → ~2.7 kHz shelf knee). Three stepped positions reproduce the original EQP-1A rotary; the CON stop engages Continuous mode and exposes the Atten CON knob.

OUTPUT Section

Character (Drive)

Range: 0 dB to 12 dB Default: 0 dB Parameter: drive / drive_R

Drives the ECC83 pre-gain on the make-up amplifier with a matching post-gain trim, so the chain's small-signal level stays unity but the valve sees a hotter input. The matching post-gain runs after the tube stage, which means Character adds harmonic colour and a touch of compression without consuming the rest of the chain's headroom — push it for grit, leave it at 0 dB for the chassis's small-signal voice. Hardware Accurate forces it to 0 dB.

Strength (Auto Gain depth)

Range: 0 % to 100 % Default: 100 % Parameter: autoGainStrength / autoGainStrength_R

The depth of the Smart Auto Gain trim — 100 % is a full perceived-loudness match, lower values keep some of a big boost's added weight so the loudness compensation does not entirely erase the gesture. Active only when Auto Gain is on; dimmed and inert otherwise.

Gain (Output)

Range: −12 dB to +12 dB Default: 0 dB Parameter: outputGain / outputGain_R

Bipolar trim at the very end of the chain, per-channel under LINK off. Use it as the master make-up after a big boost or atten, or as a per-side balance trim under M-S routing.

Auto Gain

Default: Off Parameter: autoGain / autoGain_R

A param-derived output trim that holds perceived loudness constant as the EQ is adjusted, so an A/B against bypass is honest rather than a level shootout. The trim is computed from a 64-point log-frequency scan of the current transfer function on the message thread (the audio thread only consumes the cached result) — knob drags stay smooth at any sample rate or buffer size. Hardware Accurate forces it off. The Strength knob immediately above it sets the depth of the compensation when on.

The Auto Gain chip renders inline below the Strength knob (rather than as a header chip) because Strength is the Auto Gain depth control — visually grouping the two reinforces that the chip switches the feature on, and the knob then dials in how much it does.

The LOW section is the most musically distinctive part of the chassis — the part responsible for the EQP-1A's reputation. Where most equalisers cascade their bands in series, the EQP-1A's low end runs two first-order shelves in parallel around the same dial frequency: a BOOST shelf whose corner lies just above the dial, and an ATTEN shelf whose corner lies significantly higher than the BOOST corner. The two networks share the dial but not the cap bank — the BOOST capacitors and the ATTEN capacitors are physically separate, and their values place the two turnovers at deliberately different ratios of the dial. Engaging both knobs at once therefore does not cancel: the BOOST lifts below its corner, the ATTEN cuts below its (higher) corner, and the difference between the two shelves becomes the famous lift-with-a-small-dip shape known as the Pultec trick (see section 14).

Frequency Stops

The LOW FREQUENCY rotary picks one of four engraved stops — 20, 30, 60, 100 Hz — plus a CON position for the continuous-frequency mode described below. Each stop swaps one capacitor in the BOOST bank and one capacitor in the ATTEN bank, so the two turnover ratios drift slightly across the four stops rather than holding to the single nominal value. The plugin's Hardware Accurate mode reproduces these per-band ratios from the EQP-1A3 and DIYRE EQP5 cap-bank measurements published on the GroupDIY and DIYRE forums; Hardware Accurate off uses one representative multiplier (6× boost, 13.5× cut) across every band.

The ratios above are not nominals — they are the per-band fits derived from EQP-1A3 / DIYRE EQP5 cap-bank measurements published on the GroupDIY and DIYRE forums. The 20 Hz cap pair is measurably looser than the others; the 100 Hz cap pair is measurably tighter. The mean across the four bands matches the single nominal multiplier used outside Hardware Accurate.

LF Boost

Continuous 0 to +13.5 dB. The BOOST knob is a first-order low shelf whose turnover sits at approximately six times the dial frequency — so a 60 Hz dial gives a corner around 360 Hz. Below the corner the shelf lifts toward its full dialled gain; above the corner it returns to 0 dB at 6 dB/oct. The shelf is bilinear-prototyped from H(s) = (G·s/wc + 1) / (s/wc + 1), so the transition through the corner is a single smooth curve with no overshoot.

The BOOST shelf is wired in parallel with the ATTEN shelf, not in series. Internally the two first-order transfer functions are summed as H_boost + H_cut − 1 and folded into a single second-order biquad at coefficient-update time. This is what makes the Pultec trick a mechanical certainty rather than a happy coincidence — the BOOST and the ATTEN do not stack the way two cascaded shelves would; they sum the way the passive network sums them.

LF Atten

Continuous 0 to −17.5 dB cut. The ATTEN knob is also a first-order low shelf, but its turnover sits significantly higher than the BOOST corner — approximately 13.5 times the dial frequency under the default (non-HA) ratio, rising to 15× at the 60 Hz stop under Hardware Accurate. With a 60 Hz dial, that puts the ATTEN corner near 910 Hz. The analog prototype is H(s) = (s/wc + 1/G) / (s/wc + 1): gain 1/G at DC, unity above the corner, level at the shelf floor rather than tilting the top end.

The two corners diverging — BOOST corner near 360 Hz, ATTEN corner near 910 Hz at the 60 Hz stop — is the engineering reason the Pultec trick produces a dip rather than a cancellation. Below the BOOST corner both shelves contribute lift; between the two corners the ATTEN subtracts faster than the BOOST adds, and the difference shows up as an emergent shallow dip in the upper bass / lower mid; above the ATTEN corner both shelves are done and the response is flat. Move the dial and the dip moves with it; lift the BOOST without engaging the ATTEN and the dip never forms.

Hardware Accurate Toggle

A round chip directly below the knob row, in the LOW section's bottom band. The label sits left of the chip; the dot inside the chip lights when engaged. Engaging this toggle puts the entire chassis into Hardware Accurate mode: it pins the LOW section's frequency rotary back to its four engraved stops (disabling CON for every section), swaps the single nominal turnover multipliers for the per-band fits in the table above, forces the make-up amplifier's Voltage to 117 V / Wear to 0 % / Valve to 12AX7 / Transformer Headroom to 100 %, forces Character (Drive) to 0 dB and Auto Gain off, and engages the make-up amplifier's residual flat-state LF shelf and overall lift. The toggle is one chassis-wide master mode — see the Hardware Accurate Mode chapter for the full list of locks and the audible character difference.

LF CON (Continuous-Frequency Mode)

When the LOW FREQUENCY rotary is rotated to the CON position, the section's effective frequency is no longer read from the four-stop array — it is read from a small LF CON knob that becomes active in the same row. The CON knob is log-scaled across the full LOW range (20 to 100 Hz), centred so the canonical 60 Hz position lies near the middle of the throw, and its value drives both the BOOST and the ATTEN shelves (they always share the dial). Hardware Accurate forces the rotary back to a stepped position and dims the CON knob — the per-band measured turnover ratios HA brings in are only defined at the four engraved stops, so allowing CON under HA would be a contradiction the chassis refuses to entertain.

HPF Helper Knob

A small knob between the LOW FREQUENCY rotary and the LF BOOST knob, dimmed when the HPF chip in the helpers row is off. The HPF is a 12 dB/oct Butterworth high-pass, log-scaled across 20 to 800 Hz with a default at 80 Hz. The crucial point about the HPF is its position in the signal flow: it sits ahead of the passive-EQ network, so a high-passed signal is what reaches the LF BOOST shelf — the HPF cuts the rumble before the BOOST has the chance to lift it. This is the standard mix-bus move that the original EQP-1A could not make without an outboard high-pass in front of it; the plugin builds the high-pass into the same chassis. Under LINK off the HPF has independent values per side (L/M and R/S), like the rest of the helper row. See the Helper Filters chapter for the routing detail and the per-side behaviour.

The HIGH section combines two related but mechanically distinct EQ blocks. The HF BOOST is a series-resonant LC tank — a tuned circuit whose centre frequency is set by the LC product and whose Q is set by a damping resistance, producing a peaking response with the Q and the peak amplitude mechanically coupled. The HF ATTEN is a separate first-order high shelf cut, fed from a different cap bank, with its own frequency selector that picks the dial position — not the corner — of the shelf. The two blocks share neither capacitors nor frequency arrays; each has its own rotary, and each has its own CON knob for continuous-frequency work.

HF Boost

HIGH FREQUENCY · KCS (rotary)

Seven engraved stops plus a CON position: 3, 4, 5, 8, 10, 12, 16 kHz. Each stop selects a different inductor / capacitor pair from the original tank, and the choice is therefore discrete rather than continuous on the hardware. The 8 kHz stop is the factory default — the most-used position on every EQP-1A in working studios, and the one the plugin loads with on a fresh insert.

The eighth position on the rotary is CON, which hands the section's frequency to the HF CON knob below.

Boost (HF)

Continuous 0 to +18 dB. The shape is a peaking EQ centred at the selected HF, implemented as an RBJ peaking biquad, with the peak's Q set by the BANDWIDTH knob below. The crucial coupling is mechanical, not arbitrary: the original LC tank's loaded Q and peak height move together — sharper Q rings harder and reaches the full +18 dB; broader Q is more damped and sheds peak height. The measured peak-gain swing across the BANDWIDTH range is approximately 9 dB at full boost (sharp +18 dB, mid +13.5 dB, broad +9 dB), per the CARTEC measurement reference from the 1990s. The plugin reproduces that 9 dB swing at the BANDWIDTH endpoints exactly and interpolates linearly between them.

Bandwidth (HF Boost)

Continuous 0 to 100 %. Dials the peak's Q from broad (Q ≈ 0.6 at 0 %) to sharp (Q ≈ 2.5 at 100 %). The 50 % midpoint sits at Q ≈ 1.5, a usefully musical neighbourhood between the two extremes. The coupling between BANDWIDTH and peak gain follows the LC tank's loaded Q: as Q rises the tank rings harder at resonance and the peak rises with it, so a sharp HF BOOST at +12 dB sounds different from a broad HF BOOST at +12 dB — the broad version sheds peak height even before the BANDWIDTH itself has been reduced. The 9 dB swing referenced above is the gain-vs-Q coupling preserved exactly in the plugin.

