Overview 
ARCNET-The
Hidden Real-Time Network
ARCNET, once quite popular in office automation, has reinvented
itself into an embedded networking technology that is frequently
found in applications such as industrial control, building automation,
transportation, robotics and gaming. Extremely popular in Japan
with continuing popularity in America and Europe, ARCNET is now
making inroads into China with some of the top China universities
incorporating ARCNET into its projects.
Like Ethernet and Controller Area Network (CAN), ARCNET is a data-link
layer technology with no defined application layer. Designers
write their own application layer to meet their particular needs
and frequently do not advertise the fact that ARCNET is being
used in their product. ARCNET receives no name recognition, but
is frequently the network of choice in embedded applications.
It is hidden from the user, but with over 11 million ARCNET nodes
sold gives credibility that ARCNET is indeed popular.
Originally introduced at about the same time as Ethernet, ARCNET
incorporates a token-passing protocol where media access is determined
by the station with the token. When a station receives the token,
it can either initiate a transmission to another station or it
must pass the token to its logical neighbor. All stations are
considered peers and no one station can consume all the bandwidth
since only one packet can be sent each token pass. This scheme
avoids collisions and gives ARCNET its greatest advantage in real-time
applications-it is deterministic! By being deterministic, the
designer can accurately predict the time it takes for a particular
station to gain access to the network and send a message. This
is of particular importance for control or robotic applications
where timely responses or coordinated motion are needed.
CAN was originally designed for the automotive electronics market
and has experienced success as a device level network when used
in conjunction with higher layer protocols such as DeviceNet and
CANopen. ARCNET is best considered a controller level network
because of its higher performance over CAN. CAN communication
is limited to 1 Mbps and it can only send eight data bytes per
frame. Newer ARCNET chips can transmit at 10 Mbps and data packets
can be up to 507 bytes in length. Not only is ARCNET faster than
CAN, ARCNET can send more data per transmission.
Another advantage of using ARCNET is that it supports multiple
physical layers. The original physical layer was a dipulse transceiver
which was optimized for 2.5 Mbps. Newer generation transceivers
are much smaller in size and can operate up to 10 Mbps. A designer
can select popular and lower cost EIA-485 transceivers when distances
are relatively short. For higher isolation voltage, a transformer
coupled device is available that will also operate up to 10 Mbps.
The advantage of these transceivers is that they will operate
over a bus topology. One of the complaints regarding Ethernet
is the need to operate in a star topology which requires the use
of hubs. Not only does the hub require a source of power, the
designer needs to find a place to mount the hub and it represents
a single source of failure. For embedded applications, bus topology
is much more convenient since hubs are not required.
Another problem with Ethernet is the requirement to add a transport
layer protocol. Of course, the most popular transport layer protocol
is TCP but the code size for a TCP/IP stack is on the order of
50 kB. Small microcontrollers either lack the memory capacity
for this stack or the processing power to execute both the stack
and the application. Embedded applications do not require a full
seven-layer communication model. An application layer, data-link
layer and physical layer are all that is needed. Industrial automation
protocols such as DeviceNet and CANopen are prime examples. Their
data-link layer is CAN and transport functionality, such as, guaranteed
message delivery is handled at the application layer. The same
can be done with ARCNET. In fact ARCNET controllers have some
built-in transport layer functionality. All messages are appended
with a CRC-16 frame check sequence and if the frame was received
successfully at the destination station, an acknowledgement is
automatically sent back to the originating station without any
software intervention. This feature in ARCNET simplifies software
development and is absent in both CAN and Ethernet.
ARCNET has other unique features that facilitate embedded design
development. The COM20020 family of ARCNET controllers were designed
for a simple eight-bit bus microprocessor interface. The complete
communication protocol is handled by the ARCNET controller. This
includes token passing, network configuration and error handling.
Unlike Ethernet, ARCNET has built-in flow control. A message will
not be sent to a destination node without the availability of
a receive buffer. This further off loads the software requirements
and ensures that messages need not be resent because of receiver
errors. ARCNET controllers are competitively priced and there
are no licenses or expensive development platforms that must be
purchased. Only standard embedded design tools are required to
develop a system.
It is not only robotics where ARCNET is used. ARCNET is one of
the approved data links for the BACnet protocol. BACnet-Building
Automation and Control Network-was developed by the American Society
of Heating, Refrigeration and Air-Conditioning Engineers (ASHRAE)
and has now become an ISO standard. Popular in the United States
and Europe, BACnet is becoming popular in China as well and the
thick standard has been translated into Chinese.
