2000-01 -C028
`
`Future Automotive Multimedia Subs stem
`Interconnect Techno ogies
`Visteon Automotive Systems
`
`the most
`between them. Table 1.0 lists some of
`important applications that are already present or will be
`soon introduced into the automotive environment.
`
`ABSTRACT
`
`For the past decade or so, automotive entertainment
`subsystem architectures have consisted of a simple
`Human Machine Interface (HMI), AM-FM tuner, a tape
`deck, an amplifier and a set of speakers. Over time, as
`customer demand for more entertainment
`features
`increased, automotive entertainment
`integrators made
`room for new features by allowing for
`the vertical
`integration of analog audio and adding a digital control.
`The
`new digital
`control
`came
`to
`entertainment
`subsystems via a low speed multiplexing scheme
`embedded
`into
`the
`entertainment
`subsystem
`components,
`allowing remote control of
`these new
`features. New features were typically incorporated into
`the
`entertainment
`subsystem by
`independently
`packaging functional modules. Examples of
`these
`modules are cellular telephone, Compact Disc Jockey
`(CDJ), rear-seat entertainment, Satellite Digital Audio
`Radio System (S-DAFIS) receiver, voice and navigation
`with its associated display and hardware. Figure 1.0 is a
`block diagram of typical entertainment subsystem. This
`paper discusses alternatives to the module-expansion of
`entertainment subsystem via low speed digital control
`and analog audio. Moreover, the discussion is expanded
`to cover future multimedia and infotainment subsystem
`interconnects technologies.
`
`INTRODUCTION
`
`Recently, great achievements have been reached in
`information,
`communication,
`entertainment,
`comfort,
`safety and security products. Moreover, new Intelligent
`Transport Systems (ITS) services, requiring state-of-the—
`art electronics, are appearing on the market to help
`drivers process
`information, make decisions,
`and
`operate vehicles more safely and effectively.
`
`As a consequence, our cars will be equipped more and
`more with
`digital
`systems
`communicating
`and
`exchanging information. Whenever possible, the trend is
`
`Safety
`Security
`
`Road-side
`assistance
`Mayday
`
`& Entertainment
`
`Radio
`(AM/FM/DAB)
`
`Information and
`Communication
`
`Internet access
`
`E-mail
`Audio (cassette,
`Panic call
`
`CD player, MP3)
`S-DARS
`Weather forecast
`Collision
`Head-line News
`avoidance
`Stock quotes
`
`Mobile . hone
`
`Video (TV, DVD)
`Antitheft system
`
`information
`Games
`Navigation
`
`Traffic
`information
`Tollin- s stem —
`
`Car dia- nosis
`
`Table 1.0: Near-Future Vehicle's Features
`
`This paper assesses the suitability of current mobile
`multimedia transport for the accommodation of these
`technological advances. In addition, this paper identifies
`system and functional requirements for future mobile
`multimedia transport as well as differences between
`existing networks such as Ethernet, IEEE 1394, Media
`Oriented System Transport (MOST).
`
`CURRENT MOBILE ARCHITECTURE
`
`the primary
`In the beginning of the automobile era,
`function of a vehicle was a reliable transport. Over a
`period of years. the desire for a basic transport has been
`
`

`

`features by allowing for the vertical integration of analog
`audio and adding a digital control. Despite the digital
`nature of most of
`the new added modules and the
`introduction of Digital Signal Processor (DSP) within the
`mobile multimedia system,
`the vertical
`integration of
`audio and its transport remained analog.
`
`The Current mobile architecture with its analog transport
`has the following limitations:
`
`0
`
`subsystem
`the
`complicates
`transport
`- Analog
`interconnects, decreases reliability, adds weight and
`cost. Two twisted pairs are required for cabin media,
`one twisted pair is required for voice module, one
`twisted pair is required for cellular phone module,
`one twisted pair is required for navigation module, 3-
`5 wires are required for low speed multiplex scheme
`and synchronization. Additionally, two more twisted
`pairs are required for an optional media player such
`as CDJ. In the case of rear seat entertainment, more
`wires or coaxial cables are required for video. The
`number of wires or coaxial cables required for video
`applications is proportional
`to the number of rear
`seat occupants. Moreover, these wires require wide
`connectors with more pins at
`the analog power
`amplifiers input connector.
`
`.
`
`The module level expansion strategy and vertical
`integration of analog audio resulted in a closed
`architecture with limited expansion path. The number
`of pins available at
`the power amplifiers input
`connector bounded the expansion path.
