Audio/Video Transport WG                                  Ari Lakaniemi
Internet Draft                                              Ye-Kui Wang                                                    Nokia
Intended status: Standards track                                  Nokia                            Ye-Kui Wang
Expires: October 2009                               Huawei Technologies
                                                          April 28, 2009                                    October 23, 2008

                 RTP payload format for G.718 speech/audio

Status of this Memo

   By submitting this Internet-Draft, each author represents that any
   applicable patent or other IPR claims

   This Internet-Draft is submitted to IETF in full conformance with the
   provisions of which he BCP 78 and BCP 79.  This document may contain material
   from IETF Documents or she is aware
   have been IETF Contributions published or will made publicly
   available before November 10, 2008.  The person(s) controlling the
   copyright in some of this material may not have granted the IETF
   Trust the right to allow modifications of such material outside the
   IETF Standards Process.  Without obtaining an adequate license from
   the person(s) controlling the copyright in such materials, this
   document may not be disclosed, modified outside the IETF Standards Process, and any
   derivative works of which he or she becomes
   aware will it may not be disclosed, in accordance with Section 6 of BCP 79. created outside the IETF Standards
   Process, except to format it for publication as an RFC or to
   translate it into languages other than English.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups.  Note that
   other groups may also distribute working documents as Internet-
   Drafts. Internet-Drafts.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   The list of current Internet-Drafts can be accessed at

   The list of Internet-Draft Shadow Directories can be accessed at

   This Internet-Draft will expire on April 23, October 28, 2009.

Copyright Notice

   Copyright (C) The (c) 2009 IETF Trust (2008). and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents in effect on the date of
   publication of this document (
   Please review these documents carefully, as they describe your rights
   and restrictions with respect to this document.


   This document specifies the Real-Time Transport Protocol (RTP)
   payload format for the Embedded Variable Bit-Rate (EV-VBR)
   speech/audio codec, specified in ITU-T G.718. A media type
   registration for this RTP payload format is also included.

Conventions used in this document

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in RFC 2119 [RFC2119].

Table of Contents

   1. Introduction...................................................3
   2. Background.....................................................3
      2.1. The G.718 codec...........................................3
      2.2. Benefits of layered design................................5
      2.3. Transmitting layered data.................................5
      2.4. Scaling scenarios & rate control..........................6
   3. G.718 RTP payload format.......................................7
      3.1. Payload Structure.........................................7
         3.1.1. Payload Header.......................................7
         3.1.2. G.718 transport blocks...............................8
      3.2. Handling the Encoded data................................10 data................................11
      3.3. G.718 scaling............................................13
      3.4. CRC verification.........................................13 verification.........................................14
      3.5. G.718 session............................................14
      3.6. Cross-stream/cross-layer timing synchronization..........14
      3.7. RTP Header usage.........................................14 usage.........................................15
   4. Payload Format Parameters.....................................15
      4.1. Media Type Registration..................................15
      4.2. Mapping to SDP Parameters................................17
      4.3. Offer/answer considerations..............................17 considerations..............................18
      4.4. Declarative usage of SDP.................................18
      4.5. SDP examples.............................................18
   5. Security Considerations.......................................20
   6. Congestion control............................................21
   7. IANA Considerations...........................................21 Considerations...........................................22
   APPENDIX A: Payload examples.....................................22 examples.....................................23
      A.1. Simple payload examples..................................22 examples..................................23
         A.1.1. All the layers in the same payload..................22 payload..................23
         A.1.2. Layers in separate RTP streams......................23 streams......................24
      A.2. Advanced examples........................................24 examples........................................25
         A.2.1. Different update rate for subset of layers..........24 layers..........25
         A.2.2. Redundant frames with limited set of layers.........25 layers.........26
   8. References....................................................27 References....................................................28
      8.1. Normative References.....................................27 References.....................................28
      8.2. Informative References...................................28 References...................................29
   Author's Addresses...............................................29
   Intellectual Property Statement..................................29
   Disclaimer of Validity...........................................30
   Copyright Statement..............................................30 Addresses...............................................30
   9. Open Issues...................................................30
   10. Changes Log..................................................31

1. Introduction

   The International Telecommunication Union (ITU-T) Recommendation
   G.718 [G.718] specifies the Embedded Variable Bit Rate (EV-VBR)
   speech/audio codec. This document specifies the Real-time Transport
   Protocol (RTP) [RFC3550] payload format for this codec.

2. Background

2.1. The G.718 codec

   G.718 is an embedded variable rate speech codec having a layered
   design. The bitstream of the G.718 core codec consists of a core
   layer, denoted as L1, and four enhancement layers, denoted as L2-L5.
   The bit-rates of the G.718 core codec range from 8 kbit/s (core layer
   only) to 32 kbit/s (with all layers up to L5). Furthermore, the G.718
   codec supports also discontinuous transmission (DTX) and comfort
   noise generation (CNG) by sending Silence Descriptor (SID) frames
   during periods of non-active input signal, resulting in a reduced
   bit-rate. The sampling frequency of the core codec is 16 kHz and the
   codec operates on 20 ms frames. The G.718 codec is also capable of
   narrowband operation with audio input and/or output at 8 kHz sampling

   While transmitting/receiving the core layer L1 is enough for
   successful decoding of the audio content, each of the enhancement
   layers Ln (n being 2 to 5, inclusive) provides an improvement to
   reconstructed audio quality. Thus, the core layer ensures the basic
   communication while the enhancement layers can be used to improve the
   perceptual quality. Furthermore, enhancement layers are dependent on
   all the lower layers in a sense that successful decoding of layer Ln
   requires also all the layers Lm with m<n to be available.

   The sizes, sampling rates and possible outputs of the G.718 core
   codec layers L1-L5 are summarized in Table 1 below, where "Bytes"
   column indicates the number of bytes per encoded data unit for a
   layer, NB and WB denotes narrowband and wideband, respectively. The
   "Bytes" column in other tables has the same meaning. Note that for
   layers L1 and L2, the corresponding output may either be NB or WB,
   depending on the rendering device and the application requirement,
   regardless of the sampling rate of the encoded data.