HF Atten Select

A separate three-position rotary plus CON: 5, 10, 20 kHz. The figure on the dial is not the corner of the shelf — the actual shelf corner sits at the dial frequency divided by approximately 7.5. This is the measured turnover ratio of the EQP-1A's HF-cut network (Ian Bell's GroupDIY analysis of the factory "5/10/20 KC SHELF ATTEN FAMILY" curves), and it is the same kind of constant the LOW section's 6× and 13.5× ratios are. The dial marks the frequency at which the shelf has reached its floor; the corner lies that 7.5× ratio below it.

Atten (HF Cut)

Continuous 0 to −16 dB. Sets the shelf floor depth at the corner. The shelf is first-order with the analog prototype H(s) = (G·s/wc + 1) / (s/wc + 1), G < 1: flat at DC, gain G above the corner. The response bends down from 0 dB at the corner and levels onto a floor at the dialled depth — it does not roll off toward silence the way a low-pass would. A resistor in the passive network sets that floor, and the shelf is therefore a bounded cut. ATTEN at −10 dB with the dial at 10 kHz reads as the room got a little smaller and a little softer, not as the highs disappeared.

HF CON / Atten CON

Each section has its own continuous-frequency knob that becomes active when its rotary is set to CON. The HF CON knob is log-scaled across the HF BOOST range (3 to 16 kHz, default 8 kHz); the Atten CON knob is log-scaled across the HF ATTEN range (5 to 20 kHz, default 20 kHz). Hardware Accurate snaps both rotaries back to their nearest engraved stops and dims the CON knobs — the same logic that gates the LF CON knob.

LPF Helper Knob

A small knob between the HIGH FREQUENCY rotary and the HF BOOST knob, dimmed when the LPF chip in the helpers row is off. The LPF is a 12 dB/oct Butterworth low-pass, log-scaled across 1 to 20 kHz with a default at 12 kHz. Like the HPF, the LPF sits ahead of the passive-EQ network — so the upper-mid harshness is smoothed before the make-up amplifier's tube cascade has a chance to saturate it into something harder. Under LINK off the LPF carries independent values per side. See the Helper Filters chapter for the routing detail.

The OUTPUT section is the chassis's tail. It carries CHARACTER (the make-up amplifier drive), STRENGTH (the Auto Gain depth), GAIN (the output trim) and the AUTO GAIN toggle inline below STRENGTH. The four controls share a row, with the AUTO GAIN chip docked under its STRENGTH partner so the toggle and its depth knob read as one block of control.

Character (Drive)

Continuous 0 to +12 dB. CHARACTER drives the make-up amplifier harder without altering the overall through-level: internally the chassis lifts the ECC83 pre-gain by the dialled dB value and applies a matching post-gain trim, so the level reaching the Auto Gain comp stays consistent while the tube cascade sees more grid swing and saturates accordingly. It also progressively opens the global negative-feedback loop (see section 13) — at CHARACTER 0 the loop is fully closed and the cascade is at its cleanest; toward CHARACTER 12 the loop opens up to 85 % of its full depth and the open-loop tube colour begins to dominate.

Hardware Accurate forces CHARACTER to 0 dB and locks it there. The label dims and the knob refuses input until HA is released.

Strength (Auto Gain depth)

Continuous 0 to 100 %. Active only when Auto Gain is engaged. Sets how aggressively the Auto Gain compensator pulls the make-up amplifier's output back to perceived unity against the live EQ + CHARACTER settings.

• 100 % — full compensation. A +10 dB LF BOOST is matched by a near-equivalent output trim so the chassis sounds the same loudness with and without the boost engaged. Useful for A/B work where the EQ shape is what matters and the level shift would be a distraction.
• 50 % — half compensation. The level shift of the EQ is partly preserved (you still hear that bass added energy as energy), partly compensated (you do not lose your monitoring sweet spot).
• 0 % — no compensation, but Auto Gain still measures. Equivalent to Auto Gain off for the audio path; the live readout still moves.

Hardware Accurate forces Auto Gain off and dims the STRENGTH knob — the original EQP-1A had no compensator, so HA disables both. A live readout below the STRENGTH knob shows the current Auto Gain trim in dB, so you can see what the compensator is doing even before deciding how much of it to apply.

Gain (Output)

Continuous −12 to +12 dB. A bipolar trim at the chain tail, after the make-up amplifier and the Auto Gain compensator. GAIN is purely a level move; it does not feed back into the saturation, the feedback loop or the iron — it sits at the very end of the signal path. Under LINK off the GAIN carries independent values for the L/M and R/S sides, like the rest of the per-side controls.

Auto Gain (toggle)

A gold round chip inline below the STRENGTH knob, with a dot inside that lights when engaged and an ON / OFF caption left of the chip. Engages the Auto Gain compensator; the STRENGTH knob activates with it. See the Auto Gain chapter for the pink-power weighting it uses and how the compensation differs from a LUFS measurement. Hardware Accurate forces this off.

The EQP-1A's passive EQ network inserts approximately 15 dB of broadband loss regardless of the dial position — it must, because a passive network can only attenuate; every shelf and tank in front of you is paid for with a constant loss in the rest of the band. The make-up amplifier recovers that loss with approximately 17 dB of closed-loop gain, leaving the chassis at unity at its I/O. The two extra dB account for the iron-loss and inter-stage trim under fresh-valve / nominal-mains conditions.

The signal path through the amplifier is a transformer-coupled cascade of two push-pull triode stages wrapped in a global negative-feedback loop:

EQ output ─► Input transformer ─► ECC83 push-pull ─► Interstage transformer ─► ECC82 push-pull ─► Output transformer ─► Out
                                          ▲                                                              │
                                          │                                                              │
                                          └────────────────── Global negative feedback ◄─────────────────┘
                                                              (tertiary winding to ECC83 cathodes)

Each block in that diagram is modelled as the physical thing it is, not approximated.

Input ECC83 Push-Pull

Two triode sections of one ECC83 sit either side of the transformer-coupled load; the signal arrives at the grids in anti-phase and the differential plate-current sum drives the interstage primary. A matched pair cancels every even harmonic and passes only odd-order content; real valves carry a small section-to-section transconductance mismatch, and that residual mismatch lets some 2nd harmonic through. The plugin's nominal imbalance is 3 % at Wear 0, rising linearly to 10 % at Wear 100 % — the chassis sounds slightly brighter / more honest fresh, slightly darker / fatter old.

Plate current is evaluated from the Maillet six-term polynomial fit to a measured RCA 12AX7: each of the five ln(Va)-polynomial coefficients is itself a degree-7 Horner polynomial in Vg. The model holds the real tube's low-plate-voltage knee, where the simple Koren space-charge law begins to diverge from measurement. The operating point is approximately 140 V plate and 1 mA per section — well inside Maillet's accurate region.

The load line is solved against the triode curve at bake time, off the audio thread: Newton iteration to a residual of 1e-12, warm-started from the previous solve, with the model supplying the exact slope d(Ip)/dVa. The solved transfer surface is baked into a 2048-point lookup table per supply-sag curve, with 7 sag curves spanning the programme-dependent rail droop. The audio thread reads via linear interpolation between adjacent LUT points, and the read itself is band-limited by the first-order antiderivative anti-aliasing formulation: the per-sample output is (F(Vg) − F(Vg_prev)) / (Vg − Vg_prev), the mean transfer across the sample's grid-swing span rather than the instantaneous transfer at one point. The antiderivative is kept in double precision because the numerator is a difference of two near-equal large numbers.

Interstage Transformer

A signal-path transformer modelled as its equivalent circuit: a Thevenin source resistance driving the non-linear magnetising core, with the reflected load shunting it. The core's incremental inductance falls from its unsaturated value Lm toward the air-core value Lm/25 as the flux linkage grows, with a smooth tanh knee at the saturation flux level. The polarity asymmetry of the knee (0.5 in the equivalent circuit) is what lets the core produce even harmonics alongside the odd ones the push-pull cancels — a single matched valve pair would otherwise be uncomfortably odd-harmonic-only.

The flux linkage is an integrator state, advanced one sample per audio sample by a backward-Euler step solved by Newton iteration (2 iterations, warm-started) — implicit integration so the integrator is stable at any drive level, with the exact slope of the magnetising current supplied by the core model. The magnetising current is read with the same ADAA-1 antiderivative formulation as the tubes: (F_core(λ) − F_core(λ_prev)) / (λ − λ_prev), the mean current across the sample's flux span. The antiderivative integrates to λ² / (2·L_air) + c·λ_sat²·ln(cosh(λ/λ_sat)), evaluated in double precision through a numerically-stable ln(cosh) form so a hard-driven core does not lose precision at the difference quotient. The loaded Q of the interstage affects the upper-band response — the HF BOOST tank's ring is partly damped by the interstage primary's reflected impedance.

Output ECC82 Push-Pull

The ECC82 (12AU7) is a lower-mu, higher-current twin triode than the ECC83 — designed for the buffer / output role rather than the small-signal input role. The operating point is approximately 250 V plate and 8–10 mA per section. Plate current is evaluated from the Koren 1996 closed-form space-charge equation fitted to the Philips ECC82 datasheet: 21 independent operating points cross-checked against the datasheet's self-bias load tables, 4.1 % mean error. The closed-form Koren equation is computationally lighter than the six-term Maillet polynomial — one logarithm, one softplus, one log-exp pow — and the ECC82's lower-mu / higher-current region is well inside Koren's accurate envelope, so the loss in fit accuracy is below the audible threshold.

The same pipeline applies: Newton-solved load line at bake time, baked into a 2048-point LUT per sag curve, audio-thread read by linear interpolation, output band-limited by ADAA-1 antiderivative anti-aliasing. The audio path does not run a per-sample Koren evaluation — it reads the baked transfer.

Output Transformer

The same equivalent-circuit model as the input and interstage transformers, but with Lm / saturation-knee parameters chosen to match the Peerless S-217-D family — the most-cited reference for the original Pultec output transformer. The output transformer carries the most magnetising inductance (10 H against 4 H input / 6 H interstage) and the lowest saturation knee (9 mWb against 18 / 12), so it is the iron that flat-out first when the chassis is driven hard. Its tertiary winding closes the global feedback loop, and the feedback loop linearises it almost completely under closed-loop conditions — see below.