`In addition,
`
`life and is not
`the new subsystem had a short
`compatible with
`the
`digital
`trending of
`future
`entertainment features such as digital audio. digital
`video, Digital Audio Broadcast
`(DAB) and next
`generation of compact disc technology. Digital
`Versatile Disc (DVD).
`
`The module level expansion strategy and vertical
`integration of analog audio resulted in
`costly
`subsystem architecture. Often, modules added to the
`subsystem exhibited wasteful redundant hardware
`resources in order to achieve compatibility with an
`analog architecture. An example of this hardware
`wastefulness is the addition of a digital-to-analog
`converter to the output of CDJ or CD player to
`achieve compatibility with analog integration and
`processing. Moreover,
`the hardware
`resources
`available for each module are for that modules’ own
`use and can't be shared with other subsystem’s
`modules. This will add cost to each module and will
`contribute to the overall cost of the subsystem.
`
`the module level expansion strategy and
`Presently,
`vertical integration of analog audio has reached its upper
`integration limit for an HMI, AM-FM tuner, voice, cellular
`phone. CDJ, a media player, Steering Wheel Control
`(SWC), a Bear Integrated Control Panel
`(FIICP) and
`navigation. The implementation of such a subsystem
`requires seven modules and a minimum of 31 wires. A
`saving of one module and four wires is possible if a
`media, HMI, tuner, and power amplifier is packaged in
`one module. However, a complicated heat dissipation
`
`N
`
`ti
`nig- nn
`
`Steering Wheel
`C Ito]
`
`
`Heed-phone Module
`Elm -
`
`
`
`E?
`
`
`Dlgltn] Cont-ml
`
`Rear
`Seat
`Entertalnment
`
`Figure 1.0: A Block Diagram of A Typical Entertainment System
`
`and overall power management strategy is requlred for
`the success of such integration. and may limit the audio
`performance.
`
`REQUIREMENTS FOR NEW TRANSPORT
`
`Vehicles are running out of the real estate required to
`house new modules.
`Interconnect harness thickness
`and costs are ever
`increasing.
`This section is a
`discussion of both system and functional requirements
`for a new transport:
`
`0 Open Standard: the new transport shall be an open
`standard to ensure that all vehicle and electronic
`product manufacturers have equal access to the
`standard and to the market.
`
`. Minimal Standard: the new transport standard shall
`be extensible and capable of migration to future
`technologies and different physical media with no
`impact on application software. A layered approach
`to the protocol, such as is used in the reference
`model of Open System Interconnect (OSI) model,
`shall be used.
`
`.
`
`0
`
`to allow
`digital
`shall be
`The new transport
`multiplexing of data, control, audio and video over
`the same media. In addition, digital transport enables
`open system architecture; a node can be added at
`anytime during the vehicle‘s life without modifying an
`existing node’s
`connector. Moreover,
`a digital
`transport enables the natural transport of digital data
`from one ever
`increasing digital application to
`another without
`exhibited wasteful
`redundant
`hardware resources such as Digital-to Analog (DAC)
`and Analog-to-Digital (ADC) converters.
`
`Safety & Security: the new transport shall provide an
`environment
`into which devices can be plugged,
`unplugged, and operated in a vehicle in a manner
`which does not threaten the integrity of the vehicle or
`the devices that are being used. The new transport
`shall
`include
`suitable
`security for
`transactions
`involving the exchange of money and/or proprietary
`information.
`
`the new transport compatible
`- Manufacturability:
`devices and software shall be easy to design,
`implement and integrate. This
`implies minimal
`specifications and unambiguous requirements. Ease
`of manufacture helps to assure the wide adoption of
`the standard and contributes to lowering the cost of
`the manufactured items.
`
`- Ease of Use: the addition or removal of devices shall
`
`the incremental direct material cost to
`Low Cost:
`implement a new transport node shall be a fraction of
`the cost of the application supported.
`
`Graceful Degradation: the new transport’s physical
`layer and its supported bus
`topology shall be
`designed such that any fault shall not cause any
`damage to the cable, the vehicle or any attached
`devices.
`Functional operation under any of these
`conditions may cease but shall resume within the
`boot/discovery time after the fault is removed. The
`new design shall be such that no single failure, other
`than a fault in the physical layer as described above,
`or the loss of primary power, shall cause the entire
`system to fail.
