VINAIxds Manual

01Overview

VINAIxds splits the incoming signal into four frequency bands with Linkwitz-Riley crossovers, then drives each band through its own distortion character, all in parallel, so the clean and distorted energy sit together. A 2048-point real-time analyzer overlays input and output so you can see the harmonics you create, with the three crossover handles living right on the graph.

Each band carries the same five characters (Tube, Tape, Diode, Fold and LDX) and can be switched between Analog and Digital voicing. Distort the low end into a wall while keeping the highs glassy and clean.

02The interface

• Analyzer (top): input vs. output spectrum overlay; drag the three crossover handles to set the band splits.
• Band strips (LOW · LOW-MID · HIGH-MID · HIGH) each carry their own character, Drive, Level, Dynamic and Proc/Mute/Solo/Delta.
• Global column (right): Input, Oversample, Filter Mode (Analog/Digital), Mix, Output, Low Cut and Safe Clip.

03Knobs & buttons

Per band

Crossovers & global

04The five characters

05Under the hood

Signal flow

The signal runs from IN gain into a Linkwitz-Riley crossover that splits it four ways. Each band drives in parallel through its own character, oversampled: upsample, shape with one of the five kernels, block DC on the added component only, then decimate back to base rate. The four bands recombine, pass the global Mix blend and OUT trim, then an optional pre-clip Low Cut and an optional Safe Clip brick-wall at 0 dBFS. Two FFT taps, one before the split and one after recombination, feed the analyzer overlay off the audio path.

• Crossover topology: Linkwitz-Riley filters split the spectrum into four bands. LR sections sum with flat magnitude and matched phase at each crossover, so the four bands recombine to unity when no drive is applied.
• Parallel band processing: each band runs its own waveshaper (one of five characters, Analog- or Digital-voiced) in parallel, not in series. Bands are phase-aligned before they are summed, so the recombined output stays coherent rather than smearing the splits.
• Oversampling & anti-aliasing: polyphase-IIR half-band oversampling at 1× / 2× / 4× / 8× / 16× (default 4×) runs the nonlinear stages above base rate so the harmonics generated by heavy drive stay below Nyquist. The oversampler alone carries the anti-aliasing, so higher factors matter most on aggressive characters and hot drive. There is no antiderivative or ADAA smoothing in the shapers: the transfer curve is evaluated per sample, and aliasing is set by the oversampling factor alone. Every band and the dry Mix reference pass the same oversampler, so they share its group delay and recombine coherently.
• DC block: a one-pole high-pass at 5 Hz sits inside each band and is applied to the added component only (out = in + block(shaped − in)). Asymmetric characters like Tube and Diode inject DC, and this removes it without touching the clean band, so a band at zero drive is a true null rather than a high-passed copy.
• Latency & PDC: the polyphase-IIR oversampler adds a small fixed latency. That figure is reported to the host and PDC-compensated, so the wet path stays sample-accurate against the rest of the session; it scales with the oversampling factor.
• Metering: a 2048-point FFT drives the input + output analyzer overlay, so you read the harmonic content you add directly against the source spectrum. The analyzer taps are measurement-only and do not feed back into the audio.
• Safe Clip: the optional final stage is a native-rate brick-wall clamp that pins the sample peak at 0 dBFS. It adds no latency and stays transparent below full scale, so it only ever catches an overshoot rather than coloring the level.

06Measured

Measured in-house (headless host + numpy/scipy), 48 kHz.

07Specifications