If you ask a designer why they use ARCNET, they will usually say
it is simple to use, low-cost, extremely rugged and high performing.
Newer ARCNET controller chips have an extremely small footprint
making it ideal for embedded applications. Unlike TCP/IP/Ethernet
networks with complex 48-bit addressing and assignment, ARCNET
nodes can only number 255 and can be set by a simple 8 position
DIP switch.
The opportunities for using ARCNET are enormous. From Pachinko
machines and high-speed trains in Japan to process automation
and motor drive controllers in Europe. ARCNET continues to perform
in non-stop applications.
Features 
- Deterministic
Performance - Users Can Calculate the Worst Case Node to Node
Message Time
- Logical Ring -
Nodes Automatically Find Their Neighbor to Create A Ring
- Automatic Reconfiguration
- A New Node Joins the Ring Automatically Without Software Intervention
- Broadcast and
Directed Messages
- Multi-Master with
Automatic Token Generation
- Cabling Options
- Coaxial, Fiber, EIA-485 Twisted Pair
- High Speed - Standard
2.5 Mbps, Optionally 19 kbps to
10 Mbps
- Low Cost Chips
- Low Protocol Overhead
- 3 or 4 Bytes—Good Usage of Available Bandwidth
- Variable Packet
Size - 0 to 507 Bytes
- Bit Rate Scalable up
to 10 Mbps—Grows with Your Application
- High Noise Immunity
- Easy/Simple Manageable
Technology—No Special Development Tools Required
Benefits
- Used for Industrial
High Speed Applications
- Used in Embedded
Design to Communicate Between Controllers
- A Multi-Master
Network of Nodes Allows Equal Access to All Nodes
- Simple Low Cost
Hardware Design - One Chip + Transceiver
- Software Drivers
Are Packet Driven
- High Noise Immunity
- Easy/Simple Manageable
Technology—No Special Development Tools Required
History
The History
of ARCNET
ARCNET was one of many technological front-runners from Datapoint,
a corporation founded in 1968 in San Antonio, Texas. ARCNET was
originally classified as a local area network or LAN. A LAN is
defined as a group of nodes that communicate to one another over
a geographically-limited area usually within one building or within
a campus of buildings. That was the intent of ARCNET when it was
originally introduced as an office automation LAN by Datapoint
in the late 1970s. Datapoint envisioned a network with distributed
computing power operating as one larger computer. This system
was referred to as ARC (attached resource computer) and the network,
that connected these resources, was called ARCNET.
 |
John
Murphy, ARCNET chief architect, continues to be amazed about
the diverse application for the technology he developed. |
John Murphy,
ARCNET's chief architect, said ARCNET had been developed precisely
because Datapoint's customers had expressed the need to connect
computers. Gordon Peterson, senior systems programmer at Datapoint,
who wrote ARC's innovative network operating system likens the
unveiling of ARCNET on December 1, 1977, to pushing a boulder
over a cliff. "I knew that we were changing forever the way
that business data processing would be done," he said.
Initial attempts
centered on designing a "shared disk controller"—
a method for Datapoint's customers' machines to communicate with
each other. "The disks we were using moved data to and from
the disk surface at 2.5 Mbps, so that seemed like a nice target
speed," explained Murphy.
The original
hardware was known for a while as BAIL (Bit Assembly Intercom
Link), but it was changed to RIM (Resource Interface Module).
Flow control
was built in to remove a significant burden from the software
and make transfers more efficient. Even today, built-in flow control
is a feature of ARCNET that is lacking in competing technologies
such as Ethernet.
Said Murphy,
the RIMs were designed to be external because "at that time
everything was an external box, everything from disk controllers
to modems to terminal controllers. Datapoint computers of the
time were small, stylish desktop units—with the rest of the
hardware hiding either in or behind the desk, or under a computer
room floor."
| Gordon
Peterson wrote ARC' innovate network operating system—and
indispensable part of ARCNET's success. |
 |
Later, during
1981, the RIM was converted to a LSI (large-scale integration)
chip. George Beason managed the project, which he named the "The
NEWPORT PROGRAM."
Technical
issues influenced the packet size. ARCNET's original design was
very conservative, with the goal of sustaining operation under
the most unlikely circumstances. If the design had fudged on items
like recon timeouts, ARCNET wouldn't be known for its reliability.
As for the
RIM chip, designers said no one ever requested the extension to
508 bytes. The 508-byte option made it into the chip for one reason
according to John Moschner, who played an integral part in converting
the original RIM to a LSI chip: "it was so trivial to implement,
it was hard to oppose it."
