ARC BRIEF

ARC BRIEF
By ARC Advisory Group

FEBRUARY 7, 2007
Building a Comprehensive
Industrial Wireless Future

Executive Summary
The revolution in wireless networking has huge potential to improve manu-
facturing. A major part of realizing that potential requires the development
of industrial wireless sensor networks (WSN). Efforts to
A truly comprehensive approach develop wireless sensor network technology suitable for
to industrial wireless networking manufacturing have been hampered by the immaturity
must eventually address the and rapid development of the technology, and by stan-
unique requirements factory
dardization programs (such as ZigBee) that initially
automation sensors in discrete
attempted to standardize a far broader scope than was
manufacturing operations.
feasible. A specifically industrial approach to defining
standards is now being pursued by the ISA SP100
group, but their work at present is forced to disregard the differences be-
tween sensor requirements in process vs. discrete manufacturing, and is
focused on sensor networks for process automation.

Lately an informal collaboration has developed between 3 companies with

complementary expertise in discrete manufacturing. These collaborating
companies are ABB, an automation leader and a company that has commer-
cialized wireless sensor products targeting factory automation, FESTO, a firm
highly regarded for its machine actuator components and solutions, and Mil-
lennial Net, a wireless sensor networking company recently acquired by
privately held PHC Holdings. These 3 companies plan to build on their exist-
ing work and expertise in order to further advance the use of wireless sensor
networks in factory automation and discrete manufacturing. As such, their
work should prove complementary to that of SP100, and at some point in the
future could be included in industry standards addressing discrete manufac-
turing.

2007-WP-WSN001v001
ARC Brief • February 2007

The Promise of Industrial Wireless

No development since the Internet has greater potential to advance manufac-
turing than wireless networking. As with earlier revolutions in
manufacturing technology, advances in wireless IT and consumer electronics
technology are having impact at all levels of manufacturing. The huge de-
mand for wireless products by IT and consumer markets have made a
number of powerful wireless technologies available to manufacturers at very
low cost. Globalization of the IT and telecom market has also created global
standards for many types of wireless networks. Technologies like Wi-Fi,
Bluetooth, and GSM have become global standards, while emerging wireless
technologies like WiMAX, Wireless USB, and ZigBee also have the potential
to become pervasive globally.

Future manufacturing plants will support not one but many wireless tech-
nologies at the same time. Wireless networks will provide voice
communication, mobile computing, process and factory automation, process
sensors and transmitters, building automation, personnel safety, access con-
trol, and security. These applications will employ a set of wireless
technologies, not just one, just the way wired industrial networks employ
Ethernet, fiber optic, and various industrial fieldbus solutions.

There is another important analogy between the wired industrial networks of

today and the wireless plants of the future. Industrial networks are often
classified by level (see Figure 1). In the
wired world of manufacturing there
are sensor and fieldbus networks (that
Plant Intranet or
Internet Connections capture sensor and actuator informa-
tion) automation networks (that control
and manage production equipment
Connections to
Control Network
and human interfaces), and enterprise-
level networks that provide services for
Fieldbus
the whole manufacturing firm and its
Connections
supply chain partners and customers.
The same is true for industrial wireless.
Machine and
This discussion will focus on wireless
device level
sensor/actuator networks, the lowest
Standard sensors and actuators
level, the most challenging, and poten-
tially the most revolutionary new type
Figure 1 – Automation Network of wireless network.
Functional Hierarchy

2 • Copyright © ARC Advisory Group • ARCweb.com

 ARC Brief • February 2007

The Balancing Act of Wireless Sensing

Creating any wireless product for the industrial market requires a balancing
act, in order to meet the distinct requirements of manufacturing as opposed
to IT and consumer segments. In the case of sensor-level wireless, this chal-
lenge is daunting.

Sensors used in manufacturing need to be highly reliable. The performance

of these sensors often determines the success of the manufacturing operation
itself. If the sensors do not perform their mission, production may be lost,
sometimes large amounts of valuable production or production time.

