MixBus Platinum v1.2

1. Introduction

MixBus Platinum is a comprehensive broadcast audio processor designed for professional FM transmission, HD Radio, and digital streaming delivery. It implements a complete 14-stage serial signal chain plus ProcIQ supervisory intelligence, delivering broadcast-grade audio processing for today’s online and broadcast delivery requirements.

The processor engine was engineered to address the specific challenges of modern broadcast environments where source material arrives in widely varying formats—from heavily compressed HE-AAC to pristine lossless PCM—and must be conditioned for consistent, competitive on-air presentation without sacrificing musicality or listener fatigue management.

1.1 Key Features

1.2 Signal Chain Overview

The MixBus Platinum signal chain consists of 14 serial processing stages, supervised by ProcIQ’s adaptive intelligence layer. Audio flows left-to-right through every enabled stage in the fixed order shown below.

1.3 Parameter Smoothing

All continuous parameters in MixBus Platinum are interpolated to prevent audible artifacts during real-time adjustment. The smoothing time varies by parameter category:

2. Master Settings

Master Settings govern the global operating mode of MixBus Platinum. These parameters affect how every subsequent stage in the signal chain behaves.

2.1 Controls

2.2 Input Format and Adaptive Behavior

The Input Format selection automatically configures multiple stages to optimally handle the characteristics of the incoming audio. The table below summarizes the adaptive behavior triggered by each format:

3. APRE

APRE is the first processing stage in the MixBus Platinum signal chain. It applies psychoacoustically-modeled restoration to compensate for the specific artifacts introduced by lossy audio codecs. Unlike conventional input conditioning that applies static equalization, APRE analyzes the incoming signal in real time and applies targeted corrections only where codec damage is detected.

APRE operates automatically based on the Input Format selection. There are no user-adjustable parameters.

3.1 Stage Architecture

Stage 1: Transient Micro-Restoration HE-AAC only

Low-bitrate HE-AAC encoding smears transient attacks as the codec allocates bits toward tonal reconstruction at the expense of temporal precision. Stage 1 detects transient events in the 1–4 kHz band using a dual-envelope comparator with a 1.5 ms fast attack and an 80 ms slow baseline. When the fast envelope exceeds the slow envelope by a threshold amount, a brief dynamic high-shelf boost of +0.5 to +2 dB is applied, restoring perceived attack sharpness without adding sustained brightness.

Stage 2: SBR Harmonic Stabilizer HE-AAC only

HE-AAC uses Spectral Band Replication (SBR) to reconstruct frequencies above the codec’s baseband cutoff (typically 6–14 kHz). SBR produces characteristic modulation artifacts where reconstructed harmonics fluctuate in level from frame to frame.

Stage 2 monitors the modulation depth of the SBR region and applies gentle dynamic smoothing to stabilize the fluctuating harmonics. A small amount of band-limited saturation (tanh waveshaping at 3.5% blend in the 5–10 kHz range) adds subtle harmonic density that masks residual SBR artifacts.

Stage 3: Stereo Coherence Repair HE-AAC + AAC-LC

Codec quantization introduces decorrelation between left and right channels, particularly in the high-frequency region above 8 kHz. Stage 3 measures the running cross-correlation of the HF content and gently reduces the decorrelated component while optionally reinforcing subtle mid-band width in the 2–5 kHz region.

Stage 4: Micro-Dynamic Density HE-AAC + AAC-LC

Lossy codecs sacrifice low-level detail to maintain accuracy on louder content. Stage 4 applies very light upward micro-compression (1.1–1.2:1 ratio) within the 300 Hz–4 kHz band, targeting only low-level material below −24 to −16 dBFS. Gain change is capped at 3 dB.

