San Jose, Ca. - At the Embedded Systems
Conference, Silicon Valley, April 15-18, about a dozen
classes
will present developers with details on array of alternatives for
deploying microcontroller-based sensor network designs, both wired and
wireless.
In addition to giving them a range of choices for connecting
their sensor systems, the alternatives present developers with some
tough multiple choice questions they will have to answer about their
designs including:
Do I want wired or wireless? What
are the choices and what are the
tradeoffs? Where do I put the network intelligence, centrally or in the
periphery in the devices? Do I use 8- or 16-bit devices? RISC or
digital signal controllers? And once put into the field, how to I
maintain upgrade and even perform debug operations on them without
going off-line?
Tough enough for wired networks. But for wireless control networks -
even though Zigbee Consortium has
done a good job in promoting
its protocol - there are a number of better suited protocols that will
work with the IEEE 802.15.4 spec.
This means developers will be faced with even more multiple
choice questions.
Wired USB vs Ethernet
In "Working with USB/Ethernet Software
for Distributed Sensor Networks (ESC-223)," Eric Gregori
makes
it clear that even though wireless connectivity is gaining most of the
headlines, there are a a number of cost effective choices available if
an embedded designer wants to go with a wired solution. Among the
choices, he says, are number of embedded TCP/IP Ethernet alternatives
as well as the use of the Universal Serial Bus (USB), a protocol
traditionally associated with linking peripheral devices to the desktop
computer.
"Connectivity options associated with computing, such as Ethernet
and USB, are enabling network innovations for home/industrial
automation," Gregori says. "The low cost of Ethernet-enabled silicon
and embedded TCP/IP has opened the door for small distributed
Ethernet-based sensor/control networks." In addition, he says, the low
cost of USB enabled silicon and embedded USB stacks, has opened the
door for small distributed USB based sensor/control networks.
Both USB and Ethernet have advantages when designing distributed
sensor / control networks," says Gregori. "The primary advantage for
USB is power distribution and network cost. Power distribution is built
into the USB specification, 100ma per device, or 500ma per device for
powered hubs (with software negotiation). The power distribution system
is built into the cabling, hubs, and hosts. Another advantage is low
cost USB devices. A variety of low cost USB devices designed to be
connected to a PC can be used in the sensor network."
The advantages of Ethernet, he says, are performance and
compatibility. Ethernet and TCP/IP enabled sensors do not require a
gateway to connect to a PC LAN network. This makes the sensor data
significantly easier to get to the internet for remote monitoring
around the world. WiFi compatibility is also an advantage if the sensor
network is within range of a existing WiFi network.
"When designing a sensor / control network for a device that will
have a local controller ( for instance a manufacturing machine ) a USB
network is a good option," says Gregori. "Assuming the sensors draw
less then 100ma, the USB wiring can be used as the power distribution
network, eliminating the additional cost of power wiring.
But he points out that when designing a sensor/control network for a
device that will be remotely monitored or controlled, an Ethernet
network is a good option. "In most cases, the PC LAN network in the
building can be used as a connection to either a remote node in the
same facility, or a remote node on the internet."
Wireless connectvity's multiple
choices
For a connectivity protocol that was still in an embryonic state in the
late 90s, the potential of the IEEE 802.15.4 wireless spec in embedded
systems designs has increased dramatically.
According to Matt Maupin, who will be presenting "IEEE 802.15.4: Providing the foundation of
wireless sensing and control (ESC-343)," while everyone is
looking for the nirvana of a single technology to solve all the market
needs, this rarely happens. Wireless sensing is no exception." However,
he said, with 802.15.4, it is possible to have a common base to provide
a superset of technologies to cover the majority of sensing and control
applications.
In addition, to Zigbee, he said, developers now have a choice of
wireless control alternatives, including the Freescale Synkro
networking protocol, WirelessHart, and the SP100/ISA100.11a network
stack, all of which he will cover in detail in his class.
But which one is best?
