AudioEngine

AudioEngine or (AE) refers to XBMC's new audio system.

Features
AudioEngine replaces SDL and brings some of the external dependencies into XBMC, and wraps up all the different media types for mixing, samplerate conversion, format conversion, encoding, upmix, downmix, etc.


 * Features of AE include:
 * Support for DTS-HD MA / Dolby TrueHD Bluray formats
 * Support for 24-bit and floating-point audio at up to 384,000hz
 * Mixing of all streams including GUI sounds even when transcoding audio
 * Start-up enumeration of hardware audio devices and their capabilities with log output
 * Bitstreaming support in PAPlayer (XBMC's music player)
 * Upmixing of stereo to full channel layout
 * Tighter syncing of A/V streams
 * Floating-point processing of audio
 * 24-bit and floating-point decoding/handling of mp3
 * Full support for ReplayGain
 * Built-in sample-rate conversion and transcoding

It's still early days for AE. Bugs will be found, and new and exciting features added.

Settings
Audio settings can be found in the GUI at


 * Settings -> System -> Audio output

Refer to AudioEngine/HOW-TO: Configure audio for more detail.

Supported Audio Formats for Passthrough
A summary of some of the supported formats within the AudioEngine

Test audio files
Below You can find a excellent range of flac test audio files stereo/multichannel and varied sample/bit rates and quality.

Click here to download flac test files

Click here to download wma test files

Click here to download various audio system test files

History and Team
AudioEngine, AE for short was two-year project spear-headed by Gnif which was released in v12 (Frodo), and was first merged with master for user testing as of May 15th, 2012.

Why write a new audio subsystem?
It's been an accepted fact that the existing audio code was little-touched since the early days of XBMC, and was limited by the 16-bit architecture of the original Xbox and codecs available at the time. Dolby AC3 and mp3 (or earlier formats!) ruled the media codec world. Over time HTPC's became more mainstream, more powerful and better connected. Analog connections gave way to optical SPDIF connections, which in turn were replaced with HDMI. Processors, GPU's and media formats evolved at a steady pace.

With the growth of the HTPC and new formats like those made available with Bluray technology, the audio subsystem was under scrutiny as an area for improvement. Higher definition audio like multichannel FLAC was becoming a preferred standard for those demanding higher quality audio media. It was time for an update, and what an update!

Early efforts at patching fixes and even an initial attempt at a completely new system fell by the wayside. The scope of the required changes and the complexity of the API made it a daunting task, especially for volunteer programmers working in their spare time. This was compounded by the fact that XBMC runs on a wide range of hardware and operating systems. Most of the developers focus on specific platforms or subsystems within the whole of XBMC. Any new system had to be platform-agnostic, cohesive and flexible.

In time, it became obvious the project was going to require a full re-write and replacement of the audio engine. Furthermore, XBMC's base of hardware and OS platforms continued to grow. The detailed specifications for the new audio formats were unknown, and of course the master code for XBMC was ever-changing.

Three years of work
In November of 2009 Gnif decided to bite the proverbial bullet, and after extensive consultation with the team finalized the API or structure which was to become AE. Drawing in other developers as required, the massive project slowly took shape, and reached the ready-point for user tests a year later. The forum thread for those hardy beta-testers (thank you!) reached over 1,200 posts, and slowly but surely features were added and debugged, and system stability grew.

Over the next two years, thanks to the determination of Gnif the core engines, decoders, encoders and utilities that make up AudioEngine became a reality, and other developers began to contribute to the core and especially the platform-specific sinks or output stages. Among those contributing were Gimli, Fneufneu, Memphiz, Dddamian, Anssi and others.

There was a goal to have the new audio system ready for Eden 11.0, but there was just too much left to do. Many users were eagerly looking forward to it's inclusion, but stable releases must continue, and AE wasn't ready when the Eden feature-freeze went into effect. During this time the AE code-base was solidified, bugs tracked and squashed, features added and testing increased.

Finally, on May 15th 2012, AudioEngine was merged to master. Weighing in at over 22,000 lines of code, it represents one of the single biggest code-merges ever for XBMC. Now for the first time it is readily accessible for all the development team (and users!) to see, test and improve within the greater context of XBMC's master branch.

ActiveAE
ActiveAE was a large rewrite of SoftAE and combined all platforms to one engine. Sinks are available for all platforms. ActiveAE uses an active state machine pattern, with a separate high priority sink thread, that cares for underruns. ActiveAE is completely build upon ffmpeg, which is used for resampling, channel mapping and decoding. High performance SSE algorithms - already existent in the AEUtils - were fixed up and are used to do the final conversion into formats the sinks can accept. ActiveAE uses the float format internally and it is working bit-exact.

A new feature is the upmix via AC3 transcoding support. On optical devices where only 2 pcm channels are available, we can now upmix to 5.1 by transcoding to ac3. Another new feature is the PulseAudio sink for ActiveAE, which was merged into XBMC mainline on January 9th 2014 and enables Linux desktop users to use the PulseAudio system instead of ALSA, which was an often requested feature. See PulseAudio

ActiveAE has three build-in profiles: Best Match, Optimized and Fixed. The main developers of ActiveAE are Fernetmenta and Fritsch.
 * Fixed takes care that all audio you will play is play in the very same data format. Your receiver does not need to switch modes.
 * Best Match is what previously existed as the "audiophile" mode. Whenever a new video or music playback starts, the audio engine will select the best sink for that stream.
 * Optimized is something special, as it tries to reduce the reopening of sinks a lot. One example is live-TV where it switches between 5.1 to 2.0 during commercials. In order to  not open the device anew and loosing some (milliseconds of) audio, the 2.0 commercials are played with the already open 5.1 sink by muting the non existent channels.