4. ProcIQ

ProcIQ is the adaptive supervisory intelligence of MixBus Platinum. Unlike conventional broadcast processors that react blindly to level changes, ProcIQ actively listens to your audio, performing real-time spectral analysis across 8 frequency bands on every single sample from 80 Hz through 12 kHz. From this analysis it derives spectral centroid, spectral tilt, spectral flux, zero-crossing rate, spectral flatness, and per-band energy ratios.

Content is classified as Voice, Music, or Mixed through a hysteresis-stabilized state machine. ProcIQ maintains a rolling temporal memory of spectral and stress data. A closed-loop effectiveness tracker continuously monitors whether corrections are reducing stress. All adaptation is constrained by user-defined ceiling limits.

4.1 How ProcIQ Works

ProcIQ operates through four interconnected subsystems:

8-Band Spectral Analyzer

Center frequencies at 80, 160, 400, 1000, 2500, 5000, 8000, and 12000 Hz. Computes spectral tilt and energy distribution across the full audio bandwidth on every sample.

Scene Classification

A hysteresis-stabilized state machine classifies content as Voice, Music, or Mixed. Two classification engines are available: Classic (heuristic rules) and Neural Lite (softmax classifier). ProcIQ applies bounded parameter offsets to AGC, compression, presence, de-essing, stereo, limiter, clipper, and SmartBand stages based on the current scene.

Stress Model

ProcIQ continuously computes three stress metrics:

• Artifact Stress — Measures processing artifacts such as distortion and pumping
• Density Stress — Measures over-compression and loss of dynamics
• Stereo Stress — Measures stereo image instability and excessive width

Effectiveness Tracker

A closed-loop system with a range of 0.3–1.0 that continuously monitors whether ProcIQ’s corrections are actually reducing the measured stress values. When effectiveness drops, ProcIQ automatically scales back its intervention.

4.2 Controls

4.3 Telemetry

When Verbose Telemetry is enabled, ProcIQ exposes the following real-time readouts:

4.4 When to Enable ProcIQ

ProcIQ is most beneficial in the following scenarios:

• Mixed-content broadcasts — Stations that alternate between music, voice, and mixed programming benefit from automatic adaptation.
• Live radio — Unpredictable source material requires real-time processing adjustments.
• Unattended operation — ProcIQ provides automated oversight when no engineer is monitoring the chain.

For consistent single-genre content where the processing chain has been carefully tuned, ProcIQ can be disabled. At reduced strength (20–40%), ProcIQ provides a useful safety net without significantly altering the established sound signature.

5. Phase Rotator

The Phase Rotator reduces the crest factor of asymmetric waveforms by applying frequency-dependent phase shifts in the bass region. It uses four cascaded first-order allpass filter stages, all tuned to 200 Hz. Each allpass stage shifts phase without altering the frequency response. The cascaded result produces 360° of total phase rotation at 200 Hz.

All internal processing operates at f64 (64-bit double precision) for maximum numerical accuracy.

5.1 Controls

5.2 Recommended Settings

6. Stereo Enhancer

The Stereo Enhancer is a 3-band Mid/Side processor designed to produce wide, engaging stereo presentation that remains safe under downstream limiting, clipping, and FM pre-emphasis.

6.1 Operating Principle

The incoming stereo signal is decoded into Mid (L+R) and Side (L−R) components, then split into three frequency bands:

• Low band — Split from the Bass Mono frequency (80–350 Hz)
• Mid band — Between the bass mono and HF enhance crossover points
• High band — Split from the HF Enhance Crossover (400–8000 Hz)

Each Side band receives an independent width multiplier. A transient detector operates on the high-band Mid channel with dual envelope followers (2 ms attack / 240 ms release), enabling transient-aware stereo enhancement that can widen the image on percussive events while maintaining stability on sustained material.

6.2 Bass Monophonizer

A high-pass filter applied to the Side channel collapses frequencies below the Bass Mono Freq (60–200 Hz, default 120 Hz) to mono. This prevents intermodulation distortion in FM stereo encoders and maintains tight low-frequency imaging for all delivery modes.

