Protocols For Aerospace Control Systems: A Comparison of Afdx, Arinc 429, Can, and TTP
Protocols For Aerospace Control Systems: A Comparison of Afdx, Arinc 429, Can, and TTP
Protocols For Aerospace Control Systems: A Comparison of Afdx, Arinc 429, Can, and TTP
Current developments and future trends are having enormous effects on the aerospace industry.
While the life cycle of aircraft is increasing, the life cycle of their electronic components is
decreasing. As more and more electric applications are finding their way into airplanes, the
complexity of the overall system is growing. But all-electric aircraft require higher levels of safety,
reliability and availability. Additionally, the cost pressure for development and operation is
constantly rising. This calls for reusability of software and hardware components and its
certification basis across several aircraft projects – both commercial and military.
TTTech provides a well-proven solution to these challenges. The Time-Triggered Protocol (TTP)
represents a communication platform for true modular avionics/controls (distributed or backplane).
This fault-tolerant deterministic data bus is supported by an integrated tool chain and dedicated
communication controllers. A network designed according to the principles of the time-triggered
technology guarantees new levels of safety at lower overall cost. The features of TTP translate
directly into benefits such as reduced wiring and weight, lower total life cycle costs, shorter time-to-
market, ease of system integration and system upgradeability, as well as increased reliability of the
overall system.
The multitude of embedded controllers and communications equipment found on aircraft need to
communicate with each other to collect data, pass commands, and health monitoring. Currently, a
move from largely independent, federated subsystems to more integrated system designs can be
observed throughout the industry. The increased amount of information shared on the data bus
within and between systems and subsystems, with many of the shared parameters being vital for
the operation of the aircraft, asks for airborne networks with high throughput, availability, and
reliability. The throughput, however, will always come second to the need for reliability.
More advanced data bus architectures such as ARINC 629 are bidirectional. ARINC 629 is applied
on Boeing 777 and some Boeing 737 models. Here, any member or user of the system can
transmit, receive, or both on a common, shared data bus. As the implementation of the physical
layer is comparatively expensive, market penetration did not happen. More than 30 years ago,
especially military avionic systems, which previously used discrete wiring, were upgraded to
interface with the MIL-STD-1553B bidirectional data bus. This led to a significant weight saving,
and the labor to run wiring harnesses could be reduced by thousands of hours. However, the bus
control became more complex, and the engineering effort to integrate the system was no small
task.
The advantage of the central control or command/respond approach is that only one bus
component has control of the bus traffic (master). All users are directed by the bus controller
(master). If the data bus architecture has to change, only the bus controller (master) has to be
modified to support the new configuration. But the most significant weakness of such architecture
in an aircraft environment, where the bus controller or network failure could have catastrophic
repercussions, is that the bus controller (master) represents a single point failure. When it fails, the
entire system fails. Therefore, many systems based on MIL-STD-1553B are equipped with
redundant bus controllers, only one having control of the network at a time.
In a fully distributed control system, all members of the network are in charge of their own access.
Single fault tolerance can be achieved when appropriate fault handling mechanisms are
implemented as well as true partitioning is in place (both within and between Line Replaceable
Units, LRUs) to prevent fault propagation. So if any member fails or happens to operate erratically,
the rest of the system will not be affected and will continue operating correctly. The weakness of
the distributed control is exactly the opposite of the central control strength. It generates higher
complexity in bus access sharing, and a change to the system configuration may require a change
to every user of the system.
A contention protocol like Carrier Sense Multiple Access with Collision Detection (CSMA/CD) for
Ethernet allows any bus user to start transmitting at any time, provided the bus is idle at the
moment. But if two users start transmitting at the same time, a collision occurs. This is not an
efficient use of the bandwidth and the throughput suffers during heavy traffic. The time slot
allocation, on the other hand, assigns a specific time slot for every user. The user listens to the bus
for inactivity and if the bus is inactive during the user’s assigned slot, the user may begin to
transmit. If the bus is busy, access is not attempted until the next time slot. Some protocols may
decide to proceed quickly to the next time slot, if the owner of the current time slot did not start to
transmit anything. In a token ring protocol, the user is allowed to transmit only after the node has
received a unique bit pattern called token.
Some CSMA/CD protocols like Ethernet need collision detection to determine whether or not their
attempt to transmit was successful. The users monitor their own transmissions, comparing them to
the transmitted data internally stored. If they detect data corruption, they stop transmitting. Then,
when the users detect that the bus is idle again, they wait for a random time and, provided the bus
is still idle, attempt to access it again. The reason for the random delay is to avoid permanent
collisions.
Another problem that must be addressed is propagation delays on the transmission line. One
station/user waiting to transmit may see an idle bus while another user is already transmitting.
Random time delays alleviate this problem, but the result is poor bus utilization and severe
throughput problems under heavy traffic.
The strive for better throughput of avionic networks resulted in sophisticated time allocation
methods and increase of data bandwidth. Generally, the bus control methods are so critical that
they are made transparent to the designer of a user terminal by providing him with a dedicated
chipset and couplings to build the interface. Token ring protocols are not commonly used in avionic
systems but do exist.