                           Table 1: G.718 layers

        Layer   Bytes   Cumulative bit-rate   Sampling rate   Output
         L1       20        8 kbit/s           8 or 16 kHz    NB or WB
         L2       10       12 kbit/s           8 or 16 kHz    NB or WB
         L3       10       16 kbit/s           16 kHz         WB
         L4       20       24 kbit/s           16 kHz         WB
         L5       20       32 kbit/s           16 kHz         WB

   The G.718 codec includes also an operating mode that is compatible
   with the Adaptive Multi-Rate Wideband (AMR-WB) codec [AMR-WB], for
   which the RTP payload format is specified in [RFC4867]. In this AMR-
   WB interoperable mode, layers L1, L2 are replaced by L1' consisting
   of AMR-WB encoded data. Furthermore, together with L1' modified L3'
   is used instead of L3. The usage of layers L4 and L5 is not affected
   by transmitting AMR-WB data in the lower layers. If layer L3' is
   present in the encoded bit-stream, the base layer L1' must use the
   AMR-WB mode 2 with the bit-rate of 12.65 kbits/s. Otherwise (the
   encoded bit-stream contains only the L1' layer), any of the 9 AMR-WB
   coding modes 0, 1, 2, 3, 4, 5, 6, 7, and 8 correspond to the bit-
   rates of 6.60, 8.85, 12.65, 14.25, 15.85, 18.25, 19.85, 23.05, and
   23.85 kbit/s, respectively, may be in use. Table 2 summarizes the
   AMR-WB interoperable mode when more than one layer may be present.

          Table 2: G.718 layers in the AMR-WB interoperable mode

        Layer   Bytes   Cumulative bit-rate   Sampling rate   Output
         L1'      32       12.8 kbit/s           16 kHz          WB
         L3'       9       16.4 kbit/s           16 kHz          WB
         L4       20       24.4 kbit/s           16 kHz          WB
         L5       20       32.4 kbit/s           16 kHz          WB

   Note that the bit-rate for the raw bit-stream of AMR-WB mode 2 is
   12.65 kbits/s. However, after counting the padding bits to make each
   encoded data unit byte-aligned, as in the octet-aligned mode
   specified in [RFC4867], the resulting bit-rate is then 12.8 kbits/s.

   In the AMR-WB interoperable mode, when the base layer L1' is
   transported in its own RTP packet stream, the packetisation specified
   in [RFC4867] MUST be used, to enable legacy RFC4867 receivers to
   receive the base layer L1'.

   ITU-T SG16 is currently working on a set of extension layers in order
   to provide a so-called super-wideband (SWB) audio and stereophonic
   encoding extensions on top of the G.718 core codec. Further details
   and the usage of these layers are TBD.

   The main application of the G.718 codec is telephony. Other expected
   applications include audio/video conferencing and streaming.

2.2. Benefits of layered design

   The layered design enables simple scalability of the transmitted
   stream simply by conveying a suitable number of layers. The number of
   layers used in a session may be selected for example based on the
   capacity of the transmission channel, current transmission conditions,
   characteristics of the source signal or available processing capacity.

   Another obvious benefit of the layered codec design is the
   possibility to exploit the scalability to support congestion control
   by transmitting/dropping some of the (higher) enhancement layers in
   order to alleviate congestion in the network. See more detailed
   discussion on the congestion control in section 6.

   Furthermore, the layered design also implicitly provides possibility
   for unequal error detection/protection by employing different levels
   of protection on core layer and enhancement layers.

2.3. Transmitting layered data

   In principle there are two basic approaches to carry the data from a
   layered encoder:

   1. All the layers are carried within a single RTP session.

   2. The encoded data is divided over multiple RTP sessions, each
      session carrying a subset of layers. This is also referred to as
      Multi-Session Transmission (MST).

   The first choice is the most efficient in terms of exploitation of
   transmission bandwidth. Furthermore, using only one packet to carry
   all encoded data layers of a frame requires less resources also from
   the end-systems (and intermediate systems) since the number of
   packets is kept at minimum and only single RTP packet stream needs to
   be handled. However, this option requires any intermediate network
   element performing the scaling operation to be fully media-aware
   since removing encoded layers requires modification of the payload.
   Furthermore, the intermediate network element needs to be within the
   security context to enable the meaningful manipulation of the payload,
   in case secure transport is employed. This might not be feasible in
   all systems/scenarios, but some special-purpose devices such as e.g.
   media gateways in cellular telephone systems may be able to implement
   this kind of media-aware functionality.

   The second alternative transmitting selected subsets of layers in
   separate RTP sessions facilitates simple scalability in intermediate
   network elements without the requirement of being fully media-aware.
   One use case of this alternative is layered multicast [McCanne]. On
   the other hand, this approach introduces separate packet header
   overhead for each subset of layers for those low-delay application
   scenarios wherein aggregation of data from multiple frames is not
   ideal. In this case, when the size of the encoded data block per
   single layer is in the range of 10 to 20 bytes, the packetisation may
   result in relatively high amount of protocol overhead, which might be
   an expensive solution on bandwidth-limited links. Another drawback of
   this approach is somewhat more complex session setup and the
   additional complexity associated with handling of several concurrent
   RTP sessions. However, this is a trade-off that enables simple
   scalability also by intermediate network elements that are not aware
   of the details of the transmitted media.

2.4. Scaling scenarios & rate control

   In principle there are three different ways to make use of the
   layered design to control the bandwidth usage:

   1. A sender decides to change the number of layers it is transmitting
   (for example due to congestion control constrains)

   2. A receiver or an intermediate network element instructs a sender
      to change the number of layers it is transmitting

   3. An intermediate network element passes forward only a subset of
      layers it receives

   The most appropriate mechanism depends on the application and the
   employed network topology. For example point-to-point conversational
   audio connection can easily introduce rate control by changing the
   number of transmitted layers, while in centralized audio/video
   conferencing scenario the conference server is a more appropriate
   point to implement the rate control instead of transmitting end-
   point. end-point.
   Please refer to [RFC5117] for extensive discussion on the different
   topologies and their implications to the transmission.