6.3 Per-Band Ratio Guard

The ratio guard enforces Side-to-Mid energy limits on a per-band basis to prevent excessive stereo width from causing problems in downstream processing:

• FM modes — Auto-enforced caps of 0.82 (75µs) and 0.72 (50µs) ensure compatibility with FM stereo encoding.
• Online / HD Radio — User-configurable via the Side Limit Ratio parameter.

A global guard with configurable attack and release times smoothly reduces Side energy when the ratio exceeds the limit. A Side lookahead buffer (0–3 ms, default 2 ms) allows the guard to anticipate transients.

6.4 Decorrelation Injection

Decorrelation injection creates the perception of enhanced stereo width by introducing phase-shifted Mid content into the Side channel via two cascaded allpass filters. A noise floor detector suppresses injection at approximately −60 dBFS to prevent artificial widening of silence or very quiet passages.

Two algorithm modes are available:

• Transient Focus — 45% injection, transient-gated. Decorrelation is applied primarily during transient events for a punchy, dynamic stereo image.
• Decorrelated — Full injection plus 35% HF transient width boost. Provides a consistently wider image across all content.

6.5 Controls

6.6 Advanced Controls

7 — Automatic Gain Control (AGC)

The AGC operates as a slow macro-level loudness stabilizer, maintaining consistent output levels across varying source material without audible pumping. It employs a dual-envelope follower architecture: a fast envelope running at 50% of the attack time and 30% of the release time tracks transient peaks for protection, while a slow envelope using the full attack and release values provides the stable reference for gain calculation.

Windowed target-zone gating is a key design feature. When the signal level is already within the configurable Window Size of the target, the AGC release slows dramatically by the Window Release Mult factor. This prevents the leveler from adding unnecessary density to well-controlled material, producing more natural-sounding results that preserve intended dynamic expression.

7.1 — Controls

7.2 — Crossover & Program Adaptation

The crossover uses LR4 topology (Linkwitz-Riley 4th order) at 24 dB/octave, delivering perfectly flat magnitude response and zero phase difference at the crossover point. The bands recombine after processing without comb filtering or level discontinuities.

Program Adaptation dynamically adjusts the crossover frequency based on spectral energy balance between the low and high bands. When the low band carries significantly more energy, the crossover shifts slightly upward to give more spectral range to the low-band compressor. When the high band dominates, the crossover shifts downward. This adaptation improves consistency across widely varying mixes without overfitting a single fixed split point.

7.3 — Crossover Controls

8 — SmartBand EQ

SmartBand is a fully-parametric 6-band equalizer with a 7th adaptive band that operates in tandem with ProcIQ’s spectral intelligence. It provides 7 filter shapes: Peaking Bell, Low Shelf, High Shelf, High-Pass, Low-Pass, Band-Pass, and Notch. Full frequency range (20 Hz–20 kHz), gain (±24 dB), and Q (0.1–18.0) are available on every band. The adaptive HF band receives live spectral tilt data — up to +4.5 dB boost on dark content, up to −2.0 dB cut on bright — with a 30% minimum intervention floor. The engine operates at f64 internally.

8.1 — Controls

Global

Adaptive HF

Bands 1–6

Default Band Layout

8.2 — Adaptive HF Settings

When dark content is detected, the adaptive band applies up to +4.5 dB boost. When bright content is detected, up to −2.0 dB cut is applied. The amount is scaled by Adaptive Strength and the ProcIQ effectiveness tracker (30% minimum floor).

9 — Compression

MixBus Platinum provides two selectable compression architectures, switchable with a click-free 10 ms crossfade: Dual-Band mode (LR4 crossover) and 5-Band Multiband mode.

9.1 — Dual-Band Mode

The signal is split at the Crossover frequency (120–300 Hz, default 180 Hz) using an LR4 filter with perfectly flat magnitude response. Each band features a soft-knee algorithm with a 10 dB default knee width: compression begins 5 dB below the threshold and reaches the full ratio 5 dB above it, producing transparent gain control that avoids the abrupt character of hard-knee compression.

