Print this article

Comment

01 October 2003

Wireless reliability

Dependable way to collect data in sensing, control apps.

By Scott Quillin

Wireless devices are becoming commonplace in our society. Most everyone now carries a cellular phone, pager, two-way radio, portable digital assistant, or other wireless device. The expectation from these devices is information can transfer to whoever needs it when they need it.

When messages do not get through or calls fall through, the user usually attempts to retry the data collection. Sometimes this might require a delay in time or movement to reestablish the broken link, but overall it goes unnoticed. In an industrial wireless system, you cannot move locations or "call back" later as your schedule fits; the information must get to where the user needs it.

So, just how do you increase the reliability of low-cost industrial wireless systems so required data ends up in the user's hands when required?

There will always be times when it makes more sense to run wires rather than put in a wireless system. Reasons for running wires include cost, bandwidth, very high reliability, ease, familiarity, accessibility, knowledge, tools, requirements, distance, and power. There are many times, however, that a wireless approach would make more sense, even when some of the "wired" reasons are a part of the equation.

When it comes to benchmarking, wireless systems always compare against wired systems in terms of reliability. When a wire connects two end points, you can assume the data will get through, and for the most part, it will. Using various protocols that include error correction and minimum or maximum requirements for voltages, currents, wire size, wire length, and other variables, a wired system will perform very well in its installed environment. If a system does not meet some of the requirements, the wired system has a much higher chance of failure. The good news is you can easily measure and compare all of the requirements of a wired system against installation requirements. If any of these values were to change due to unforeseen circumstances, the system would again be vulnerable. Bad connections and broken wires are the most common faults in a wired system.

Some applications will never use a wireless link. When the wire runs are easy and cheap, as well as over rather short distances, it only makes sense to use wired systems. Also, when the amount of data being taken is near or greater than the allowable bandwidth (data rate) of the radios available, a wired approach only makes sense.

Many of the sensors currently used in industrial environments, however, fall outside those descriptions. Typically, wire runs for industrial sensor networks cost much more than the sensors themselves. Cost models for labor and materials for many industrial installations put the figure between $50 per foot up to hundreds of dollars per foot just for the wire, conduit, fittings, barriers, and labor to install. The wired sensors become just a small portion of the cost to collect the data deemed important enough to monitor. As well, the actual amount of data transmitted by typical sensors is very low, on the order of bytes per second or less, while the wires themselves could handle kilobytes or megabytes of data per second. Sound like overkill? Still, other sensors could connect to the installed wire as needs arise in the future.

A RELIABLE SYSTEM

The design of wireless systems should be within certain limits for expected performance. When the system does not meet those requirements, you have to expect similar results as compared to a wired system. The difference is a wireless system's requirements are harder to measure. An example with a wired system having a maximum cable length of 100 feet would compare to a maximum path loss in a wireless system. Path loss sometimes compares to distance in a wireless system, but with wireless, line-of-sight (LOS) qualifies as a distance. In a wireless system installation, there may be several obstructions be-tween the transmitting end and the receiving end. Each of these obstructions degrades the radio path due to reflections, absorption, scattering, and other factors.

Measuring the exact path loss when designing a wireless system is difficult without the proper tools. Even with the proper tools, changing conditions mean you have to leave some headroom in the system to compensate for those changes. If a wireless system goes in without extra headroom, there will be dropouts in coverage due to fading, weather, vehicles, or any number of other reasons. Depending on the wireless system installed, the headroom required for consistent performance varies. If the radios are robust, the overall headroom, or signal margin, can be as little as 14 decibels (dB). If not, there may be as much as 30 dB or more of signal margin required to guarantee consistent performance.

A robust radio system should have mul-tiple paths for the information to flow. You can accomplish this by using repeaters that just pass the information on while increasing effective power. Repeaters are two-way radios that receive information from like devices and "forward" the information on using their own radio transmitters. A user can accomplish multiple paths by using multiple receivers, located apart from one another, that are directly connected to the system. This is spatial diversity.

