Reference manual

MEY 640 "Gravity" is a surround-capable mastering compressor built around three serial compression stages with a coupled, gain-linked detector, a hard-brickwall safety stage, and a parallel dry (MIX) path. It is designed for classical and acoustic repertoire: transparent by default, with character living in the ballistics of each stage rather than in added harmonics. This page describes MEY 640 v0.1.0.

Serial stages, two controls beside the chain

Signal enters Stage I (Glue), passes into Stage II (Density), and finally through Stage III (Peak), which can be switched into BLOCK mode for a true brickwall ceiling. Two unit-wide controls sit alongside the chain rather than inside it: SC HPF feeds a high-pass–filtered copy of the signal to the detectors of stages I and II only (the audio path itself is untouched), and MIX blends the fully processed signal back against a phase-aligned dry copy at the very end of the chain. Stage on/off is a transparent bypass with a short 5–10 ms crossfade: no click, no gap. Turning ENGAGE off bypasses the whole unit cleanly, with latency compensation, so an A/B against the untouched signal is always level- and time-aligned.

Mono to 7.1.4, detected automatically

MEY 640 auto-detects the host's bus layout and adapts: mono, stereo, 5.1, 7.1, 5.1.4 and 7.1.4 are all supported, with a fixed internal channel order of L R C LFE LS RS LB RB TFL TFR TRL TRR. Channels outside the detected layout are dimmed and inactive in the interface: nothing to configure by hand. One optional external sidechain bus, in the same layout, lets any channel switch its detector source between internal (its own signal) and external (the sidechain bus, same channel index); external falls back to internal if no sidechain is connected.

One image, or twelve channels of your own

By default (LINKED), one gain-reduction value drives every channel in the bus: the loudest channel controls the whole stereo or surround image, so the image never tilts under compression. LFE follows that shared gain but never drives it: a held organ pedal on the LFE channel won't pull the rest of the mix down with it. Switching to PER CH gives every channel its own parameter set and detector, edited one channel at a time from the channel column on the left of the panel.

Glue, Density, Peak

Three compressors in series, each with a distinct character, tuned together on real orchestral and classical material rather than borrowed from a general-purpose preset. Factory defaults are deliberately gentle (soft ratios, slow times, shallow reduction) because Stage III is the safety net, not the workhorse.

Stages I and II use RMS detection with a smoothing pole on the gain itself: the compressor responds to average level, not to individual wave cycles, which is what keeps low frequencies (a held double-bass note, an organ pedal) from being tracked as if they were the signal to compress. Stage III intentionally keeps a plain peak detector with no smoothing: it is the ceiling, and a ceiling that responds to average level instead of the actual peak is not a ceiling.

Turning up always means "more"

All three THRESHOLD knobs read in the direction of compression, not the direction of dB. Fully closed (turned all the way left) means +12 dBFS (no compression at all) and opening the knob clockwise lowers the threshold and brings in more compression. This keeps every knob on the panel consistent: turning up always means "more," rather than requiring you to remember that one knob works backwards from the rest.

Release that listens to what came before it

Release time is not fixed to the RELEASE knob value alone. A short, shallow reduction recovers on a fraction of the set time; a sustained, held-down passage recovers on the full set time (or longer). Measured examples on real programme material show recovery around 161 ms after a short peak against roughly 789 ms after a sustained passage, from the same release setting. This keeps hall reverb and room tone from audibly "breathing" with the compressor, while brief transients still recover quickly. The behaviour is driven by a slow-moving memory of the reduction the detector is asking for, not by the smoothed gain itself, so it responds to what the source is actually doing rather than to the compressor's own output.

Stages I and II only

Stages I and II each carry a MAKEUP control (−6…+12 dB) with its own AUTO toggle, unit-wide rather than per channel. Per-channel makeup would tilt the surround image. With AUTO engaged (the default), makeup is computed from that stage's current threshold and ratio: a fixed formula, not a metering feedback loop, so moving the threshold or ratio shifts the makeup with it and A/B'ing stays loudness-fair, but the gain path itself never moves on its own during playback. Stage III carries no makeup control by design: its threshold is the output ceiling, and makeup gain after that stage would defeat it.

A true safety ceiling, Stage III only

BLOCK overrides Stage III with an effectively infinite ratio, an instant (0 ms) attack and a hard knee: a brickwall at the Stage III threshold. The ratio readout switches to "∞:1" and the ratio control dims, since it no longer does anything. A true sample-accurate brickwall depends on the effective processing rate (session rate × oversampling factor), not on the oversampling factor by itself: at 44.1 or 48 kHz sessions, set OVERSAMPLING to 8× if you're relying on BLOCK as a hard ceiling for that pass. At higher session rates (88.2/96 kHz and above), 4× is already enough.

Two unit-wide controls beside the chain

SC HPF is a high-pass filter (12 dB/octave) inserted only in front of the Stage I and Stage II detectors: the audio path itself is never filtered. Range is off, then 20–400 Hz logarithmic; the factory default is 60 Hz. Its purpose is to stop low fundamentals (a held double-bass note, a pipe-organ pedal) from single-handedly driving the gain reduction for the whole mix. Stage III's detector stays broadband on purpose: a safety ceiling has to see everything, low end included.