Global Negative Feedback

The loop wraps the two valve stages and the interstage transformer, and returns a small portion of the output transformer's secondary back to the ECC83 cathodes in opposing phase. The closed-loop gain settles to A_open / (1 + β·A_open). With A_open of order 50 and β chosen for a closed-loop depth of 10, that is approximately 17 dB closed-loop against approximately 34 dB open-loop — the loop suppresses 17 dB of nonlinear behaviour from the cascade.

Audibly, the loop linearises the cascade. Open-loop the cascade carries about 3 to 4 dB of nonlinear character before the saturation knee — second and third harmonics from the push-pull mismatch and the iron asymmetry, plus the soft compression from the supply sag. Closed-loop the same cascade is clean to within a fraction of a dB across the full programme range up to the saturation knee; above the knee both paths saturate together. CHARACTER (Drive) progressively opens the loop, letting more of the open-loop colour through; Hardware Accurate forces the loop fully closed. The feedback β is computed at prepare time from the measured open-loop gain — which itself depends on Voltage, Wear and Transformer Headroom — so the closed-loop gain stays at design unity regardless of where the make-up amplifier's character knobs are set.

Make-Up Amplifier Panel

The TUBE chip in the OUTPUT row opens the make-up amplifier modal. The modal pauses the chassis behind it (the chassis dims to 30 % and stops responding to clicks) and surfaces the per-side make-up amplifier state on two stacked cards:

┌──────────────────────────────────────────────────────────────────┐
│  Tube                                       L / M side card       │
│  ─────                              ┌──────────────────────────┐ │
│                                     │   Voltage   Wear         │ │
│                                     │   Valve     Transformer  │ │
│                                     └──────────────────────────┘ │
│                                                                   │
│                                              R / S side card     │
│                                     ┌──────────────────────────┐ │
│                                     │   Voltage   Wear         │ │
│                                     │   Valve     Transformer  │ │
│                                     └──────────────────────────┘ │
│                                                  [Close]          │
└──────────────────────────────────────────────────────────────────┘

Each side carries its own independent state for the four make-up amplifier controls — Voltage, Wear, Valve, Transformer Headroom. With LINK engaged, the two sides are coupled and a move on one mirrors on the other; with LINK off, the two sides diverge and the L/M and R/S amplifiers can be set to different valve types, ages, voltages and iron headrooms. The modal is dismissible by click outside the card area, by the Escape key, or by the explicit Close button; the chassis behind is non-interactive while open and re-engages on dismiss.

Hardware Accurate forces the make-up amplifier to its nominal-spec state and locks every control until HA is released:

The Pultec trick is the most-cited engineering technique on the EQP-1A, and the reason the chassis sits on so many mix-buses and so many bass channels. It is what gave the unit its reputation in the first place — and it is one of the very few EQ moves that works the same way on a vocal, a kick, a bass and a full mix-bus, because the mechanism is the same in every case.

What Is It?

Set LOW FREQUENCY to 60 Hz. Turn LF BOOST up to about +5 dB. Turn LF ATTEN up to about +3 to +5 dB of cut. Listen.

What you hear is not a cancellation. You hear a clear lift around 60 Hz, a small dip in the upper bass around 200 to 400 Hz, and a flat response above. The low end is bigger — the chest of the voice or the body of the bass fills out — and tighter — the lower midrange that would otherwise get woolly is cleaned out by the same move that adds the weight. Both knobs are up; the response is not zero.

How It Works

The lift-with-dip shape is the emergent result of the parallel-shelves topology described in section 10:

• The BOOST shelf has its corner at approximately 6× the dial frequency — so a 60 Hz dial puts the BOOST corner near 360 Hz. Below that corner, the shelf lifts toward its dialled gain.
• The ATTEN shelf has its corner at approximately 13.5× the dial frequency — so the same 60 Hz dial puts the ATTEN corner near 910 Hz. Below that corner, the shelf cuts toward its dialled depth.
• Below 360 Hz, both shelves are lifting / cutting. The BOOST is faster than the ATTEN (the BOOST is closer to its corner, so it has already done most of its work), so the net response is a lift — the deep low end.
• Between 360 Hz and 910 Hz, the BOOST is rolling off toward flat (its corner is behind), while the ATTEN is still cutting (its corner is ahead). The ATTEN therefore subtracts faster than the BOOST adds, and the difference shows up as an emergent shallow dip — the upper bass tightening.
• Above 910 Hz, both shelves are done. The response is flat.

The two corners being different is what makes the trick work. If both shelves turned over at the same ratio of the dial, the BOOST and the ATTEN would track each other and there would be no dip. They turn over at different ratios because they are fed from different cap banks in the passive network — the BOOST capacitors and the ATTEN capacitors are deliberately scaled to different values. The plugin's LF section preserves both corners exactly, so the dip is mechanical rather than scripted.

The Result

Typical Settings

Why It Works

The Pultec trick is a happy accident of the EQP-1A's passive LC topology. The two shelves are mechanically tied to slightly different ratios of the dial frequency because they are wired to different capacitor banks in the network, and those banks were chosen to produce the shelves their designers wanted — not to produce a trick. The trick is what emerges when both knobs are turned up at once, and it is the reason the EQP-1A turned out to be a more musical equaliser than the schematic alone would suggest.

The shape cannot be exactly reproduced on a parametric EQ without explicitly setting up two shelves at the correct corners with the correct gains — and even then the depth of the dip and its tracking with the dial frequency is fiddly to dial in. On the EQP-1A it is one move: set the dial, turn both knobs up. The Niviem plugin's LOW section preserves the corners exactly, so the trick lands the same way it does on the original chassis. The factory preset Helper - Pultec + Clean Sub puts the chassis directly at a tested Pultec-trick state with the HPF helper engaged below it, so the subsonic content the trick would otherwise lift is removed at the input.

Hardware Accurate (HA) is the chassis's master switch — engage it and every Niviem extension snaps back to its schematic-faithful value, instantly. The toggle sits below the knob row in the LOW section, sandwiched between the EQ controls and the make-up amp panel so a glance at the chassis tells you which mode the unit is in. HA was the answer to a single question: "what does this preset sound like on a 1962 EQP-1A?" One click answers it.

What HA forces

Every Niviem extension is pinned to a hardware-original value while HA is on, and its UI control is dimmed / disabled so the locked state is visible at a glance. The table below is exhaustive — if it is not here, HA does not touch it.

What HA leaves alone

The EQ section proper, the workflow controls, and the output trim are all free under HA — they are not extensions, they are the original equaliser. Output Gain sits in this list because it is a purely post-amp trim: it does not alter the make-up amp's character, it just rescales what comes out, so leaving it free under HA does not break the schematic story.

What HA preserves across the cycle

HA is non-destructive. Engaging it does not zero the stored values for the extensions it forces off — the chassis applies the override at signal-path resolution time and leaves the parameters' stored values alone. Disengage HA and every extension wakes up at whatever it was set to before. The user can dial a heavily-extended sound (M-S routing, CON freqs, 5751 valve, 70 % drive), click HA on to A/B against the strict 1962 chassis, click HA off again, and land back exactly where they were. This is the deliberate opposite of the LINK toggle, which destroys sister-side values on the off → on edge (see § 19).

Click-free engagement

HA flips many sub-flags at once — the CON snap-back on three sections, the tube forced on, AG forced off, LINK override, route forced to Stereo, every per-side character pin. The per-parameter coefficient smoothers cannot keep all of that click-free; even with 8 ms biquad ramps everywhere, the combined topology shift across a single instant is audible. The chassis raises a click-suppression flag every time HA toggles, and the wet path drops through an 11 ms mute envelope (3 ms fade-down + 5 ms hold + 3 ms fade-up) across the switch. The dry path the bypass mixer holds is untouched — only the processed signal is muted, so a non-bypassed chassis crossing an HA toggle hears a brief dip rather than a click.

When to use HA

Use HA when:

• A/B-ing a pure-Pultec voicing against the Niviem extensions on the same source — the cleanest possible comparison of "what does CON / M-S / tube-swap / drive add?"
• Mastering sessions that need a reference 1962 chassis on the bus, with the option to dial in extensions later from the same preset.
• Confirming a mix-bus instance is not leaking creative colour you forgot you dialled in three sessions ago.

Use unlocked when:

• The LF section needs an intermediate frequency the engraved stops do not cover (CON).
• The signal needs the HPF / LPF helpers ahead of the EQ section.
• You are voicing through different valves, mains voltages, or tube wear amounts.
• Mid and Side need to diverge (M-S routing + LINK off).
• The bypass A/B needs to be loudness-honest (Auto Gain).

Both filters are opt-in Niviem extensions: a 12 dB / octave Butterworth high-pass and a 12 dB / octave Butterworth low-pass, sitting ahead of the EQ chain. Both are off by default; both engage from a chip in the header; Hardware Accurate forces both off.

Why include them

The EQP-1A is a programme equaliser. It was designed to shape the broadcast / mastering tone of a balanced source, not to remove unwanted content from one. Modern programme material almost always wants two things the EQP-1A cannot do on its own: strip the lowest octave (rumble, mic stand vibration, DC offset, sub-bass content that eats headroom on the limiter) and smooth the upper-mid harshness on hot digital sources. Engineers in 1962 reached for a separate HPF / LPF unit for exactly this — Pultec's own HLF-3C is the canonical companion. The Niviem chassis bundles the equivalent into one strip so the typical chain (clean up → equalise → make-up gain) lives in one window. Hardware Accurate forces both filters off because the original does not have them — the helpers are a workflow extension.

HPF (High-Pass Filter)

12 dB / octave Butterworth, cutoff range 20 Hz – 800 Hz log, default 80 Hz. The filter sits at the very top of the chain — before the input transformer, before the passive EQ network, before the make-up amplifier — so the EQ and the valve never see content the user has decided to throw away.

Per-side under LINK off — useful in M-S where the Mid channel carries a tight HPF (kick / vocal centre) while the Sides stay open (room ambience, panned overheads).

LPF (Low-Pass Filter)

12 dB / octave Butterworth, cutoff range 1 kHz – 20 kHz log, default 12 kHz. Same chain position as the HPF — ahead of every coloured stage so the valve does not see content that gets thrown away anyway.

Per-side under LINK off.