Specialized antennas for the transmitters and/or the receivers will also increase system reliability.

Typically, specialized antennas would point the transmitting signal in a certain direction, or similarly point a receiver to listen in a certain direction. These directional antennas also have the added benefit of increasing transmitted power or receiver sensitivity, but with added cost and complexity to the system.

Multiple redundant transmissions further increase the probability of receiving a message of importance.

If the redundant messages randomly disperse in time (time diversity) and there is a local radio-jamming source intermittently present, the redundant random transmissions have a higher probability of being received compared to just a single transmission. The use of historical data in each transmission also helps with system reliability. If the system misses a single message, the next message would statistically get through and would contain data for the current readings as well as some number of past readings.

This is especially useful for systems that do trending and does not include the redundant message scheme mentioned above.

NOT ALL EQUAL

There are quite a few radio companies and even more individual radio types currently on the market. Many of these radios will not work well in an industrial environment. There are several questions you have to ask when determining what wireless system to design and install at a specific site.

How much data will transmit, and how often? This question determines the data rate of individual radios as well as the bandwidth of the system. If the monitored sensors have slow update rates such as temperature, level, or flow where the information cannot physically change very fast, a lower data rate system would usually be better. If you are monitoring fast-changing parameters such as vibration (also high bandwidth) or there are many sensors directly connected to a programmable logic controller (PLC), and all of the sensor information also needs to get to a remote location, a higher data rate radio system would be a better choice. With the higher data rate systems, the radios usually establish a continuous link to transfer information between them, using two-way devices to acknowledge reception of data. Lower data rate systems can many times be one-way systems where the transmitters send out a little bit of information occasionally, on some preprogrammed time intervals.

How many devices do you need to monitor? This is actually an extension of the last question, because as there are more and more end points, the effective bandwidth of the system increases. If there are many devices, but each is only reporting a few bytes of measured information every few minutes, the overall bandwidth of the system is still quite low. Monitoring many high-data rate devices can overload the system, though, because of the nature of those types of systems. This is not to say you cannot do it, it is just that you have to do more planning to get the system to work effectively.

What environment will the system have to work in? Radio system choices become limited in hazardous environments. Many of the radio manufacturers have not pursued certification in Underwriters Laboratories (UL) Hazardous Locations. If you choose a radio not allowed in these environments, additional installation or certification costs would result. Also, the temperature rating on many radios is not fully industrial rated (–40°C to +85°C). A user should also examine data sheets for shock and vibration tests if the installed location calls for it, as well as environmental issues. Industrial radios appear in some of the most severe environments ever found. The radio and enclosure must also be up to the task by being weatherproof, shock resistant, chemical resistant, and temperature rated.

HOW FAR?

What is the range requirement? There are some installations that only require the data to transmit a few feet in the same room. Vibration monitoring on large rotating machinery is an example. Wireless devices see use in this case (even though the distance is very short), because it would be impossible to connect wires to the rotating machine. Also, the radio does not have to be very powerful and would normally operate by battery. Battery life would be short when gathering continuous measurements from a device such as this.

Other installations need to get data from 20 or 30 miles away. The radios used in these installations are higher power transmitting devices with very good external antennas and very sensitive receivers. The cost of these radios is the greatest, and the design and installation of the system is also the most expensive. These radios typically connect to several sensors or PLCs and can send continuous update information back to a location that is useful.

The rest of the installations are somewhere between these extremes. Sensors appear all over campus-sized installations or on unmanned rigs within a mile or two of a working platform. The parameters measured are typically lower data rate items such as temperature, pressure, or flow.

Update rates for each sensor are every few minutes to every few hours. Many times prior to wireless system installation, the parameters either did not undergo measurement or users recorded them by hand when they visited the location. Often the data was incorrect or not read at all, and would then have to go into the system to keep historical data on the device. With the addition of the wireless sensor, users can now accurately measure the data when they are supposed to measure it. Also, it automatically logs into the system for future analysis.