`
`It shall be possible to attach
`Hot Plug and Play:
`devices to, and remove devices from,
`the new
`transport system at any time, whether power is on or
`off.
`
`Self-Identification: devices attached to the new
`transport shall be able to identify themselves to other
`system devices and shall self configure to obtain
`unique addresses on the new transport. No user
`intervention shall be required to complete this
`configuration other
`than the
`provision
`of
`the
`appropriate application software.
`
`Short Boot/Discovery Time: the new transport and all
`attached nodes shall complete self-configuration
`within one second after device initialization. When a
`new node is added to the system while it
`is
`operating,
`detection
`of
`the
`new node
`and
`reconfiguration of the system to include it shall be
`completed within 2 seconds.
`
`the set of devices
`Peer-to-Peer Communications:
`that is likely to be attached to the new transport is
`unpredictable and no single device is guaranteed
`always to be present. Therefore, a device on the new
`transport shall be able to communicate directly with
`any other device in the system and the vehicle
`without need for any additional device. An application
`that
`is implemented across multiple devices may
`choose to implement a "master controller" for that
`application but this shall not be a requirement for all
`applications.
`
`Automotive Physical and Electrical Specifications:
`the new transports physical
`layer
`shall meet
`automotive environmental
`(temperature, vlbration.
`shock,
`EMI.
`etc.)
`and electrical
`specifications
`(reverse
`voltage.
`load
`dump,
`etc.)
`for
`the
`
`
`
`

`

`new
`and
`technologies
`evolving
`accommodate
`applications. A layered approach to the protocol is
`required to guarantee minimum impact on existing
`designs as new physical layers and new applications
`are developed.
`
`Security and Authentication Services: The new
`transport
`protocol
`shall
`provide
`security
`and
`authentication
`sen/ices
`for access
`to
`vehicle
`functions. It is anticipated that additional services will
`require
`additional
`security
`measures
`to
`accommodate
`applications or
`transactions
`that
`require
`billing,
`authentication,
`confidentiality,
`confirmation, non-repudiation, etc.
`
`Bit Error Rate: the bit error rate shall be less than 1 x
`10*. Applications requiring better than this shall be
`able to implement appropriate measures at higher
`layers of the protocol.
`
`Tlme-Crltlcal Delivery of Packets- Deterministic
`Latency: Some
`applications may require
`that
`messages be transmitted within a given time period.
`The new transport shall be deterministic and it shall
`be possible to determine the maximum latency for
`any message in a given system configuration
`
`Private Message Service: Equipment manufacturers
`wish to be able to develop applications that span
`their own suite of products and provide competitive
`functions and features not achievable when products
`from different manufacturers are interconnected. The
`new transport
`protocol
`shall
`support
`the
`implementation of private messages that will allow
`such applications to be developed.
`
`the new transport gateway shall
`Power Loading:
`make power available for all devices connected to
`the new transport system. The total operating current
`drain of all devices connected to the new transport
`shall not exceed the capacity of the gateway unless
`power is routed directly to the device from another
`source or the device is self-powered (e.g.,
`internal
`batteries).
`
`lntemetworking: it is anticipated that wireless access
`to and from the Internet will be required for many
`devices attached to the new transport system. The
`new protocol
`shall
`not
`preclude
`the
`future
`implementation of intemetworking services across
`multiple gateways or bridges between the new
`transport and other subnets such as Intelligent
`Transportation Data Bus (IDB).
`
`Explicit Device Addressability: it shall be possible to
`
`address to deliver this message and all other devices
`shall ignore it.
`
`it shall be possible for an
`Broadcast Messages:
`application to generate a broadcast message to all
`devices connected to the new transport without
`having to address each one explicitly. A broadcast
`message may or may not require acknowledgment
`or
`confirmation of delivery. For
`example,
`an
`application
`may
`require
`confirmation
`(acknowledgment) that at least one receiving device
`capable of acting on the message has received the
`message.
`
`Consumer-Friendly Device Connection (No Special
`Tools Required): in most cases, it shall be possible
`for a consumer to install new transport nodes with
`common hand tools. There will be cases where
`professional
`installation may be required, but this
`should be the exception, not the rule.
`
`Wake/Sleep: any node shall have the ability to wake
`up the new transport system or pot it to sleep.
`It
`shall be possible to wake up the nodes by sending a
`wake-up message, such as a pager.