Workflow combinations

The factory bank ships an eight-preset Helper Filters category that walks through the most useful HPF / LPF / EQ combinations:

• Helper - Tight Kick — HPF at 40 Hz strips the inaudible sub; the LF Boost at 60 Hz lifts the kick fundamental.
• Helper - Pultec + Clean Sub — gentle HPF + the classic Pultec trick at 60 Hz; the sub stays clean while the low-end gets its trick lift.
• Helper - Vocal Cleanup — HPF at 100 Hz strips plosives and rumble; HF Boost adds presence.
• Helper - Smooth Mix Bus — gentle HPF + gentle LPF + small HF lift; the textbook mastering chain ahead of the make-up amp.
• Helper - Smooth Saturation — HPF + LPF braces the tube path so the valve only sees the band of content that benefits from saturation.
• Helper - M-S Bass Tighten — M-S routing + tight HPF on the Sides only; cleans up the low end without losing the centre fundamental.
• Helper - Lo-Mid Cleanup — HPF at 200 – 400 Hz on a muddy source; reaches where LF Atten cannot.
• Helper - Lo-Fi Dub — heavy LPF at 1 – 2 kHz + TUBE + LOW Boost; saturated dark dub-bass voice.

Click-free toggling

Both chips engage and disengage through a 30 ms wet / dry crossfade smoother. Off means the biquad is skipped entirely — zero CPU. On means the filter runs with its coefficients smoothed across cutoff changes: a knob drag walks the cutoff through a multiplicative (log-domain) glide rather than block-aligned jumps, which would crackle a direct-form biquad. The mix smoother hides the toggle engagement; the cutoff smoother hides drag-induced coefficient swaps. Both are inaudible in practice.

CON is a Niviem extension that releases the chassis's frequency selectors from their stepped rotaries onto continuous knobs. Each of the three frequency selectors — LOW FREQUENCY CPS, HIGH FREQUENCY KCS, and ATTEN SEL KCS — carries an opt-in CON stop at the end of its rotary list.

Why CON exists

The EQP-1A's rotary stops are exact LC component values — 20 / 30 / 60 / 100 Hz for LF, 3 / 4 / 5 / 8 / 10 / 12 / 16 kHz for HF, 5 / 10 / 20 kHz for HF Atten. Every other frequency is unreachable on the real hardware, because the inductors and capacitors that set those corners are physical components on a switch wafer. Modern programme material occasionally wants intermediate frequencies — a 70 Hz vocal mid-warmth that sits between the 60 and 100 Hz stops, a 9 kHz sibilance-band cut that sits between 8 and 10 kHz, a 7.5 kHz snare-stick presence lift. CON opens those frequencies up without changing the chassis character: the EQ topology is identical, the only difference is that the corner frequency is dialled freely rather than picked from a stop.

Engaging CON

Each rotary's last stop is labelled CON. Click that stop and the section switches to continuous mode: the matching continuous-frequency knob (LF CON, HF CON, or Atten CON) illuminates and becomes the live control, while the stepped rotary stops dim to show the section is no longer stepped. Click any stepped stop on the rotary to return to stepped mode. Hardware Accurate forces every section back to its stepped state and disables the CON stop on each rotary; the user's stored CON values are preserved across the HA cycle and restored when HA toggles back off.

Per-section behaviour

Each CON range spans the stepped extremes — the continuous knob reaches every frequency the rotary covers, plus every frequency between. The log scaling means every octave of sweep occupies the same arc on the knob, which matches the way the ear perceives a frequency change.

Per-side mode under LINK off

Each section's CON mode is a per-side state. Under LINK off, the bare side and the sister side can carry independent CON / stepped states — the Mid channel can sit on stepped 60 Hz while the Side channel runs on continuous 75 Hz, for example. The combination is unusual but supported. Under LINK on, the two sides are coupled and CON engagement mirrors between them.

The Route chip in the header flips the entire EQ chain between two domains — STEREO (the default, a linked-pair) and M-S (matrix-encoded Mid + Side processed independently). The Route chip sits next to the LINK chip in the header; the two of them together decide whether the chassis runs as a single stereo unit, two independent stereo channels, or two independent Mid / Side channels.

Why M-S

M-S is a standard mastering trick for opening the stereo image without re-mixing the source. The Mid channel carries everything that is identical on L and R — the bass, the lead vocal, the kick, the snare, anything panned to centre. The Side channel carries the L – R difference — the room reverb, the panned overheads, the panned synths, every component of the stereo width. Processing Mid and Side independently lets the engineer do things a single linked equaliser cannot: tighten the Mid's bass without thinning the Sides, lift the Side high-frequencies for air without making the Mid harsh, de-ess the Sides only (the Mid vocal is untouched), or run a sharper Pultec trick on the Mid while the Sides stay open in the low-end. The original EQP-1A is a mono unit — Niviem runs two complete chassis instances internally, one per side, to make M-S work on a single plugin instance. That is the thing the original chassis cannot do physically; it is the most powerful M-S Pultec emulation available because both halves are complete chassis, not a single chassis fed an encoded signal.

How the matrix works

Encode:    M = (L + R) / 2
           S = (L - R) / 2

Decode:    L = M + S
           R = M - S

The encode uses a 0.5 scale on both channels; the decode uses a 1× scale. The round-trip is bit-perfect identity at unity gain, so a chassis in M-S mode with both halves flat is sample-equivalent to the dry input. The 0.5 / 1× asymmetry is deliberate: it leaves the Side channel sitting at half the original L – R amplitude during processing, so a Side EQ boost reacts to the same input level as a Stereo-mode boost on the same input. The user dials against what the EQ plot says, not against an internally-gained difference channel.

Per-side parameter divergence under LINK off

Under M-S with LINK off, the Mid channel carries the "bare" parameters (LF Boost, HF Boost, HPF, LPF, Drive, Voltage, etc.) and the Side channel carries the "_R" sibling parameters (LF Boost R, HF Boost R, HPF R, LPF R, Drive R, Voltage R, etc.). The two sides are entirely independent — different EQ curves, different helper filters, different make-up amp valves, different mains voltages, different tube wear. The L-side knob row drives the Mid; a second small "ms-sister" row that appears next to each main knob under LINK off drives the Side. Under LINK on, the two sides are coupled and the sister parameters mirror the bare parameters automatically.

Click-free engagement

Engaging or disengaging M-S is a topology change — the encode and decode matrices flip in and out between blocks. The chassis raises its 11 ms wet-path mute envelope across the toggle (the same envelope HA uses), so the routing swap is hidden under a silence window rather than a click.

Hardware Accurate

HA forces the routing to STEREO and disables the Route chip. The chassis returns to its 1961 mono-amp-twice-over linked-pair behaviour regardless of what the Route control stores; the user's M-S preference is preserved across the HA cycle.

The LINK chip in the header controls whether the two sides of the chassis (L / R under STEREO routing, M / S under M-S routing) carry coupled or independent parameter values. LINK is the dial that decides whether the user sees one set of knobs or two.

LINK on (default)

Every L / R parameter pair stays equal. Moving LF Boost on the L side mirrors instantly to LF Boost R; moving the make-up amp voltage on the centre instance mirrors to the side instance; moving the HPF cutoff on one side mirrors to the other. The user sees a single set of knobs; the chassis maintains a linked L / R pair internally. This is the canonical 1961 behaviour — the original EQP-1A is a mono unit, so a stereo pair was patched as two identical units with the same settings on both. Under LINK on, parameter mirroring goes through the host's automation lane (so DAW undo, automation pickup, and host metering all stay coherent).

LINK off

The two sides decouple. The sister L / R knobs become visible — small "ms-sister" duplicates appear next to each main knob — and each side carries its own values; moving one knob does not affect the other side. LINK off is what unlocks Niviem's full M-S engineering: with both LINK off and Route on M-S, the centre and the sides can carry different EQ curves, different helper filters, different make-up amplifiers (each with its own valve, voltage, wear, and transformer).

Hardware Accurate

HA forces LINK on and disables the LINK chip. The chassis runs as a linked stereo pair regardless of the stored LINK state. The user's per-side parameter values are preserved across the HA cycle — HA's override is non-destructive (§ 15).

Engagement is destructive (deliberately)

Toggling LINK off does not change the sister parameters — they read whatever was last stored, so the divergent state appears immediately. Toggling LINK back on is destructive: the chassis copies every bare parameter onto its sister on the rising edge, so any per-side values the user dialled under LINK off are discarded. This destructive-on-edge behaviour is deliberate: when the user wants the two sides to be equal again, they want them equal now, not slightly different because the sister side was 1.7 dB off. This is the deliberate opposite of HA, which is non-destructive. Be intentional with the LINK on transition — your divergent per-side work is lost in that direction.

Auto Gain is a Niviem extension that automatically compensates the chassis output level for the perceived loudness change of an EQ move. With AG on, the bypass A / B answers the right question — "is this EQ move good?" — honestly, because engaging and disengaging the chassis no longer changes the perceived level, only the spectrum.

Why Auto Gain

The loudness war happens at the engineer's bypass button. A 6 dB boost at the right frequency makes the signal sound "better" almost regardless of whether the EQ move is any good — louder reads as bigger, fuller, and more impressive, and the ear cannot easily separate the spectral change from the level change. The only way to A / B an EQ move honestly is to keep the perceived level constant on both sides of the toggle. The traditional fix is a hand-trimmed output gain, applied by the engineer who tries to match levels with their ear. It works, but it takes thought and it is slow. Auto Gain compensates automatically: the plugin measures the heard EQ curve, integrates a pink-weighted loudness across it, and applies a compensating trim at the output. The bypass A / B is loudness-honest from the first dial move.

How the compensation is computed

1. Grid: 64 log-spaced frequency points across 20 Hz – 20 kHz. Every octave occupies the same number of grid points — a pink (flat-octave) measure.
2. Heard magnitude per point: the passive EQ shape (LF + HF + Atten sections) + the make-up amp character contribution (a constant +1.13 dB lift, plus a small −1 dB shelf below 150 Hz, but only under Hardware Accurate; without HA the character contribution is 0 dB because the Drive control replaces the schematic character with the user's own) + the HPF / LPF contribution if engaged.
3. Mean-power integration: each magnitude is converted from dB to linear power (10^(dB/10)); the mean of those 64 values is the section's pink-weighted mean-power gain.
4. Compensation: in dB, compDb = −10 × log10(meanPower). A boost-heavy curve gives a negative compensation (the chassis is louder, the trim cuts); a cut-heavy curve gives a positive compensation (the chassis is quieter, the trim lifts).
5. Per-channel: under LINK off, the Mid and Side sides each compute their own mean-power and apply independent trims, so the two sides hold their own perceived level even when their EQ curves diverge.
6. Strength: the Strength knob (0 – 100 %) scales the compensation depth; the final trim in dB is compDb × strength.
7. Smoothing: the trim is smoothed into a per-side smoothedAutoGain ramp that matches the EQ coefficient smoother (the same 8 ms ramp), so the trim glides into place alongside the new curve, with no perceptible level swell across the difference.