Wireless sensors need to be designed specifically to fulfill an industrial mis-

sion. A large number of design variables need to be controlled so that the
wireless sensor can perform. Some of these variables are:

• Power supply – The sensor will need reliable battery power (with its dis-
advantage of additional service intervals) or some other source of energy.
• Range – the sensor network will need to communicate reliably over a
specified distance. Some production equipment is compact, and some is
very large.
• Latency – The time required for a sensor reading to reach the automation
system must be bounded if the sensor data is used in a closed-loop con-
trol system. The upper limit of this bound depends upon the application.
It can range from seconds to just a few milliseconds.
• Data types – There are both analog and discrete output sensors, and in
addition some “smart” sensors may supply diagnostic information that
can also be used to improve the reliability or quality of production.
• Production cycle – Some production operations shut down every day,
providing a window for maintenance activities. Others have operating
cycles that last for several months or even more than a year.
• Control loop cycle times – Process automation control loop cycle times
are normally in the range of seconds to minutes, while machine control
loop cycle times are normally in the range of milliseconds.
• Scalability – The number of sensors in a single production unit can vary
from a few to several thousand.
• Immunity to interference – Radio interference is certain, and the sensor
network must be able to perform its mission in the presence of many
types of radio interference.
• Radio signaling – Wireless sensors can use a variety of signaling and
modulation techniques, and the choice of these may limit some other as-
pects of sensor network performance.

Copyright © ARC Advisory Group • ARCweb.com • 3

ARC Brief • February 2007

• Network Protocols – Sensor networks use a variety of different commu-

nication protocols and are not suited for TCP/IP, the most common set of
protocols used in wired networks.

Different industrial applications will have very different needs and will
choose different solutions, values, or bounds for each of these variables. The
solution developer must focus intently on what the manufacturing operation
requires. Then the challenge is to choose a set of network properties that will
satisfy all the application requirements. The real objectives for any wireless
industrial product are determined by the manufacturing operation in which
it must perform, not by the technology employed.

Meeting the Requirements of Factory Automation Applications

While wireless sensing is very broadly applicable and is certainly getting its
share of publicity, the number of commercial products that have been intro-
duced is disappointingly low. Even applications where concept and
prototype products have been shown publicly have not
Commercial products using seen full-scale commercialization of new wireless sensing
wireless sensor networks have products. However in some industrial applications prod-
been slow in coming to market,
ucts are now beginning to appear. Process field
even in areas such as building
transmitters are one area where major product introduc-
lighting where prototype products
have been shown to the public.
tions have occurred. But why has it taken so long, and
what happened to non-industrial areas such as building
lighting control that were supposed to be in the market 1-2
years ago? To answer this question it helps to review some recent history
concerning the development of standards for wireless sensor technology.

ZigBee
The ZigBee standard, developed by a broad consortium and made public in
2004, was the first major attempt to standardize all types of wireless sensor
networks. ZigBee focused originally on networking, network formation, and
network management. The idea behind ZigBee was to create a single tech-
nology replacing many services provided to wired networks by the TCP/IP
suite and the global Internet. Originally, ZigBee was envisioned as a lingua
franca for all types of wireless devices from thermostats, to cardiac monitors,
to industrial sensors, to light fixtures. At such an early stage of WSN technol-
ogy, this turned out to be far too broad a scope. Many end users had trouble
getting ZigBee to meet their application requirements satisfactorily. This was
especially true in industrial applications and when applications were scaled
up to normal size. They also encountered significant structural issues when

4 • Copyright © ARC Advisory Group • ARCweb.com

 ARC Brief • February 2007

trying to introduce a disruptive yet immature technology into established

business channels.

In response, the ZigBee organization has further refined its underlying tech-
nology and has also narrowed its focus. It is now addressing applications
like building lighting and has a longer term roadmap that gradually expands
its target market. The failure of developers to commercialize many ZigBee
products speaks volumes about the difficulties and the risks of trying to
standardize on too broad a scale. The large-scale deployment of ZigBee will
take far longer than its proponents predicted just a few years ago.

HART Wireless
If ZigBee tried to do too much, that would certainly not cause any problem
for the wireless working group of the HART Communication Foundation.
The HART wireless group is now working to specify a wireless implementa-
tion of HART for the new generations of wireless field devices. HART is a
well-established 20-year-old protocol that is available on millions of analog
field devices. HART modulates the 4-20 mA DC current signal put out by
these field devices with a low-level sine-wave, without affecting the average
DC signal.

