IEEE 1722 Media on AVB Networks

IEEE 1722 Media on AVB Networks
IEEE 1722 Media on AVB Networks
Presentation to the AVnu Alliance Broadcast Advisory Council
Rob Silfvast
Principal System Architect, Avid
rob.silfvast@avid.com
6-June-2013
Outline
• Brief overview of IEEE 1722 AV Transport Protocol
• Description of AVTP Professional Video Format
• Comparison and co-existence with ST-2022
• Clock Reference Streams on AVB
• Q&A
AVB Terminology
AVB Cloud (or AVB Domain)
A collection of interconnected network nodes, and the links between them, all of
which support the 802.1 AVB standards
Stream
A regular flow of packets in an AVB cloud that contains one or more channels of
media (or other real-time) data. A stream has a unique ID and is reserved using
802.1Q-2011 Stream Reservation Protocol.
Talker
An entity in the AVB cloud that can send a stream
Listener
An entity in the AVB cloud that can receive a stream
Controller
An entity on the network which configures and connects Talkers and Listeners in an
AVB network
Media Synchronization across a shared LAN
• Maintain A/V lip-sync between rendering devices
– No loss of sync due to separate transport flows across network
– Separate streams can have different delivery times
• Maintain tight audio sync between multiple transducers
– Need < 2 usec to support phased arrays (mics or loudspeakers)
– Tight video sync also useful for multi-display (compound) screens
• Convey media clocks from master nodes to slave nodes
– Across a fundamentally asynchronous network infrastructure (!)
– Support multiple, independent clock domains on same network
Audio Video Transport Protocol (AVTP)
IEEE 1722 Basics
IEEE 1722 Basics
•
•
•
•
•
•
•
•
Ethernet frame format
Carried at the Data-Link layer (OSI Layer 2)
Ethertype = 0x22F0
Max Frame size = 1542 Bytes (on the wire)
Priority Tagging used to differentiate from non-AVB traffic
VLAN Tagging supported
Destination Address is (typically) a Multicast Address
Unique ID per Stream
IEEE 1722-2011 Data Format
•
•
•
•
•
Based on IEC 61883
Used for media transport on IEEE 1394 (FireWire)
Supports myriad audio and video formats
Even supports timecode and MIDI
Has some “legacy baggage”
IEEE 1722 Presentation Timestamp
• Tells Listener the exact time to present media samples
– Using the common 802.1AS clock as a “measuring stick”
– De-couples media play-out time from network transit time
Stream X
Stream X
Talker
10G
1G
Stream X
Listener
A
Listener
B
Both nodes render
Stream X payload
at the same time
IEEE P1722a – new features for AVTP
• AVTP Video Formats
– “APVF” = AVTP Professional Video Format
– RTP Video encapsulated in an AVB stream
• AVTP Audio Formats
– Lose the baggage from IEC 61883 formatting
• Clock Reference Streams
– Convey media clocks across the network
– Leverages Presentation Time concept of 1722
APVF (AVTP Pro Video Format)
• Designed to carry SDI-encoded bitstreams, including
all HANC and VANC data
– 270 Mbit (SMPTE 259)
– 1.5 Gbit (SMPTE 292)
– 3 Gbit (SMPTE 425)
• Packetization scheme correlates to raster scanning
– Start of video line always starts a new packet
– AVTP timestamp for every horizontal blanking event
• Marker flags to indicate key events in the stream
– Vertical blanking
– Interlaced: field 0 vs field 1
– metadata frames
SMPTE Standards for Video over IP
• ST-2022-1 and ST-2022-2
– MPEG-2 Transport Stream (TS) in Constant Bit Rate (CBR) with
Forward Error Correction (FEC)
• ST-2022-3 and ST-2022-4
– As above but with Variable Bit Rate (VBR)
• ST-2022-5 and ST-2022-6
– Uncompressed video over RTP/UDP IP from 270 Mb/s to 3 Gb/s
– Payload encoding uses SDI formats (ST-259/292/495)
– Optional FEC (ST-2022-5)
Comparison of ST-2022 and AVB
Attribute
SMPTE 2022 (over IP)
AVB Pro Video Format *
Compression
MPEG, H.264, J2K, none
None
FEC
Optional
None
Link types
WAN, MAN (Layer-3 datagrams)
LAN (Layer-2 datagrams)
End-to-end Sync
External to the standard
802.1AS and 1722 presentation time
QoS
External to the standard
802.1Q SRP and FQTSS
Expected Latency
Medium (~1 video frame)
Low (~1 video line)