   However, the fundamental difference between these choices is that
   method 1 does not necessarily need any feedback from the receiver(s),
   while methods 2 and 3 require a signaling mechanism to support rate

3. G.718 RTP payload format

   The basic G.718 source data unit is one layer of an encoded frame.
   Since generally the term layer refers to time series of data
   representing certain encoding layer, in this specification we use the
   term Encoded Data Unit (EDU) to refer to a single layer of data from
   single encoded frame. Thus, each EDU has a (conceptual) frame number
   indicating its location in encoding/decoding order and a layer number
   indicating the encoding layer the EDU represents.

3.1. Payload Structure

   The G.718 payload format consists of a payload header, followed by
   one or more transport blocks (TB) forming the actual payload data.

    +-----------------+----------+----------+- /// -+----------+
    | Payload header  |  TB(1)   |  TB(2)   |          TB(n)   |
    +-----------------+----------+----------+- /// -+----------+

3.1.1. Payload Header

   The payload header consists of an 8-bit payload CRC checksum:

    |     CRC       |

   In the transmitting end the payload checksum is computed over the
   primary transport block (specified in section 3.1.2. ) of the payload
   using the generator polynomial

      C(z) = z^8 + z^4 + z^3 + z^2 + 1.

   Subsequent transport blocks are prepared in such a way that the
   payload checksum is valid for any integer number of contiguous
   transport blocks within one RTP packet starting from the beginning of
   the primary transport block.

   In the receiving end the payload CRC checksum can be used to verify
   the correct reception of any contiguous subset of transport blocks
   within one RTP packet starting from the beginning of the primary
   transport block (see section 3.4.  for detailed description).

3.1.2. G.718 transport blocks

   The basic building block of the G.718 RTP payload data is an G.718
   transport block (TB). There are two types of transport blocks:
   primary transport block and secondary transport block.

   The structure of the primary transport block is depicted below.

     0 1 2 3 4 5 6 7
    |   L-ID    |NF | Encoded data               |

   The structure of the secondary transport block is depicted below.

     0 1 2 3 4 5 6 7                              0 1 2 3 4 5 6 7
    |   L-ID    |NF | Encoded data               |     Tail      |

   The layer ID (L-ID) and the NF fields form the transport block header.
   The L-ID field is used to identify the layer structure of the encoded
   data carried in this G.718 transport block, and the NF field
   indicates the number of encoded frames with this layer structure
   carried in the Encoded data part following the transport block header.
   The Tail field of the secondary transport block carries a modified 8-bit 8-
   bit CRC checksum computed over the transport block, as specified

             Author's note: For streaming or other applications that
             allow for relatively long end-to-end delay, sometimes it
             would be beneficial to aggregate more than 4 frames in one
             TB. Should the length of NF be larger?
   An G.718 RTP packet payload SHALL include exactly one primary
   transport block, which MAY be followed by one or more secondary
   transport blocks. The data fields of both transport block types are
   described below.

   L-ID Identification (6 bits) of the encoded data carried in this
        transport block. Table 3 below specifies the mapping between L-
        ID and the encoded data.

                Table 3: Layer identification (L-ID) values

          L-ID    Encoded data
            0     Empty frame
            1     L1
            2     L1-L2
            3     L1-L3
            4     L1-L4
            5     L1-L5
            6     L2
            7     L2-L3
            8     L2-L4
            9     L2-L5
           10     L3
           11     L3-L4
           12     L3-L5
           13     L4
           14     L4-L5
           15     L5
           16     L1'
           17     L1', L3'
           18     L1', L3', L4
           19     L1', L3', L4-L5
           20     G.718 SID
           21     AMR-WB SID
           22-63  Reserved

             Author's note: The current approach provides maximum
             flexibility in terms of layer configuration. However,
             limiting choices would be one way to leave more bits for
             stereo & SWB layer configurations.

             Author's note: One suggested way to make sure we do not
             run out of L-ID values with the extension modes has been
             to make the mapping between L-ID and layer configuration
             dynamic (to be specified using SDP in session set-up).
             While this would provide effective usage of L-ID bits, it
             would require all elements processing the payload to be
             signaling-aware. A compromise solution would be to provide
             static mapping for selected layer configurations and leave
             'more exotic' cases to be dynamically mapped on session
             basis. The usage of this type of approach is FFS.

             Author's note: Yet another possible way is to do similar
             as in the SVC RTP payload format draft, i.e. to signal the
             bitrate etc. parameters for an operation point, and signal
             dependency between sessions using the MMUSIC decoding
             dependency draft. This way should be generic enough and
             applicable to future versions of scalable codecs. However,
             the above methods (using detailed layer configuration may
             provide more useful information as the bitrate etc. of
             each layer is fixed, not as flexible as in SVC.)

             Author's note: Yet another approach is to allocate L-ID
             according to different mode. For example, the mode with L1
             being present and the AMR-WB compatible mode (with L1'
             being present) use different value spaces of L-ID.

   NF   Number of frames in this transport block (2 bits) decreased by
        one. The number of frames is equal to the value of NF
        incremented by one. For example, value NF=0 indicates that the
        transport block carries one frame, and value NF=3 indicate that
        the transport block carries four frames. If the sender wants to
        encapsulate more than four frames per payload, several
        transport blocks need to be used.

   Encoded data

        Encoded data consists of EDUs as specified by the values L-ID
        and NF fields, arranged according to rules given in section 3.2.
        When L-ID is equal to 0 (empty frame), the encoded data field
        is not present (i.e. it consists of zero octet).

   Tail The 8-bit tail field of the secondary transport block carries a
        bit field that is needed to modify the partial CRC checksum
        over the payload data up to the end of this TB to match the
        payload CRC field value carried in the payload header.

        In the transmitter the Tail bits for a secondary TB(n) are
        computed by first computing the CRC checksum CRC(n) over the
        payload data from the beginning of the primary TB up to the end
        of TB(n) using the generator polynomial C(z) given above. The
        bits of the Tail field of TB(n) are set to zero value for the
        CRC computation. The transmitted value of the Tail field in
        TB(n) is obtained by bitwise XOR operation between the payload
        CRC field value carried in the payload header and the CRC(n)
        computed for TB(n).