Many of the sensors in a campus setting or on an oil rig need to operate via battery. Several reasons exist for battery-powered devices, including no additional wiring and ease of installation, up to the fact that it might be near impossible to get power to some of the monitored locations.

Depending on the design of the transmitter and/or sensor, battery life for these units ranges from several days to many years. Battery life of the transmitter is highly dependant on the update rate of the device, as well as its current when asleep, and finally, the output power of the device. More power gives longer range but drains batteries more quickly.

There are many different types of wireless devices currently on the market. Technologies such as 802.11 and Bluetooth come to mind when describing wireless networks. These are wireless standards originally designed for computers and peripherals so that connection was as simple as being within range of the network. This is a great idea for an industrial wireless network except that for an industrial sensor network, the amount of data transferred is now on the order of bytes instead of kilobytes—not what the standards' founders designed them for. As well, the amount of overhead (extra protocol information) to send the data greatly outweighs the data itself. Proprietary wireless sensor protocols and the yet-to-be IEEE 1451 Wireless Sensor standard address many, if not all, of the above concerns simply by the fact they are most efficient when used in an industrial sensor network. Bluetooth and 802.11 devices may work in this environment, but not as efficiently.

SPREAD SPECTRUM

The Federal Communications Commission (FCC) regulates all the radio frequencies in the U.S. They have set aside several frequency bands for use without having to purchase a license. The most common are the Industrial, Scientific, and Medical (ISM) bands. The ISM rules generally state that transmitted power has a maximum of 1 watt when using spread-spectrum modulation and 1 milliwatt with narrowband modulation methods. The reason for the great increase in power (1,000 times the power, or +30 dB) when using a spread-spectrum system is the ability to have many different users in the same geographic location without large system degradation.

Spread-spectrum modulation works by either spreading the transmitted information over a much wider frequency band than just the information itself would take up (direct sequence spread spectrum [DSSS]), or by changing the transmitting frequency several times during a single message (frequency hopping spread spectrum [FHSS]). To other systems, either DSSS and FHSS look like momentary low-power jamming sources at worst, or they do not affect other systems by any measurable means. Spread-spectrum radios must meet all of the requirements of the FCC's Part 15.247 of the Code of Federal Regulations (CFR).

Spread-spectrum radios are very common. They include cordless phones, wireless local-area networks, and millions of industrial sensors. The FCC allows spread-spectrum radios to operate in the 902- to 928-megahertz (MHz) band, the 2.45-gigahertz (GHz) band, or the 5.6-GHz band. The lower frequency radios are typically more inexpensive, but they are also lower data rate devices because of the limited frequency bandwidth available. The 2.45-GHz and 5.6-GHz radios are able to handle much more information per second because of the larger allowable frequency band. Spread-spectrum radios operating within these bands must only get approval for the general operation of the tested device, which usually occurs when they make the radio or when the company integrates the radio module into a final product. With most other radio types, a user must get individual licenses for each radio installation. These licensing issues are time consuming and expensive, if they are even available.

SELECTION ALTERNATIVES

The choice of which radio to install depends on many factors. If there is a system already installed at the location of interest and it can handle the extra information that new radios would add to the system, using like products would be the logical choice. Radios that do not require licenses for each site, such as the ISM band radios, are less of a hassle to deal with. The benefit of the licensed radios is they can be much higher power than the 1-watt limit for DSSS or FHSS radios in the ISM band. The extra power means longer reliable LOS distances, but at the cost of increased power demands and higher cost per device. Choosing between FHSS and DSSS is not quite as easy as determining if a license is available. Both spread-spectrum radios should perform in a similar fashion.

Michael Moore of Oak Ridge National Labs wrote in his article "The Next Step—Wireless IEEE 1451 Smart Sensor Networks" that DSSS radios are more reliable, require less power, are less interfering, and more spectrally efficient than FHSS radios. You can attribute the performance benefit of direct sequence modulation to processing gain, where the information spreads out in frequency at the transmitter and then compresses back into its original bandwidth at the receiver. The receiver attenuates anything that was not part of the original signal by its processing gain. A frequency hopping system does not generate processing gain, but does move the signal around in frequency during the transmission. This moving around in frequency also requires much longer synchronization times than direct sequence systems, because the receiver in a FHSS system must continually search though its list of frequencies looking for a transmitter that happens to be sending a message. The transmitter must include enough synchronization time in its transmitted message for the receiver to find it. This adds precious time to messages when systems are operating near their peak data rates.

