MPX Prime Studio

Version: 0.42

MPX Prime Studio is an FM composite (MPX) generator written in Swift. It takes live audio input or a test tone, applies optional broadcast-style processing, generates stereo FM baseband with pilot and optional RDS, and sends the MPX (plus optional decoded monitor audio) to the sound hardware.

It runs on two platforms:

• macOS — a native SwiftUI application: the MPX Prime Studio encoder plus its companion MPX Prime Meter analyzer, both shipped in one DMG (Core Audio, full GUI).
• Linux — the encoder runs headless from the command line (ALSA output, no GUI), with the embedded REST API + web dashboard as its interface. SIMD-accelerated so the full chain runs in real time on low-power hardware (a fanless Celeron J4105 handles it); shipped as a Debian/Ubuntu package with a systemd service.

The same DSP core drives both: a full broadcast-style processing chain — phase rotator, wideband AGC, 4-band parametric EQ, 3-/5-band multiband compressor, stereo widener, PrimeBass, bass and audio-band clippers, L/R pre-emphasis, pre-encode true-peak limiter, BS.412 power limiting, and an oversampled composite clipper — ahead of a pilot-locked stereo encoder, keeping the pilot and RDS subcarriers out of all peak control (post-clipper injection). The Linux build is bit-for-bit the same processing; only the audio backend and the front end differ.

Intended use and status. MPX Prime Studio is for experimental, hobby, and small-budget broadcast — community / LPFM stations, pirate and SDR-fed exciters, prosumer encoding, and study of FM signal processing. It implements core behavior from EN 50067 / IEC 62106 and common FM-stereo practice, but it is experimental and not certified — no conformity or compliance is promised. Do not rely on it for regulated production broadcast.

App structure

The macOS GUI is organised into these sections. The Linux build has no GUI — the web dashboard (see Remote control) mirrors this same layout (Monitoring, per-stage Processing, RDS, Interfaces, Test Tone), so the two front ends feel the same.

• Monitoring: live status, transport, interfaces summary, DSP status, RDS snapshot
• Processing: Overview, Core, Phase Rotator, AGC, Parametric EQ, Multiband (with optional transient-aware attack + inter-band gain coupling), Expander, MB Limiter, Stereo Widener, PrimeBass, Bass Clipper, DC Clipper, Audio Limiter, Composite Clipper (optional look-ahead peak control and experimental multiband composite clipping on top of the soft-clipper), BS.412, Final Stage
• RDS: status (master enable + live snapshot), identity (PI / PTY / PTYN / ECC + PS banks + runtime flags TP / TA / MS / DI), radiotext (RT / RT+ / Now Playing), long PS, alt. frequencies (AF), schedule (group sequence + clock-time), subcarrier (injection level + frequency + Gaussian shaping)
• Tools: Test Tone (sine / pink / white, four stereo modes, frequency presets, dBFS level — replaces the audio input live when enabled, ⌘T)
• Settings: configuration path, interfaces, output mode (MPX composite vs processed audio), audio engine, spectrum options
• Separate windows: Scopes, Spectrum (composite spectrum with FM band captions -- Mono L+R, 19 kHz Pilot, Stereo L-R, 57 kHz RDS, SCA), Levels, Help

The RDS detail tabs are organised per UECP message-class taxonomy (AF is a peer of PS, RT+ lives under ODA, etc.). Every operationally toggled RDS setting applies live without restarting the transport — PI, PTY, PTYN, TP/TA/MS/DI flags, AF list, group sequence, CT enable, all RT/PS/Long PS text. Only physical-layer settings (rds_level, Gaussian shaping FIR taps/BW) require a transport restart since they reconfigure the modulator.

MPX Prime Studio — the encoder (macOS + Linux)

Makes the FM multiplex. Runs as a macOS GUI app or headless on macOS/Linux.