ARINC 429
ARINC 429 was developed and is maintained by the Airlines Electronic Engineering Committee
(AEEC) comprising members that represent airlines, government, and Aeronautical Radio,
Incorporated (ARINC). The latter is a major company that develops and operates systems and
services to ensure the efficiency, operation, and performance of the aviation and travel industries.
It was organized in 1929 by four major airlines to provide a single licensee and coordinator of radio
communications outside the government. Only airlines and aviation-related companies can be
shareholders, although all airlines and aircraft can use ARINC’s services.
ARINC 429 is a specification that defines, among many other things, how avionics equipment and
systems should communicate with each other. They are interconnected by twisted pair wiring. The
specification defines the electrical and data characteristics and the protocols that are used.
ARINC 429 employs a unidirectional data bus standard known as Mark 33 Digital Information
Transfer System (DITS) and can be used in an unidirectional multidrop system topology (see
Fig. 1). Messages are transmitted at a bit rate of either 12.5 or 100 kbit/s to other system elements,
which are monitoring the bus messages. Transmission and reception are on separate ports so that
many wires are needed on aircraft equipped with a large number of avionics systems.
ARINC 429 has been installed on most commercial transport aircraft of leading manufacturers
including Boeing and Airbus. A newer system specified as ARINC 629 was installed on the
Boeing 777 for flight controls, and some aircraft use alternate systems in an attempt to reduce the
weight of wire needed and to exchange data at a higher rate than is possible with ARINC 429. The
unidirectional ARINC 429 system provides high reliability at the cost of wire weight and limited data
rates.
Key Attributes
Unidirectional data bus standard MARK 33 Digital Information Transfer System (DITS)
Bit rates: high speed 100 kbit/s, low speed 12.5 – 14.5 kbit/s
Encoding: return to zero bipolar tri-state modulation
Message length: 32-bit word, 255 word data block in block transfer mode
Classes of service: periodic, sporadic and file transfer
Media access: simplex single source multiple sink plus full duplex RTS/CTS handshake
Topology: single source multiple sink
Media: 78 Ohm unbalanced shielded twisted pair copper cable
Number of nodes: 20 sinks, 1 source
CAN
Controller Area Network (CAN) is a serial data communications bus developed in the mid-eighties
by Robert Bosch GmbH for the German car industry. It has since been adopted worldwide for
automobile data communications, and in a wide range of other applications. The principal needs
which drove the development of CAN were similar to those which spawned avionic data buses, i.e.
to provide real-time data communication, with reduced cabling size and weight, and standardized
I/O specification.
CAN belongs to the class of event-triggered protocols where the temporal control signals are
derived primarily from events occurring outside or inside the computer system. Messages are sent
if the host computer requests the transmission of a message, the channel is idle and the message
priority wins over the messages that other nodes intend to send at the same time. CAN is deployed
in a bus topology (see Fig. 2).
There are two versions of CAN: CAN 1.0 and CAN 2.0. Most CAN controllers now in production
are CAN 2.0, which is divided into two parts, known as A and B. Part A has a standard 11-bit
identifier field, no specification for message filtering, and layered architecture based on Bosch’s
internal model. Part B extends the identifier field to 29 bits, has some message filtering, and
supports the layer description of the Open Systems International (OSI) model. CAN 2.0 has been
formalized in two ISO standards, ISO-11898 for high-speed applications (125 kbit/s to 1 Mbit/s)
and ISO-11519 Part 2 for low-speed applications (125 kbit/s).
CAN is subdivided into different layers: the object layer, the transfer layer, and the physical layer.
The CAN specification focuses on the first and second layers of the OSI model (data link layer and
physical layer). A variety of protocols are used between the CAN data link layer and the various
application layers. CAN implements five error detection mechanisms: Cyclic redundancy checks
(CRC), frame checks, acknowledgment error checks, bit monitoring, and bit stuffing.
In the aerospace industry many different flavors of CAN are being used in order to compensate for
the lack of determinism and fault tolerance.
Key Attributes
Multimaster priority-based serial communications protocol supporting distributed real-time control
and multiplexing using non-destructive contention-based arbitration
Bit rates: 1 Mbit/s (with 40m bus), 100 kbit/s (with 500m bus)
Encoding: non return to zero (NRZ)
Message length: 0 to 8 bytes
Classes of service: periodic and sporadic
Media access: carrier sense multiple access with collision avoidance (CSMA/CA)
Topology: terminated differential two wire bus
Media: screened or unscreened twisted pair or flat pair telephone cable
TTP
Time-Triggered Protocol (TTP) has initially been developed more than 20 years ago at the Vienna
University of Technology and is now being maintained by the international cross-industry
consortium TTA-Group since 2001. The protocol is at the heart of the Time-Triggered Architecture
(TTA), which is a distributed computer platform for highly dependable real-time systems in the
automotive, aerospace and other transport industry applications. A TTA system has effective
consistency services and error detection mechanisms implemented to provide a maximum degree
of fault tolerance, safety, and availability. Fault tolerance is dependent on the network topology
used.