`It shall be
`possible to put any node and all connected devices
`back to sleep with a sleep message. Absence of
`message traffic on the new transport for more than
`30 minutes shall cause the new transport and all
`attached devices to go to sleep.
`
`Priority Sensitive Flow (lsochrono‘us): consumer
`electronics devices such as video games. DVD and
`MP3 players. Dolby AC-s audio components, etc.,
`may require support for high speed isochronous data
`communications (i.e.. data packets delivered at a
`guaranteed rate in a guaranteed order).
`
`Data Types: The new transport protocol-shall allow
`any data type (ASCII. binary, bulk, etc.)
`to be
`transmitted in a message without need for any
`special escape characters or other similar artifacts
`added by the application.
`
`Fair Access to the new transport system: no single
`device shall be allowed to monopolize the new
`transport system.
`
`the new transport
`Priority Flagging:
`Message
`protocol shall provide a means to specify that the
`current message is a high or normal priority
`message.
`The
`protocol
`simply
`provides
`a
`mechanism to identify the message as a high priority
`message.
`It
`is up to'the device manufacturer to
`determine whether support is provided and up to the
`
`required: a
`if
`0 Confirmation of Message Delivery,
`device sending a message to another device shall be
`able to explicitly request confirmation of error-free
`delivery of that message from the receiving device.
`
`EXISTSING PROTOCOLS: ETHERNET, IEEE 1394
`AND MOST
`
`Fast Ethernet
`
`Ethernet is a term commonly used to describe a variety
`of network implementations that share the same basic
`technology. Some early varieties of Ethernet are
`1OBase-2 and 1OBase-5 which are also called ‘thin nef
`and ‘thick net' respectively. All nodes on such network
`tap into a single cable. A later version of Ethernet,
`toBase-T, introduces the concept of a hub or a repeater.
`All nodes are connected directly to a single repeater,
`which simplifies cabling and provides buffering of
`electrical signals.
`
`A newer version of Ethernet, called Fast Ethemet,
`operates at 10 times the speed of 1OBase-T or 100 Mbit
`per second. It has the same star topology as toBase-T
`and comes in a few different versions called tooBase-
`TX,
`1OOBase-FX, and 1008ase-T4. The difference
`between these versions is the physical layer which is
`electrical for TX and T4, and optical for FX. In this paper,
`Fast Ethernet will refer to the most common version
`100Base-TX.
`
`Ethernet Topology
`
`Thick net Ethernet uses a single cable as a backbone for
`the network. Each node in the network taps into this
`cable through what is called a T connector. In an office
`environment,
`this cable could be routed through the
`ceiling with taps dropping into each office. This works
`well except It is difficult to add new users to the network
`and signal quality is sometimes difficult to control.
`
`A Fast Ethernet uses a central repeater which connects
`directly to each node. This star topology enables the
`signal quality on the transmission line between a
`computer and the repeater to be well controlled and
`provides a relatively simple means to add users. The
`repeater has a number of ports which connect to each
`computer on the network. To add another computer, a
`wire is run from a free port on the repeater to the new
`computer.
`if there are no free ports,
`typically, another
`hub (or switch) can be connected to an uplink port to
`expand the network to virtually any size.
`
`In a thick Ethernet. all computers are connected to one
`coaxial cable. This cable is used for sending and
`receiving messages. When the bus is idle, the voltage on
`the coax cable remains in a high impedance state at an
`intermediate level. This level is not a one or a zero, so
`that all nodes can easily determine if the bus is idle.
`When a node is transmitting, the voltage on the bus is
`pulled to high and low voltages depending on the data to
`be transmitted. Only one computer can send information
`at any one time.
`If multiple nodes try to transmit at the
`same time, a collision occurs,
`the data from both
`computers is corrupted, and both computers have to stop
`transmitting and try again when the bus is idle.
`
`In Fast Ethernet. all nodes connect to a central repeater
`through two sets of
`twisted pairs. One pair
`is
`for
`transmitting and the other for receiving. Although each
`node has its own cable. the network operates exactly like
`thick Ethernet at a higher level. When a computer sends
`a message to the central repeater, the repeater sends
`the data exactly as received to all other computers
`connected to it. Again, if two computers try to send at the
`same time, a collision occurs. and both computers must
`try again later.
`
`Ethernet Arbitration
`
`the computers on an Ethernet share the
`Since all
`transmission media, only one computer can send
`information at any one time.