3.2. Handling the Encoded data

   In order to provide unique mapping of EDUs to encoded frames, the
   following rules on sequence of frames and sequence of layers need to
   be followed when creating a payload:

   o  The frames within a payload MUST form a set of contiguous frames
      in decoding order, i.e. if a payload carries frames n and n+N, all
      frames between n and n+N in decoding order MUST also be present in
      the payload.

   o  The layers within a frame MUST form a contiguous set of layers,
      i.e. if layers Lx and Ly of a frame are included in the payload,
      all layers between Lx and Ly layers MUST also be present.

   The EDUs within a transport block are arranged according to the
   following rules:

   o  The EDUs within a transport block MUST be arranged in increasing
      order of layer number

   o  The EDUs with the same layer number within a transport block MUST
      be arranged in decoding order

   Explicit timing information for the transport blocks is not needed,
   since the ordering of EDUs in the payload and their mapping to
   transport blocks can be used to implicitly carry this information.
   The following rules apply:

   o  If the highest layer carried in transport block k is n, and the
      lowest layer carried by transport block k+1 is n+1, then the EDUs
      of transport block k and k+1 belong to the same encoded frame.
      Furthermore, if transport blocks k and k+1 carry EDUs belonging to
      the same encoded frame(s), these transport blocks MUST include the
      same number of EDUs.

   o  If the highest layer carried in transport block k is n, and the
      lowest layer carried by transport block k+1 is smaller than or
      equal to n, the EDUs of transport block k and k+1 belong to the
      two separate encoded frames, which are contiguous in decoding

   o  Multiple copies of an EDU MUST NOT be included in the payload.

   A set of EDUs can be allocated to transport blocks in several ways.
   For example each EDU can be encapsulated in its own transport block,
   all EDUs can be carried in single transport block, EDUs belonging to
   the same encoded frame can be encapsulated in dedicated transport
   block, or EDUs representing the same layer can be carried in their
   own transport blocks. Three examples on this with two frames with
   layers L1-L3 are given below. The first example illustrates the case
   using a single transport block for the whole payload, while the
   second payload example introduces separate transport blocks for each
   of the EDUs. The third example shows an approach where all layers are
   carried in dedicated transport blocks. The notation Fx-Ly is used to
   denote layer y of frame x.

   Example 1: All EDUs in a single transport block

     | L-ID=3  |NF=1 | F1-L1 | F2-L1 | F1-L2 | F2-L2 | F1-L3 | F2-L3  |

             Author's note: Currently, it is mandated that lower layer
             EDUs of later frames go before higher layer EDUs of
             earlier frames. This way is friendlier to adaptation
             (dropping of higher layers). However, if all layers are
             received, then the depacketizer needs to reorder the EDUs
             to their decoding order before feeding them to the decoder.
             Therefore, the other way around (i.e. lower layer EDUs of
             later frames go after higher layer EDUs of earlier frames,
             or EDUs in transport blocks are placed in decoding order)
             is more friendly to the depacketizer. Another benefit of
             the latter is that it does not introduce any end-to-end
             delay. Which way to be specified (or both allowed if
             needed) is FFS.

   Example 2: All EDUs in separate transport blocks
     | L-ID=1  |NF=0 | F1-L1 | L-ID=1  |NF=0 | F2-L1 |
     | L-ID=8  |NF=0 | F1-L2 | L-ID=8  |NF=0 | F2-L2 |
     | L-ID=14 |NF=0 | F1-L3 | L-ID=14 |NF=0 | F2-L3 |

   Example 3: Dedicated transport for EDUs of each layer

     | L-ID=1  |NF=1 | F1-L1 | F2-L1 | L-ID=6  |NF=1 | F1-L2 | F2-L2 |
     | L-ID=10 |NF=1 | F1-L3 | F2-L3 |

   While the first example carrying data from all layers in the same
   transport block obviously consumes less bandwidth, the second example
   using separate transport block for each EDU, and the third example
   using dedicated transport blocks for each layer provide simple
   scaling possibility: while in the first case the removal of e.g.
   layer L3 (from each frame in the payload) would require changing the
   value of the L-ID in addition to removing the corresponding EDU(s),
   in the second and third options it is enough to just remove all
   transport blocks carrying L3 data and the remaining part of the
   payload can be left untouched (however the packet size information in
   high-layer protocol headers needs change).

3.3. G.718 scaling

   Some media-aware network elements (MANEs) MAY modify the G.718
   bitstream by dropping some of the layers in case congestion control
   or e.g. access link bandwidth requires such scaling to take place.
   Such MANEs are RTP translators (with the topology Topo-Translator as
   described in [RFC5117]), for which the rules for RTP translators
   specified in [RFC3550] apply.

   A payload can be either completely dropped or some of the transport
   blocks it carries can be discarded. In case full payloads are dropped
   to implement scaling, a packet containing the core layer L1 SHOULD
   NOT be discarded, since the decoding of higher layers of the same
   encoded frame is not possible without the core layer data being
   available. This means that payloads with L-ID values equal to 1 to 5,
   inclusive and 16 to 19, inclusive, SHOULD NOT be completely discarded.

             Author's note: To be checked whether the case of dropping
             a subset of the transport blocks in one packet also
             strictly follows the topology Topo-Translator.

   In case the payload is forwarded with modified content, at least the
   primary transport block MUST be preserved in the payload, while some
   of the secondary transport blocks at the end of the payload MAY be

3.4. CRC verification

   Both UDP-Lite [RFC3828] and DCCP [4340] provide partial checksum
   options, in which partially damaged payloads can be delivered to the
   application layer.  In cases wherein such a transport layer operation
   is in use, and the partial checksum service by the transport layer
   protects up to the RTP header and the payload header, the CRC
   checksum provided in the payload header can be used to verify whether
   an RTP packet payload contains corrupt transport blocks.