TTA can be implemented in a bus or star topology with controllers as bus interface units (see
Fig. 3). The bus topology uses a broadcast bus where all nodes are electrically connected to each
other. The star topology has all nodes connected to each of the replicated channels of the
interconnection network via bidirectional links. In the bus topology each node is equipped with a
local bus guardian; in the star topology two redundant central bus guardians (CBG) are
implemented. A central bus guardian is a unit that acts as a failure mode converter to protect the
communication channels from temporal transmission errors.
TTP belongs to the class of the time-triggered protocols, where the temporal control signals are
solely derived from the progression of time. TTP was specifically designed for high-dependability
hard real-time applications, where timely error detection, fault isolation, and fault tolerance must be
provided. Media access is controlled by a conflict-free Time Division Multiple Access (TDMA)
strategy. Every node is assigned to a sending slot in a TDMA round. Additional slots can be
defined for nodes to be added later. Every TTP controller contains the dispatching table (MEssage
Descriptor List, MEDL) that contains the information which node is allowed to send which message
at a particular point in time. The MEDLs of a cluster are constructed before run time and are
common knowledge to all nodes. Every TTP controller contains two replicated communication
channels in order that the loss of one channel can be tolerated, but can also operate only on one
channel.
With TTP both event-triggered and time-triggered messages can be sent. The transmission of
event-triggered messages is performed over an event channel. Bandwidth is reserved for event
transmissions inside the TDMA slots, and the messages use identifiers. A typical event channel
mechanism is a CAN emulation, in which a CAN-compatible interface is provided and the CAN
messages are transmitted inside TTP data frames. Modularity of the TTP system is maintained
when using event transmission in this way because bandwidth is not arbitrated among different
nodes but only among different functions within a node. Timing and bandwidth analysis for event
transmissions is therefore done on a per-node basis and does not need system-level design.
With its deployment in a variety of aerospace applications, TTP is a key contender for next-
generation aircraft. Honeywell uses TTP for General Electric's F110 full authority digital engine
control (FADEC) system on the Lockheed Martin F-16 fighter aircraft and for Honeywell’s F124 full
authority digital engine control system on the Aermacchi M-346 fighter trainer aircraft. Honeywell's
APEX integrated cockpit with Nord-Micro has selected TTP as communication protocol for the
Airbus A380 cabin pressure control system. Alcatel uses TTP as field bus protocol for their railway
signaling system ELEKTRA 2, which will be used in Switzerland, Austria and Hungary.
Key Attributes
Real-time communication protocol for interconnection of LRUs in distributed real-time fault-
tolerant systems but also backplane (see FADEC)
Bit rates: currently available TTP controllers support up to 25 Mbit/s synchronous and up to
5 Mbit/s asynchronous transmission. Upper range is not limited by TTP itself and can be
extended with new implementations of the TTP communication controller.
Encoding: modified frequency modulation (MFM) and Manchester (both up to 5 Mbit/s), and MII
(up to 25 Mbit/s) with currently available TTP controllers
Message length: up to 240 byte with 4-8 bit header and 24-bit CRC
Media access: distributed Time Division Multiple Access
Topology: dual channel linear bus and star topology or mixed
Media: copper and optical fiber
Number of nodes: up to 64 nodes
Switched Ethernet is based on a star topology with full-duplex transmission, i.e. a node may send
and receive at the same time (see Fig. 4). All nodes are connected to a switch that forms the
center of the star. This switch distributes the messages in the system. If two or more nodes start to
transmit at approximately the same time, their messages will be buffered in the switch and relayed
sequentially. When the buffers are overrun and packages are dropped, timeliness can no longer be
guaranteed.
Key Attributes
Full-Duplex Switched Ethernet
Bit rates: 10/100 Mbit/s cross data rate
Encoding: PAM5x5
Message length: 64 – 1518 bytes
Classes of service: station to station, multicast and broadcast
Topology: dual redundant star network
Media: copper and fiber
Number of nodes: up to 1024 (without bridges)
Summary
This paper gives a comparative overview of four communication protocols used in avionics and
aerospace controls systems. From an assessment of their key features it can be concluded that
ARINC 429 is a well-defined, reliable, easy-to-implement, inexpensive protocol whose throughput
and reliability have been adequate for most applications in the past. More sophisticated protocols
are needed when large amounts of data have to be moved around quickly and reliably. CAN
seems to be most appropriate for avionic systems that have less stringent timeliness and reliability
requirements. If avionic systems require more throughput and reliability, then the TTP protocol
appears to be the more suitable choice at subsystem level. Future airborne system applications
will be most likely based on TTP and AFDX. These two data buses complement each other
because they concentrate on different application areas, with TTP focusing on aircraft subsystems
and general aviation and AFDX being optimized for higher level data transfer functions in large
transport aircraft. TTP is ideally suited for low-speed (up to 25 Mbit/s), low-cost, fault-tolerant,
robust applications and AFDX is best used for high-speed (up to 100 Mbit/s), fault-tolerant, robust
applications.