• Real-time MPX generation with 19 kHz pilot and 38 kHz stereo subcarrier
• Premium receiver-side stereo separation at the default config (0.28): 65 dB at 1 kHz, 50.5 dB at 10 kHz, 43.4 dB at 14 kHz, measured by --verify-receiver through the reusable MPXDecoder (matches Optimod 8x00 / Stereotool published numbers)
• Optional RDS generation with pilot-locked 57 kHz subcarrier
• Live input source or built-in Test Tone generator (sine / pink / white, mono / L=−R / left-only / right-only modes, frequency presets, −60..0 dBFS level slider, live Enable toggle that replaces the audio input without restarting the engine)
• Optional wideband AGC, HPF, program lowpass, HF trim, PrimeBass, mono bass, stereo widener, and multiband processing (including 0.28 opt-in transient-aware attack + inter-band gain coupling)
• Broadcast-style Final Stage (Broadcast Preset + Final Drive + Composite Deviation + Final-MPX safety limiter with look-ahead) and a separate Audio Limiter tab (pre-encode 4× oversampled stereo-linked true-peak limiter with default-on look-ahead and Dolby HF-subband-aware detector — US 5,579,404, expired 2013 — for audibly cleaner HF transients and preserved LF punch), feeding the 16× oversampled composite clipper (with optional OS-rate sliding-window-max look-ahead peak control and experimental default-off multiband composite clipping) with live clipper telemetry
• Calibrated MPX line output in dBFS for repeatable exciter drive
• TX-path engine toggles on the Core tab: linear-phase FIR encoder lowpass and FIR multiband splitters (latency vs. quality choices, restart-required)
• Composite budget telemetry with pilot/RDS/audio visibility, safety-limiter readout, and a composite budget governor that holds the audio path under the post-injection clamp so pilot/RDS subcarrier amplitude stays constant for sane configs (over-budget flag for impossible configs)
• Broadcast preset picker for AGC/final-stage tuning (Balanced Music, CHR / Dance, Punchy Music, Speech / Talk)
• Italo / disco / dance multiband presets (5B Italo, 3B Italo) with pumped low-band character
• Decoded MPX monitor output on a selectable monitor device
• Processed-audio output mode (Settings - Output Mode): emit processed stereo L/R instead of the FM composite, to feed an external stereo coder + RDS encoder on transmitters that only accept L/R / AES3 audio. Runs the full audio chain (no composite clipper / BS.412 / pilot / RDS), with selectable pre-emphasis (apply it here, or stay flat if the coder does); runs at the audio device rate (48 kHz / 24-bit recommended). Composite-only and RDS controls hide while it is active.
• Scopes, spectrum, levels, sticky peaks, and live monitoring views (macOS GUI; mirrored by the web dashboard)
• Remote control — an embedded, default-off REST API + web dashboard for local or remote operation (see below); it is the primary interface on the headless Linux build
• Linux command-line build (experimental): the encoder runs headless with ALSA output, SIMD-accelerated so the full chain fits low-power hardware; shipped as Debian/Ubuntu packages with a systemd service. No GUI, no Meter.
• Config persisted to the INI (~/Library/Application Support/MPX Prime Studio/MPX Prime Studio.ini on macOS; ~/.local/share/... or /var/lib/mpxprime/ on Linux)

MPX Prime Meter — the analyzer (macOS only)

The receive/analyze counterpart, shipped as MPX Prime Meter.app in the same DMG. Where Studio makes the composite, the Meter measures it: feed it an MPX composite (a Core Audio input device, or a live station via an in-process RTL-SDR / SDRplay tuner) and it decodes stereo + full RDS on one dashboard window. Its deviation, MPX-power, and SM.1268 readings were cross-validated against a Profline SFP-X measuring receiver (pilot / RDS matched exactly, peak deviation within the SM.1268 +/-2 kHz tolerance). macOS only, Apple Silicon (the SDR tuner is arm64) — there is no Linux or Intel Meter.

