Avoid Common Wireless Network Blunders

“Wireless reliability is not addressed as often as it should be,” agreed Bruce Hofmann, product marketing manager – advanced connectivity, Weidmuller North America. “Wireless signals are easy to block, especially by accident, and easy to disrupt or tap into. Reliability is a function of physics as well as what you pick for your radio or wireless controller.”

“Despite the inherent weakness of wireless, though,” continued Hofmann, “good reasons to use a wireless network include to replace cables, to cover long distance communication needs, setup a temporary or backup application, or get a process up and running quickly. However, I would convert to wired at the first opportunity.”

Tools. This pitfall should be common sense, but engineers often don’t have the right tools. “Typically these tools include a spectrum analyzer, network analyzer, and perhaps an RF generator,” noted John Schwartz, technical support manager Maxstream, Digi International. “It’s also a good idea to have some expertise in RF too.”

Remember, RF is inherently analog. “Even though you hear terms like digital modulation and digital radios, transmission and data reception travel along analog paths,” continued Schwartz. “In particular, the need for modulation devices is often overlooked in wireless systems signaling over the Ethernet protocol.”

Chips or cards. Select the specific wireless technology for an application and choose how you will incorporate it in the device or equipment. The options include chip sets, wireless cards, and network modules.

At the microprocessor level, wireless chip sets are a cost effective choice for small lot quantities. Plus, they help tailor the device to specific application needs. One pitfall to watch, though, is device driver integration. Often, the selected drivers are incompatible with the bus structure. Don’t assume that you are working with a PCI interface. Another pitfall is the pace of technological change. The life cycle of a chip set may be only two or three years, and therefore obsolete before the product goes to market.

Rather than chip sets, consider wireless cards, such as PCMCIA or CompactFlash cards. “We’ve had people integrate PCMCIA cards into a machine, said Gary Marrs, senior field application engineer, Lantronix. “You buy an off the shelf PCMCIA card, and embed them into a machine. You design backwards to make your real-time operating system work.” An advantage of this option is that it is easy to achieve higher power output and the card manufacturer must obtain the necessary FCC approvals. Plus, the card’s life cycle is the responsibility of the manufacturer. A disadvantage is that the driver must be ported to the operating system.

A third option is to use wireless network modules. You need not worry about end of life notices, and FCC approval is often a matter of paperwork rather than testing. Several companies are starting to offer wireless modules. “They make it easy to add wireless,” continued Marrs. “Some will provide a simple serial interface for receipt and transmission of serial data off the microprocessor.” Depending on the module, it can replace the RS232 port and transmit UART data for 802.11 connection.

“If you are working for a small, targeted customer base of one-half a million users or less, consider RF modules. RF circuit board components change so fast that a module saves the headache of trying to be current,” noted Schwartz.

Know the environment. A common blunder is the failure to ask important questions, such as: Where will the wireless devices and network be deployed? What are the obstructions? What is the location of metal machinery? “The location determines the frequencies you can use,” said Schwartz. “For example, if you need to penetrate through several walls in a building, in the US you can select the 900 MHz band.”

Site surveys can be crucial and eliminate many potential problems that could obstruct, limit, degrade, or bounce RF signals.

Match impedance. “With RF frequencies, printed circuit board traces become a significant percentage of a wavelength,” said Schwartz. “Wavelength equals the speed of light divided by the frequency. One wavelength of a 2.4 GHz signal is about 1/3 meter or 1 foot long. Therefore, a 2-in. trace on the board equals 1/6 wavelength, which could reflect or emit signals like an antenna. Consequently, board trace dimensions are considered to be transmission lines and have characteristic impedances, so every input and output port or coupling in the RF path should be impedance matched.”

Watch antenna placement. Any metal placed close to an antenna, to within some fraction of a wavelength, will tend to couple to that antenna and act as part of it. “It’s the near field effect,” added Schwartz.

Another problem arises with multipath interference when the signal from a transmitter takes more than one path to the receiver. “Directional antennas will clear this problem,” added Marrs. Other options are an antenna hierarchy that focuses the radiated signals or repeaters.