`If multiple nodes try to
`transmit at the same time, they must arbitrate for use of
`the bus. The rules that every node follows is called
`Carrler Sense Multiple Access with Collision Detection
`(CSMA/CD). Before a node sends a message on the
`bus.
`It sends a stream of one‘s and zero's called a
`'carrier‘. All other nodes on the network sense this carrier
`and do not attempt to send their own message until the
`original node completes its message.
`
`There is a finite amount of time from when a node begins
`sending a carrier to when all nodes detect this carrier.
`During this time, other nodes may attempt to send a
`carrier.
`If this happens. a ‘collision’ will occur which
`corrupts the data on the bus. Transmitting nodes will
`both ‘detect’ this condition and stop transmitting. The
`rules then specify that each node must wait a random
`amount of time before attempting to transmit again. The
`probability that another collision will occur is low.
`
`Ethernet Switches
`
`Ethernet switches have become more popular than
`repeaters in recent years. An Ethernet switch is a more
`
`

`

`If
`operates just like the coaxial cable in thick Ethernet.
`two computers try to send messages at the same time,
`the messages collide.
`
`switch divides the network into many
`An Ethernet
`collision domains. Each port on a switch is a different
`collision domain. For example,
`if one computer
`is
`connected to a port on an Ethernet switch, the collision
`domain consists of two nodes;
`the computer and the
`switch. If a port on a repeater is connected to a port on a
`switch, the collision domain consists of the switch and all
`computers connected to the repeater.
`
`The switch has intelligence which learns the addresses
`of
`the computers connected to each pelt. When a
`message is received at one port, the destination address
`is determined and the message is
`sent out
`the
`appropriate port. Switches are typically much more
`efficient than repeaters. However. they cost more.
`
`Ethernet Communication Mechanism
`
`Information in Ethernet networks is communicated in
`packets. Each packet consists of a header, usable data
`and a checksum. The header contains information such
`as source and destination address, the length of usable
`data and possibly information about the message type.
`The checksum is a code sent at the end of the message
`so that the receiving node can determine if the packet
`was corrupted during transmission.
`
`Since the header and checksum are only used to send
`the packet safely from the transmitting node to the
`receiving node,
`it is considered network overhead.
`It is
`not
`information usable by the application.
`In Fast
`Ethernet,
`this consumes 18 bytes.
`If you include the
`arbitration time,
`the total overhead is 38 bytes.
`In
`addition,
`the minimum usable data is 46 bytes per
`packet. Even if you only wish to send one byte, you still
`must send the 1B—byte header, 46 bytes of data, and wait
`20 bytes worth of time for the bus.
`
`The efficiency of the network can be defined as the
`number of user data bytes per packet divided by the
`number of bytes in the packet plus the overhead of
`waiting for the bus. If only one byte of user data is sent
`per packet the efficiency is 1/(64+20) X 100% = 1.2%.
`Since the maximum user data per packet is 1500 bytes
`in 1008aseT,
`the theoretic maximum efficiency is
`1500/(1518+20) X 100% = 97.5%.
`
`particular Wav file from the server can be played on the
`sound card in the following way. The client software on
`the PC manages a FIFO (First-In First-Out) which
`continually outputs audio data to the sound card. When
`the FIFO gets close to being empty, the client sends a
`message to the sewer to send more data. The server
`sends another packet to the client to fill the FIFO up
`again. As long as the server sends the new packet
`before the FIFO empties, audio can be heard.
`If the
`server responds slowly or the network is very busy, the
`packet may not arrive in time, the FIFO will empty and
`sounds will momentarily stop. This is unacceptable for
`most audio applications.
`
`There is a trade off between FIFO size, the frequency of
`requests for more data and packet size. Since the large
`packets are more efficient
`than small packets
`(%
`overhead lrorn header, etc), let’s assume we will use the
`largest packet size; 1500 bytes of user data. If the audio
`sample rate is 43 kHz, and the audio is 16 bits/sample
`stereo, then we need an average of 192K bytes/second
`or 128 packets/sec. The overhead for
`the header,
`checksum and the required idle between packets is 38
`bytes. Since the client software on the PC with the sound
`card must inform the sewer when the FIFO is nearly
`empty, there are another 84 bytes of overhead to send
`this message to the server.
