Hi HN,
If you’ve ever plugged an external monitor (HDMI/DisplayPort) or a USB DAC into a Mac, you probably know the frustration: the volume keys instantly die, the volume slider greys out, and macOS throws a icon at you. Apple strictly treats digital outputs as "Fixed Volume" devices and refuses to apply software gain.
I got tired of this, so I built SoundBridge – an open-source, ultra-lightweight system audio router that natively restores volume controls for any external device.
Repo: https://github.com/chenjy16/SoundBridge
The Problem with Existing Solutions:
DDC/CI tools (like MonitorControl): They are great when they work, but they frequently fail if you route through certain Type-C docks, KVM switches, or if the monitor firmware is just buggy. Plus, they don't work for external DACs.
Existing Software Mixers (eqMac, SoundSource): They are either heavy (Electron/web-based UI), expensive, or require intrusive Accessibility permissions to intercept the F11/F12 media keys using CGEventTap.
My Architecture (How SoundBridge works):
I wanted a solution that feels 100% native to macOS with near-zero latency. It consists of three parts:
The C++ HAL Plugin (Driver): Built on top of libASPL, it creates a virtual "Proxy Device". The magic trick here is that by properly implementing the kAudioDevicePropertyVolumeScalar and kAudioDevicePropertyMute properties at the driver level, macOS natively handles the F11/F12 keys and pops up the native volume HUD. No Accessibility permissions or keyloggers required.
Shared Memory (IPC): The C++ driver streams the intercepted PCM audio and the volume scalar state to the host process via a lock-free ring buffer (using std::atomic and mmap) in /tmp/.
The Swift Host (AVAudioEngine): A lightweight background daemon reads from the ring buffer, applies the software gain (using linear scaling to match macOS's perceived logarithmic volume curve), and outputs the bit-perfect audio to the actual physical HDMI device.
Request for Roast:
While I’m comfortable with Swift, writing lock-free concurrency in C++ for real-time audio (where a single lock or allocation can cause a glitch) was a steep learning curve.
I would absolutely love it if the C++ and CoreAudio veterans here could roast my driver architecture, specifically the RFSharedAudio memory barriers and the IPC synchronization.