The physical layer convergence procedure (PLCP) header seems to be the
problem. The 32-bit cyclic redundancy check (CRC) guarantees that
packets arrive at the destination uncorrupted. The PLCP header, though,
has only one bit for error checking, making it more vulnerable to
corruption, as though there is a “blind spot” where retries are not
seen by the protocol.

“If your network will remain within a building, 900MHz may be just fine,” noted Schwartz. “At this frequency, you will have about 28 MHz of available bandwidth. The 2.4 GHz in the US goes from 2.4 to 2.4835, so you have 83.5 MHz available.”

The 802.11 protocol is usually placed at the 2.4GHz level. If there are multiple 802.11 networks in the application, it may not be desirable to add more, especially if channel control is not available.

Part of the choice is based on whether the application needs to move small amounts of information a long distance or large amounts of information a short distance. “The speed desired on an 802.11 application,” added McCurdy, “isn’t available with 900MHz because of the modulation techniques used for the 802 protocol.”

An option is to use the 5GHz band. It sits above all the noise of the 2.4GHz range. “If you need the speed of a fast network but not the interference, try the 5GHz frequency range,” added McCurdy. “However, there are more installation challenges at the higher frequencies. You won’t necessarily obtain the same distances, and line of sight is important.”

But the 5 GHz band can be used internationally. Other factors to consider with these frequencies include:
• Most 900MHz devices are proprietary technology.
• The 802.11 protocol has fewer interoperability concerns.
• The law limits wireless applications to 100 mW of power.

The 802.11g provides more than 20 Mbps in the 2.4 GHz band.

Also, consider the hand off among multiple wireless devices. In automotive applications, for example, transponders on vehicles record data at each assembly station. As the vehicle moves from one station to the next, it may encounter a signal dead zone and stop the line. To prevent this, you can use an optional but little known feature of IEEE 802.11i identified as fast roaming. “As a vehicle moves down the line, a client radio or device talks to the access points. As it moves away from one point, it starts the authentication process to identify itself with the next radio, eliminating lost time,” said Ira Sharp, wireless product specialist, Phoenix Contact. “Anytime you have devices moving, consider fast roaming.”

Power. Wireless devices tend to consume more power. This is a particular issue with wireless sensor networks (WSN), which are often used in predictive maintenance applications that cover large distances. Most WSN devices operate from battery power. The attached sensors conserve that power by transmitting a signal only at a state change. A widely used protocol for this is IEEE 802.15.4.

Plan for growth. Design the ideal system. “You may not deploy all the components immediately,” said Hofmann, “but set the foundation for them. You’re going to add to it once you work with wireless and see how cool it is. You will find a range of projects that wireless can handle. Fortunately, these networks are flexible and can be expanded easily.”  

Topography. In industrial applications, the preferred topology is master slave. However, in wireless, a mesh layout works best. Mesh layouts allow all wireless devices to communicate with each other. “The setup is time consuming and an effort in reasoning,” said Hofmann. “If done right, though, you can set up self-healing rings where a signal can travel around a downed segment.”

The FCC. A large blunder is not obtaining required agency approvals. FCC approval, which is not guaranteed, is still needed even with the license free ISM bands. “If there’s too much RF noise, too much power output, then you have to go back to the drawing board,” noted Marrs. “They still police the spectrum, and they try to keep a level playing field to avoid the introduction of devices that overpower others.”

The government also imposes restrictions such as the maximum effective rate of power, which is limited to 4W. “Technically, you could push through more, but legally, the FCC won’t allow it,” noted Hofmann. So far, engineers are working around the regulatory imposed limitations with repeaters, antennas and gateways among other devices.

Often overlooked, serial-based legacy network systems can disrupt a wireless implementation. The wireless system may need to transmit that serial data, which not all RF modules handle.

Other tips include, have a good ground plane and control power routing. Use a star configuration where possible, not a daisy chain. On a circuit board for example, the power module would be central and connect individually to each component needing power.

And, in all of these recommendations, be aware of the limitations of commercial grade RF products, which were not suited for noisy industrial environments.