MOST and Ethernet: Payload efficiency and network considerations (cont.)
In comparison to MOST, AVB provides similar QoS when the above introduced sub-standards are supported by the devices of the AVB network. Their bandwidth consumption is considered similarly to the MOST administrative data. IEEE 802.1Qat utilizes reservation messages that lead to a negligible background load. IEEE 802.1Qav does not utilize any messages. The time synchronization frames of IEEE 802.1AS lead to a background load of about 0.1 Mbit/s.
On top of the AVB protocols, communication for setting up and controlling media streams is required. Therefore, a general background load of 1.5 Mbit/s is assumed, as this is also offered by the MOST control channel. Hence, the actually available streaming bandwidth of a 100 Mbit/s AVB network is reduced to BAVB = 98.4 Mbit/s. Note that the Ethernet AVB standard limits the bandwidth of QoS supported data to 75 % of the totally available bandwidth, since 25 % is reserved for conventional best effort traffic.
As the preferred AVB transport protocol for media streams, P1722 has been examined. The P1722 frame structure is depicted in, P1722 embeds several media stream formats, like e.g. the IEC 61883 based formats, to an IEEE 802 network.
AVB supports two QoS classes. Class A provides a maximum latency of 2 ms and Class B provides a maximum latency of 50 ms over 7 hops. For Class A, P1722 frames are sent each 125 µs, which e.g. allows collecting six stereo audio samples at a sampling rate of 48 kHz. For Class B, P1722 frames are sent every 250 µs.
Figure 2: IEEEP1722 frame structure.
The streaming use cases defined in the left column of Table 1a) were used for the comparison of MOST and AVB. The Blu-ray disc is specified with a maximum data rate of 53.95 Mbit/s for audio and video and will soon enter the infotainment domain.
Due to the maximum support of 50 Mbit/s for a single transport stream of the MOST INIC Transceiver (9), we used a reduced media content for the calculation composed of Dolby Digital Plus audio (1.7 Mbit/s) and video (38.5 Mbit/s). This leads to an actual transport stream Payload PBlu-ray of 41.1 Mbit/s considering the transport stream header.
The payload efficiency PE for the use cases is shown in column PE of the respective network in Table 1a) and was calculated using (Eq 1).
Table 1: Payload efficiency of streaming use cases.
As mentioned above, MOST150 has a payload efficiency of 100%. When transmitting an audio stream with P1722, the overhead (including CRC) is about 62 bytes. This causes bad payload efficiency when transmitting a small amount of data, as is the case with CD-Audio streams. Since the P1722 payload for a single packet can be up to 1472 bytes long, the payload efficiency also reaches values beyond 90%, as visible in a).
Unlike audio, video packets sent via P1722 have a better balance between payload and header information. Applications needing lots of bandwidth generate better payload efficiency because larger frames need to be used. The substantial overhead of each frame is spread out over more payload bytes.
However, just looking at the payload efficiency is not sufficient. Additionally, the network utilization NU needs to be considered. NU is defined by the ratio of the required to the available bandwidth of the complete network.
NU = D/B (Eq. 2)
Mapping the use cases above to a MOST (10) and Ethernet based vehicle architecture, as shown in Figure 3, leads to network utilizations as summarized in Table 1b. Looking into the infotainment domain, this leads to one exemplary combination of streaming use cases to
NUINFO = (DCD-A + DDVD-A + 2 x DBlu-ray) / B (Eq 3)
i.e. actually to NUMOST150_INFO = 63.3% and NUAVB_INFO = 37.9%.
Looking into the Advanced Driver Assistance Systems (ADAS) domain, with a Top-View system as a possible application generating a kind of birds' eye view of the vehicle, this requires at least 4 live video streams. For MOST, this leads to
NUMOST150INFO_ADAS = (4 x DMOST150_LiveVideo) / BMOST150 = 92.4%. (Eq 4)
since the available streaming bandwidth is shared. For Ethernet AVB this leads to
NUAVB_ADAS = (4 x DAVB_LiveVideo) / 4 x BAVB = 40.4% (Eq 5)
since each link also multiplies the available streaming bandwidth. Because shows a 5th camera used, e.g. for driver observation, birds' eye, traffic sign recognition, etc, the NUMOST150_ADAS is more than 100 %, which is not the case for NUAVB_ADAS - that remains at 40.4%.
Figure 3: Possible MOST and Ethernet based vehicle architecture.