MIX is a single dry/wet blend applied once, at the very end of the chain, across the whole unit: 0% is the fully untouched input, 100% (the factory default) is fully processed. The dry signal is carried through the same lookahead delay as the wet path, so the two stay phase-aligned at every MIX setting, and because wet equals dry below every stage's threshold, the blend is loudness-neutral by construction. There is nothing extra to compensate as you move it. This replaced the plug-in's original two per-stage bleed taps, which were uncompensated parallel adds that could raise level by up to 6 dB and made a fair A/B impossible.

Global, and reported to your host

Lookahead is a fixed 2 ms delay on the audio path (not the detectors), on by default. It gives Stage III enough runway to catch a peak before it arrives rather than after, which matters most when BLOCK is engaged. Toggling it changes the plug-in's reported latency; your DAW compensates automatically. Oversampling runs at 1×, 2×, 4× or 8× (default 2×), applied across all three stages together and reported to the host as added latency.

Built against broadcast standards

Input and output ladders share one design: a 48-cell grid built from two commensurate scales that meet exactly on the cell boundaries, so every printed scale mark lines up with the edge of a lit cell rather than falling inside one: 24 cells above −24 dBFS at 1.0 dB per cell, 24 below at 1.5 dB per cell. Cells light in whole steps, never a fractional cell; a lit cell rises the instant the signal reaches it and falls back only after a 300 ms debounce, at a PPM Type IIa rate of 8.6 dB/s. A peak-hold marker sits above the lit portion of the ladder and decays on its own, slower timer.

Each cell is coloured by its own dB value, in four named zones rather than one continuous tint ramp. A real over (≥ −0.1 dBFS) latches that ladder red until you click the meter bank to clear it: a missed clip can't hide behind the next quiet passage. The output ladder additionally shows a faint ghost mark of that channel's input level, so the level change through the whole chain is visible on the output meter alone.

Each stage's GR bar is drawn from the top of that stage's own level bar downward (the reduction actually being applied to the signal in front of you, at that moment, on that bar) rather than as a separate bar measured from 0 dBFS. A short-term loudness column runs beside each meter bank, fed by a BS.1770-4 loudness measurement (K-weighted, 3 second rolling window) taken independently pre and post the chain. Surround and height channels are weighted +1.5 dB relative to front channels per the BS.1770-4 channel-weighting rule; LFE is excluded from the loudness measurement entirely, though it still passes through the audio path and the level meters as normal.

A starting point, not a suggestion engine

One button, in the right-hand column: it reads Stage I's current threshold, ratio, attack and release, and writes a computed starting point into stages II and III. The calculation is a fixed formula: no listening, no adaptation, no state, no machine learning. The same Stage I setting always produces the same advice, and you tune from there. It acts once, when clicked: it does not listen to the audio at all, does not change Stage I, and does not turn Stage II or III on if they're currently off. The parameters are set either way, but you decide whether to engage the stage. Click it again after changing Stage I to recompute.

Stage III's advice is always the same fixed safety-ceiling setting (−2 dBFS, 8:1, 1 ms, 80 ms), independent of Stage I: its threshold is the output ceiling, so it is never derived from anything else. Stage II's advice scales with how hard Stage I is already working, with headroom built in for the soft-knee zone Stage I deliberately lets through.

LINKED vs. PER CH, copy/paste, save/load

LINKED (the default) means two things at once: one shared parameter set, automatable from your host, and one coupled detector. The loudest channel drives the gain applied to every channel, which is what keeps the surround image from tilting under compression. LFE follows that shared gain but never drives it. Switching to PER CH hands both of those to the selected channel individually: its own twelve values, four each for stages I–III, become plugin state rather than host-automatable parameters, edited one channel at a time.

COPY snapshots the selected channel's full parameter set to an in-instance clipboard. PASTE applies it to the selected channel while PER CH, or to every channel at once while LINKED. SAVE PRESET stores the whole unit (every channel, every flag, every global) both into the plugin's own state (so it recalls with your session) and to a standalone file at ~/Documents/MEY/640/user.mey640preset. LOAD restores it, and stays disabled until a preset has been saved at least once. There is one user preset slot per installation, not a library.

Installers, no choices to make

This release ships as a signed macOS installer (.pkg) for Apple silicon, containing VST3, Audio Unit and standalone builds, and as a Windows download containing the VST3 and the standalone. Double-click the installer and follow the standard macOS flow: it places the VST3 in ~/Library/Audio/Plug-Ins/VST3, the AU in ~/Library/Audio/Plug-Ins/Components, and the standalone app in /Applications, with no choices to make. If macOS shows a Gatekeeper warning anyway, right-click the installer and choose Open once, then confirm in the dialog that follows. This is only needed the first time. On Windows, unzip the download and copy the VST3 into C:\Program Files\Common Files\VST3. After installing: restart your DAW (or rescan plug-ins) so it picks up the new version, insert MEY 640 on a stereo or surround bus, and confirm ENGAGE is on and the AUTO badge shows the layout you expect (e.g. "AUTO · 7.1.4").

Free, no key, no account

MEY 640 is free and needs no licence key and no account. Press the button for your platform on theproduct page and the download starts; installing it is all it takes to run. This is an early release (v0.1.0); the sound and the controls may still change as it settles, and updates land on the same download link.

Something not working?

Write to info@wearemey.com and describe what you did, what you expected, and what happened instead. That's usually enough to find it.