mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-10-31 22:25:38 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

bluez5: clarify OPUS-A2DP spec

Browse source 
Also remove DRAFT tag.
This commit is contained in:
Pauli Virtanen 2022-09-03 13:58:06 +03:00 • committed by Wim Taymans
parent b70f32a259
commit df858cae7c
1 changed files with 92 additions and 78 deletions
Download patch file Download diff file Expand all files Collapse all files

170
spa/plugins/bluez5/README-OPUS-A2DP.md
View file

@ -6,31 +6,18 @@ date: Jun 4, 2022
# OPUS-A2DP-0.5 specification # OPUS-A2DP-0.5 specification
DRAFT In this file, a way to use Opus as an A2DP vendor codec is specified.
In this file, we specify how to use Opus as an A2DP vendor codec. We We will call this "OPUS-A2DP-0.5". There is no previous public
will call this "OPUS-A2DP-0.5". There is no previous public
specification for using Opus as an A2DP vendor codec (to my specification for using Opus as an A2DP vendor codec (to my
knowledge), which is why we need this one. knowledge), which is why we need this one.
[[_TOC_]] [[_TOC_]]
# A2DP Codec Capabilities # Media Codec Capabilities
The A2DP capability structure is as follows. The Media Codec Specific Information Elements ([AVDTP v1.3], §8.21.5)
capability and configuration structure is as follows:
Integer fields and multi-byte bitfields are laid out in **little
endian** order. All integer fields are unsigned.
Each entry may have different meaning when present as a capability.
Below, we indicate this by abbreviations CAP/SNK for sink capability,
CAP/SRC for source capability, CAP for capability as either, and SEL
for the selected value by SRC.
Bits in fields marked RFA (Reserved For Additions) shall be set to
zero.
The capability and configuration structure is as follows:
| Octet | Bits | Meaning | | Octet | Bits | Meaning |
|-------|------|-----------------------------------------------| |-------|------|-----------------------------------------------|
@ -42,7 +29,19 @@ The capability and configuration structure is as follows:
| 17-20 | 0-7 | Return Direction Audio Location Configuration | | 17-20 | 0-7 | Return Direction Audio Location Configuration |
| 21-23 | 0-7 | Return Direction Limits Configuration | | 21-23 | 0-7 | Return Direction Limits Configuration |
See `a2dp-codec-caps.h` for definition as C structs. All integer fields and multi-byte bitfields are laid out in **little
endian** order. All integer fields are unsigned.
Each entry may have different meaning when present as a capability.
Below, we indicate this by abbreviations CAP for capability and SEL
for the value selected by SRC.
Bits in fields marked RFA (Reserved For Additions) shall be set to
zero.
> **Note**
>
> See `a2dp-codec-caps.h` for definition as C structs.
## Vendor ID Part ## Vendor ID Part
@ -53,13 +52,16 @@ The fixed value
| 0-3 | 0-7 | A2DP Vendor ID (0x05F1) | | 0-3 | 0-7 | A2DP Vendor ID (0x05F1) |
| 4-5 | 0-7 | A2DP Vendor Codec ID (0x1005) | | 4-5 | 0-7 | A2DP Vendor Codec ID (0x1005) |
The Vendor ID is that of the Linux Foundation, and we are using it > **Note**
here unofficially. >
> The Vendor ID is that of the Linux Foundation, and we are using it
> here unofficially.
## Channel Configuration ## Channel Configuration
The channel configuration consists of the channel count and a bitfield The channel configuration consists of the channel count, and the count
indicating which of them are encoded in coupled streams. of coupled streams. The latter indicates which channels are encoded as
left/right pairs, as defined in Sec. 5.1.1 of Opus Ogg Encapsulation [RFC7845].
| Octet | Bits | Meaning | | Octet | Bits | Meaning |
|-------|------|------------------------------------------------------------| |-------|------|------------------------------------------------------------|
@ -74,26 +76,23 @@ coupled (left & right) channel pair. The count shall satisfy
`(Channel Count) >= 2*(Coupled Stream Count)`. `(Channel Count) >= 2*(Coupled Stream Count)`.
The Stream Count is `(Channel Count) - (Coupled Stream Count)`. The Stream Count is `(Channel Count) - (Coupled Stream Count)`.
Streams and Coupled Streams have the same meaning as in Sec. 5.1.1 of
Opus Multistream [RFC7845].
The logical Channels are identified by a Channel Index *j* such that `0 <= j The logical Channels are identified by a Channel Index *j* such that `0 <= j
< (Channel Count)`. The channels `0 <= j < 2*(Coupled Stream Count)` < (Channel Count)`. The channels `0 <= j < 2*(Coupled Stream Count)`
are encoded in the *k*-th stream of the payload, where `k = floor(j/2)` and are encoded in the *k*-th stream of the payload, where `k = floor(j/2)` and
`j mod 2` determines which of the two channels of the stream the logical `j mod 2` determines which of the two channels of the stream the logical
channel is. The channels `2*(Coupled Stream Count) <= j < (Channel Count)` channel is. The channels `2*(Coupled Stream Count) <= j < (Channel Count)`
are encoded in the *k*-th stream of the payload, where `k = j - (Coupled Stream Count)`. are encoded in the *k*-th stream of the payload, where `k = j - (Coupled Stream Count)`.