Long haul, back haul,
contribution links, primary
distribution
SDI-like transport of live media
within a facility
(based on encoding)
Best Application
*IEEE P1722A Draft
Packetization schema: ST-2022-6 vs IEEE P1722A APVF
SDI
Payload
ST2022-6
Packet
Stream
line 1
line 2
line 3
line 1125
T
T
T
T
T
T
T
T
T
T
T
T
T
T
P
P
P
P
P
P
P
P
P
P
P
P
P
P
line 1
T
P
T
T
T
T
P
P
P
P
CRC
CRC
CRC
CRC
Pad
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
1 Video Frame
4497 packets
Marker Flags
last packet of frame
first packet of line
SDI
Payload
line 1
line 2
T
APVF
Packet
Stream
line 3
T
line 1125
T
line 1
T
T
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
CRC
1 Video Frame
4500 packets
Timestamps in ST2022-6 vs 1722 APVF
• ST-2022-6 has two timestamp fields per packet
– RTP timestamp correlates to “first byte in the RTP datagram”
• This first byte is the media payload header
• Unrelated to the actual media
– Video timestamp
•
•
•
•
Is optional
Correlates to first sample in media payload to a video clock at the sender
Most often this is not a particular point in the raster
No mandate for sender’s video clock to be replicated at all nodes
• APVF has one timestamp per video line (every Nth packet)
– Presentation time of first pixel in the line
Why Operate at Layer 2 ?
• Limiting the domain size can be a good thing
– Don’t try to “boil the ocean”
• L3 (IP) datagrams can go anywhere
• An all-IP solution would need different modes for LAN vs WAN routing
– Expectations of QoS and Latency more deterministic on a LAN
– Low-latency maps to Local Area (Long haul is tolerant to latency)
– Tighter user control => more secure
• Media packets handled lower in the network stack
– Less header/overhead processing (3G SDI: 360,000 packets/sec!)
– Filter at L2 => avoid implementing a HW-based IP parser
• L2-only is lower cost than L3+L2
ST-2022 and AVB together
Diagram by
Axon Digital
“We think s2022 and AVB will work hand in hand: AVB for a closed, layer-2
production environment, and s2022 for connecting venues with the network as
an alternative to satellite uplinks.” -- Peter Schut, CTO Axon Digital
AVTP Clock Reference Streams
• Convey timing of “clock tick” events across an AVB LAN
• Establish common timing grid for all participating nodes
– Same concept as House Sync
•
•
•
•
No media payload => low bandwidth usage
Talker = Clock Master (CRS Master)
Multicast to all listeners in same clock domain
Supports multiple clock domains on same LAN
– CRS (and media) timing is decoupled from 802.1AS clock
– Any number of CRS can be carried on a LAN
AVB End Node
Listener
System
Diagram
with CRS
media
rx
data in
data out
media stream(s)
media
tx
Internal
Media
function
aligned*
clock
recovery
Talker
CRS
clk in
* Listener expects media timestamps to
be aligned to the common timing grid
established by the CRS (within a nominal
tolerance). This implies a requirement for
the Talker to align timestamps to the grid.
Network Infrastructure
Talker
sampler &
timestamper
CRS
originating
source
add
Td
CRS tx
A master clock can reside in any AVB
end node, and should be ostensibly
designated as “CRS Master”
AVB End Node
Listener
media
rx
data in
Internal
Media
function
aligned*
AVB End Node
Listener
media
rx
data in
aligned*
clock
recovery
data out
Internal
Media
function
clk in
clock
recovery
Talker
media
tx
data out
clk in
Talker
media
tx
AVTP CRS details
• Common timing grid is typically the line clock for video or the
sampling clock for audio
• CRS payload = timestamps
– corresponding to every Nth tick on the common grid
• Listeners can up-sample the timestamps to re-establish the grid
clock
– Use a PLL when very low jitter is required
• Simple logic at listener: compare CRS timestamps to locally
running 802.1AS clock
Early Adoption of IEEE 1722a Pro Video Format
Axon Digital CEO Jan Eveleens
at NAB show, April 2013
Demonstrating a working prototype
of 3G SDI running bi-directionally
across 10G Ethernet/AVB, through
an Extreme Networks switch
Thank You
(Q&A)
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising