Helge Hornis, Ph.D.
Dr. Thomas Sebastiany
A conveyor control system activates a motor once every 12 sec over its 10-yr service life. An automotive carrier system must work 24/7/365 days a year. And a high-speed RFID system must operate near drives, solenoids, and power lines that generate electrical interference.
As these examples illustrate, industrial automation equipment must often perform reliably under some of the harshest conditions imaginable. One of the factors that make industrial conditions harsh is electronic noise pollution. Industrial devices and sensors must operate reliably in a caucophony of electromagnetic emissions both intentional and unintentional.
The key to eliminating electromagnetic noise is to first understand its origin. The right-hand rule relating to electric currents and magnetic fields — the one that every engineer learned in school — is more than just a nice final-exam question: It is the basis for our day-to day-problems with electromagnetic plant noise. Fortunately, it also provides the tools we need to design systems that operate well in those environments.
As a brief review, consider an electric charge traveling along a conductor. It creates a magnetic field that expands outward at right angles to the conductor. The strength of the magnetic field is proportional to the amount of current flowing through the conductor. Conversely, when a varying magnetic field crosses a conductor at right angles, it generates an electrical potential in that conductor. Thus a conductor that carries a changing current produces a constantly changing magnetic field that would induce a voltage in a parallel conductor. This is why control and power signals should not run too close together in parallel and only cross at right angles.
Keep emissions in check
It sounds almost too trivial to mention but well-designed hardware should have low field emissions. Unfortunately, price pressures make it too easy to abandon good design practice to save a few dollars. Properly designed PCBs with good component layout have few problems with emissions. PCBs without continuous ground planes, long wire traces, and extensive use of blue-wire jumpers are likely candidates for emission problems. A board that looks bad, with numerous extra leads evident, is likely to create a great deal of electromagnetic dirt.
Of course, these practices are unacceptable in equipment that must carry the CE mark. To obtain that mark, systems must meet IEC61000-6-2 and IEC61000-6-4 that regulate maximum emitted field strength, at what frequencies a device radiates, and how much noise at given frequencies a device can accept without failure.
To allow for devices that do not abide by the rules, digital I/O networks should follow a symmetrical designed so any noise pulse influences the I/O plus and minus sides by an equal amount. By working with differential signals instead of absolute, ground-referenced levels, these networks effectively cancel out external noise.
AS-Interface is an example of such an industrial I/O network taking this idea to the extreme. It enables bulletproof transmission of data and power over an unshielded, unterminated, completely topology-free, two-conductor cable several hundred feet long. RS-232 is a well-known counter example, where signal levels are ground referenced, making noisy plant environments problematic.
Shielding is another common strategy. Unfortunately, improperly executed shielding may actually force the hardware to deal with more noise, not less. Problems arise because not all so-called shielded cable is built the same. Traditionally, many shielded cables do not provide a physical ground connection into the coupling nuts. This means that a noise pulse is not diverted to a machine ground but instead travels along the cable, possibly jumping into the signal leads at one of the cable ends. At best these cables add mechanical strength and protection but have no positive effect in terms of noise protection.
An example of a well thought-out shielding/grounding concept is Pepperl+Fuchs’ Ident Control, an RFID sensing system heavily used in automotive, material-handling, and assembly applications. The shield is continuos from the R/W heads to the control interface. The metal housing of the control interface offers dedicated grounding lugs. As a result, noise that makes it onto the cable shield has no opportunity to get into the cable or interface housing but instead travels directly to machine ground. The 24-Vdc power input uses a metal feed-through designed to filter out and divert any noise coming in over the supply line.
Noise detection and error correction
There is a particular problem with external sources putting out noise at the operating frequency of communicating devices. Suppressing the noise at this frequency would prevent the device from functioning. An RFID system operating at 13.56 MHz, for instance, cannot suppress reception at this frequency or it could not read tags.
In a sense, the operating frequency of the hardware is its Achilles heel. To deal with this difficulty there must be a means to detect data corruption and ensure the intended information still gets through to the recipient. The solution is to combine error-detection methods with automatic transmission retries. Error-detection methods can include long checksums, signal-shape monitoring, and simple parity bit checks.