The prescription here is identical to [RFC7845] with channel mapping
`mapping[j] = j`.
The semantic meaning for each channel is determined by their Audio > **Note**
Location. >
> The prescription here is identical to [RFC7845] with channel mapping
> `mapping[j] = j`. We do not want to include the mapping table in the
> A2DP capabilities, so it is assumed to be trivial.
## Audio Location Configuration ## Audio Location Configuration
The channel audio location specification is similar to the location The semantic meaning for each channel is determined by their Audio
bitfield of the `Audio_Channel_Allocation` LTV structure in Bluetooth Location bitfield.
SIG [Assigned Numbers, Generic Audio] used in the LE Audio.
| Octet | Bits | Meaning | | Octet | Bits | Meaning |
|-------|------|------------------------------------------------------| |-------|------|------------------------------------------------------|
@ -103,11 +102,9 @@ The values specified in CAP are informative, and SEL may contain bits
that were not set in CAP. SNK shall handle unsupported audio that were not set in CAP. SNK shall handle unsupported audio
locations. It may do this for example by ignoring unsupported channels locations. It may do this for example by ignoring unsupported channels
or via suitable up/downmixing. Hence, SRC may transmit channels with or via suitable up/downmixing. Hence, SRC may transmit channels with
audio locations that are not marked supported by SNK. The maximum audio locations that are not marked supported by SNK.
Channel Count however shall not be exceeded.
The audio location bitfield values defined in [Assigned Numbers, The audio location bit values are:
Generic Audio] are:
| Channel Order | Bitmask | Audio Location | | Channel Order | Bitmask | Audio Location |
|---------------|------------|-------------------------| |---------------|------------|-------------------------|
@ -144,8 +141,8 @@ Generic Audio] are:
| 30 | 0x40000000 | RFA | | 30 | 0x40000000 | RFA |
| 31 | 0x80000000 | RFA | | 31 | 0x80000000 | RFA |
In addition, we define a specific Channel Order for each. The bits Each bit value is associated with a Channel Order. The bits set in
set in the bitfield define audio locations for the streams present in the the bitfield define audio locations for the streams present in the
payload. The set bit with the smallest Channel Order value defines the payload. The set bit with the smallest Channel Order value defines the
audio location for the Channel Index *j=0*, the bit with the next audio location for the Channel Index *j=0*, the bit with the next
lowest Channel Order value defines the audio location for the Channel lowest Channel Order value defines the audio location for the Channel
@ -154,46 +151,53 @@ Index *j=1*, and so forth.
When the Channel Count is larger than the number of bits set in the When the Channel Count is larger than the number of bits set in the
Audio Location bitfield, the audio locations of the remaining channels Audio Location bitfield, the audio locations of the remaining channels
are unspecified. Implementations may handle them as appropriate for are unspecified. Implementations may handle them as appropriate for
their use case, considering them as AUX0-AUXN, or in the case of their use case, considering them as AUX0–AUXN, or in the case of
Channel Count = 1, as the single mono audio channel. Channel Count = 1, as the single mono audio channel.
When the Channel Count is smaller than the number of bits set in the When the Channel Count is smaller than the number of bits set in the
Audio Location bitfield, the audio locations for the channels are Audio Location bitfield, the audio locations for the channels are
assigned as above, and remaining excess bits shall be ignored. assigned as above, and remaining excess bits shall be ignored.
The channel ordering defined here is compatible with the internal > **Note**
stream ordering in the reference Opus Multistream surround encoder >
Mapping Family 0 and 1 output. This allows making use of its surround > The channel audio location specification is similar to the location
masking and LFE handling capabilities. The stream ordering of the > bitfield of the `Audio_Channel_Allocation` LTV structure in Bluetooth
reference Opus surround encoder, although being unchanged since its > SIG [Assigned Numbers, Generic Audio] used in the LE Audio, and the
addition in 2013, is an internal detail of the > bitmasks defined above are the same.
encoder. Implementations using the surround encoder shall check that >
the mapping table used by the encoder corresponds to the above channel > The channel ordering differs from LE Audio, and is defined here to be
ordering. > compatible with the internal stream ordering in the reference Opus
> Multistream surround encoder Mapping Family 0 and 1 output. This
For reference, we list the Audio Location bitfield values > allows making use of its surround masking and LFE handling
corresponding to the different channel counts in Opus Mapping Family 0 > capabilities. The stream ordering of the reference Opus surround
and 1 surround encoder output, and the expected mapping table: > encoder, although being unchanged since its addition in 2013, is an
> internal detail of the encoder. Implementations using the surround
| Mapping Family | Channel Count | Audio Location Value | Stream Ordering | Mapping Table | > encoder need to check that the mapping table used by the encoder
|----------------|---------------|----------------------|---------------------------------|--------------------------| > corresponds to the above channel ordering.