`
`The minimum total bandwidth required for one audio
`channel is:
`
`1500 + 38 + 84 = 1622 bytes/packet
`
`1622 bytes/packet * 128 packets/sec = 207616 bytes/sec
`
`207616 bytes/sec ' B Bits/bytes = 1.66 Mbit/sec
`
`Since the packet size in this example is the largest
`allowed by Fast Ethernet, the network overhead is small
`compared
`to
`the
`audio
`data
`throughput. The
`disadvantage of the large packet size is the buffer size
`requirement in the Client.
`It must be 1500 bytes deep
`plus more for handshaking. If the extra depth is not large
`enough for the network to guarantee another packet will
`arrive prior to the buffer emptying, a loss of audio quality
`may occur.
`If the buffer empties, the audio stops.
`In a
`Fast Ethernet,
`it
`is
`impossible to guarantee any
`bandwidth. If the network has lots of traffic, you may not
`even be able to get the 1.66 Mbit/sec average throughput
`that is required. More commonly, at times of high traffic
`the buffer may empty no matter how large it may be.
`
`(CSFt) Architecture for Microcomputer Buses formally
`adopted as ISO/IEC 13213 (ANSI/IEEE 1212).
`
`This architecture defines a set of core features such as
`node architecture, address space, common transaction
`types, Control and Status Registers (CSR), configuration
`ROM format
`and
`content. message
`broadcast
`mechanism to all nodes and interrupt broadcast to all
`nodes.
`IEEE 1394 specifies how units attached to a
`serial bus can talk to each other, but does not define the
`protocols used to communicate between the nodes.
`
`IEEE 1394 is similar to Fast Ethernet in many ways. Data
`is always communicated between nodes in packets.
`If
`multiple nodes try to send packets at the same time, they
`must arbitrate for the bus. The information in the packet
`headers, the packet sizes and the arbitration method, are
`different. However,
`the fundamental mechanisms are
`similar.
`
`The most significant feature that IEEE 1394 provides
`(which Fast Ethernet does not) is guaranteed bandwidth
`for
`real
`time
`applications. These applications are
`allocated isochronous bandwidth which enable real time
`data to be communicated in packets sent at regular time
`intervals. This is an improvement over Fast Ethernet.
`However,
`it will be shown that there are still serious
`limitations.
`
`The raw bit rate for IEEE 1394 is defined to be selectable
`between approximately 100, 200, and 400 Mbit/sec.
`Work is currently being done on an 800 Mbit specification
`as well. Silicon is currently being advertised to run at
`both 100 and 200 Mbit. However, most implementations
`are now at 100 Mbit/sec which is the same data rate as
`the large installed base of 1OOBaseT. Gigabit Ethernet is
`currently under development as well.
`
`IEEE 1394 Physical Layer
`
`The physical layer for IEEE 1394 consists of two sets of
`twisted pair wire for signals and two wires for power and
`ground connected between each pair of nodes. One set
`of twisted pairs is called data and the other is called
`strobe. When one node begins to send a packet, it sends
`Nonretum-to-Zero (NFIZ) data on the data line and
`transitions the strobe line only between consecutive 1's
`or 0’s. Both sets of twisted pairs are bi-directional. Each
`node sends and receives data on the same sets of wires.
`When neither node is sending data, the twisted pairs are
`held in a high impedance state.
`
`that the transmission line can be longer. The maximum
`cable length for Fast Ethernet is 100meters, while it
`is
`only 4.5 meters for IEEE 1394.
`
`Each node on an IEEE 1394 bus (or in a Fast Ethernet)
`has its own timing source which is typically a crystal
`oscillator. This timing source is used by a node to
`transmit data and is used by a node to over sample the
`received data and strobe lines to recover the data. This
`means that all IEEE 1394 nodes are asynchronous at the
`lowest
`level. The accuracy of the timing reference in
`IEEE 1394 is specified to be i 100 PPM which is
`typically the frequency tolerance of widely available
`crystal oscillators.
`
`The nominal data rate in 100 Mbit IEEE 1394 is 98.304
`Mbit.
`If the crystal oscillator at a particular node is
`operating at the high end of its frequency tolerance, it will
`be able to transmit data at 98.304 Mbit +100 PPM or
`98.314 Mbit/sec.
`If it is operating at the low end of the
`frequency tolerance,
`it will
`transmit data at 98.294
`Mbit/sec. This may seem like a trivial issue. However,
`the section on system timing will
`illustrate
`some
`important consequences for real time applications.
`
`IEEE 1394 Topology
`
`The physical topology for a typical IEEE 1394 network is
`a tree structure. Typically, a node will have a least two
`ports which enables multiple nodes to be daisy chained
`together.