In wired HART, the device transmits the measured value using the 4-20mA
analog signal and the HART information using a modulated 2200 Hz signal.
In a wireless implementation of HART, the only access to the device is
through the wireless signal using the HART protocol. This means that the
wireless sensor network must be highly reliable so that measurement infor-
mation is always available. Wireless HART devices would also have to form
networks of a certain size so that network management is simplified. Finally
these devices will need to perform well in areas where Wi-Fi networks are
also deployed. So defining a successful wireless HART solution depends in
part upon the other types of wireless networks that are operating near the
HART devices. A broader approach, one that looks at all types of wireless
devices in a plant, is needed. That is the task now being worked on by the
ISA’s SP100 working group.

SP100
Since its creation, the SP100 group has worked diligently to manage the large
task it has been assigned. The group recognized that industrial applications
had a very broad range of requirements and so different technologies would
fit in different parts of this vast application space. One of the first pieces of
work that the group developed was a classification scheme for industrial ap-

Copyright © ARC Advisory Group • ARCweb.com • 5

ARC Brief • February 2007

plications (see Figure 2). All of the group’s work is done in the context of
these classes of applications. In addition, the SP100 group has employed the
OSI 7-layer communication model as a 2nd classification to define which
communication “layers” are addressed by their work. This comes at a cost of
some complexity, and the group cannot further subdivide the application

Category Class Application Description

Safety 0 Emergency action (always critical )

Closed loop regulatory

1 (often critical )

Importance of message
control

timeliness increases
Closed loop supervisory
Control 2 (usually non -critical )
control

3 Open loop control (human in the loop )

Short-term operational consequence

4 Alerting
(e.g., event-based maintenance )
Monitoring No immediate operational consequence
Logging and
5 (e.g., history collection , sequence -of-
downloading /uploading
events, preventive maintenance )

Figure 2 – ISA SP100 Classification of Wireless Automation Applications

space without having far too many classifications to manage. As a result, the
distinctions between process and discrete manufacturing cannot be ad-
dressed by this group at present.

The SP100 group is being managed very much like the IEEE 802 committees,
and has focused on the sensor network layer because there is broad agree-
ment among the committee about higher level wireless networks. So while
the group has managed to avoid mission creep, they have chosen not to focus
on the substantial differences between process and discrete manufacturing
and the resulting differences in the sensor-level devices used by these appli-
cations.

The Unique Requirements of Discrete Manufacturing

Most actual manufacturing operations exhibit some characteristics of both
continuous and discrete. However these 2 types have very different charac-
teristics that determine the way they are operated and the way they are
automated. For example an automotive assembly operation must start and
stop every day, while some chemical manufacturing operations must run
night and day for several years without stopping. The table nearby (Figure 3)
shows some of the different characteristics of continuous and discrete manu-
facturing and automation. Note that for many of these application
requirements, the two types of manufacturing differ by orders of magnitude.

6 • Copyright © ARC Advisory Group • ARCweb.com

 ARC Brief • February 2007

The discrete manufacturing world has not seen the same level of frenzy for
developing wireless industrial products. Ethernet is a very common network
in such automation systems, so the use of industrial Wi-Fi products is grow-
ing rapidly. Some wireless bridges for discrete fieldbus technologies (such as
DeviceNet) have been introduced. But at the industrial sensor/actuator
level, the only major product introduction has been the WISA product intro-
Continuous Discrete
Process Attribute Manufacturing Manufacturing
Plant Cell
Sensor types Predominately Analog Predominantly Discrete
Sensor count per unit 1000 100
Unit Physical Size 1000 Meter 10 Meter
Units per Plant 10s 100s
Production Cycle Length 100 Days 1 Day
Unit Startup Time Hours Minutes
Control Loop Time .1-1000 Sec .001-.5 Sec
Field Device Cost $1000 $100
Installation Cost/device
10 X 4X
cost
Automation Technology DCS PAC/PLC
Profibus-DP,
HART,
DeviceNet,
Commonly Used Device Foundation Fieldbus,
Interbus,
Networks Profibus-PA,
AS-i,
Ethernet
Ethernet
Figure 3 – Characteristics of Continuous and Discrete
Manufacturing Operations

duced by ABB (WISA stands for Wireless Interface for Sensors and
Actuators).