The Strength knob

0 – 100 %. Lower values let some of the EQ-induced loudness change through, which is useful when you want the EQ to add presence without fully levelling itself.

• 100 % — full compensation, the bypass A / B holds level perfectly.
• 50 % — half compensation, the EQ keeps half of its natural loudness rise (a useful default for "add presence" preset work).
• 0 % — effectively off, the trim collapses to unity.

HA forces AG off and dims the Strength knob.

The live trim readout

A small chip below the Strength knob reads the current trim in dB. It reads 0.0 dB when AG is off; it reads the negative dB of the computed compensation when AG is on (a boost-heavy EQ shows a negative trim, e.g. −3.2 dB, indicating the chassis is cutting 3.2 dB at the output to balance the boost; a cut-heavy EQ shows a positive trim). Under LINK off, the readout splits to per-side values (L / R under STEREO, M / S under M-S).

What Auto Gain does NOT do

Auto Gain is not a loudness normaliser. It does not target a fixed LUFS level (−23, −14, or anything else); it compensates the change in pink-weighted power across the bypass toggle and stops there. A song that came in at −10 LUFS and a song that came in at −20 LUFS will both pass through with a chassis trim derived from the EQ shape alone — the levels they leave at are whatever level the host delivered, plus or minus the EQ-shape compensation.

Auto Gain is not a brick-wall limiter. It can boost the output above the input if the EQ section is cut-heavy. A signal that was clipping at the input can still clip at the output; AG does not protect against clipping, it only compensates for spectral change.

When to use Auto Gain

• Mix-bus EQ moves, where any unintended loudness drift would bias every downstream decision.
• A / B preset comparison, where presets are dialled at different overall gain levels and the user needs to compare what they do rather than how loud they are.
• Mastering A / B, where the bypass toggle must be loudness-honest by definition — anything else is self-deception.

When to turn Auto Gain off

• Tracking with a target tone, where the engineer wants the loudness rise of an EQ boost as part of the desired delivery.
• Surgical cuts, where the loudness change is the move (notching out a resonance, for example, where the resulting quieter signal is the whole point).
• Hardware Accurate, which forces AG off automatically — the trim is a Niviem extension, not a 1962 feature.

The Make-Up Amplifier panel exposes a per-side Valve selector with three options for the ECC83 input pair: 12AX7, 5751, 12AY7. This is a Niviem creative extension. The original EQP-1A was designed for the 12AX7 (its ECC83 equivalent) — there is no internal switch, and no documented hardware support for a different valve. The plugin's swap exists because tube-rolling (physically swapping the input valve in a hardware EQP-1A) is a documented modification in hi-fi and pro-audio communities, and the alternative valves' transfer characteristics are well-published.

The three valves

Each option is a real DSP model — a different transfer function, a different load line, a different Newton solve, a separately baked LUT. The Koren space-charge model is fitted to the 5751 from the GE GL-5751 datasheet and to the 12AY7 from the Tung-Sol 12AY7 datasheet, each verified within roughly 2–3 % against the published operating points.

Effect on the chassis's sound

A lower µ means a lower gain through the input stage, which means more headroom before the nonlinear region. On a hot signal, the 12AX7 saturates first (audible harmonic content at lower input level), the 5751 holds clean longer, and the 12AY7 is the cleanest of the three. On a normal-level signal sitting close to the chassis's nominal operating point all three sound very similar — the saturation knee lives below the signal level on each, and the three transfer curves are near-linear there. The valves diverge as the input gets hotter.

Per-side under LINK off

Under M-S routing with LINK off the centre and sides can run different valves. The mid channel can carry the standard 12AX7 with its classic character while the sides run the 5751 with cleaner headroom — useful for stereo programme material that would sound harsh under the heavy 12AX7 colour across the sides.

Hardware Accurate

Hardware Accurate forces the Valve selector to 12AX7 and disables the control. The chassis runs the 1961 stock complement.

Workflow notes

Tube swap is subtle by design. Don't expect the 5751 to "sound dramatically different" from the 12AX7 — the chassis runs at a low operating point and the valve change only shows up in the saturation region. The swap is for the engineer who has dialled in everything else and wants to fine-tune the make-up amplifier's character for a specific source. The factory preset M-S - Independent Tubes demonstrates per-side valve divergence.

The Make-Up Amplifier panel exposes a per-side Voltage knob — 100 V to 127 V continuous, default 117 V. The knob scales the make-up amplifier's plate supply voltage. The 1961 EQP-1A was designed for 117 V US nominal mains, regulating its rails through a 6X4 rectifier to a ±150 V plate supply. Real-world mains varies — 110 V on a long cable run, 125 V on a fresh circuit, 105 V during a brownout. The Voltage knob sweeps the chassis through that range.

Effect on the chassis's sound

Lower mains drops the plate supply, which drops headroom, which means the valve sags faster under signal peaks — a browner, fatter character. Higher mains lifts the plate supply, giving more headroom and a cleaner character with more clean range before saturation. The supply also affects the valve's operating point — a lower plate voltage shifts the load line, changes the entry point into the nonlinear region, and slightly changes the harmonic spectrum. The plugin recomputes the load-line solve and re-bakes the valve LUT on every Voltage change. The audible result is a true voltage-dependent valve response, not a level scaling.

Per-side under LINK off

Independent Mid / Side voltage settings — useful for stereo character work where the centre image needs a different operating point from the sides.

Hardware Accurate

Hardware Accurate forces Voltage to 117 V (US nominal) and disables the knob.

When to use

• Lower voltage (100–110 V) — fatter, browner saturation on bass, drums, and any source that wants the "vintage-sagged" character.
• 117 V — the design operating point. The clean valve character that matches the schematic.
• Higher voltage (120–127 V) — cleaner, more open valve colour with more headroom. Useful on a mix bus where you want make-up amplifier character but cleaner than the schematic gives.

The Make-Up Amplifier panel exposes a per-side Wear knob (chassis label "Wear"; labelled "Age" in pre-release versions and still carrying the tubeAge parameter ID for backwards-compatible session reload) — 0 to 100 %, default 0 %. This is a Niviem creative-modification control. Wear scales three aspects of the make-up amplifier together:

1. Cathode emission drops with wear. At 100 % the cathode coating's emissivity is at 65 % of fresh, reducing the valve's plate-current output. This is a documented real phenomenon — vacuum tubes age primarily through cathode emission loss.
2. Push-pull section imbalance rises with wear. The two halves of the twin-triode age slightly differently, drifting apart in transconductance. The plugin's imbalance rises from 3 % fresh to 10 % at full wear, letting 2nd-harmonic content through that a perfectly matched pair would cancel.
3. Rail sag deepens with wear. A worn valve draws more current trying to hold its operating point, sagging the rectifier-supplied rail harder under programme-dependent load.

Effect on the chassis's sound

Wear produces a vintage-condition character — softer, slightly looser, slightly warmer. At low settings (0–30 %) it is subtle; at higher settings (70–100 %) it is recognisable as the sound of a well-played vintage piece. The 2nd-harmonic rise from the imbalance is the most audible component — it adds a gentle even-order warmth under the signal.

Per-side under LINK off

Independent Mid / Side wear — useful for stereo character work and asymmetric mastering moves.

Hardware Accurate

Hardware Accurate forces Wear to 0 % (fresh valves) and disables the knob.

Wear is a vintage-condition character knob

The cathode-emission loss is a real measured phenomenon — the ~35 % loss at full wear matches typical aging studies of dual-triode glass valves under continuous service. The imbalance rise and the rail-sag rise are tuned to sound musical rather than measured from a specific worn unit. Think of the knob as "how vintage do you want this" rather than "how many hours has this tube run".

The Make-Up Amplifier panel exposes a per-side Transformer knob — 0 to 100 % continuous, default 100 % (spec-authentic core). The knob controls the saturation knee of the three transformer cores (input, interstage, output) in the make-up amplifier path.

• 100 % — saturation knee at the spec-authentic level. The transformers sit well below their saturation point at any normal programme level; the chassis sounds clean.
• 0 % — saturation knee dropped roughly 48× lower. The transformers saturate immediately on any input — the chassis produces audible 2nd-harmonic warmth and low-frequency thickness from the iron at all times.

Effect on the chassis's sound

Lower Transformer Headroom means earlier iron saturation, which means more 2nd-harmonic content and low-frequency thickness. The effect is most audible on the low end of the chassis, where the input and output transformers' nonlinearities multiply across the bass band. The voice is closer to a tape saturator than a tube saturator — warm, slightly compressed, with the particular character that suggests vintage iron under load.

Per-side under LINK off

Independent Mid / Side transformer headroom settings.

Hardware Accurate

Hardware Accurate forces Transformer to 100 % (spec-authentic core) and disables the knob.

Workflow notes

• The factory preset Creative - Iron Push demonstrates a low-headroom (40 %) setting with normal valve drive — a warm vintage colour with iron contribution but without the heavier valve character of an explicit Character / Drive move.
• The Transformer knob is independent of the Character knob. Character drives the valves; Transformer drives the iron. The two can be combined for a layered vintage character.

The chassis header carries a three-chip A/B settings group — A / A→B / B→A.

What A/B does

The chassis carries two complete snapshots of itself — slot A and slot B — that you can flip between to A/B compare two settings without losing either. Each slot holds the full parameter state: every knob position, every chip's on / off, every frequency selector, every modal-panel knob. The group is workflow-only — it does not affect the audible chain's behaviour beyond letting you compare the two snapshots against each other. Hardware Accurate has no effect on A/B.

