Offline Applications

In offline or file-based applications the compressed file size is the most important measure related to the bit rate or quality setting. Hence, for these applications the instantaneous bit rate fluctuation and varying packet sizes are less important. Ultimately, the goal here is to achieve the best audio quality for a given file size, or conversely, to get the smallest file size for a given quality.

The target file size can be controlled in ABR mode, with the bit rate approximation of file size divided by the audio duration. The target quality can be controlled using VBR mode. The VBR_constrained mode provides a compromise between ABR and VBR modes. The average bit rate in this mode will be at least as high as in ABR mode but it is permitted to increase if the quality would otherwise drop.

Real-time Applications

For real-time applications such as streaming or communications, the instantaneous packet sizes and temporally varying bit rates can be more important depending on the properties of the transmission channel. While the ABR mode is tailored to control file size in offline applications, the VBR and CBR modes are more appropriate for real time.

The encoder generates packets of varying size representing constant-size blocks of the input audio signal. The packet size can vary significantly depending on the audio content up to 6144 bits per channel. In CBR mode, the packet sizes are constrained so that a true constant bit rate can be achieved with a specified end-to-end delay. This mode conforms to the CBR mode as defined in the MPEG-4 standard. In this case the end-to-end delay is the codec delay plus the delay due to the bitstream buffering. The bitstream buffering is used to enable a constant bit rate transmission of variable packet sizes.

Fig. 2 illustrates the bitstream buffering at the encoder and decoder site to smooth out the variable packet sizes of the encoder. The transmission rate in this scenario is perfectly constant and the buffer fullness at the encoder and decoder always add up to 100%.

First appearing in iOS 6.0, applications that need to limit the maximum packet size can force the encoder to maintain that limit. This is only supported in VBR mode. The packet size limit will automatically result in the appropriate VBR quality setting that can maintain the packet size limit. The equivalent bit rate limit is usually about 20% below the virtual bit rate necessary to continuously transmit packets with the maximum size.

Fig. 3 shows the packet sizes of a castanet recording in VBR mode. The packet sizes vary by a large factor. Fig. 4 shows the resulting packet sizes with the same encoder settings but additional packet size limitation to 4000 bits.

Given a packet size limit, the average bit rate can be estimated by:

with

R: estimated average bit rate

D: margin (usually 0.2)

S: maximum packet size in bits

Fs: output audio sample rate

N: codec block size in samples per channel

Example, for the low-complexity AAC encoder at a sample rate of 44.1kHz with a packet size limit of 4000 bit/s the result is:

Command Line Settings

afconvert is a command line tool for audio file conversion including encoding and decoding. It supports all bit rate control options. Table 2 lists the relevant command line options for bit rate control of afconvert in the first column.

afconvert PCMin.wav AACout.caf –d aac –s 1 –b 128000

afconvert PCMin.wav AACout.caf –d aac –s 3 –u vbrq 64

afconvert PCMin.wav AACout.caf –d aac –s 2 –b 128000

afconvert -h

-u vbrq <n>

Bit rate control via API

When using the API, the properties available for bit rate control are listed in Table 3. The encoder will be configured according to these property values. The properties can be read as well. This way the user can find out for instance which VBR quality is used when a certain packet size limit is set.

Constant Bit Rate (CBR)

This mode achieves a constant target bit rate and is compliant with the CBR mode specified in the MPEG-4 standard. This mode is suitable for constant-bit-rate network transmission when decoding in real-time with a fixed end-to-end audio delay. However, due to the strict constant bit rate constraint, this mode results in lower audio quality and higher complexity than other encoding modes.

Average Bit Rate (ABR) - Default Mode

-- Recommended for controlling file size --

A target bit rate is achieved over a long-term average (typically after the first few seconds of encoding). Unlike CBR mode, this mode does not provide constant delay when using constant bit rate transmission, but this mode provides almost best global quality while still being able to strictly control the resulting file size.

Variable Bit Rate But Constrained (VBR_Constrained)

-- Recommended as a compromise between VBR and ABR --

This mode is similar to VBR but limits the average bit rate variation. The lower limit is the user-selected bit rate. Higher bit rate is adapted for difficult audio.

Variable Bit Rate (VBR)

-- Recommended for controlling the audio quality --

The audio signal is encoded with constant quality and virtually no bit rate constraints. This is the best mode to achieve consistent audio quality with the lowest overall bit rate.

A packet size limit can be applied in this mode if required for real-time transmission.