Each compressor uses a dual-time-constant release system: a fast release provides quick recovery from transient peaks, while a slow release maintains smooth sustained gain reduction. The Release Blend control sets the static mix between fast and slow. An adaptive algorithm automatically increases the slow release contribution when the signal is far above the threshold, providing program-dependent behavior that sounds natural across widely varying content.

9.2 — Dual-Band Controls

9.3 — Multi-Band Mode

Five bands with split frequencies at 120 / 700 / 2800 / 8000 Hz.

10 — Presence

The Presence stage combines a 60% parametric bell with a 40% high-shelf at the target frequency, providing articulation and top-end energy that cuts through broadcast processing without brittleness.

The Adaptive HF system continuously monitors the broadband-to-HF energy ratio of the input signal using 500 ms smoothed envelopes. When the input material is spectrally dull, the system automatically increases presence gain. When material is already bright, it reduces gain. This keeps tonal balance consistent across wildly different masters without manual intervention. An optional Dynamic mode adds HF containment above a configurable threshold.

10.1 — Controls

11 — De-Esser

The De-Esser provides post-presence high-frequency control for sibilance and density management. Placed after the Presence processor and before the limiter, it prevents high-band aggressiveness from the presence stage from being amplified by downstream limiting and clipping. The de-esser operates as a split-band compressor: the signal is split into a low band (passed through unaffected) and a high band (compressed when sibilant energy exceeds the threshold).

The threshold is auto-offset by Input Format: +2 dB for AAC-LC and +3.5 dB for Lossless. This reflects the fact that higher-quality sources tend to have more intact sibilant energy and therefore require less aggressive de-essing to achieve a natural result.

11.1 — Controls

12 — TruePeak Limiter

Broadcast-grade peak-protection with optional ITU-R BS.1770 true-peak detection. The Master TruePeak Limiting toggle upgrades to 4× oversampled polyphase FIR. The limiter comprises a primary lookahead limiter plus a separate TruePeak Guard.

12.1 — Primary Limiter

Uses a circular delay buffer for true brick-wall limiting at the configured ceiling without overshoots. The configurable Lookahead (0.5–5 ms, default 2.5 ms) determines how far ahead the limiter can see; longer lookahead produces smoother, less audible gain reduction at the cost of slightly more latency.

Program-dependent release extends the release time proportionally to the depth of current gain reduction. When enabled, the effective release is calculated as: release × (1 + (1 − gain) × 2), meaning deeper limiting produces proportionally slower recovery. This makes the limiter less audible on heavily limited material without requiring manual adjustment.

When True Peak Limiting is enabled, the lookahead detector upgrades from sample-rate detection to 4× oversampled polyphase FIR peak detection. A 48-tap Kaiser-windowed sinc filter (β = 7.857, ~80 dB stop-band rejection) is decomposed into twelve-tap polyphase sub-filters evaluating the signal at ¼, ½, and ¾ sample offsets. The lookahead buffer sees these FIR-detected peaks before they arrive and reduces gain smoothly, with zero overshoot at the ceiling.

12.2 — TruePeak Guard

The TruePeak Guard operates as a safety net at the very end of the signal chain, catching any inter-sample peaks that the upstream limiter, clipper, or pre-emphasis stages may have introduced. It uses the same 4× polyphase FIR detection with a fixed 0.15 ms attack when inter-sample peaks exceed the configured ceiling. In FM modes, this stage serves as a post-emphasis overshoot guard, controlling the peak excursions that FM pre-emphasis produces in high-frequency content.

