RTP Buffering - Frame Based Buffering

Introduction

In Song module version 8 a new RTP buffering method called frame based buffering was introduced. The algorithm calculates the audio buffer level in milliseconds rather than in bytes.

Features

Frame based buffering allows:

• configurable decoding delay with one frame accuracy
• synchronisation of several decoders to the same stream (just by configuring them to the same initial delay)
• stable delay over long period of time
• automatic correction of clock difference between encoder and decoder

Applications

The following applications use frame based buffering:

Application Name	Version
Streaming Client	2.17
Annuncicom Full Duplex	0.21
RTP STL	2.01

Configuration

The only configuration parameter for the RTP decoder is the delay in milliseconds.

The delay parameter is the desired processing delay of the decoder (between the network input and the audio output). Please note that the end-to-end delay between the encoder and the decoder might be (significantly) different to the value configured.

In an ideal case the delay parameter would be 0 ms, however due to device's internal buffers a small delay (depending on the hardware) is inevitable. The delay value should also cover possible temporary network hick-ups (jitter). E.g. if the network might sometimes delay the packet delivery by 20ms due to a temporary load, the configured parameter should not be less than 20ms.

The maximum configurable delay is limited by the device's internal buffer (64, 32 or 16kB).

The following sections show recommended delay values for various audio formats.

Recommended Settings

MP3 CBR

The following table shows the minimum and the maximum possible delay with MP3 constant bitrate. The maximum delay differs between the Streaming Client, which has 64kB audio buffer available, and ABCL (Annuncicom FDX, STL), which features only 32kB buffer. The minimum delay includes 100ms network jitter.

MP3 CBR bitrate	Min delay	Max delay (SC)	Max delay (ABCL)
320kbps	150ms	1,588ms	769ms
256kbps	163ms	2,011ms	987ms
192kbps	183ms	2,741ms	1,349ms
160kbps	200ms	3,277ms	1,638ms
128kbps	225ms	4,121ms	2,073ms
64kbps	350ms	8,342ms	4,246ms
32kbps	600ms	16,784ms	8,592ms

MP3 VBR and ABR

Variable or average bitrate the minimum and delay depends on the bitrate variation interval. The minimum delay is taken from the CBR table for the low end of the interval, whereas the maximum delay is the CBR value for the high end of the interval.

Please note that most MP3 encoders use the whole bitrate range starting from the lowest bitrate 32kbps. E.g. VBR 128kbps varies from 32 to 128kbps

MP3 Format	Min delay	Max delay (SC)	Max delay (ABCL)
32-320kbps	600ms	1,588ms	769ms
32-256kbps	600ms	2,011ms	987ms
32-192kbps	600ms	2,741ms	1,349ms
32-160kbps	600ms	3,277ms	1,638ms
32-128kbps	600ms	4,121ms	2,073ms
32-64kbps	600ms	8,342ms	4,246ms

PCM

In uncompressed audio (PCM, uLaw or ALaw) the minimum and maximum delay depend on the bit rate and on the hardware.

The following table lists minimum and maximum settings for all standard RTP audio formats:

Format	Min delay VLSI	Min delay MAS	Max delay (SC)	Max delay (ABCL)	Max delay (ABCL full duplex)
uLaw 8kHz mono ALaw 8kHz mono	349ms	80ms	8170ms	4057ms	2027ms
PCM 8kHz mono	323ms	60ms	4075ms	2027ms	1003ms
uLaw 12kHz mono ALaw 12kHz mono	246ms	67ms	5441ms	2710ms	1345ms
PCM 12kHz mono	229ms	54ms	2710ms	1345ms	662ms
uLaw 24kHz mono ALaw 24kHz mono
PCM 24kHz mono
uLaw 32kHz mono ALaw 32kHz mono
PCM 32kHz mono
PCM 44.1kHz stereo
PCM 44.1kHz mono
PCM 48kHz stereo

Multiple Device Synchronisation

Multiple devices receiving the same RTP stream can be configured to play in sync by entering the same delay parameter.

Barix recommends to use broadcast or multicast together with synchronisation, otherwise a small inaccuracy (few milliseconds) might be caused by the network delivery to different locations.

Deliberate Delays

In some applications it is desired to artificially delay the audio. E.g. in a tunnel to eliminate the delay caused by the distance between the devices.

An artificial delay can be introduced by configuring the devices to different delay values. E.g. 100ms, 120ms, 140ms, 160ms, etc.