• Decoded scopes (composite, decoded L, decoded R) and a stereo vectorscope
• MPX spectrum (0-100 kHz) with band captions (Mono L+R, 19 kHz Pilot, Stereo L-R, 57 kHz RDS, 67.65 kHz Direct Band, 92 kHz SCA)
• Measurement-grade deviation + modulation metering to ITU-R SM.1268-5 / BS.412-9: IN / L / R / M / S levels + correlation; pilot / RDS / total (MAX) deviation; MPX power (ITU-R BS.412 uniform sliding 60 s window) with the worst-window MPX MAX; trailing-60 s peak +/- deviation; the OVER 77 kHz SM.1268 exceedance statistic; best stereo separation; SIGNAL (SDR RSSI); and deviation / MPX-power trend graphs
• Full RDS decode: PI / PS / PTY / RT / RT+ / Long PS / CT / AF / group histogram + live BER
• WAV recording: the decoded stereo audio, or the raw MPX composite (24-bit, capture rate), for later re-analysis
• Input: an audio device, or in-process SDR tuning (Source -> SDR) -- RTL-SDR and SDRplay RSP (auto-preferred when present) are decoded by a linked-in tuner library (no helper process, no Homebrew for end users), just a connected dongle. Frequency, IF bandwidth, gain / auto gain, LNA, antenna, Bias-T, PPM, and RTL AGC all retune live (no restart). Headless terminal modes also exist (./run-meter.sh --device <n> / --stdin)

See the MPX Prime Meter manual for details.

Remote control

An embedded, default-off REST API + web dashboard ([CONTROL] in the INI, or the GUI Settings tab) controls the encoder locally or remotely: transport, live meters, RDS text/TA, sound presets, and every INI setting with live-apply where the engine supports it. Localhost needs no authentication; any wider bind requires an API key. See the user manual.

This is a MPX Prime Studio (encoder) feature only — on both macOS and Linux. The MPX Prime Meter has no REST API or web interface for now; it is operated from its own window (or its headless terminal modes).

Output modes

MPX Prime Studio drives its main output device in one of two modes, chosen in Settings → Output Mode (restart-required). What you need from your hardware depends on which you use:

• MPX Composite (default) — the finished FM multiplex: mono sum + 38 kHz stereo subcarrier + 19 kHz pilot + optional 57 kHz RDS, for a transmitter / exciter that accepts a composite ("MPX" / "wideband" / baseband) input. Needs a 192 kHz output device so the ~59 kHz upper RDS sideband stays below Nyquist. A 96 kHz device can carry stereo but not RDS.
• Processed Audio — the processed stereo L/R audio only (no pilot / subcarrier / RDS / composite clipper / BS.412), for transmitters that accept only L/R analog or AES3 audio and have their own stereo coder + RDS encoder. 48 kHz / 24-bit is all you need — the high composite sample rates do not apply. Pre-emphasis is selectable (apply it here, or stay flat if the coder does), with an optional final loudness clipper. See the user manual.

A separate, optional Decoded Monitor output (any sample rate) demodulates the internal composite back to L/R for headphone monitoring on a second device — a listening aid, not the on-air signal.

Requirements

• macOS 15+ (primary platform). An experimental Linux command-line port of the encoder (headless --nogui into an ALSA device, verifier, benchmark; no GUI, no Meter) builds from the same source tree -- see docs/BUILDING.md.
• Platform support tiers: Apple Silicon (arm64) is Tier 1 — the primary, fully-supported target. Intel (x86_64) is Tier 2, best-effort — the universal binary runs and the audio chain is identical, but performance tuning (e.g. the GUI refresh profile) targets Apple Silicon first; Intel gets lighter-weight fallbacks where they help but is not the optimization priority.
• Xcode command line tools / Swift 6 toolchain (only needed for building from source — download the DMG below if you just want to run it)
• Audio output device — depends on the output mode (see above):
  • MPX Composite with RDS: an external USB / Thunderbolt interface that runs 192 kHz natively. Built-in Mac audio tops out at 96 kHz, which cannot carry the 57 kHz RDS subcarrier; 96 kHz can do stereo-without-RDS.
  • Processed Audio (feeding an external coder): 48 kHz / 24-bit is sufficient — built-in audio or any interface works.
• The output device's format in Audio MIDI Setup must match the configured sample_rate, or Core Audio's implicit resampling starves the render thread.
• Input devices may run at any rate; the app converts internally.