`If a node has more than two ports, multiple
`branches can be created. During initialization, one node
`is defined to be the root node with all nodes extending
`down different branches. The topology can have any
`number of branches if no loops are created.
`
`The tree topology, with the ability to daisy chain nodes,
`has the advantage of simplicity for small networks. If you
`have a few devices, it is easy to plug them together in a
`daisy
`chain. However,
`large
`networks
`can
`be
`cumbersome, particularly if network performance is
`optimized. To improve performance,
`it
`is desirable to
`minimize the propagation delay of data between any two
`nodes in the network. Long daisy chains can be split into
`many branches to reduce the delay; however, care
`should be taken to balance the length of the branches.
`
`In a Fast Ethernet, a repeater or switch is required for a
`network with more than two nodes. This makes small
`networks complicated. However, it makes large networks
`simpler.
`
`The maximum efficiency is never achieved since many
`collisions will occur if there is a lot of activity on the bus.
`The effect on efficiency is difficult to predict.
`
`IEEE 1394
`
`In contrast, the physical layer for Fast Ethernet consists
`0f
`two sets of
`twisted pairs; one pair
`is used for
`
`IEEE 1394 Arbitration
`
`
`
`

`

`this
`condition is detected, and the nodes begin to arbitrate for
`the bus. Likewise, all nodes on an IEEE 1394 network
`have the same collision domain. Only one node can send
`a message at one time.
`If multiple nodes try to send
`messages at the same time, only one node will gain
`control of the bus.
`
`if a collision between two nodes
`In a Fast Ethernet,
`occurs, both nodes will stop transmitting and wait a
`variable amount of time before another attempt.
`If they
`both happen to wait the same amount of time, another
`collision will occur. Mechanisms are built into the network
`to minimize the probability of nodes colliding more than
`once or twice.
`
` the same time, the transmitted data is corrupted,
`
`An IEEE 1394 network operates differently. Nodes which
`are closer to the root node have a higher natural priority.
`When two nodes attempt to transmit at the same time,
`the node with the higher priority wins arbitration and
`control of the bus.
`in order to prevent higher priority
`nodes from monopolizing the bus a fairness interval is
`defined. During a faimess interval, all nodes are given
`the opportunity to send one message.
`
`Unlike Fast Ethernet the arbitration mechanism in IEEE
`1394 is deterministic. There is zero probability that nodes
`will collide many times before successfully sending a
`message. This is important for the delivery of real time
`data since any unpredicted delay, no matter how
`unlikely, may cause buffers to overflow or underflow.
`The arbitration process in IEEE 1394 consists of bus
`requesthrant handshaklng between child and parent
`nodes. A parent node is defined as the node on a 1394
`cable which is closer to the root. The node which is
`further from the root, is called the child.
`If a child and a
`parent both request the bus at the same time, the parent
`will block the child's request and send its request to its
`parent. The request continues down the tree until
`it
`reaches the root node. The root node issues a bus grant
`which travels back up the tree to the node requesting
`control of the bus. Once a node receives a bus grant, it
`can begin sending a message.
`
`The time that it takes to arbitrate for the bus depends on
`the size of the network. A bus request from a node at the
`end of a number of branches must propagate down the
`tree to the root and back up the tree to the requesting
`node. lnfonnation must be sent down all other branches
`to prevent any other node from driving the bus (through a
`bus request), until the granted message is sent. The total
`arbitration delay for a network with N hops (from parent
`to child or child to parent) between the furthest two
`nodes,
`is about N x 80ns in a 100 Mbit IEEE 1394
`
`IEEE
`While Ethernet treats all packet data the same,
`1394 provides different types of packets. The primary
`packet
`types are
`asynchronous
`and lsochronous.
`Asynchronous packets are functionally equivalent
`to
`Ethernet packets. lsochronous packets are only available
`in IEEE 1394, and provide guaranteed bandwidth to time
`critical applications.
`
`IEEE 1394 Asynchronous Packets
`
`An asynchronous packet consists of a header, a header
`checksum, user data and a user data checksum. The
`header contains information such as source address,
`destination address, message length, message type, etc.
`The size of the header and the checksums is typically 24
`bytes long. The user data can be up to 512 bytes long in
`a 1394 network operating at 100 Mbit/sec.
`
`When a particular node receives a message, an
`acknowledgement signal
`is automatically sent back to
`the sending node. If the