"The answer, of course, is it depends on the application," says
Maupin. "The two key factors to look at are cost and the target market
(consumer, commercial and industrial) as these factors tend to be at
opposite ends. For example, if a solution is built for the industrial
needs, the cost tends to go up. At the same time, to keep the consumer
device cost down, features and flexibility are often stripped out.
Synkro technology, he says, targets consumer devices and has been
developed with cost in mind. From a silicon standpoint, the ICs can be
simplified to remove much of the I/O while memory is the least required
of all stacks discussed. This allows smaller ICs with less memory to be
developed to address this specific market.
ZigBee technology is the next market up, and in general has the
largest market appeal. "However, it fits best in the commercial space,
" says Maupin, "but does not fit as well into the low end consumer due
to cost, and is generally not perceived as an industrial fit due to
lack of features such as channel hopping."
ISA100.11a and WirelessHART technology are both targeted toward the
industrial market, but the mandatory 10ms time slot and 10dBm power
output for WirelessHART specification will slightly increase the cost
over ISA SP100.11a technology.
In both cases, he says, another cost is likely to be the stack
itself. While ZigBee and Synkro networking protocol are developed by
the IC makers and usually provided at no cost, it is likely that
SP100.11a and WirelessHART specification will have additional cost
associated with them for the stack, adding to the overall product cost.
"While there is still no 'One Size Fits All'," says Maupin, "the
various protocols still benefit from many of the key technical and
market advantages of 802.15.4, such as lower cost, increased vendor
selection, and competitive enhancements"
How much intelligence, and where?
In "Intelligent Sensor Signal
Processing (ESC-523)," Priyabrada Sinha, will discuss the
fundamental problem facing developers of embedded sensor networks:
where to you do you put the intelligence needed to process the data
received and to make decisions on what actions to take?
"As sensor-based applications rapidly increase in complexity, it
becomes imperative to embed a greater degree of intelligence to the
sensor interface," says Sinha. "Many applications utilize multiple
sensors to obtain a variety of measurements and process them in highly
innovative ways. In some cases, the signals from multiple types of
sensors must be processed simultaneously (and therefore by the same
MCU), a scenario that can be termed "sensor fusion."
Each type of sensor has its own signal characteristics and requires
a different set of post-processing to extract useful information from
it, which increases the amount of CPU computations and peripheral data
handling.
Also, in many applications, says Sinha, sensors are physically
dispersed over a wide area, such as in a large building or factory, or
in various parts of an automobile. A centralized processing/control
approach often proves ineffective, or inefficient at best, for such
distributed systems.
"To offload some of the processing and data-storage requirements
from the central control unit, it is beneficial to spread the
processing capability over multiple MCUs located close to, or even
integrated with, the sensors," he says. "This 'distributed sensor
processing' approach requires a variety of powerful signal-conversion
and communication peripherals."
More connectivity options
Other useful ESC classes on embedded wired and wireless networking and
sensor design include:
1) "Embedding
TCP/IP (ESC-103),"
"USB for embedded
systems (ESC-220)," and "IP
version 6 Overview and Transition Mechanisms (ESC-240/260),"
taught by Christian Legare.
2) "CAN " A secure high speed data
communications bus (ESC-302)," presented by Carl Stenquist.
3) "Building and operating robust and reliable
Zigbee networks (ESC-322)," taught by Zachary Smith.
4) "Security in a wireless embedded world
(ESC-341)," which will be presented by Timothy Stapko and
Owen
Magee.
5) "RFID as a networking technology (ESC-363),"
which will be taught by Martin Payne.
6) "Choosing a wireless protocol: 802.15.4 vs
Zigbee vs. proprietary (ESC-403)," to be taught by Miguel
Morales and Kevin Belnap.
7) "Wireless sensor network system engineer's
rules of thumb (ESC-423 )," to be presented by Ilya Bagrak.
8) "What a Mesh! The Ins and Outs of Mesh
Networking Technologies (ESC-443)," taught by Joel Young
To attend these and other informative classes, sign up now on the ESC registration page.
