Designing-In Low-Cost Noise Immunity

How do manufacturers of retail products achieve low emissions, low sensitivity to external interference, and a competitive price? To find out, we asked several companies how they dealt with noise immunity test requirements. While they didn’t reveal trade secrets, they did confirm that dealing with EMC is a design issue, not a matter of luck.

Design is an iterative, compromise solution to a number of incompatible requirements. Nothing new there, but with EMC added to other technical and economic constraints, finding the best solution is even harder.

A common approach incorporates EMC techniques into the design as early as possible. This means that as the design progresses, EMC-related changes can be incorporated along with other PCB layout modifications. Following this course should result in better EMC performance without last-minute surprises, but by implication you will need to test prototypes or parts of prototypes much earlier in the design cycle. Changing how you deal with EMC will affect your test strategy.

Mobile Motor Controller

As an example of a product that must pass rigorous RF immunity tests, consider a battery-powered wheelchair. Penny & Giles Drives Technology of Christchurch, England, develops and produces motor drives for manufacturers of wheelchairs throughout the world. You may not think this is a terribly demanding application, but it represents a unique combination of design challenges: user safety, cost, efficiency, size, weight, performance, maintenance, and EMC compliance.

According to Jo Crane, the chief engineer at Penny & Giles, plastic cases often used for the motor drives have cost, weight, and environmental-protection advantages and require little or no maintenance. On the other hand, they are not much help in an RF immunity test. Because injury or death could occur due to an out-of-control wheelchair, testing is done from 26 MHz to 1 GHz at very high field-strength levels—initially 40 V/m or more, but recently reduced to 20 V/m. No one wants to contemplate the specter of a citizen’s-band radio or cell phone taking control of a wheelchair halfway across a busy intersection.

It must be stated that powered wheelchairs have an excellent safety record even when compared to manually operated chairs. Nevertheless, if the immunity test levels have been set unrealistically high, it’s because regulators of powered wheelchairs are erring on the side of safety.

“Radiated immunity testing was the main problem we started with because of these very high levels—higher than most military and civil aviation levels,” said Mr. Crane. “When we were testing above 50 V/m for a Federal Drug Administration (FDA) 510(k) file submission, we caused a fire in the chamber at a large test facility. One of the conical foam absorbers caught fire because of the very high field strength.”

Although the motor drives are tested separately, the entire wheelchair must be tested to ensure that pickup on motor cables and the wheelchair frame doesn’t affect the controller operation (Figure 1). Mr. Crane continued, “The manufacturers should get fully accredited tests, so if they choose to use our facility, it’s really to improve their confidence of passing the actual test. If they get through here, they’ve got at least an 80% chance that there won’t be a problem. Also, if we make changes, we can advise them then whether or not they have to retest.”

Dealing with high field strengths is a matter of understanding where the energy is going and making sure that’s where you want it to go. Controller inputs and outputs are filtered to shunt interference around sensitive circuitry, ferrites dissipate noise signals in other cases, and good board design ensures minimal pickup of the noise that remains.

Mr. Crane elaborated on design practices, including the use of shielding in low-cost products. “Shields are applicable in some areas, but not in most of our work. You have to remember that the unscreened cables that go in and out of the controller are antennas. It doesn’t matter how good the enclosure is, if you’ve got a very good antenna taking all the noise into the enclosure, it’s pointless.

“It comes down to good techniques,” he concluded. “You may use ferrites to limit the amount of energy coming in. In our case, we try to shunt or reflect the energy straight back out so it doesn’t cause any havoc once it gets in there.”

As far as self-induced noise goes, Penny & Giles deliberately restricts the slew rates of its pulse-width modulated drives to minimize emissions. Wheelchair motors are rated for 220 to 300 W, but the controllers can deliver up to 4 kW peak. Good layout practices ensure that the microcontroller clock does not propagate outside the controller.

“The key is having our own facility where we can test the first iteration of board design as soon as it’s available,” Mr. Crane said. “If there’s a need for an iteration, we can put the changes straight into the next board design because there will have been other changes as well as EMC-related ones. Then we can test that new design as soon as it becomes available. This capability contrasts strongly with when we used to travel to a separate test facility. Then we hoped there wouldn’t be any board changes required because we really couldn’t investigate them.”

White Goods Control Panel

A second example of immunity testing involves domestic laundry equipment—washing machines and dryers. Within the United States, these appliances are not regulated for EMC to the degree they are in Europe. Nevertheless, as David Tarnowski, a senior engineer working in product development at Whirlpool, said, “We pretty much self-regulate ourselves. Even though our EMC measures don’t add customer value directly, it’s really a means of preventing customer dissatisfaction.”

