Skip to content

Asterisk 10 Codecs and Audio Formats

Overview

As a part of the project for Asterisk 10, changes have been made to Asterisk to increase the number of codecs it's capable of supporting, to handle codecs with custom formats, and to support audio sampling rates greater than 16kHz. This has resulted in several practical changes to Asterisk that will benefit its users.

SIP Only Note that the additional codecs discussed here are available for use in Asterisk's SIP channel driver, only. Asterisk 10 does not make them available for IAX2, MGCP, SSCP, H.323, UniSTIM, etc.

Expanded Signed Linear Support

Versions of Asterisk prior to 10 supported 16-bit Signed Linear sampled at 8kHz and at 16kHz (versions 1.6.0 - 1.8). New to Asterisk 10 is support for a much wider range of sampling rates. Asterisk can resample between any of these sampling rates and can read/write raw 16-bit signed linear audio files from/to disk. The complete list of supported sampling rates and file format extensions is:

Sampling Rate Asterisk File format
8kHz .sln
12kHz .sln12
16kHz .sln16
24kHz .sln24
32kHz .sln32
44.1kHz .sln44
48kHz .sln48
96kHz .sln96
192kHz .sln192

Asterisk 10 removes the format_sln16 file format in favor of expanded support in the main format_sln file format for all sampling rates. So, users who notice the absence of format_sln16 from their Asterisk 10 builds should not panic.

Users can create 16-bit Signed Linear files of varying sampling rates from WAV files using the sox command-line audio utility.

  • SOX example

sox input.wav -t raw -b 16 -r 32000 output.sln mv output.sln output.sln32

In this example, an input WAV file has been converted to Signed Linear at a depth of 16-bits and at a rate of 32kHz. The resulting output.sln file is then renamed output.sln32 so that it can be processed correctly by Asterisk.

32kHz Speex Support

Asterisk versions prior to 1.8 supported 8kHz Speex. Asterisk 1.8 supports 8 and 16kHz Speex. Asterisk 10 now supports 8, 16, and 32kHz Speex. Use of the 32kHz Speex mode is, like the other modes, controlled in the respective channel driver's configuration file, e.g. chan_sip's sip.conf.

  • Speex Example

sip.conf

[mypeer] type=peer secret=mysupersecret!!! host=dynamic context=fancycalls disallow=all allow=speex

[mypeer2] type=peer secret=myothers3cr3t host=dynamic context=fancycalls disallow=all allow=speex16

[mypeer3] type=peer secret=passwordisaterriblepassword host=dynamic context=fancycalls disallow=all allow=speex32

In this example, we have created three SIP peers for 3 different devices. The first, mypeer, supports only the 8kHz sampling of Speex; the second, mypeer2, supports only the 16kHz sampling of Speex; and the third, mypeer3, supports the new 32kHz sampling of Speex.

For comparison, here are some Speex samples, saved as WAV files in .mov containers, for ease-of-playback.

8kHz 20100 16kHz 20100 32kHz 20100

CELT Pass-through Support

Asterisk 10 adds pass-through support for the CELT codec. CELT provides low-delay transmission of high-quality audio. Unlike many other codecs that are focused on the transmission of human speech only, CELT is suitable for the transmission of both speech and audio, e.g. music.

Because the CELT codec is being folded, along with SILK, into a future codec called OPUS, and because the CELT bitstream isn't finalized, we have chosen not to add transcoding support for CELT as this time. CELT is configured in codecs.conf with the following parameters.

Option Values Description
type celt Sets the CELT codec as the type of codec being configured
samprate 32000, 441000, 48000 Defines the sampling rate in Hz to be used for the defined codec
framesize factors of 2 Represents the duration of each frame in samples. Defaults to 480 and should only be defined if a client does not use the default size. This option allows the codec to split 20ms frames into multiple frames in an anticipatory way. Thus, with 20ms frames at 48kHz are 960 samples, the packet is large. So setting framesize to 480, 20ms frames are transmitted in two 480 sample packets.
CELT codecs.conf example

[celt32] type=celt samprate=32000

[celt44] type=celt samprate=44100

[celt48] type=celt samprate=48000

In this example, three different CELT codecs are created: one for 32kHz mode, one for 44kHz mode, and another for 48kHz mode.

These codecs cannot be dynamically changed while Asterisk is running. In order to make changes, an Asterisk restart is required.