In an effort to further develop solutions that can specifically address discrete

manufacturing, ABB is now working in an informal collaboration with
FESTO of Germany and Millennial Net of the United States. Swiss-based
ABB is one of the world’s largest automation companies. FESTO is a large
privately-held automation company that is well known for its actuator prod-
ucts that are used in many types of production machinery. Millennial Net is
a wireless sensor network company that was recently acquired by privately
held PHC Holdings. This collaboration brings together ABB’s experience in
developing wireless sensor products for discrete manufacturing, FESTO’s
leadership in machine actuation, and Millennial Net’s deep knowledge of
low-power sensor networks.

Copyright © ARC Advisory Group • ARCweb.com • 7

ARC Brief • February 2007

WISA and MeshScape

ABB’s development of WISA is a case study in adapting standard wireless
networks to very demanding industrial applications. The target was to de-
velop wireless replacements for common sensors like inductive proximity
sensors. In these applications, the sensors are located on a single machine in
a fairly dense configuration, but must be read rapidly and reliably. Because
there can be more than 100 sensors on a single machine, ABB rejected battery
power as an option for powering the sensors – it would decrease overall reli-
ability or increase maintenance to have that many batteries on each machine.

WISA uses IEEE standard radios of the same type used in Bluetooth devices,
but without any of the Bluetooth networking. Bluetooth networks can only
extend to 8 devices, whereas ABB needed a network scalable to well over 100
devices that could be polled in a few milliseconds. In place of the Bluetooth
network stack, ABB developed a protocol that used a duplex Time Division
Multiple Access (TDMA) technique and also used the frequency hopping
characteristic of the Bluetooth radios. This was combined with an innovative
wireless power supply for the sensors that employs a rotating magnetic field
created around the machine by 2 sets of coils. The result is a set of hundreds
of sensors that operate without batteries and can be read in milliseconds.

WISA alone is not a complete solution for wireless factory automation. Ana-
log machine sensors and larger sensor networks will require a different type
of networking. This is where Millennial Net’s MeshScape solution comes
into play. MeshScape uses a hybrid approach to sensor networking that en-
ables large and adaptive networks to form but still operate at the very low
power levels that enable battery-powered devices to have long life, though
MeshScape networks will have far longer latency than WISA. The greater
scalability comes from using localized network routing, formation, and syn-
chronization and an Ethernet-like CSMA rule for me-
ABB, FESTO, and Millennial Net, dium access that allows large networks to respond faster
the 3 companies that are now than is possible under TDMA rules. In many ways the
engaging in this wireless sensor
design choices taken by MeshScape are quite different
collaboration, have very different
than WISA, because the target applications are much
and complementary expertise.
different than the simpler sensors served by WISA.

While ABB and Millennial Net each have distinct wireless sensor networking
expertise, FESTO’s core competence is in its actuation products, for which the
firm is highly regarded. At present all FESTO’s products use wired networks,
but with so much expertise and experience in these products, FESTO can

8 • Copyright © ARC Advisory Group • ARCweb.com

 ARC Brief • February 2007

contribute application expertise and deep insight into the target market, cus-
tomer needs, and use-cases.

The collaboration between these 3 firms remains informal and no written

agreement is in place. The stated purpose of their joint work is to:

• Further define the requirements and applications of wireless sensors in

factory automation
• Improve the existing technologies in factory automation wireless sensor
networks
• Harmonize their work with other ongoing developments in industrial
wireless sensor networks
• Provide a common voice to standards bodies such as SP100 to help en-
sure that the needs of factory automation are being addressed in the
standardization efforts

Recommendations
• Suppliers of sensors and sensor networks to factory automation applica-
tions should consider participation in future work of the
ABB/FESTO/Millennial Net collaborative.

• The collaborative should report its results and progress regularly and
publicly to discrete manufacturing industry groups, and to various SP100
teams.

• End users of future products, especially machine builders, should be-

come familiar with this work now in order to prepare their products to
take advantage of developments in this area.

Analyst: Harry Forbes

Editor: Wil Chin

Copyright © ARC Advisory Group • ARCweb.com • 9

2007-WP-WSN001v001
3 ALLIED DRIVE DEDHAM MA 02026 USA 781-471-1000

BOSTON, MA | WASHINGTON, D.C. | PITTSBURGH, PA | PHOENIX, AZ | SAN FRANCISCO, CA

CAMBRIDGE, U.K. | DÜSSELDORF, GERMANY | MUNICH, GERMANY | HAMBURG, GERMANY | TOKYO, JAPAN | BANGALORE, INDIA | SHANGHAI, CHINA

2007-WP-WSN001v001