RFID system designers can do a few simple things to help out. Because noise frequently appears in short bursts, it is best to avoid sending large data blocks over the air-interface — the air gap between a tag and the read/write head. The longer the data block, the higher the probability that a noise burst comes at just the right (or rather wrong) time, rendering the entire data block useless. On the other hand, splitting the data into smaller units increases the chances several of them will get through before noise interferes with just one. It takes much less time to retry a single small packet than to repeat a large packet. Overall throughput becomes much better even with the added overhead of handling more packets.
Bus systems also need such detection and retry procedures. Getting back to AS-Interface, the Manchester II coding mechanism combined with a few additional protection bits creates a network with interesting performance qualities. Manchester coding transmits data serially as phase changes in the middle of the data bits. For Manchester II, a change from lowto- high represents a zero data bit, while a high-to-low change represents a data bit of one. This gives the signal a self-clocking ability that assures data synchronization by the receiver.
The most common error created by electrical noise are substitution errors. This means a zero data bit is interpreted as a one data bit by the receiver, or vice versa. Single bit substitution errors are typically detected using parity checking, a method where the number of one data bits are counted as an even or odd number. For example, if even parity checking is used and an odd number of one bits arrive, then an error has occurred and the packet is asked to be resent. Of course, if two substitutions occur in the same packet, then the parity check is maintained and the bad packet is passed along as valid. The AS-Interface is far more robust. Assuming typical plant-noise conditions and 24/7/365-days-a-year operation, it will take over 2,300 years for an undetected substitution error to occur in the AS-Interface.
In cases where retries are simply not possible, data must transmit via a method that includes forward-error correction. Data with forward-error correction includes redundant information so errors that occur in transmission can be corrected by the receiver without the need to resend the data packet. An example that uses forward-error correction is the optical readers that evaluate 2D DataMatrix codes. Frequently, these codes are read in high-speed applications where the object literally flies by so fast that the camera can only capture a single image. Consequently, there is no chance to retake the image. CDs, DVDs, and satellite data transmitted from probes orbiting far-away celestial bodies are other applications for forward-error correction.
The DataMatrix optical readers use a Reed-Solomon algorithm to evaluate the code. The Reed-Solomon algorithm is powerful enough to mathematically recreate missing data in cases where up to 30% of the original image is unreadable.
Simple measures that are inexpensive can also keep noise down. For example, the main control cabinet can be a source of many problems. It typically holds components that create strong interference fields, such as power supplies, contactors, and fluorescent bulbs. It also contains devices susceptible to noise interference, such as PLCs, I/O cards, signal converters, and HMI screens. It is critical that noise emitters and susceptible devices stay well separated. A metal barrier that’s neither painted nor anodized placed between the devices can help. And do not discount fluorescent bulbs. Fluorescent fixtures emit over a wide spectrum and should always switch off when the enclosure doors close.
When routing cables inside the enclosure, it is a common practice to separate the plus leads from the minus leads running each in separate plastic cable channels. This is actually not a good idea. Keeping both signal paths together reduces the emissions at a noise source as well as the susceptibility of the more delicate devices.
Outside the controls enclosure, cables frequently route through the plant in cable trays. Plastic is never a good choice. It is best to use a metal tray with a metal cover. Obviously, a cable tray with low-voltage control cables must not carry high-voltage power lines — especially if the cable is unshielded. Frequently, open metal cable trays without covers are used to bridge large distances. Placing cables in the corner of the metal tray offers some additional shielding from external noise sources by the sidewalls of the tray. But it does not offer the same protection as a totally enclosed conduit.
All in all, most component designers expend time, money, and effort to come up with products that work well in industrial applications. It is sometimes necessary to verify that installers have not negated design goals by taking unnecessary shortcuts.
Pepperl+Fuchs, (330) 486-0001,
How to thwart electrical Interference
Keep sensor cables short: The cable between a field-mounted sensor and a PLC I/O card is an antenna that can let noise into the system. The longer the cable the more it acts like an antenna. Instead of using long (and usually expensive) cable runs, install a highly distributed I/O system with field mounted connection modules. This keeps cable runs short and limits noise problems.