   In the receiving end the CRC verification is made in such a way that
   the CRC computation is started from the beginning of the primary TB,
   i.e. from the MSB of the first octet of the TB(1), and the
   computation is continued until the end of the payload data or until
   an erroneous TB is encountered. At the end of each TB a check MAY be
   performed: if the CRC value at the end of TB(n) matches the payload
   CRC value received in the payload header, the verification is
   successful and the data up TB(n) is valid. If the CRC value at the
   end of TB(n) does not match the payload CRC value received in the
   payload header, there is an error in the TB(n) and it MUST be
   discarded as corrupted. Furthermore, if the verification indicates
   corrupted TB(n), all subsequent transport blocks TB(m) with m>n MUST
   also be discarded.

3.5. G.718 session

   An G.718 session consists of one or several RTP sessions carrying
   encoded G.718 data according the payload format specified in section

3.6. Cross-stream/cross-layer timing synchronization

   In case an G.718 session consists of multiple RTP sessions, the RTP
   packets transmitted on separate RTP sessions need to be synchronized
   in order to enable reconstruction of the frames in the receiving end.
   Since each of the RTP sessions uses its own random initial value for
   the RTP timestamp, there is also a random offset between the RTP
   timestamps values carrying the EDUs belonging to the same encoded
   frame in different RTP sessions.

   The receiver MUST use the traditional RTCP based mechanism to
   synchronize streams by using the RTP and NTP timestamps of the RTCP
   Sender Reports (SR) it receives.

3.7. RTP Header usage

   This section specifies the usage of some fields of the RTP header
   (specified in section 5 of [RFC3550]) with the G.718 RTP payload
   format.  Setting of other RTP header fields is as specified in

   The RTP timestamp corresponds to the sampling instant of the first
   encoded sample of the earliest frame in the payload. The timestamp
   clock frequency is 32 kHz.

   The marker bit (M) of each of the RTP streams of the session SHALL be
   set to value 1 if the payload carries an EDU belonging to the first
   frame after an inactive period, i.e. an EDU from the first frame of a
   talkspurt. For all other packets the marker bit is set to value 0.

4. Payload Format Parameters

   This section defines the parameters that may be used to configure
   optional features in the G.718 RTP transmission.

   The parameters are defined here as part of the media subtype
   registration for the G.718 codec.  Mapping of the parameters into the
   Session Description Protocol (SDP) [RFC4566] is also provided for
   those applications that use SDP.  In control protocols that do not
   use MIME or SDP, the media type parameters must be mapped to the
   appropriate format used with that control protocol.

4.1. Media Type Registration

   This registration is done using the template defined in RFC 4288
   [RFC4288] and following RFC 4855 [RFC4855].

   Type name:  audio

   Subtype name:  G718

   Required parameters:  none

   Optional parameters:

      mode:      This parameter MAY be used to indicate whether the
                 mode with layer L1 being present or the AMR-WB
                 compatible mode (with layer L1' being present) is in
                 use. If this parameter is not present or the value of
                 this parameter is equal to 0, the mode with layer L1
                 being present is in use. Otherwise, the AMR-WB
                 compatible mode is in use. When this parameter is
                 present, the value MUST be either 0 or 1.

      Author's note: When the upcoming stereo and SWB options are
      present, the semantics of this parameter may change.

      layers:    The numbers of the layers (in range from 1 to 5,
                 denoting layers from L1 to L5, respectively)
                 transmitted in this session, expressed as comma-
                 separated list of layer numbers. If the parameter is
                 present, at least layer L1 or L1' MUST be included in
                 the list of layers in one of the RTP sessions included
                 in the G.718 session. If the parameter is not present,
                 all layers up to layer L5 MAY be used in the session.

      Author's note: Why not use semantics similarly as L-ID?

      ptime:     The recommended length of time (in milliseconds)
                 represented by the media in a packet.  See Section 6
                 of [RFC4566].

      maxptime:  The maximum length of time (in milliseconds) that can
                 be encapsulated in a packet.  See Section 6 of

      Author's note: Some further study is needed to see if separate
      parameters for sending and receiving capabilities/preferences are
      needed -- especially for upcoming stereo and SWB options.

      Author's note: The support for upcoming SWB and stereo options
      needs to be taken into account. Basically we can either 1) extend
      the parameter "layers" to cover also this aspect, or 2) define
      separate parameter(s) for these new options when more details on
      the stereo/SWB support are available.

   Encoding considerations:

     This media type is framed and contains binary data; see Section 4.8
     of [RFC4288].

   Security considerations:  See Section 6 of RFC xxxx
   Interoperability considerations:  none

   Published specification:  RFC xxxx

   Applications which use this media type:

     For example Voice over IP, audio and video conferencing, audio
     streaming and voice messaging.

   Additional information:  none

   Person & email address to contact for further information:

     Ari Lakaniemi,

   Intended usage:  COMMON

   Restrictions on usage:

     This media type depends on RTP framing, and hence is only defined
     for transfer via RTP [RFC3550]


     Ari Lakaniemi,

   Change controller:

     IETF Audio/Video Transport working group delegated from the IESG

4.2. Mapping to SDP Parameters

   The information carried in the media type specification has a
   specific mapping to fields of the SDP [RFC4566], which is commonly
   used to describe RTP sessions.  When SDP is used to specify sessions
   employing the G.718 codec, the mapping is as follows:

   o  The media type ("audio") goes in SDP "m=" as the media name.

   o  The media subtype ("G718") goes in SDP "a=rtpmap" as the encoding
      name.  The RTP clock rate in "a=rtpmap" MUST be 32000 for G.718.

      Author's note: The current choice for the RTP clock rate is a
      'placeholder'. The clock rate needs to be set according to SWB
      sampling rate, which is still T.B.D. Since the core codec employs
      16000 Hz sampling rate, an integer multiple of 16000 Hz seems to
      be a preferable choice.

   o  The parameters "ptime" and "maxptime" go in the SDP "a=ptime" and
      "a=maxptime" attributes, respectively.

   o  Any remaining parameters go in the SDP "a=fmtp" attribute by
      copying them directly from the media type string as a semicolon
      separated list of parameter=value pairs.