Communicating
within the network was based on a token-passing protocol, rather
than a carrier sense arbitration method. Murphy noted, "in
the beginning, the team had hit on the idea of distributing the
polling task among the nodes, a self-polling system. Years later,
this became known as token-passing."
Limited resources,
the need to keep it small and working with the extreme simplicity
of the hardware were key challenges in developing the ARCNET microcode.
Nevertheless, the microcode project was completed in three months
and developers had it working by late summer of 1976.
According
to the team, "Murphy's code was a miracle of bit-twiddling,"
noting that the code interfaced with a single pair of 512x4-bit
ROMs. "Some basic protocol decisions were made to accommodate
the limited code resources."
Murphy explained
that the outstanding feature of ARCNET was its star (or distributed
star) topology connected to a hub.
When it came
to node-to-node wiring, Murphy knew that wires were funny critters
at high frequencies. And Moschner knew that anyone who put up
a Christmas tree, knew better than to try and connect several
computers with a bus.
Murphy realized
that active hubs could pose a burden to users of very tiny networks.
To counteract this, he designed the passive hub—a simple
resistive matching pad."The
original idea was to use the passive hub in a network of up to
four nodes," said Murphy. "But no sooner had we built
the first active hub than a programmer managed to combine a passive
hub with the active hub in a way that crashed the system."
So rather
than try to educate the world about passive hubs, Murphy established
one simple rule: active and passive hubs cannot be mixed.
"Until
Standard Microsystems Corporation (SMSC) entered the picture,
every attempt was made to keep ARCNET proprietary so that it would
be a unique selling point for Datapoint computer systems,"
Moschner said."You
have to remember that we didn't begin to design a separate entity
called ARCNET, we were building a high-speed I/O bus to facilitate
more powerful computer systems," noted Murphy. "Once
we had committed the hardware to silicon, it made sense to try
to lower our costs by increasing the volume of silicon being manufactured.
A lot of effort was expended in trying to popularize the hardware.
The software was connected with our operating system, and there
was no real thought of making any of the protocols public until
after the development of RMS, Datapoint's Resource Management
System, an operating system designed from the ground up with networking
in mind."
Overall,
ARCNET stands up well to the test of time. For certain, Murphy
would make one minor change:"I would ask for a royalty of
a few cents on every node installed."
The "NEWPORT
PROGRAM" Chip
Some individuals in the industry may wonder why George Beason,
manager of the LSI (large-scale integrated circuits) chip project
and now the executive director emeritis of the ARCNET Trade Association
(ATA) named his program "NEWPORT" back in the late 1970s.
He had various assignments taking him to various semiconductor
companies where the chip design was happening. When he visited
Silicon Systems Inc (SSI) in Tustin, CA, Beason stayed in Newport
Beach, CA...just next door to Tustin. Beason liked the name "NEWPORT"
and it sounded good as any other name, probably better. So the
project was coined the "NEWPORT PROGRAM," simply based
on a location, nothing complicated within the LAN world.
 |
George
Beason, Newport Project Leader, has been associated with ARCNET
and the ARCNET chip and protocol and Standard in some capacity
since 1978. |
In no time,
the Vice President of Engineering at Datapoint Corporation hired
Beason in June of 1978 based on his experience in design programs
exploiting LSI. According to Beason, his assignment was to construct
a RIM (Resource Interface Module) chip. Beason felt his most important
contribution to the program was hiring John Moschner from NCR
Corporation to be the Project Engineer. Moschner wrote the chip
specifications with the advice and review of John Murphy and others,
giving way to a PC board inside the Datapoint 6600 that served
as the basis for the chip design and testing.
Beason visited
a few silicon design houses before making any decision on chip
manufacturing, but he elected SSI for two reasons: 1) Beason had
prior experience with SSI and his colleagues were impressed with
SSI's expertise and 2) Beason was quite satisfied with their design
automation process. SSI now became Datapoint's proprietary chip
vendor.
However,
problems incurred. SSI didn't have their foundry running. Beason
said the compromises made to accommodate the different manufacturing
processes almost wrecked the program. Datapoint was paying for
wafer runs and sometimes getting only one working chip per wafer.
SSI's process and foundry were not available unless one could
commit to buying one million chips per year.
 |
Paul
Richman, Chairman of Standard Microsystems, was influential
in moving ARCNET to silicon. |
Beason needed
help and he turned to Paul Richman of Standard Microsystems Corporation
(SMSC). Richman committed all of SMSC's resources to "fixing"
the chip. Now Beason negotiated a quit-claim with SSI and transferred
design, artwork and all to SMSC. Richman's word was "good
as gold." SMSC was able to make the chip yields. The chip
was customer specific (CSIC) for Datapoint. But when it came to
finalizing a legal agreement, SMSC asked for the right to sell
the chip to others. And Datapoint management agreed...a deal was
set!
