Developer FAQ

1   Overview

1.1   What is AVB?

Audio Video Bridging is a term for a collection of IEEE Standards for time sensitive networking which facilitates the transport of high performance audio and video on a LAN:

1.2   How is media transported with AVB?

IEEE Std 1722-2011 builds on the following standards for media packetization:

1.3   What are the minimum requirements for an AVB switch?

An AVB Switch is required to support:

  • IEEE Std. 802.1Q-2011 FQTSS and SRP.
  • IEEE Std. 802.1AS-2011 on each AVB enabled ethernet port

1.4   What are the minimum requirements for an AVB device?

An AVB Device that is both an AVDECC Talker and AVDECC Listener is required to support:

  • IEEE Std. 802.1Q-2011 FQTSS and SRP.
  • IEEE Std. 802.1AS-2011 on each AVB enabled ethernet port
  • IEEE Std. 1722.1-2013

1.5   What are 802.1Qav and 802.1Qat?

802.1Qav and 802.1Qat are two of the amendments to IEEE Std 802.1Q-2005 that were rolled into IEEE Std 802.1Q-2011.

  • What was known as 802.1Qav is now known as IEEE Std 802.1Q-2011 Clause 34, "Forwarding and Queuing for time-sensitive streams".
  • What was known as 802.1Qat is now known as IEEE Std 802.1Q-2011 Clause 35, "Stream Reservation Protocol".

1.6   Someone said that an AVB endpoint does not have to support Qav, Qat - the AVB switch can take care of it

That is incorrect. All end points are required to implement 802.1AS-2011, 802.1Qav AND 802.1Qat. This is all defined in IEEE Std 802.1BA-2011.

Specifically see the PICS requirements for talkers defined in IEEE Std 802.1BA-2011 Clause 7.4.7,"Common requirements - Talker end stations" where it shows in Clause 7.4.7.1 the item "T-AS" requires 802.1AS, "T-SRP" requires 802.1Qat, and item "T-FQ" which requires forwarding and queuing of time sensitive streams (FQTSS) aka 802.1Qav. Further details of these requirements are in 7.4.7.2, and 7.4.7.3.

1.7   What is the difference between PTP, PTPv2, gPTP, 802.1AS and IEEE 1588?

  • IEEE Std 1588-2008 defines the PTPv2 (Precision Time Protocol Version 2) framework and is also known as PTPv2.
  • IEEE Std 802.1AS-2011 defines a specific profile of IEEE 1588-2008 with additional timing features which greatly improve timing accuracy and lock time. This is called gPTP (Generalized Precision Time Protocol).

1.8   Can an AVB device just use IEEE 1588v2 (PTP) instead of 802.1AS?

No, IEEE 802.1AS is required to maintain the required synchronization compatibility and accuracy.

1.9   My device's ethernet port says it supports IEEE 1588. Can it do AVB?

IEEE 1588v2 is a framework for time synchronization.

Profiles of 1588v2 are not necessarily compatible or even require the same hardware.

For instance, once profile of IEEE 1588v2 is used for synchronization of Cellular Telephone towers and that profile has very different hardware and protocol requirements compared to the default 1588v2 profile seen in generic switches, and is different again compared to the profile used in industrial switches.

So there is no clear definition of what “1588-2008 capable hardware” means. I have seen one ethernet chip that says that but only timestamps incoming packets, not outgoing packets.

802.1AS has no options like that. If your device supports 802.1AS then it will provide AVB quality synchronization better than the default 1588v2 profile. You need to have the support in the ethernet hardware to provide timestamps of specific types of both incoming and outgoing packets. These timestamps need to represent the time that the packet hit the wire. Many ethernet hardware systems provide timestamps instead at the MAC level which means your software needs to compensate for the PHY latencies for your specific PHY. The PHY latencies are typically different for 100baseT and GigE and different for TX and RX.

802.1AS needs timestamping of Pdelay, PdelayFollowUp, PdelayRequest, PdelayRequestFollowUp, Sync, and FollowUp messages for both directions.

1.10   Can FQTSS be implemented in pure software?

IEEE Std 802.1Q-2011 Clause 34, "Forwarding and Queuing for time-sensitive streams (FQTSS)" is important to implement on Talkers. For devices that need to transmit many streams at the same time, you really want hardware support in the ethernet interface to assist in scheduling the time-sensitive packets for transmission.

If you have only one stream, it is typically feasible to implement FQTSS in software - Specifically for a class A stream you need to send one packet every 125 microseconds and no faster. You can do this on a CPU with a high frequency timer interrupt.

But if you were needing to provide 4 streams simultaneously, then it is much harder to do in software as you then need to send a time sensitive stream packet every 31 microseconds (and no sooner).

Since this can be expensive and inefficient to implement in software so ideally you want an Ethernet interface like the intel i210 ( with the driver at https://github.com/intel-ethernet/Open-AVB/tree/master/kmod/igb ) or the equivalent Broadcom ethernet chip which the Apple Mac OS X uses.

The priority queues are needed for the traffic shaping as defined by IEEE 802.1Q Clause 34 (also known as IEEE 802.1Qav). Media packets have to be properly traffic shaped and need higher priority than non media packets. Ideally you need 4 separate Queues in each direction, in priority order:

  • Highest Priority: #1 802.1AS messages
  • Next Priortity : #2 IEEE 1722 media packet messages
  • Next Priority : #3 IEEE 802.1Qat (SRP) stream reservation messages
  • Lowest priority : #4 all other control messages (IEEE 1722.1, MAAP) and IPv4 and IPv6.

1.11   How many AVB streams of audio can you have on an entire network?

The maximum number of streams on an AVB network depends on the streams properties.

  • If the streams properties are not packable, there can be 318 streams on a network - each with 1 or more channels of audio.
  • If each talker device provides streams with stream properties that are packable, there can be 318 talkers on a network each with as many streams as they can fit on their links.
  • If all the stream properties across the entire network are packable, there can be much more than 318 talkers on a network.

1.12   What makes stream properties packable?

A stream has the following properties:

  • Stream ID
  • Destination MAC Address
  • Stream Class
  • Bandwidth
  • MSRP Accumulated Latency

In order for two streams to be "Packable," the two streams are required to have the same Stream Class, Bandwidth, MSRP Accumulated Latency. The two streams must have consecutive Stream ID's and Destination MAC addresses.

A Talker that provides multiple streams of the same size would typically automatically have packable stream properties.

In order to have packable streams from multiple talkers on the network the AVDECC Controller is required to use the SET_STREAM_INFO command to manually set the Stream ID and Destination MAC address for each talker stream source.

1.13   What additional work is being done for AVB standardization?

The IEEE 1722 working group is currently working on an amendment to IEEE Std 1722-2011 called IEEE P1722A which will include support for additional media streaming formats:

  • AVTP Audio Format (allowing for more flexibility of bit widths, samples per packet, and channel counts.
  • Compressed Video Format (allowing for H.264, MJPEG, and JPEG2000 formats)
  • AVTP Control Format (allowing for real time transport of FlexRay™, CAN FD, LIN®, and MOST® messages as well as serial, parallel, sensor data, vendor specific data, and AVDECC AECPDU messages)
  • Clock Reference Format (for transporting arbitrary clocks including video vertical and horizontal sync clocks)
  • SDI Video Format (for transporting high definition uncompressed video via SMPTE Serial Digital Interface )
  • Raw Video Format (for transporting generic video frame buffers)
  • Elliptic Curve Encryption and Signing of control messages
  • AES Encryption of control messages and stream data.

Comments

Comments powered by Disqus