LYNCH MASTER LIMITER — private0.1 An original stereo mastering limiter with independently switchable Tube and Tape color stages, followed by real lookahead peak control. These are musical analog-inspired algorithms, not measured replicas of a particular unit. CHAIN Input -> Tube -> Tape -> Output -> lookahead Limiter -> wet mix -> Bypass. Optional vintage noise enters before Output/limiting. Tube and Tape can both be on; Tube always precedes Tape. Output is before limiting so it pushes into the chosen ceiling. The limiter is not an ordinary saturator renamed as a limiter. STARTING Init disables Tube, Tape, Limiter and Noise and passes the exact delayed input. Enable Limiter for peak control or select a grouped starting preset. Input and Output add level; lower Ceiling sets a stricter output sample bound. Drive and fast timing can create deliberate heavy coloration. No forced loudness target or automatic gain matching is applied. TUBE Drive increases excitation through an asymmetric soft stage with level-dependent sag/recovery. Bias changes the asymmetry and harmonic balance. Tube saturation is processed at8x at44.1/48kHz and4x at88.2/96/192kHz. The Tube readout estimates instantaneous nonlinear transfer compression; it is not the final limiter GR. TAPE A separate saturating flux-memory path and high-frequency-loss contour. Drive sets magnetic excitation and density; Tone adjusts the high-frequency loss from dark to open. It is an original reduced tape-inspired effect, not a measured magnetic tape or tape-machine simulation. It uses the same oversampled path. LIMITING A stereo-linked max-filter/clipping-control envelope anticipates peaks before they reach the output. Attack/Release are nominal time controls; Clean, Glue and Crush shape their actual response differently. Clean favors controlled recovery, Glue is slower and more program-dependent, and Crush uses faster timing. Program Release increases recovery time after heavier sustained reduction. Stereo Link100% applies equal gain to both channels; lower values permit more independent reduction and can move the image. Intersample Aware adds an eight-phase129-tap-per-phase sinc reconstruction estimate to the detector, with0.35dB additional detector headroom. In this mode the active audio also passes through the bandlimited oversampling FIR, even when Tube/Tape are off. Measured small-signal gain differs by at most0.072dB from 20Hz to20kHz across the five tested rates, with aligned linear phase. Its64-sample analysis delay is accounted for inside the fixed lookahead. It reduces peaks between samples and has been tested against a separate reconstruction reference. This is not a claim of certified true-peak compliance on every possible signal. The control remains labeled Sample Ceiling. A downstream reconstruction filter, conversion, lossy codec or further processing can alter peaks. The sample-ceiling bound applies when Limiter is on and Wet Mix is100%. Lower Wet Mix reintroduces unlimited dry audio. Mix0 and Bypass reproduce the exact delayed input, including its original peaks. Ceiling and other sound controls smooth during automation. A final sample guard handles numerical/automation overshoot and has its own protection meter; severe activation can be audible. LATENCY Audio always reports608 samples:96 samples for oversampling alignment plus512 samples of lookahead. Active oversampled audio can have small pre-ringing before the impulse maximum at608; the declared value is its group delay. Init with all processing off, global Mix0 and Bypass retain the exact608-sample delayed raw signal. This is about13.79ms at44.1k,12.67ms at48k,6.89ms at88.2k, 6.33ms at96k and3.17ms at192k. Fixed PDC avoids latency jumps when modules change. Mix0/Bypass have the same delay. Intersample analysis uses part of the existing lookahead and does not add latency. VINTAGE NOISE Noise is optional and defaults Off. Choose Tape, Valve or Cassette and set its approximate RMS level in dBFS. Input above-90dBFS arms it; after silence it holds for6 seconds and fades over2 seconds. Cold silence produces no hiss, even if Noise is enabled. The conservative host tail is9 seconds including delay. Noise is deterministic after reset for reproducible renders and tests. METERS AND PRESETS Input/outputRMS and held sample peak describe actual audio. Tube/Tape readouts show color-stage transfer compression; Limiter GR shows applied envelope gain. Their sum is a useful activity indication, not a loudness or dBTP measurement. Presets are curated starting points, including deliberately aggressive and parallel examples. Global Mix and Bypass retain their documented dry behavior. Every sound control is automatable; ? explains controls and Guide is integrated. Published context: Hamalainen, DAFx2002, Smoothing of the Control Signal without Clipped Output in Digital Peak Limiters; ITU-R BS.1770-5 Annex2 discusses why sample peaks differ from reconstructed peaks. Original implementation, no vendor DSP or audio samples. Actual Logic-session listening remains Robin's QA.