4.3. Offer/answer considerations

   The following considerations apply when using the SDP offer/answer
   [RFC3264] mechanism to negotiate the G.718 transport. The parameter
   "layers" MAY be used to indicate the layer configuration for the each
   RTP session belonging to current G.718 session an end-point making
   the offer is ready to transmit and wishes to receive.

   o  In case the G.718 session consists of a single RTP session, it is
      RECOMMENDED not to impose any layer restrictions for the session
      but to use the rate control functionality to set possible
      restrictions on usage of the higher or highest layers. If the
      offer includes a layer configuration parameter, the answer MAY use
      different configuration, but the highest layer in the answer MUST
      NOT be higher than the highest layer of the offered configuration.

      Author's note: Support for answer modifying the layer
      configuration is FFS.

   In case the G.718 session consists of multiple RTP sessions, the
   answer MUST use the layer configurations provided in the offer for
   the sessions it accepts.

4.4. Declarative usage of SDP

   In declarative usage, such as SDP in RTSP [RFC2326] or SAP [RFC2974],
   the parameter "layers" SHALL be interpreted to provide a set of
   layers that the sender may use in the session.

4.5. SDP examples

   Some example SDP session descriptions utilizing G.718 encodings are
   provided below.

   The first example illustrates the simple case where the G.718 session
   employing a single RTP session and the AVPF profile is offered, and
   the answer accepts the offer without any changes.


     m=audio 49120 RTP/AVPF 97
     a=rtpmap:97 G718/32000/1


     m=audio 49120 RTP/AVPF 97
     a=rtpmap:97 G718/32000/1

   The second example shows a bit more complex case where the G.718
   session using a single RTP session and the AVPF profile is offered
   with restriction to send/receive only with layers L1 and L2. The
   answer indicates that the other end-point is happy to receive (and
   send) layers up to L5.


     m=audio 49120 RTP/AVPF 97
     a=rtpmap:97 G718/32000/1
     a=fmtp:97 layers=1,2


     m=audio 49120 RTP/AVPF 97
     a=rtpmap:97 G718/32000/1
     a=fmtp:97 layers=1,2,3,4,5

   The third example shows an G.718 session using multiple RTP sessions
   with the AVPF profile. The answerer wishes to use only layers up to


     m=audio 49120 RTP/AVPF 97
     a=rtpmap:97 G718/32000/1
     a=fmtp:97 layers=1,2

     m=audio 49122 RTP/AVPF 98
     a=rtpmap:98 G718/32000/1
     a=fmtp:98 layers=3
     a=depend:lay 1

     m=audio 49124 RTP/AVPF 99
     a=rtpmap:99 G718/32000/1
     a=fmtp:99 layers=4,5
     a=depend:lay 1 2


     m=audio 49120 RTP/AVPF 97
     a=rtpmap:97 G718/32000/1
     a=fmtp:97 layers=1,2

     m=audio 49120 RTP/AVPF 98
     a=rtpmap:98 G718/32000/1
     a=fmtp:98 layers=3
     a=depend:lay 1

   Note that the dependency signaling according to [smd-sdp] is used in
   the third example above to indicate the relationship between the
   layers distributed into separate RTP sessions.

5. Security Considerations

   RTP packets using the payload format defined in this specification
   are subject to the security considerations discussed in the RTP
   specification [RFC3550], and in any appropriate RTP profile (for
   example [RFC3551] or [RFC4585]).  This implies that confidentiality
   of the media streams is achieved by encryption; for example, through
   the application of SRTP [RFC3711].  Because the data compression used
   with this payload format is applied end-to-end, any encryption needs
   to be performed after compression.

   A potential denial-of-service threat exists for data encodings using
   compression techniques that have non-uniform receiver-end
   computational load.  The attacker can inject pathological datagrams
   into the stream that will increase the processing load of the decoder
   and may cause the receiver to be overloaded. For example inserting
   additional EDUs representing the higher enhancement layers on top of
   the ones actually transmitted may increase the decoder load. However,
   the G.718 codec is not particularly vulnerable to such an attack,
   since the majority of the computational load in an G.718 session is
   associated to the encoder.  Another form of possible attach might be
   forging of codec bit-rate control messages, which may result in
   encoder operating employing higher number of enhancement layers than
   originally intended and thereby requiring larger amount of
   computation resources. Therefore, the usage of data origin
   authentication and data integrity protection of at least the RTP
   packet is RECOMMENDED; for example, with SRTP [RFC3711].

   Note that the appropriate mechanism to ensure confidentiality and
   integrity of RTP packets and their payloads is very dependent on the
   application and on the transport and signaling protocols employed.
   Thus, although SRTP is given as an example above, other possible
   choices exist.

   Note that end-to-end security with either authentication, integrity
   or confidentiality protection will prevent a network element not
   within the security context from performing media-aware operations
   other than discarding complete packets.  To allow any (media-aware)
   intermediate network element to perform its operations, it is
   required to be a trusted entity which is included in the security
   context establishment.

6. Congestion control

   As scalable codec G.718 implicitly provides means for congestion
   control by providing a possibility for 'thinning' the bitstream. The
   RTP payload format according to this specification provides several
   different means for reducing the G.718 session bandwidth. The most
   appropriate mechanism (in terms of impact to the user experience)
   depends on the employed payload structure and also on the employed
   session configuration (single RTP session or multiple RTP sessions).
   The following means (in no particular order) can be used to assist
   congestion control procedures -- either by the sender or by the
   intermediate node.

   o  The transport blocks carrying the EDUs representing the highest
      layers within the payload may be dropped.

   o  The payloads carrying the EDUs representing the highest layers in
      an G.718 session are dropped.

   o  Transport blocks or payloads carrying EDUs belonging to redundant
      frames included in the payload are dropped.

7. IANA Considerations

   IANA is kindly requested to register a media type for the G.718 codec
   for RTP transport, as specified in section 4.1.  of this document.

APPENDIX A: Payload examples

   The G.718 payload structure enables flexible transport either by
   carrying all layers in the same payload or separating the layers into
   separate payloads. The following subsections illustrate different
   possibilities for transport by simple examples. Note that examples do
   not show the full payload structure to keep the illustration simple.