Workflow

The canonical copy-and-branch flow:

1. Dial in a setting (slot A holds it — A is active at load).
2. Click A→B to copy the current state into slot B (both slots now agree).
3. Click the A chip to flip to slot B (the chip label now reads "B"; the chassis shows slot B).
4. Modify the chassis (the changes live in slot B; A is unchanged).
5. Click the B chip to flip back to slot A (you hear the original).
6. If the slot B variation wins, click B→A to make A hold it too.

Click-free flipping

Each A/B flip raises an 11 ms wet-path mute envelope at the moment of the switch. A parameter switch can change many knobs at once and the per-parameter smoothers cannot keep that click-free on their own — the mute window hides the cumulative transient. The mute is applied to the wet path only; the bypass path is unaffected.

Per-session, not per-project

The A/B state lives for the duration of the editor session. Closing the project does not persist both slots — only the active slot is saved through the standard preset / project state pipeline. Reopening the project lands you on the slot that was active at the time of close; the other slot is fresh. Treat A/B as a fast in-session compare, not a persistent two-preset memory. Use Save Preset for permanent storage.

A/B vs Bypass

The A/B slots compose with the chassis's bypass / AutoGain pair:

• Bypass (Power) turns the chassis off completely → compares "EQ on" vs "EQ off" (AutoGain keeps that comparison level-matched).
• A/B compares two different chassis settings against each other (and, optionally, against the dry signal through the Power switch).

Use Bypass to answer "is the chassis helping?". Use A/B to answer "which of these two settings is better?". Use both together to answer "is the better of these two settings actually better than the dry signal?".

The chassis's central display is a single CRT-styled gold scope that overlays the live programme spectrum and the chassis's EQ curve as two layers in one view.

The display

The display is a 128-bin log-spaced rendering from 20 Hz to 20 kHz on the X axis. The Y axis spans 72 dB downward from a 0 dBFS full-scale sine, with six grid divisions at 12 dB each. A subtle scanline pattern overlays the surface for the 1970s analog-scope feel.

Spectrum (gold bars)

The spectrum is the live programme content at the chassis's output. The FFT pipeline is four stages:

1. The audio thread mixes L + R into a mono sum and writes 1024 samples to a lock-free ring buffer (spectrumFifo) at the end of every processBlock.
2. The editor's 60 Hz timer drains the FIFO, runs the FFT (1024-point, Hann window) on every complete frame, and converts to dB.
3. Per-bin spatial smoothing across recent frames damps noise; a peak hold with slow decay tracks transients.
4. The result is painted as bars with a subtle glow that suggests CRT phosphor.

A programme-material display tilt of +4.5 dB/octave is applied about the geometric centre of the display range — typical mixed audio falls roughly that fast at high frequencies, so a balanced mix reads as a near-flat line and the highs use the full panel depth. The tilt is a display convenience only; the audio path is untouched.

The spectrum is the chassis's output, so the EQ moves you make are reflected in the bars. Bypass the chassis (Power off) and the bars reflect the dry input.

EQ curve (gold line)

The EQ curve is the current chassis transfer function in dB, evaluated at the same 128 log-spaced points as the spectrum bars. The C++ side computes it directly from the live DSP coefficients — the passive EQ shape (LF boost / LF cut / HF boost / HF cut), plus the make-up amplifier's small-signal contribution, plus any HPF / LPF helper-filter response. The curve refreshes on every parameter change — the editor's 60 Hz timer flags a redraw the moment a curve-shaping parameter moves. The curve overlays the spectrum so you see in one view what the chassis is doing and what the programme is doing.

Sister curve under LINK off

When LINK is off the spectrum overlay paints two EQ curves — the gold line for the bare side, and a contrasting hue for the sister side:

• Cyan under STEREO routing (the R side).
• Magenta under M-S routing (the Side channel).

The two diverge live as you move the sister knobs. Under LINK on, only the single gold curve paints.

What the spectrum doesn't show

The spectrum reads the chassis's output after the full chain — EQ + tube + transformers + AutoGain + Output Gain + the bypass mixer. The pre-EQ content is not directly displayed. To see the input, bypass the chassis — the curve flattens to 0 dB and the spectrum reads the dry signal.

CPU and threading

The FFT is the heaviest visualisation cost and runs strictly off the audio thread. The audio thread only fills the ring buffer (a few hundred ns per block); the editor's timer drains and processes. There is no perceptible audio-thread cost from the analyser.

The chassis's nonlinear blocks — the make-up amplifier's two valve stages and the three transformer cores — produce harmonics. Without anti-aliasing those harmonics would fold back into the audible band as aliasing distortion, particularly at low host sample rates. The plugin uses a two-level anti-aliasing strategy:

1. A polyphase half-band oversampling cascade raises the internal sample rate to roughly 352 kHz. The nonlinear blocks process at the high rate, then a matched decimation cascade brings the signal back down.
2. ADAA-1 (first-order antiderivative anti-aliasing) reads inside each nonlinear block. The block's transfer function is integrated once into an antiderivative; the per-sample output is the difference quotient of the antiderivative across the input swing — band-limited by construction.

Combined, the aliasing floor sits below −120 dB even at extreme drive — well below audibility for any realistic level.

The cascade

Three identical 2× half-band FIR stages stack to give the 1× / 2× / 4× / 8× factors. The cascade selects the right factor based on the host sample rate.

The internal rate stays roughly 352 kHz across the table, so the make-up amplifier always sees the same headroom for its harmonic spectrum regardless of the host config. The audible character is consistent across sample rates — the chassis does not "sound different at 96 kHz" the way some plugins do.

The half-band filter

Each stage is a symmetric FIR with every other coefficient zero by design (the half-band property), so only the non-zero taps need computing per sample. The tap counts are 127 / 61 / 25 for the three cascade stages — the first stage is the steepest because it protects the base Nyquist where audio fills the band right to the edge; later stages only ever see an already band-limited signal, so their transition bands may be wide and they can be short. The window is a Kaiser at β = 12, giving roughly −120 dB stopband attenuation. The cascade carries per-phase DC normalisation so the chain introduces no DC level shift. Each stage's polyphase implementation reads exactly half the coefficients for the upsample (non-zero phase) and half for the downsample (other phase), giving the cascade's effective per-sample cost a factor of roughly 4× over a direct FIR convolution.

ADAA-1 in the make-up amplifier

ADAA-1 reads inside the baked tube cascade and inside each transformer core. The mechanic is three steps:

1. At prepare time the per-sample input → output transfer f(x) is integrated once to its antiderivative F(x) = ∫f(x)dx. For the tube cascade the antiderivative is baked alongside the transfer LUT; for the transformer core it is evaluated in closed form from ln(cosh(x)).
2. At audio time the per-sample output is the difference quotient (F(Vg) − F(Vg_prev)) / (Vg − Vg_prev) — the difference of the antiderivative across consecutive samples, divided by the input change.
3. The difference quotient is mathematically equivalent to integrating f(x) over the grid swing across the sample — the band-limited average of the transfer function across the input change between adjacent samples.

The output's spectral content is therefore bandwidth-limited by the input bandwidth, not by the transfer function's nonlinearity — harmonics above the input Nyquist cannot appear. The aliasing floor drops 40–60 dB below a direct ZDF read; combined with the half-band cascade above, total suppression is below −120 dB.

Latency reporting

The cascade's half-band FIRs introduce group delay. The plugin reports its full round-trip latency via setLatencySamples so the host delay-compensates other tracks; the bypass mixer's dry path is also delay-compensated internally so the bypassed signal stays sample-aligned with the wet.

Why not user-configurable oversampling

Three reasons, in order:

1. The internal rate is part of the sound. The nonlinear character at 8× is slightly different from the same block at 1× because the harmonics are positioned differently relative to Nyquist. Letting the user pick the factor changes the sound, not just the CPU cost.
2. The cascade depth is selected automatically from the host sample rate. The chassis always targets ~352 kHz internally regardless of where the host runs. No user thinking is required.
3. The CPU cost is low enough that always-on is the right default for this plugin tier. The cascade adds roughly 10–15 % CPU at its heaviest setting (8×, 44.1 kHz host).

If you need a low-CPU mode, run the host at 96 kHz — the cascade drops to 4× automatically and the per-sample cost halves.

The plugin ships with 52 factory presets across 9 categories, every one of them calibrated against the Auto Gain reference so the bank reads at a consistent perceived level — peak-to-peak variation between any two presets stays inside 3 dB at the canonical −12 dBFS reference. You can step from a vocal preset to a creative tube crush without the bus level jumping, which is the whole point of working through a curated bank in the first place. Settings cited in the tables below are the literal factory values written into the APVTS on load — the same XML you would dump from a saved user copy.

Init

Vocal

Drum

Bass

Guitar

Mix Bus

Mid / Side

All Mid/Side presets are Route = M-S, LINK = OFF so Mid and Side carry independent curves.

Creative

Helper Filters

Loading a Preset

Click the preset name plate in the header to open the browser modal. Factory presets are grouped by the nine categories above (Init, Vocal, Drum, Bass, Guitar, Mix Bus, M-S, Creative, Helper Filters) with a tenth User group at the bottom that fills in as soon as you save a preset. Click any preset row to load it — the modal dismisses on the same gesture and the chassis state animates to the new values via a host-honest gesture bracket so your DAW's automation lane records the load as a discrete edit.

Alternatively the prev / next arrows flanking the name plate step through the flattened combined list — they walk Init → Vocal → Drum → Bass → Guitar → Mix Bus → M-S → Creative → Helper Filters → User in that order and wrap at both ends, so pressing next on the last user preset loops you back to Init. Arrow keys Left / Right are wired to the same prev / next behaviour whenever the chassis carries keyboard focus.

Saving a User Preset

Click the SAVE button to open the save dialog. The text input is pre-filled with the current preset name (or Untitled if nothing is loaded). Type your name — between 1 and 64 characters — and press Save or hit Enter. Illegal filename characters (/ \ : * ? " < > |) are silently replaced with underscores so the disk write never fails, and a leading dot or control character is stripped. If a preset of the same name already exists on disk you are prompted to confirm the overwrite.

Files are plain XML — the same APVTS state the host would round-trip in a session file, with the plugin's root tag NiviemPultec.

Loading a User Preset

User presets appear at the bottom of the browser in the User group, sorted alphabetically. The prev / next arrows step through factory and user presets together as one combined list, so once you have your own bank you never need to open the modal to recall it.