To make sure a codec or format is setup correctly, you can execute:

core show codecs

from the Asterisk CLI

CELT sip.conf example

Corresponding SIP peer entries to use the CELT codec would look like:

[myceltpeer1] type=peer secret=passwordisstillaterriblepassword host=dynamic context=fancycalls disallow=all allow=celt32

[myceltpeer2] type=peer secret=momnowaitdontmakemomyourpassword host=dynamic context=fancycalls disallow=all allow=celt44

[myceltpeer3] type=peer secret=daddontmakedadyourpasswordeither host=dynamic context=fancycalls disallow=all allow=celt48

In this case, we have defined 3 peers, each with a different CELT sampling rate. Thus, you'd probably want to set at least two of them to the same CELT rate, so they could call each other.

For CELT-calling, there are not a host of options on the client side. One could try Ekiga or SFLphone as softclients to make CELT calls.

SILK Support

Asterisk 10 provides full support for Skype's SILK codec. SILK is an extremely flexible codec for the transmission of speech. It operates in low bitrate narrow-band modes as well as higher (but still very low, otherwise) bitrate super wide-band modes. With respect to CPU complexity, its consumption is roughly three times that of G.729a at comparable bitrates.

SILK is configured in codecs.conf with the following parameters

Option Values Description
type silk Sets the SILK codec as the type of codec being configured
samprate 8000, 12000, 16000, 24000 Defines the sampling rate in Hz to be used for the defined codec
fec true, false Sets the use of Forward Error Correction by the codec. Off by default.
packetloss_percentage Integer as a percent Defines the estimated packetloss in the uplink direction. This parameter affects the amount of redundancy built into SILK when fec is enabled. The larger the amount, the higher the consumed bandwidth. Default is 0. 10 is recommended when fec is enabled
maxbitrate 8kHz: 5000-20000, 12kHz: 7000-25000, 16kHz: 8000-30000, 24kHz: 20000-40000 Defines, in bps and per the sampling rate being used, the maximum bitrate that will be consumed by the codec
dtx true, false Defines whether encoding is done in discontinuous transmission mode. If enabled, bandwidth will be reduced during periods of silence, but additional CPU complexity will be required. Off by default
SILK codecs.conf example

[silk8] type=silk samprate=8000 fec=true packetloss_percentage=10 maxbitrate=20000 dtx=false

[silk12] type=silk samprate=12000 fec=true packetloss_percentage=10 maxbitrate=25000 dtx=false

[silk16] type=silk samprate=16000 fec=true packetloss_percentage=10 maxbitrate=30000 dtx=false

[silk24] type=silk samprate=24000 fec=true packetloss_percentage=10 maxbitrate=40000 dtx=false

In this example, four different SILK codecs are created: one each for 8 (silk8), 12 (silk12), 16 (silk16), and 24kHz (silk24).

These codecs cannot be dynamically changed while Asterisk is running. In order to make changes, an Asterisk restart is required.

To make sure a codec or format is setup correctly, you can execute:

core show codecs

from the Asterisk CLI

SILK sip.conf example

Corresponding SIP peer entries to use the SILK codec would look like:

[mysilkpeer1] type=peer secret=thanksdigium host=dynamic context=fancycalls disallow=all allow=silk8

[mysilkpeer2] type=peer secret=forgivingme host=dynamic context=fancycalls disallow=all allow=silk12

[mysilkpeer3] type=peer secret=suchexcellentsoftware host=dynamic context=fancycalls disallow=all allow=silk16

[mysilkpeer4] type=peer secret=touse host=dynamic context=fancycalls disallow=all allow=silk24

In this case, we have defined 4 peers, each with a different SILK codec.

The generally available SIP softphones that support SILK are, to our knowledge, CSIPSimple and nightly builds of Jitsi beginning with build 3648 (so that, and anything newer than that).

The SILK licensing, like the licensing for Polycom's Siren 7 G.722.1 and Siren 14 G.722.1C codecs, requires that the distribution of binary codec modules that can be used by Asterisk. To download the SILK codec module for Asterisk, browse to http://downloads.digium.com/pub/telephony/codec_silk/ and drop the untar'd .so file into /usr/lib/asterisk/modules and issue an Asterisk restart, or simply load the codec module from the Asterisk CLI