---

Synthesis, Processing, and Data Format Conversion Code

Audio units synthesize, process, or transform audio data. You can do anything you want here, according to the desired function of your audio unit. The digital audio code that does this is right at the heart of why you create audio units. Yet such code is largely independent of the plug-in architecture that it lives within. You’d use the same or similar algorithms and data structures for audio units or other audio plug-in architectures. For this reason, this programming guide focuses on creating audio units as containers and interfaces for audio DSP code—not on how to write the DSP code.

At the same time, the way that digital audio code fits into, and interacts with, a plug-in does vary across architectures. This section describes how audio units built with the Core Audio SDK support digital audio code. The chapter “Tutorial: Building a Simple Effect Unit with a Generic View” includes some non-trivial DSP code to help illustrate how it works for effect units.

In this section:

Signal Processing
Music Synthesis
Music Effects
Data Format Conversion

Signal Processing

To perform DSP, you use an effect unit (of type 'aufx'), typically built as a subclass of the AUEffectBase class. AUEffectBase uses a helper class to handle the DSP, AUKernelBase, and instantiates one kernel object (AUKernelBase) for each audio channel.

Kernel objects are specific to n-to-n channel effect units subclassed from the AUEffectBase class. They are not part of other types of audio units.

The AUEffectBase class is strictly for building n-to-n channel effect units. If you are building an effect unit that does not employ a direct mapping of input to output channels, you subclass the AUBase superclass instead.

As described in “Processing: The Heart of the Matter,” there are two primary methods for audio unit DSP code: Process and Reset. You override the Process method to define the DSP for your audio unit. You override the Reset method to define the cleanup to perform when a user takes an action to end signal processing, such as moving the playback point in a sound editor window. For example, you ensure with Reset that a reverberation decay doesn’t interfere with the start of play at a new point in a sound file.

“Tutorial: Building a Simple Effect Unit with a Generic View” provides a step-by-step example of implementing a Process method.

While an audio unit is rendering, a user can make realtime adjustments using the audio unit’s view. Processing code typically takes into account the current values of parameters and properties that are relevant to the processing. For example, the processing code for a high-pass filter effect unit would perform its calculations based on the current corner frequency as set in the audio unit’s view. The processing code gets this value by reading the appropriate parameter, as described in “Defining and Using Parameters.”

Audio units built using the classes in the Core Audio SDK work only with constant bit rate (CBR) audio data. When a host application reads variable bit rate (VBR) data, it converts it to a CBR representation, in the form of linear PCM, before sending it to an audio unit.

Music Synthesis

An instrument unit (of type 'aumu'), in contrast to effect unit, renders audio in terms of notes. It acts as a virtual music synthesizer. An instrument unit employs a bank of sounds and responds to MIDI control data, typically initiated by a keyboard.

You subclass the AUMonotimbralInstrumentBase class for most instrument units. This class supports monophonic and polyphonic instrument units that can play one voice (also known as a patch or an instrument sound) at a time. For example, if a user chooses a piano voice, the instrument unit acts like a virtual piano, with every key pressed on a musical keyboard invoking a piano note.

The Core Audio SDK class hierarchy also provides the AUMultitimbralInstrumentBase class. This class supports monophonic and polyphonic instrument units that can play more than one voice at a time. For example, you could create a multimbral instrument unit that would let a user play a virtual bass guitar with their left hand while playing virtual trumpet with their right hand, using a single keyboard.

Music Effects

A music effect unit (of type 'aumf') provides DSP, like an effect unit, but also responds to MIDI data, like an instrument unit. You build a music effect unit by subclassing the AUMIDIEffectBase superclass from the SDK. For example, you would do this to create an audio unit that provides a filtering effect that is tuned according to the note pressed on a keyboard.

Data Format Conversion

Audio data transformations include such operations as sample rate conversion, sending a signal to multiple destinations, and altering time or pitch. To transform audio data in ways such as these, you build a format converter unit (of type 'aufc') as a subclass of the AUBase superclass in the SDK.

Audio units are not intended to work with variable bitrate (VBR) data, so audio units are not generally suited for converting to or from lossy compression formats such as MP3. For working with lossy compression formats, use Core Audio’s Audio Converter API, declared in the AudioConverter.h header file in the Audio Toolbox framework.

---