| 0 | 1 | 0x00000000 | mono | {0} | >
| 0 | 2 | 0x00000003 | FL, FR | {0, 1} | > For reference, we list the Audio Location bitfield values
| 1 | 1 | 0x00000000 | mono | {0} | > corresponding to the different channel counts in Opus Mapping Family 0
| 1 | 2 | 0x00000003 | FL, FR | {0, 1} | > and 1 surround encoder output, and the expected mapping table:
| 1 | 3 | 0x00000007 | FL, FR, FC | {0, 2, 1} | >
| 1 | 4 | 0x00000033 | FL, FR, BL, BR | {0, 1, 2, 3} | > | Mapping Family | Channel Count | Audio Location Value | Stream Ordering | Mapping Table |
| 1 | 5 | 0x00000037 | FL, FR, BL, BR, FC | {0, 4, 1, 2, 3} | > |----------------|---------------|----------------------|---------------------------------|--------------------------|
| 1 | 6 | 0x0000003f | FL, FR, BL, BR, FC, LFE | {0, 4, 1, 2, 3, 5} | > | 0 | 1 | 0x00000000 | mono | {0} |
| 1 | 7 | 0x00000d0f | FL, FR, SL, SR, FC, BC, LFE | {0, 4, 1, 2, 3, 5, 6} | > | 0 | 2 | 0x00000003 | FL, FR | {0, 1} |
| 1 | 8 | 0x00000c3f | FL, FR, SL, SR, BL, BR, FC, LFE | {0, 6, 1, 2, 3, 4, 5, 7} | > | 1 | 1 | 0x00000000 | mono | {0} |
> | 1 | 2 | 0x00000003 | FL, FR | {0, 1} |
The Mapping Table in the table indicates the mapping table selected by > | 1 | 3 | 0x00000007 | FL, FR, FC | {0, 2, 1} |
`opus_multistream_surround_encoder_create` (Opus 1.3.1). If the > | 1 | 4 | 0x00000033 | FL, FR, BL, BR | {0, 1, 2, 3} |
encoder outputs a different mapping table in a future Opus encoder > | 1 | 5 | 0x00000037 | FL, FR, BL, BR, FC | {0, 4, 1, 2, 3} |
release, the channel ordering will be incorrect, and the surround > | 1 | 6 | 0x0000003f | FL, FR, BL, BR, FC, LFE | {0, 4, 1, 2, 3, 5} |
encoder can not be used. We expect that the probability of the Opus > | 1 | 7 | 0x00000d0f | FL, FR, SL, SR, FC, BC, LFE | {0, 4, 1, 2, 3, 5, 6} |
encoder authors making such changes is negligible. > | 1 | 8 | 0x00000c3f | FL, FR, SL, SR, BL, BR, FC, LFE | {0, 6, 1, 2, 3, 4, 5, 7} |
>
> The Mapping Table in the table indicates the mapping table selected by
> `opus_multistream_surround_encoder_create` (Opus 1.3.1). If the
> encoder outputs a different mapping table in a future Opus encoder
> release, the channel ordering will be incorrect, and the surround
> encoder can not be used. We expect that the probability of the Opus
> encoder authors making such changes is negligible.
## Limits Configuration ## Limits Configuration
@ -236,6 +240,12 @@ Namely:
If no return channel is supported or selected, the number of channels If no return channel is supported or selected, the number of channels
is set to 0 in CAP or SEL. is set to 0 in CAP or SEL.
> **Note**
>
> This is a nonstandard extension to A2DP. The return direction audio
> data is simply sent back via the underlying L2CAP connection, which
> is bidirectional, in the same format as the forward direction audio.
> This is similar to what aptX-LL and FastStream do.
# Packet Structure # Packet Structure
@ -255,8 +265,9 @@ may be fragmented to several consecutive Bluetooth packets.
The format of the Multistream data is the same as in the audio packets The format of the Multistream data is the same as in the audio packets
of [RFC7845], or, as produced/consumed by the Opus Multistream API. of [RFC7845], or, as produced/consumed by the Opus Multistream API.
(Note that we DO NOT follow [RFC7587], as we want fragmentation and > **Note**
multichannel support.) >
> We DO NOT follow [RFC7587], as we want fragmentation and multichannel support.
## RTP Header ## RTP Header
@ -311,10 +322,13 @@ correction instead of PLC.
# References # References
1. IETF RFC 3550: [RFC3550] 1. Bluetooth [AVDTP v1.3]
2. IETF RFC 7587: [RFC7587] 2. IETF [RFC3550]
3. IETF RFC 7845: [RFC7845] 3. IETF [RFC7587]
4. IETF [RFC7845]
5. Bluetooth [Assigned Numbers, Generic Audio]
[AVDTP v1.3]: https://www.bluetooth.com/specifications/specs/a-v-distribution-transport-protocol-1-3/
[RFC3550]: https://datatracker.ietf.org/doc/html/rfc3550 [RFC3550]: https://datatracker.ietf.org/doc/html/rfc3550
[RFC7587]: https://datatracker.ietf.org/doc/html/rfc7587 [RFC7587]: https://datatracker.ietf.org/doc/html/rfc7587
[RFC7845]: https://datatracker.ietf.org/doc/html/rfc7845 [RFC7845]: https://datatracker.ietf.org/doc/html/rfc7845