DSSS systems, on the other hand, require much shorter synchronization times, because they already know the frequency. The receiver only has to correlate to the message to gather the useful information. A short message from a contact closure transmitted by an inexpensive FHSS radio has a message time of near 250 milliseconds, while the same information transmitted by a similar DSSS radio only takes 8 milliseconds. When dealing with low-cost sensor systems, the additional time used by the FHSS radios causes radio frequency collisions and missed data. The more time each device is on and the more devices there are in the system, the less chance of receiving an individual message. Kuriacose Joseph statistically calculated this and explained it further in his paper "Calculation of Message Length Transition Probabilities in Selective Reject ALOHA Channels." Therefore, the shorter messages provided even by low-cost DSSS radios increase system performance by increasing the probability of each message getting to the proper recipient.

HOSTILE ENVIRONMENT

Most equipment installed in an industrial environment undergoes rigorous design and testing so it can survive extreme temperature, severe shock and vibration, caustic chemicals, weather extremes, and other factors. Radios used in these environments must also meet these requirements.

They must be able to operate in temperatures from –40ºC to +85ºC, so they do not drift too much in frequency when subjected to these extremes. They must also coat the radios in sealed enclosures for protection from weather and/or harsh chemicals. The enclosures and the radios themselves must be able to withstand falls, various banging around, and other "mistakes" caused by installers and other people in the field. If they cannot survive, the system will not perform for long.

UL certification or the equivalent is also a must for operation in hazardous environments. At a minimum, the device should have UL Class 1 Division 2 Groups C and D listing to be compatible with most locations in the field.

The radios must also be as technically robust as cost allows. The receivers should have very good sensitivity, but not allow information from unwanted bands to interfere with reception of desired signals, meaning they should have good selectivity and jam resistance. Both frequency hopping and direct sequence radios have proven they are able to perform very well in hostile environments.

Other forms of radio modulation have not been as successful at penetrating walls and dealing with the many potential obstructions encountered in industrial environments.

The choice of which ISM band to use is also difficult. Many countries do not allow 900-MHz ISM band radios as they do 2.4-GHz radios, but building penetration is better at lower frequencies. The influx of popular radio technologies such as Bluetooth and 802.11 brings up questions of overcrowding and increased background noise in the 2.4-MHz band where these devices operate. Bluetooth radios seem to be especially susceptible to jamming sources, especially when near high-power microwave devices (yes, even microwave ovens), said Charles Buffler in his article "Compatibility Issues Between Bluetooth and High Power Systems in the ISM Bands."

END RESULT

Will radios ever replace wires? Yes, they have in many applications, and will continue to grow into more applications when people understand data can reliably and cost effectively move from the end point to the controller. Designing, installing, and using a reliable low-cost industrial wireless data system takes no more effort than its wired counterpart does. The only differences are the questions a user must ask during the process.

Wireless system design must take into account the intended use of the product so a user does not install it in a location where it would work marginally at best. Users must adhere to electrical classification and FCC certifications. They must also meet power requirements for each end point. The electrical and mechanical sensor interfaces, mounting, and shock and vibration requirements must undergo scrutiny for each product.

The only difficult parameters to consider when designing a wireless system are the range of the radios and the physical operation of the devices. Radio vendors typically state the range performance of their products in one or more environments, and the physical requirements are the frequency band of use and the modulation scheme employed. These parameters are very similar to wired systems' maximum wire lengths, data speed, and wire protocols. At the end of the day, designing, installing, and maintaining a reliable industrial wireless system may just be easier than doing the same with its wired brother. IT

Behind the byline

Scott Quillin is a senior project engineer at New Orleans–based Axonn LLC.