A.1. Simple payload examples

A.1.1. All the layers in the same payload

   The illustration below shows layers L1-L3 from two encoded frames
   encapsulated into separate payloads using single transport block.

    | RTP1  | L-ID=3 |NF=0 |F1-L1 |F1-L2 |F1-L3 |

    | RTP2  | L-ID=3 |NF=0 |F2-L1 |F2-L2 |F2-L3 |

   In case the same layers from two input frames are encapsulated into
   one payload using single transport block, the structure is as shown

    | RTP1  | L-ID=3 |NF=1 |F1-L1 |F2-L1 |F1-L2 |F2-L2 |F3-L3 |F2-L3 |

   The third example illustrates the case where the layers L1-L3 from
   two input frames are encapsulated into one payload using two separate
   transport blocks, the first one carrying L1 and the other one
   containing L2 and L3.

    | RTP1  | L-ID=1 |NF=1 |F1-L1 |F2-L1 |
            | L-ID=7 |NF=1 |F1-L2 |F2-L2 |F2-L2 |F2-L3 |

A.1.2. Layers in separate RTP streams

   In this case the data for each layer is transmitted in its own

   In the first example each transport block including a single EDU is
   carried in its own RTP payload.

    +-------+--------+-----+-----+    +-------+--------+-----+-----+
    | RTP1a | L-ID=1 |NF=0 |F1-L1|    | RTP1b | L-ID=6 |NF=0 |F1-L2|
    +-------+--------+-----+-----+    +-------+--------+-----+-----+

    +-------+--------+-----+-----+    +-------+--------+-----+-----+
    | RTP1c |L-ID=10 |NF=0 |F1-L3|    | RTP2a | L-ID=1 |NF=0 |F2-L1|
    +-------+--------+-----+-----+    +-------+--------+-----+-----+

    +-------+--------+-----+-----+    +-------+--------+-----+-----+
    | RTP2b | L-ID=6 |NF=0 |F2-L2|    | RTP2c |L-ID=10 |NF=0 |F2-L3|
    +-------+--------+-----+-----+    +-------+--------+-----+-----+

   If the payloads carry data from two consecutive input frames, the
   same encoded data as in the previous example is arranged as follows.

    | RTP1a | L-ID=1 |NF=1 |F1-L1|F2-L1|

    | RTP1b | L-ID=6 |NF=1 |F1-L2|F2-L2|

    | RTP1c |L-ID=10 |NF=1 |F1-L3|F2-L3|

A.2. Advanced examples

A.2.1. Different update rate for subset of layers

   An example employing different update rates (i.e. different number of
   frames per packet) for selected subsets of layers. In these examples
   all core codec layers L1-L5 are shown.

    | RTP1  | L-ID=1 |NF=3 |F1-L1|F2-L1|F3-L1|F4-L1|

    | RTP2a | L-ID=7 |NF=1 |F1-L2|F2-L2|F1-L3|F2-L3|

    | RTP3a |L-ID=14 |NF=0 |F1-L4|F1-L5|

    | RTP3b |L-ID=14 |NF=0 |F2-L4|F2-L5|

    | RTP2b | L-ID=7 |NF=1 |F3-L2|F4-L2|F3-L3|F4-L3|

    | RTP3c |L-ID=14 |NF=0 |F3-L4|F3-L5|

    | RTP3d |L-ID=14 |NF=0 |F4-L4|F4-L5|

A.2.2. Redundant frames with limited set of layers

   An example transmitting layers L1-L3 as primary data and L1 (of the
   previous frame) as redundant data is shown below. Each payload
   carries one primary (i.e. new) frame in one transport block and one
   redundant frame, which in this example is the frame preceding the
   primary frame, in another transport block.

    | RTP1  | L-ID=1 |NF=0 |F0-L1| L-ID=3 |NF=0 |F1-L1|F1-L2|F1-L3|

    | RTP2  | L-ID=1 |NF=0 |F1-L1| L-ID=3 |NF=0 |F2-L1|F2-L2|F2-L3|

    | RTP3  | L-ID=1 |NF=0 |F2-L1| L-ID=3 |NF=0 |F3-L1|F3-L2|F3-L3|

   Alternatively, the payload carrying also redundant data for a subset
   of layers can be arranged differently, as shown in the example below.

    | RTP1  | L-ID=3 |NF=0 |F0-L1|F0-L2|F0-L3| L-ID=1 |NF=0 |F1-L1|

    | RTP2  | L-ID=3 |NF=0 |F1-L1|F1-L2|F1-L3| L-ID=1 |NF=0 |F2-L1|

    | RTP3  | L-ID=3 |NF=0 |F2-L1|F2-L2|F2-L3| L-ID=1 |NF=0 |F3-L1|

   Now the first transport block carries the primary data and the second
   transport block carries the redundant data, which in this case covers
   the frame following the primary frame. The benefit of this approach
   is that the redundant data is included in the last (secondary)
   transport block of the payload, which might be beneficial for
   possible payload scaling operation within the network.

8. References

8.1. Normative References

   [AMR-WB]  3GPP TS 26.171, "Adaptive Multi-Rate Wideband (AMR-WB)
             speech codec; General description (Release 7)", v7.0.0,
             September 2006.

   [G.718]   ITU-T Recommendation G.718, "Frame Error Robust Narrowband
             and Wideband Embedded Variable Bit-Rate Coding of Speech
             and Audio from 8-32 Kbit/s", (consented) May 2008.

   [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
             Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3264] Rosenberg, J., Schulzrinne, H., "An Offer/Answer Model with
             Session Description Protocol (SDP)", RFC 3264, June 2002.

   [RFC3550] Schulzrinne,

   [RFC3550]Schulzrinne, H., Casner, S., Frederick, R. and Jacobson, V.,
             "RTP: A Transport Protocol for Real-Time Applications", STD
             64, RFC 3550, July 2003.

   [RFC3551] Schulzrinne, H., Casner, S., "RTP Profile for Audio and
             Video Conferences with Minimal Control", STD 65, RFC 3551,
             July 2003.