Deleting a User Preset

Hover over a user preset in the browser to reveal a small × delete icon at the row's trailing edge. Click it and confirm to remove the file from disk. Factory presets carry no delete icon — the bank is read-only and survives any user-level edit.

Sharing User Presets

Because preset files are plain XML, sharing them across machines is a copy-paste operation: drop the .preset file into the same user-preset directory on the target machine and restart the plugin (or simply reopen the preset browser) to refresh the list. There is no proprietary container and no machine-locked field — what you saved is exactly what your collaborator loads.

The Modified-State Indicator

A small dot / asterisk appears next to the preset name plate whenever the chassis state diverges from the last-loaded preset (the indicator is driven by an APVTS listener so it lights on the first touched parameter, regardless of how the change arrived — knob move, host automation, even chip click). Saving the preset clears the indicator immediately; loading another preset clears it as well.

Reset to Default

Every knob's right-click context menu carries a Reset to Default entry that sends the parameter to its factory default through a proper host gesture bracket — so the host's automation lane logs a real edit you can undo, replay, or write into a take. Toggle chips (HPF on/off, M-S route, Tube on/off etc.) carry no Reset entry because their default already lives on the chip itself: click it.

Every interactive surface on the chassis — every knob, every chip, every value readout — carries a custom right-click menu. The browser's native menu (Inspect Element / Reload / View Source) is suppressed across the entire WebView so you never see a developer dropdown over your mixing console.

Menu entries

Keyboard accessibility

The menu is fully keyboard-navigable. Arrow keys move the highlighted entry, Enter or Space activate it, and Escape dismisses without action. The menu opens at the mouse cursor for pointer-driven invocations and at the screen-centre of the focused control for keyboard invocations (right-click via the keyboard menu key or Shift-F10).

Dismissal

Clicking anywhere outside the menu dismisses it. Escape dismisses it. Activating an entry dismisses it. Any chassis interaction — even one that does nothing, like dragging an empty area — auto-dismisses the menu so a stray contextual popup can never block a mix move.

Nivipedia is a 65-entry built-in encyclopedia covering every control, mode, theoretical concept, valve model, historical reference and workflow on the chassis. It lives inside the plugin so you can look something up without leaving the session.

Opening Nivipedia

• Click the NIVI chip in the header to open the welcome screen
• Right-click any control → What is this? to open Nivipedia jumped directly to that entry
• Within any entry, blue wiki-link spans are click-through cross-references to related entries

The modal carries a collapsible sidebar with the nine category groups on the left and the main article pane on the right. A search field at the top of the sidebar filters entries by title and brief — categories auto-expand to reveal matching entries as you type.

The 9 categories

Search and navigation

The search field filters by both title and the entry's brief subtitle, so typing koren matches both the dedicated Koren space-charge model entry and the 5751 / 12AY7 / ECC82 entries whose briefs reference it. Categories auto-expand whenever an entry matches so you never have to manually open a section to see hits. Inside an article, wiki-link spans jump to the target entry and a breadcrumb at the article top lets you walk back to the originating page.

Dismissal

Click anywhere outside the modal, press Escape, or use the Close button in the modal header. The chassis state is untouched — Nivipedia is read-only and carries no side-effects on the audio path.

The Pultec trick on a mix bus

The classic move: dial the same LF frequency (60 Hz for most material, 30 Hz for sub-heavy mixes) on both LF Boost and LF Atten, push the boost to +2 to +4 dB and the atten to roughly half that. The hardware's parallel-shelves topology folds these into a single resonant lift just above the corner with a gentle cut below — you get the felt weight without the inaudible sub-rumble that would eat your headroom. Start small. The trick lands harder than it looks on a meter, so +2 dB / −1 dB on a master bus is plenty before you start chasing peak.

The Pultec trick on bass

On a bass DI or sub-bass track, push the trick further: 30 Hz LF Freq, +5 dB boost, −3 dB atten, with the HF section adding a touch of 3 kHz for pluck definition. The lift sits at ~50–80 Hz where the felt body of the bass guitar lives, while the cut below 30 Hz protects your sub-bass headroom (and your woofers). Engage Auto Gain so the bass doesn't surge in level when you A/B the move — you want to hear the shape, not the level.

HF Boost at 8 kHz vs 12 kHz

These two HF stops cover the boundary between presence (8 kHz — where a vocal's intelligibility lives, where a snare's crack sits) and air (12 kHz — where a cymbal's sheen and a vocal's halo live). 8 kHz with a narrower bandwidth (BW 0.5–0.7) carves a forward, present sound; 12 kHz with a broad bandwidth (BW 0.2–0.3) opens the top end like a window. Try the same source through both with matched boost — they sound like different effects, not different frequencies.

Bandwidth narrow vs broad

The Bandwidth control is the EQP-1A's secret weapon — it's the loaded-Q of the HF series-resonant tank, and unlike a modern EQ where Q and gain are independent the Pultec couples them mechanically. Higher boost naturally tightens the peak, but the Bandwidth knob lets you push that further: BW 0 (broadest) for shelving-style air, BW 1 (sharpest) for surgical de-essing or specific-frequency presence carves. The narrow end is where the EQP-1A's character comes alive — it's the sound on every classic rock vocal.

Auto Gain for A/B sanity

Always turn Auto Gain on when you're auditioning a move. The 64-point loudness scan compensates for the entire heard curve — boosts, cuts, helper filters, drive, transformer headroom, everything that affects perceived level — so your A/B against bypass measures tone change, not level change. Once you have the tone you want, you can turn AG off and dial the Output knob to taste — but never make an EQ decision with AG off if the boost is more than ~2 dB, because your ear will follow the level and your hand will undo the move.

Use Hardware Accurate as a reference

When you're unsure whether a creative choice has wandered too far from the source, engage Hardware Accurate. It pins Drive to 0 dB, Voltage to 117 V, Wear to 0 %, Valve to 12AX7, Transformer to 100 % (the measured spec headroom), AG off and helpers off — exactly the 1961 Pultec, no extensions. A/B against your tweaked state and you'll hear precisely what your additions are contributing. Use it like a tuning fork: not necessarily the end state, but the unambiguous reference point.

M-S diverged for surgical mastering

Switch Route to M-S, turn LINK off, and the Mid and Side paths become two independent EQP-1As. A typical mastering move: subtle 60 Hz Pultec trick on the Mid for centre weight, a narrow Atten Sel 5 kHz shelf cut on the Side to soften wide cymbals, a broad 12 kHz lift on the Side only for stereo air. The Mid stays focused, the Side opens up. Without M-S the same moves would either over-process the centre or under-process the width.

A/B-and-branch workflow

The A/B slots let you carry two complete chassis states (every parameter, every chip) and switch between them with a single click. Use them to branch: dial a candidate move into A, copy to B, then push B further. Now you can A/B the original move against the more aggressive variant rather than against bypass — which is the harder, more interesting comparison most engineers never bother to make.

Helper HPF and LPF for shape, not surgery

The opt-in HPF and LPF helpers are 12 dB/oct Butterworth — gentle slopes, no resonance, designed to shape rather than carve. Use HPF to clean DC and sub-rumble before the LOW section lifts (so the boost isn't fighting an inaudible 25 Hz hum), and LPF to gently tame a bright source before the tube stage drives it (so the saturation isn't sharpening fizz you didn't want). The helpers are off by default and Hardware Accurate forces them off — they're never required, but when you need them they're the right tool.

Tube swap as a "fourth knob"

Beyond the LF, HF and Atten sections, the Valve chip is the chassis's fourth tonal control. 12AX7 is the stock high-mu triode: aggressive, classic, the sound of every original Pultec. 5751 is a lower-gain swap with more headroom and a slower, more polished saturation knee — the audiophile's choice for a master bus. 12AY7 is even lower gain, with a softer, sagged character — try it on a vocal where the 12AX7 would be too forward. Swap the valve and listen — the EQ curve does not change, but the colour does.

Per-side under LINK off

Most users stay in LINK on most of the time (matched stereo, like the hardware). But turn LINK off and the right channel — or the side channel under M-S — carries its own complete chassis: its own EQ, its own helper filters, its own valve, its own drive, its own AG. This is how you do single-channel de-essing without an M-S detour, or single-channel air on the strumming-hand side of a stereo acoustic, or independent tube colour between Mid and Side. The right-side state survives LINK on/off toggles so you can flip the link, audition, and flip back without losing your divergent state.

Drive vs Transformer for vintage colour

Drive and Transformer are two different flavours of EQP-1A character. Drive pushes the input tube past its quiet operating point — you get more harmonic distortion, more grid-bias asymmetry, more of the valve's active voice. Transformer raises the iron's saturation contribution — you get more low-mid bloom, more high-end softening, more of the unit's passive voice. Drive sounds forward; Transformer sounds glued. Mix them for a fuller hybrid colour: +3 dB Drive plus 80 % Transformer is the classic API-or-Neve-or-Pultec sound on a bus.

Why a passive LC equaliser

A passive equaliser sets its frequency response from the values of its inductors, capacitors and resistors — not from a feedback loop around an active gain stage. The consequence is that the EQ curve carries no slew limit at low signal, has very low THD inside the linear range (the network has no internal noise from a feedback amplifier and no slew-rate-limited transient response), and no time-varying parasitic that would shift the corner under load. The cost is insertion loss — typically 15 dB of attenuation through the LC network even at flat-EQ — which the make-up amplifier downstream recovers with a moderately-fed-back triode pair.

The EQP-1A's signature transparency is a direct consequence of this passive front-end: the EQ shape is purely linear, the colour comes entirely from the recovery amp. Active EQs (parametric peaking filters built around op-amp gyrators or state-variable topologies) cannot match that separation — their EQ shape and their noise floor are inseparable.

The parallel-shelves topology

The LOW section's Boost and Atten controls are not series-connected, they are parallel-summed — both shelves feed into the same summing node, and the LF Freq stop sets a common turnover that both shares. Folded together, the two first-order shelves produce a second-order response with a pole-zero geometry that depends on both controls' settings.

When both controls are dialled to roughly the same setting at the same turnover, the pole and the zero from each shelf approach each other but do not cancel — they're slightly offset in frequency because the Boost shelf and the Atten shelf use different tank elements that resonate at slightly different points. The result is a narrow resonant lift just above the turnover with a gentle cut below it — the famous Pultec trick, a curve that cannot be drawn on a single parametric EQ without two interacting bands.