Download

Pre-built universal binaries (Apple Silicon + Intel) ship as macOS .dmg files on the project's GitHub Releases page:

github.com/bkram/MPXPrime/releases

Each release is built and signed by GitHub Actions from the matching tag. Pick the latest version, download MPX_Prime-<version>.dmg, and drag the apps into /Applications (or any folder you prefer). The DMG contains two apps: MPX Prime Studio (the encoder) and MPX Prime Meter (the companion analyzer, below) — install whichever you need.

Linux (encoder only): the same releases attach Debian/Ubuntu packages mpxprime_<version>-ubuntu24.04_amd64.deb and ...-ubuntu26.04_amd64.deb. Install with sudo dpkg -i mpxprime_*.deb (or sudo apt install ./mpxprime_*.deb to pull dependencies); it provides the headless encoder as a mpxprime systemd service with the web dashboard. This is the CLI encoder only — no GUI, no Meter. See docs/BUILDING.md for setup.

First-launch security note

MPX Prime Studio is ad-hoc signed, not Apple-notarized. The DMG is built and signed by an automated GitHub Actions workflow with a self-managed signing identity — it is not enrolled in the Apple Developer Notary Service. As a result, macOS Gatekeeper will refuse to open the app on first launch with a message similar to:

"MPX Prime Studio" cannot be opened because Apple cannot check it for malicious software.

This is the standard macOS warning for any app distributed outside the Mac App Store / Apple Notarization. To approve the app once:

1. Open System Settings → Privacy & Security.
2. Scroll to the Security section near the bottom. You will see a message like "MPX Prime Studio was blocked from use because it is not from an identified developer".
3. Click Open Anyway next to that message.
4. The next time you launch MPX Prime Studio, macOS will prompt one more time — click Open.

After the first approval, MPX Prime Studio launches normally on subsequent runs. This is a one-time per-version operation; updating to a new release will trigger the prompt again on first launch.

If you would rather skip the Gatekeeper dialog entirely, build from source (see docs/BUILDING.md) — locally built binaries are not subject to the same check.

Documentation

• docs/manual.md — MPX Prime Studio manual: usage, first-time setup, configuration, RDS text, monitoring windows, verification, and the RDS PI/ECC + PTY reference tables
• docs/manual-meter.md — MPX Prime Meter manual: SDR/audio input, the measurement readouts, WAV recording, and calibration
• docs/BUILDING.md — build, run, verify, test, and package from source
• docs/ARCHITECTURE.md — detailed DSP chain and stage descriptions
• AGENTS.md — contributor / agent workflow guidance and release checklist
• plan.md — roadmap
• CHANGELOG.md — version history

References

• Standards PDFs and notes live in documents/ (EN 50067 / IEC 62106-2 / IEC 62106-6 / UECP SPB 490 / ITU-R BS.450)

Acknowledgements

The block-level RDS bit encoder in BasicRDSCoder — CRC (0x5B9), offset words, and the four-block group assembly shared by groups 0/2/3A/4A/10A/11A/15A — was initially ported from the Python RDSHelper in ryanginn/rds-master. Everything around it is MPX Prime Studio's own work: the 1187.5 bit/s biphase + Gaussian shaping FIR, the pilot-locked 57 kHz subcarrier generation, the audio-thread real-time pipeline (pre-allocated bit buffer, atomic CT cache, monotonic-clock timing), the RDSRuntimeConfig live-apply path, AF Method B, RT+ ODA (AID 0x4BD7), Group 4A clock-time with MJD + TZ, Group 10A PTYN, Group 15A Long PS, Group 1A ECC/LIC, the Stereotool-compatible text grammar, and the full FM composite chain that the encoder feeds into.

Trademarks

MPX Prime Studio is an independent open-source project and is not affiliated with, endorsed by, or sponsored by any of the companies named in this documentation. Product and company names — including Orban, Optimod, Omnia, Stereo Tool / Stereotool, Aphex, Waves, and others — are trademarks of their respective owners and are used here descriptively only, to identify published behavior, prior art, and platform APIs for comparison.

License

GPL-3.0. See LICENSE.