   [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., Norrman,
             K., "The Secure Real-Time Transport Protocol (SRTP)", RFC
             3711, March 2004.

   [RFC4288] Freed, N., Klensin, J., "Media Type Specifications and
             Registration Procedures", BCP 13, RFC 4288, December 2005.

   [RFC4566] Handley, M., Jacobson, V. and Perkins, C., "SDP: Session
             Description Protocol", RFC 4566, July 2006.

   [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., Rey, J.,
             "Extended RTP Profile for Real-Time Transport Control
             Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, July

   [RFC4855] Casner, S., "Media Type Registration of RTP Payload
             Formats", RFC 4855, February 2007.

   [RFC4867] Sjoberg, J., Westerlund, M., Lakaniemi, A., Xie, Q., "RTP
             Payload Format and File Storage Format fort he Adaptive
             Multi-Rate (AMR) and Adaptive Multi-Rate Wideband (AMR-WB)
             Audio Codecs", RFC 4867, April 2007.

   [RFC5104] Wenger, S., Chandra, U., Westerlund, M., Burman, B., "Codec
             Control Messages in the RTP Audio-Visual Profile with
             Feedback (AVPF)", RFC 5104, Feburary 2008.

   [smd-sdp] Schierl, T., Wenger, S., "Signaling media decoding
             dependency in Session Description Protocol (SDP)", draft-
             schierl-mmusic-layered-codec-04 (work in progress), June

8.2. Informative References

   [McCanne] McCanne, S., Jacobson, V., and Vetterli, M., "Receiver-
             driven layered multicast", in Proc. of ACM SIGCOMM'96,
             pages 117--130, Stanford, CA, August 1996.

   [RFC2326] Schulzrinne, H., Rao, A., Lanphier, R., "Real Time
             Streaming Protocol (RTSP)", RFC 2326, April 1998.

   [RFC2974] Handley, M., Perkins, C., Whelan, E., "Session Announcement
             Protocol", RFC 2974, October 2000.

   [RFC3828] Larzon, L-A., Degermark, M., Pink, S., Jonsson, L-E.,
             Fairhurst, G., "The Lightweight User Datagram Protocol
             (UDP-Lite)", RFC 3828, July 2004.

   [RFC4340] Kohler, E., Handley, M., Floyd, S., "Data Congestion
             Control Protocol (DCCP)", RFC 4340, March 2006.

   [RFC5117] Westerlund, M., Wenger, S., "RTP Topologies", RFC 5117,
             January 2008.

Author's Addresses

   Ari Lakaniemi
   P.O.Box 407
   FIN-00045 Nokia Group, FINLAND

   Phone: +358-71-8008000

   Ye-Kui Wang
   Nokia Research Center
   P.O. Box 1000
   33721 Tampere
   Huawei Technologies
   400 Somerset Corp Blvd, Suite 602
   Bridgewater, NJ 08807, USA

   Phone: +358-50-466-7004 +1-908-541-3518

Intellectual Property Statement

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at

Disclaimer of Validity

   This document and the information contained herein are provided on an

Copyright Statement

   Copyright (C) The IETF Trust (2008).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.


   Funding for the RFC Editor function is currently provided by the
   Internet Society.

9. Open Issues

   1) Support of super-wideband (SWB) audio and stereophonic encoding
      extensions to ITU-T G.718 currently being worked on by ITU-T is to
      be specified after ITU-T completes the work in that regards.

        a. Some further study is needed to see if separate parameters
          for sending and receiving capabilities/preferences are needed
          -- especially for upcoming stereo and SWB options.

        b. The support for upcoming SWB and stereo options needs to be
          taken into account. Basically we can either 1) extend the
          parameter "layers" to cover also this aspect, or 2) define
          separate parameter(s) for these new options when more details
          on the stereo/SWB support are available.

   2) For streaming or other applications that allow for relatively long
      end-to-end delay, sometimes it would be beneficial to aggregate
      more than 4 frames in one Transport Block (TB). Should the length
      of the NF field be larger?

   3) On layer structure and configuration signalling. Currently, a
      unique layer ID is assigned for any possible layer combinations.

      See the editing notes below Table 3 for other possible approaches.
      One of the alternative ways may be chosen in the final draft.

   4) Currently, it is mandated that lower layer EDUs of later frames go
      before higher layer EDUs of earlier frames in a transport block.
      This way is friendlier to adaptation (dropping of higher layers).
      However, if all layers are received, then the depacketizer needs
      to reorder the EDUs to their decoding order before feeding them to
      the decoder. Therefore, the other way around (i.e. lower layer
      EDUs of later frames go after higher layer EDUs of earlier frames,
      or EDUs in transport blocks are placed in decoding order) is more
      friendly to the depacketizer. Another benefit of the latter is
      that it does not introduce any end-to-end delay. Which way to be
      specified (or both allowed if needed) is FFS.

   5) MANEs dropping RTP packets are RTP translators. But are those
      MANEs dropping a subset of the transport blocks in one packet also
      RTP translators?

   6) The RTCP based cross-session synchronization is not possible until
      the first RTCP SRs are received in all sessions. This implies that
      decoding only a subset of layers may be possible until RTCP SRs in
      all sessions have been received. This may imposes higher end-to-
      end delay or higher bandwidth for RTCP data, and the approach may
      not work perfectly for some multicast topologies. There is a study
      ongoing by some AVT members. Once there is an acceptable solution
      fouthe draft documenting that solution may be referenced in this

   7) It might be better to change the semantics of the media type
      parameter 'layers' to be similar as that for L-ID.

   8) Offer/answer with answer being capable of modifying the layer
      configuration is FFS.

   9) Some references need to be updated in the final draft.

10. Changes Log

   From draft-lakaniemi-avt-rtp-evbr-04 to draft-ietf-avt-rtp-g718-00 to draft-ietf-avt-rtp-g718-01

   - Changed Updated the media sub-type name "EV-VBR" to "G718", boiler template.

   - Changed Ye-Kui Wang's affiliation and replaced
     "EV-VBR" with "G.718" in all text. address.