12.3 — Controls

12.4 — Standards Compliance

The FIR true-peak detection follows the measurement methodology outlined in ITU-R BS.1770 Annex 2. The 48-tap Kaiser-windowed sinc filter (β = 7.857) provides approximately 80 dB of stop-band rejection. The filter is decomposed into four polyphase phases of twelve taps each, enabling efficient 4× oversampled peak detection without running the entire signal path at 4× the sample rate.

Computational overhead: 72 multiply-accumulate operations per stereo sample pair per stage. At 48 kHz this is approximately 3.5 million MACs/second—negligible compared to the rest of the signal chain. When True Peak Limiting is OFF, the limiter and guard fall back to simple sample-rate peak detection with effectively zero additional overhead.

13 — Shaping

Three selectable waveshaping algorithms add harmonic density and character to the processed signal.

13.1 — PurestDrive

A sin(x) waveshaper with an audio-rate apply factor. Rather than applying a fixed transfer curve to every sample, PurestDrive dynamically calculates its intensity based on the relationship between the current and previous samples. Sustained, body-rich content receives saturation, while attacks and airy high-frequency content pass through with minimal coloring.

The Amount control directly scales the waveshaper intensity from 0 (clean) to 1 (full saturation). The Drive Trim provides level compensation. All processing uses f64 precision (64-bit double) with TPDF dither to maintain low-level accuracy even under heavy saturation.

13.2 — PurestSaturation

A mantissa-preserving polynomial waveshaper:

x − x³/8 + x⁵/128 − x⁷/4096 + x⁹/262144 − x¹¹/33554432

The polynomial uses intentional power-of-two coefficients for mantissa-preserving arithmetic—clean at low input levels, increasingly thick and present harmonics when driven harder.

The Amount control drives the input gain into the waveshaper, scaling from 1× to 10× and determining how deeply the signal enters the nonlinear region. The Character control morphs the output from thick (0.0) to clean (1.0), providing tonal flexibility independent of the drive amount. The Saturation Trim provides output level compensation.

13.3 — Clipper GX

A 4-stage broadcast clipping pipeline with 2× oversampling.

The entire Clipper GX pipeline runs at 2× internal sample rate using midpoint interpolation between adjacent samples, with anti-alias reconstruction filtering (60%/40% blend with 16 kHz lowpass) on the output to minimize aliasing artifacts.

13.4 — FM Path Behavior

When the Output Delivery Mode is set to FM 75µs or FM 50µs, the Shaping stage operates in a special FM mode. A dedicated post-emphasis GX clipping stage processes the pre-emphasized signal, providing the final peak control needed for FM transmission. The FM overshoot guard follows the Final TruePeak Guard toggle in the Limiter tab—when the TruePeak Limiting master toggle is ON, the overshoot guard uses FIR-interpolated peak detection.

14 — Output & Delivery

14.1 — Delivery Modes

14.2 — FM Signal Path

The pre-emphasis filter is implemented as a bilinear-transformed analog prototype: H(s) = (1 + s·τ1) / (1 + s·τ2), where τ₁ is the emphasis time constant and τ₂ = τ₁ / 7.94, capping maximum boost at +18 dB. The output is the pre-emphasized signal; de-emphasis is a receiver function and is not included in the transmit path.

14.3 — LUFS Metering

K-weighted momentary loudness measurement per ITU-R BS.1770-4 with 400 ms integration. K-weighting applies a +4 dB high-shelf filter at 1681 Hz followed by a 38 Hz high-pass filter, modeling human hearing sensitivity where bass energy is de-weighted and presence-band energy is emphasized.

14.4 — TruePeak Metering

Measured in dBTP (decibels relative to True Peak). When the True Peak Limiting master toggle is ON, this meter reflects the 4× FIR oversampled peak level for Online/HD paths using limiter/guard FIR detectors, while FM delivery reports the post-chain sample peak after FM overshoot control. When the master toggle is OFF, the meter reports sample-rate peak levels.

15 — Technical Specifications

Appendix A — Quick Start Guide

Get up and running with MixBus Platinum in ten steps.