Whirlpool has an anechoic chamber at the company’s Dayton, Ohio, technology center, where radiated immunity tests could be run. But they have found that electrostatic discharge (ESD) and electrical fast transient (EFT) events are the most prevalent causes of problems. The ESD and EFT tests are easier to apply than RF immunity tests and, in any case, are required by Sears, one of Whirlpool’s major customers.

Mark Herman, a lead hardware/software engineer at Whirlpool, described some EMC problems presented by a membrane-switch control panel on a new combined washer and dryer model: “The metal chassis under the membrane overlay and the printed circuitry in the overlay formed a good capacitor. We found that the energy induced into the capacitor radiated onto our PCBs very easily.

“For both ESD and EFT events, the microcontroller would reset, which is not an acceptable response,” he continued. “In addition, we have many LEDs on the overlay, and we don’t want customers to see a flashing LED during an ESD or EFT event.”

Because of the company’s emphasis on both quality and low cost, there are no quick fixes. As Mr. Herman explained, based on first-hand knowledge, “You have to go back and dig through the wires and [PCB] traces to figure out how the noise is getting in and how to minimize its effect.” See Figure 2.

Mr. Herman first tries to determine the areas of the PCB that are being affected and inserts inductors and capacitors as appropriate to exaggerate or eliminate the problem. Once he understands how the radiated energy is affecting the circuitry, it’s a case of re-examining the board layout in detail.

“Usually, no matter how well you’ve done a layout, there are loops you just didn’t picture,” Mr. Herman said. “Correcting the problem is made more difficult because most of our boards are single-sided to reduce cost. For example, you can’t position tracks on opposite sides of the board so that currents will cancel as you do in a switching power supply design on double-sided boards.

“Nine times out of 10 it’s a grounding problem,” he summarized. “You’ve made a mistake with the grounding, or you’ve forced the noise energy back through a path you didn’t intend to.”

Whirlpool designers make a point of sharing the best practices they have developed on their separate projects. They hold peer-based design reviews to ensure that the same field-proven PCB design methods are applied to each new generation of machines.

In the domestic appliance business, the largest change in control technology was made several years ago when electronic designs replaced entirely mechanical controllers. However, the older electronic technology still being used by the industry won’t be available forever. Based on their hard-won practical experience, Whirlpool engineers can advise semiconductor manufacturers about the noise-immunity features required on new ICs.

Areas that are particularly sensitive to noise include power supply inputs and clock lines. For example, any clock inputs should have a Schmitt trigger or similar circuit to provide hysteresis to reduce the effects of noise entering the chip. And power supplies aren’t likely to be well regulated. It must be assumed that transients on the AC supply will affect logic-device voltages.

Computer Peripherals

Low-cost plastic enclosures often are exploited to give mice, joysticks, and keyboards distinctive appearances, but they don’t do much to counteract the effects of ESD and EFT events. According to Steve Stegner, a compliance engineer at Microsoft, ESD and EFT are the important tests used during the development of these peripheral devices. Radiated RF immunity requirements usually are met without difficulty.

“The problem is in trying to get repeatability among our test locations,” he said. “For consistency, we use the same host computers, monitors, and peripherals in all of our labs. We also use the same models of ESD generators and EFT simulators to improve consistency of results. And there is special software to monitor all the functions of the devices so the testers can see if anything’s been activated or moved.

“Because there’s a big radiated effect from EFT testing, we keep all our cables lined up in the same way using plastic clamps,” he explained. “For ESD testing, if we go to high voltages that are likely to arc to the coupling plane, we use test fixtures to keep our devices above the plane by a certain amount.”

In EFT testing, an amount of the pulse energy radiates from the power cable and couples to the DUT cable. Some of the energy also enters the host computer system and travels down the DUT cable. To address the radiated effects, Mr. Stegner first tries a shielded DUT cable. He said that shielded cables generally work better than unshielded ones, but with the appropriate use of ferrites and layout changes, peripherals with unshielded cables will pass their immunity tests.

Conclusion

The three examples show just how seriously EMC performance is taken by major companies in a range of industries. Testing confirms that the EMC precautions already designed in are performing correctly. On-site testing gives each of the companies immediate access to the necessary test equipment should an unexpected problem be reported by a user or to expedite new designs as quickly as possible.

All three companies cite attention to design details including cable routing and PCB track layout which highlights the impracticality of trying to fix EMC problems at the final test stage. Many times, problems can be fixed, but not as well as if they had been addressed earlier in the design process.

As Mr. Herman of Whirlpool said, “Companies don’t like to throw a great deal of money at making controls, but yet you must have a quality control that’s safe. You can’t just put a barrage of inductors on a board or a big choke on the input of the power supply. Those approaches don’t go down too well.”

Copyright 2000 Nelson Publishing Inc.

April 2000

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!