Now Datapoint
management directed Beason to develop a second source. Beason
went to his contacts at NCR Corporation. NCR became second only
to SMSC in volumes shipped until they withdrew from the market.
In retrospect,
the Newport Program was more than he imagined. Beason confessed
that he never dreamed that he would still be working on the same
project more than 20 years later.
Companies/Applications
The ARCNET Trade Association
(ATA) lists companies and their ARCNET applications. ARCNET is embedded
in many companies' products and it not advertised as ARCNET. Below
are some of these applications in alphabetical order by company
name:
- ABB Kent-Taylor
- Data acquisition
- ABB Combustion
Engineering - Nuclear plant monitoring and control
- Allen-Bradley
- Broadband communications
- American Dynamics
- Closed circuit cameras
- Andover Controls
- Commercial heating and air conditioning
- Army Corp. of
Engineers - Waterways controls
- Automated Logic
- Building automation
- Autotote - Race
track wagering terminals
- Avtec - Programmable
logic controllers
- B&R Industrial
Automation - Programmable logic controllers
- BMW - Automotive
navigation system
- Barber-Coleman
- Data acquisition and control
- Bailey Controls
- Data acquisition and control
- Bristol Babcock
- Data acquisition and control
- Cimetrics Technology
- Building Automation
- Citicorp TTI -
Automated teller machines
- Comprehensive Computer
Solutions - Computers
- Consolidated Packaging
- Packaging machinery
- Control Techniques
- Motor drive communications
- Cummins Engine
- Data acquisition and control
- Disney Imagineering
- Programming terminal
- Dow Chemical -
Data acquisition and control
- Dresser Industries
- Process control
- Eastman Kodak
- Printing equipment
- Emhart Glass -
Process control
- Eurotherm Process
Controls - Process controllers and recorders
- Fastheat - Molding
machine temperature control
- Fermi National
Laboratory - Data acquisition
- Ferranti Sciaky
- Welding controls
- Festo - Programmable
logic controllers
- Fillon Paint Mix
Systems - Process control
- Frito-Lay - Packaging
equipment
- General Electric/Medical
Systems - Communications in XRAY equipment
- General Electric/Transportation
- Locomotive communications
- General Electric/Power
Plant - High reliable control communications
- Giddings and Lewis
- Motion control
- Grayhill Inc.
- Digital/analog I/O control
- Harris Graphics
- Printing press monitoring and controls
- Hi-Speed Checkweigher
- Weigh scales
- Hollister - Medical
electronics
- Hughes Avacom
- In-flight entertainment systems
- Iconics - MMI
software
- Ingersoll Rand
- Compressor controls
- Intellicall -
Phone switching system
- Johnson Controls
- Commercial heating and air conditioning
- Lauer - Text display
panels
- MTS Systems -
Automotive test system
- MagneTek - Motor
drive communications
- McDonald's - Point
of sale terminals
- Medar - Welding
controls
- Mettler-Toledo
- Weigh scale communication
- Motorola - Paging
control system
- National Aeronautics
and Space Administration - Space travel simulation
- National Optronics
- Optical Industry
- Nellcor - Medical
electronics
- Notifier Fire-Lite
Alarms - Fire alarm systems
- Novell - Network
Operating System
- OPTO22 - Control
systems and I/O for industrial control
- ORSI Automazione
- Data acquisition and control
- Pacific Scientific
- Motion control
- Panasonic - Point
of sale terminals
- Quadtech - Industrial
I/O
- Quincy Compressor
- Compressor controls
- Red Barn Computers
- Computers
- Rose Electronics
- Server
- Sartorius Corporation
- Weigh scale communication
- Semitool - Semiconductor
processing equipment
- Siemens - Programmable
logic controllers
- Softrol Systems
- Automation systems
- Systems Chemistry
- Chemical processing
- Teletrol Systems,
Inc. - Commercial heating and air conditioning
- Toyota - Automotive
navigation system
- Trane - Commercial
heating and air Conditioning
- Unisys - Mail
handling systems
- United Parcel
Service - Package tracking
- United States
Navy - Shipboard communications
- V-Band - Stock
exchange terminals
- Viskase Corporation
- Machine control
- Weidmueller -
Data acquisition and control
- White Castle Hamburgers
- POS terminals
- Wisconsin Electrical
Mfg. - Weigh batch controls
- Wizdom Controls
- Programmable logic controllers