Separate control cables from power distribution: Any power wire is a potential noise source so it is a good idea to always separate low-voltage (≤24-Vdc) control leads from highvoltage power cables. Depending on the voltages and currents involved, experts suggest cable separations between 4 and 20 in. When power and control cables must cross paths, make sure that they cross at right angles.
Use ground fault detection: Any system or technology that uses differential signals should use some kind of ground fault detection. In many cases the hardware is designed well enough to run with a ground fault. But in these cases secondary correction methods are working hard to get the data where it needs to be despite the ground fault, making use of internal retries that cost time and limit system performance. More importantly, there will be a point where even the best corrective measures no longer work. The result is the automation equivalent of a cardiac arrest.
Create solid machine grounds: The right-hand rule makes it clear how an external, varying magnetic field induces stray currents. These currents must be given an easy path to ground, emphasizing the importance of a solid machine ground. When several cable trays span long distances it is important to create good electrical connections between them.
Install shielded cables: The use of shielded cables is typically a good idea and engineers should always follow manufacturer recommendations. But simply installing a shielded cable is not enough. Two systems can be at different potentials relative to ground if they are not properly grounded. If this happens, the shield acts as a connection to equalize the different potentials. This produces a high electric current flow over relatively thin shields and drain wires. Solid, heavy gauge grounding straps equalize ground potentials without large currents traveling through the shield. Conveyor belts that build up electrostatic changes fall into this category as well. Discharge brushes that dissipate the static buildup are an absolute must.
Forward-error correction and Reed-Solomon
While the idea behind any forward-error correction is quite simple, the details are not. The sender not only sends the data but adds additional, redundant information. More precisely, before transmitting the information, the sender creates redundant data based on a specific algorithm.
Once the receiver gets the combined data set (the user and redundant data) it “reverses” the algorithm to extract the message. The extra data gives the receiver the ability to reconstruct the correct message, even if the information is partially destroyed.
Reed-Solomon is just one such method with applications in many fields of industry and commerce. Much more rudimentary methods include a two-out-of-three majority process where each data bit is simply sent three times. Clearly, this is not efficient as it creates three times the necessary amount of data traffic. When Irving S. Reed and Gustave Solomon published their groundbreaking paper in 1960 they laid the foundation for CDs, DVDs, cell-phone communication, image transmission from other planets, and RAID systems holding vital business data.
The process of constructing the redundant data is mathematically complex. It is based on oversampling a polynomial function representing the data. Interested readers can get an idea of what is involved by reading one of several articles that are available on the Internet. For a quick explanation of how this idea works, look at the polynomial equation: 3x7 + 5x3 + 9x + 1. This is an example of a polynomial with degree 7 because the highest power of x is 7. Now here comes the tricky part. It is possible to completely define this function using eight unique data points. Sending those eight points lets the receiver reconstruct the polynomial and therefore reconstruct the data. By sending additional points the receiver is given the opportunity to correctly determine the data even if the transmission was less than perfect and some points were lost in transmission.
Now “all one need do” is find a polynomial function that describes the data to be sent, oversample it, and send those sample points to the receiver. Clearly, this is an oversimplification.
Reed-Solomon, as used in the common 2D Data Matrix code, has the capability of correctly evaluating a symbol where approximately 20 to 30% of the image is destroyed. The exact corrective power depends on how much user data is encoded using this method. For instance, a 16 16 code that holds 24 numbers can still be correctly reconstructed when six of those 24 numbers are in error.
In another example similar to Reed-Solomon, our brain has the ability to reconstruct words and sentences even if a significant number of letters are missing. Look at this statement in an apartment ad: Nclds hi spd Ntrnt. It does not take long to identify the real meaning to be “Includes high speed Internet.” By counting the number of letters and spaces in both versions (18 and 28, respectively) the degree of redundancy can be calculated as the number of unnecessary letters divided by the number of letters in the full version, or 10/28, which is approximately 36%.