Series-resonant LC tank for HF Boost

The HF Boost circuit is fundamentally different from the LF section: it's a series-resonant LC tank loaded with a variable resistor (the HF Boost pot itself), driving the summing node through a coupling capacitor. The tank's impedance dips at its resonant frequency f₀ = 1/(2π√LC) — at f₀ the tank acts like a short, allowing more signal through the coupling cap, and the result is a peaking boost.

The loaded-Q — and therefore the bandwidth of the peak — is set by the ratio of the tank's reactance at f₀ to the loading resistance (the HF Boost pot). Higher boost positions present less resistance to the tank, raising loaded Q and sharpening the peak. The Bandwidth knob on the chassis exposes a second resistive element in parallel with the boost pot, decoupling Q from gain at the user's discretion. The mechanical Q-gain coupling that gives the EQP-1A its musical character is preserved in the digital model exactly as the schematic dictates.

Tube push-pull and even-harmonic cancellation

The make-up amplifier is a push-pull pair — two triodes (a single ECC83 envelope holds both sections) wired to amplify opposite halves of the signal. Section A handles the positive swing, section B handles the negative; their plate currents sum (through the output transformer's primary) to reconstruct the full waveform.

The mathematical beauty of push-pull: if the two sections' transfer functions are perfectly matched, all even-order harmonics cancel at the summing point. Whatever 2nd, 4th, 6th harmonic each section generates, its counterpart in the other section is the same waveform shifted 180° — they sum to zero. Only odd harmonics (3rd, 5th, 7th) survive, and the pair's response is symmetric and well-behaved.

But no two triode sections are perfectly matched in the real world. A typical 1961 ECC83 carries ~3 % transconductance mismatch between its sections, which means ~3 % of each section's even-harmonic content leaks through to the output and gives the chain its characteristic warmth. The plugin's Wear control widens that mismatch (from 3 % at 0 % wear up to 10 % at 100 %) — more 2nd-harmonic, more 4th, a richer, more vintage colour.

Global negative feedback as linearisation

The amplifier's output is sampled, scaled by the feedback ratio β, and returned to the input grid of the first triode in opposite phase. The closed-loop gain becomes A_cl = A_ol / (1 + β·A_ol) where A_ol is the open-loop gain. When β·A_ol is much greater than 1, the closed-loop gain approaches 1/β — entirely independent of the open-loop nonlinearities.

In the EQP-1A this is dialled at a moderate ~10 dB of feedback: enough to linearise the chain to the saturation knee (so small signals see a near-linear amplifier and the EQ curve reads as designed) but not so much that the valve character is destroyed (so larger signals still hear the colour the tubes were chosen for). It's the engineering sweet spot — feedback as a spec stabilisation tool, not as a colour-killer — and it's why the EQP-1A holds its EQ shape precisely while still sounding unmistakably like a tube unit when pushed.

The chassis loads as a blank window on Windows

The Niviem EQP-1 uses a WebView UI under JUCE on every platform. On Windows the WebView host is Microsoft Edge WebView2, which ships pre-installed on Windows 11 but must be installed manually on older Windows 10 builds. Download the WebView2 Evergreen Bootstrapper from Microsoft's developer site and run the installer — no DAW restart required, the next plugin-instance launch picks up the runtime automatically.

The plugin won't activate

Check your account at niviem.net — every license carries a finite number of activation slots, and a fresh install on a new machine consumes a slot. If you've used all slots, deactivate an older machine from the dashboard's My Licenses page first, then re-attempt activation. If the activation request fails on the network, verify your firewall allows outbound HTTPS to niviem.net.

Demo mute keeps interrupting playback

The demo path runs a 60 s on / 5 s off cycle as a designed restriction — this is not a bug. Activate your license through the ACTIVATE button in the header (or via the activation modal that opens on first launch) and the demo cycle is permanently disabled. The activation is JWT-cached so subsequent sessions do not require network access.

CPU usage too high

At its peak setting (host SR 44.1 / 48 kHz → 8× cascade to reach 352 kHz internal) the plugin adds roughly 10–15 % per instance on a modern CPU. The fastest mitigation is to run your host at 96 kHz, which drops the internal cascade from 8× to 4× automatically — same internal target, half the per-block work. The plugin's cascade auto-selection means you do not need to change any plugin setting; just bounce the host's sample rate.

Latency reported is non-zero

This is expected. The oversampling cascade introduces ~106 samples (≈ 2.21 ms at 48 kHz host) of group delay, which the plugin reports to the host via PDC. Every modern DAW handles PDC automatically, so other tracks are time-aligned in playback. If you hear a phase issue on stems mixed through a non-PDC-aware bus, check the host's PDC status — it should never be disabled on a mix bus carrying processed plugins.

Auto Gain trim looks "wrong"

Auto Gain compensates for the entire heard curve — the EQ shape, the helper filters' attenuation, the tube-stage saturation gain, the transformer headroom — not just the EQ boost. A move that looks like +3 dB of boost on the curve might come back as −1 dB of AG trim because the saturation stage also lifted the signal. This is the point of AG: your A/B against bypass measures tone, not level.

A/B slot lost on project reload

The A/B slots are in-session only — only the actively-selected slot's state persists in the host's saved project file. The other slot is a runtime convenience for A/B comparison during a single session; if you want to keep two distinct chassis states across sessions, save each as a user preset.

HPF/LPF chip won't engage under Hardware Accurate

By design. Hardware Accurate forces the helper filters off because they are not part of the 1961 EQP-1A — they're a modern extension. To use the helpers, turn HA off first; the helper chips become responsive immediately.

DAW shows "Niviem EQP1" with no hyphen

This is intentional. The plugin's binary filename is Niviem EQP1.vst3 / Niviem EQP1.component without a hyphen because some Windows DAWs (Cubase, Studio One) have historically refused to load plugin binaries with non-alphanumeric characters in the filename. The brand name in the UI and all marketing remains EQP-1. This is a filename-safety convention, not a typo.

Where do I report a bug

Log in to your account at niviem.net and use the Reports tab — every bug or feature thread is two-way, you'll see the developer's response in the same place. For urgent issues you can also email support@niviem.net.

Design & Engineering

NIVIEM EQP-1 was designed and engineered by NIVIEM AUDIO LTD (UK Company No. 17217761). DSP architecture, product design, UI engineering and audit-and-polish were carried out by Milan Vasiljev. The plugin is built on the JUCE framework — gratitude to the JUCE team for the consistently excellent host-side audio plugin infrastructure that made the WebView UI, the ADAA-1 nonlinear stages, and the per-block ADAA accumulators possible without leaving the framework.

Tube modelling references

• Pascal Maillet, Modeling of Vacuum Tubes Using Polynomial Approximations — the polynomial-fit method used for the 12AX7 / ECC83 cascade, derived from datasheet plate-curves and verified against the chassis's measured response
• Norman Koren, Improved Vacuum Tube Models for SPICE Simulations (1996) — the space-charge model used for the 5751, 12AY7 and ECC82 alternates; the foundation of every modern accurate triode model in software
• Philips ECC82 datasheet — Koren fit, mean error ≈ 4.1 % across the small-signal operating box
• GE GL-5751 datasheet — Koren fit, used for the 5751 alternate-valve emulation
• Tung-Sol 12AY7 datasheet — Koren fit, used for the 12AY7 alternate-valve emulation
• RCA 12AX7 / ECC83 manual — primary source for the Maillet polynomial fit

Pultec chassis references

• Pulse Techniques EQP-1A schematic (1961) — the canonical schematic transcription that drives every component-value choice in the digital model
• Ian Bell, EQP-1A frequency response measurements (2000s) — closed-loop gain reference for the LF, HF and Atten sections; the spec-authentic target curve
• CARTEC measurement reference (1990s) — the HF Boost peak-amplitude swing data, used to verify the loaded-Q vs gain mechanical coupling
• DIYRE EQP5 cap-bank measurements (GroupDIY + DIYRE forum threads) — per-band turnover ratios that informed the Hardware Accurate band-dependent voicing
• Peerless S-217-D output transformer datasheet — magnetisation-curve data for the transformer saturation model

Acknowledgements

A debt of thanks to the GroupDIY and DIYRE forum communities, and to the broader DIY tube-electronics community whose decades of measurement, transcription and component sourcing made an open, accurate emulation of the EQP-1A possible. The hobbyist's archive is the professional engineer's library.

v1.0.0 — 2026-05-25 (initial release)

• Component-level emulation of the 1961 Pulse Techniques EQP-1A
• Three sections: LOW (4-stop frequency + CON continuous mode), HIGH (7-stop boost frequency + 3-stop atten frequency + CON on both), OUTPUT
• Make-up amp panel: Tube / Clean switch, Voltage (100 – 127 V), Wear (0 – 100 %), Valve (12AX7 / 5751 / 12AY7), Transformer Headroom (0 – 100 %) — per side under LINK off
• Hardware Accurate master mode — pins Drive, Voltage, Wear, Valve, Transformer to spec; forces AG, helpers, CON and LINK to their spec-authentic states
• CON continuous-frequency mode on all three rotaries (LF boost, HF boost, HF cut)
• Opt-in HPF + LPF Butterworth helper filters (pre-EQ, 12 dB/oct)
• M-S routing with full per-side parameter divergence under LINK off (independent EQ, helpers, drive, valve, transformer per Mid and Side)
• Smart Auto Gain — 64-point loudness scan, per-side compensation, with continuous Strength control
• A/B settings slots for in-session candidate-state comparison
• Spectrum analyser with EQ curve overlay; sister curves drawn under LINK off for both sides
• Polyphase halfband oversampling cascade targeting ≈ 352 kHz internal sample rate, Kaiser β = 12, ~120 dB stopband
• ADAA-1 antiderivative anti-aliasing on every nonlinear stage (tube transfer, transformer saturation)
• 52 factory presets across 9 categories
• 65-entry Nivipedia built-in encyclopedia with hyperlinked cross-references
• AU + VST3 on macOS Universal (arm64 + x86_64); VST3 on Windows 64-bit
• Niviem License Manager — Ed25519 JWT verification with machine fingerprint, offline use after activation, MuseHub marketplace coexistence
• WebView2 static linking on Windows to prevent legacy IE fallback and the Navigation-canceled ieframe.dll error