Design teams are trying to find ways to overcome the apparently random nature of electronic interference.
At the end of October last year, the US Federal Aviation Authority decided to relent on its ban on electronic devices in the passenger cabins of commercial airliners. Although mobile phones will not be allowed to place or take calls, passengers will be allowed – if operators agree – to read e-books, play games and even surf the Internet if the aircraft has a Wi-Fi service during takeoff and landing as well as at cruising altitude.
Air regulators and industry players have taken a cautious approach to dealing with electronic devices. Testing has revealed how outwardly benign equipment can cause problems for other electronic systems.
In the mid-1990s, Boeing encountered several occasions when an onboard electronic device caused problems. In 1995, the company bought a laptop computer from a passenger on a 737 flight after the crew suspected it of causing the autopilot to disconnect while cruising. Although the machine emitted interference above the limits for aircraft at the time, subsequent tests could not reproduce the problem. Experiments after similar incidents proved inconclusive although tests have revealed potential problems when devices are designed to generate radio signals.
For example, a type of cockpit display made by Honeywell could go blank for several minutes when subjected to interference from Wi-Fi signals, albeit at higher levels than those supported by many consumer devices. That led to a ban on the use of Wi-Fi devices by pilots – who would otherwise have used tablets and similar devices to get real-time weather updates – as well as passengers. Before the aircraft that use the display can carry Wi-Fi services, the displays will be upgraded.
The ongoing uncertainty around using electronic devices in-flight reflects the problem of determining electromagnetic compatibility (EMC) – whether a device will affect other systems adversely or will itself generate errors because of a strong source of interference nearby.
The EMC Directive
The European Union's EMC Directive will have been in force for 20 years by the end of this year. Although the original EMC Directive dates back to 1989, it did not come into force until January 1992 and even then EU member states used transitional rules for close to four years. Since then, the generic directive has been through one major update, which has tied up some of the loose ends of the original as well as easing the job of declaring compliance for smaller manufacturers – largely by taking away the requirement for third-party EMC testing.
The legislation concentrates on the interference between systems – to ensure that complete products from different manufacturers do not interfere with each other when placed close by. But the increasing amount of electronics content in motor vehicles and the energy-supply grid is forcing manufacturers to go beyond what the legislation currently tells them to do.
The problem for any electrical or electronic system is that it simply cannot help but generate interference – the only question is one of degree. Modern electronics, particularly when it comes to digital design, treats circuits as discrete entities. But the motion of electrons through the circuits creates fields – described elegantly by James Clerk Maxwell in the 19th Century by his famous set of equations – that will interact at some level with neighbouring elements.
The effect of accidental antennas
Keith Armstrong, founder of Cherry Clough Consultants, describes electrical conductors as "accidental antennas" – delivering electromagnetic energy through fields that couple with other accidental antennas, creating stray currents and voltages. If strong enough, these voltages and currents can flip over logic bits on a bus or cause significant measurement errors in an analogue circuit.
Potentially, the fields can have a wide-ranging effect, particularly when dealing with the magnetic fields created by large current flows – the conditions found in power systems that go into vehicles and the smart grid.
"We definitely see growth in automotive," says Chris Aden, product marketing manager for the Saber system-level simulation tools at Synopsys. "We are seeing the increased content of power systems and switching power systems driving customers. Switching power is the main issue. With those, the signals are a lot louder."
While the EMC aggressors are becoming more strident, other parts of the vehicle are becoming more sensitive to noise. "There is more electrical content in cars overall now. You have low-power networks that are close enough to the power buses being used for high-power switching."
Andreas Barchanski, senior applications engineer and EMC market development manager for EMC simulation specialist CST, says: "Automotive systems designers are fighting with increased issues in EMC. There are much higher amounts of electronics packed inside their cars compared to 20 years ago.
"You now have a lot of steering assistance and other systems operating in a closed space, which result in more and more EMC issues. It's why the automotive OEMs are putting in rules for EMC that are more strict than the legal requirements. On the other side, the increasing electrification of vehicles through hybrids and electric traction also poses more EMC problems because you have much higher currents to deal with inside the car."
A further issue, which is also likely to affect smart-grid systems, is the way in which different subsystems interact. Barchanski says: "If you consider a car being a system, it might happen that different parts are delivered by different suppliers. You do not have the ability to design the complete system yourself. So you have to rely that the other components of the car are holding to these EMC standards. This is why companies are working to standards that are much tougher than the legacy standards: to make sure that everything when assembled will still work."
Brad Brim, senior staff product engineer at Cadence Design Systems, says another sector focusing on EMC issues is medical, an area where several electrical safety standards have been upgraded recently. "Customers here have approached us with more emphasis on EMC compliance. In medical, you can say someone's life is at stake. They are being driven by the compliance to do better.
Brim adds: "There is also the concern that new compliance standards are on the horizon and not just that but, as compliance enforcement hasn't been that strict in the past, that the rules will be enforced more rigidly."
Shielding the interference
Testing remains the key way in which electronics design teams discover how well their equipment will conform to standards, although it may not uncover potential issues when it comes to immunity from interference. And it is a procedure that comes late in the project, beyond the point at which it is cost-effective to make changes. A further issue that for lightweight, space-constrained systems such as those used in aerospace and automotive designs may not be able to add the shielding that could be used elsewhere.
Barchanski says: "It's not just about passing or failing the test but how can you pass the test with the lowest effort possible? Say you have an input filter that allows you to pass the test. A revised design with fewer components will save you more money. For mobile phones, consumer products and cars, just saving one or two capacitors on a board is a huge saving in the end because of the volume of production."
Waiting until the testing stage can be risky, because of the added complexity of dealing with an entire system, especially one that involves internal or external cabling. Movement of the cables can lead to changes in EMC performance.
"Often you face a problem where the standalone components work well but once the system is assembled an EMC problem arises. When you plug a cable in, the system starts emitting more than expected. You may have something geometrically complex to deal with and something quite electrically complex. The combination of these parameters – physical geometry, electrical drivers and terminations – makes up the EMC performance. That's what makes EMC difficult to analyse," Barchanski says.
The overall complexity of the system makes it difficult to take a PCB design and feed that to a tool that will tell the engineering team whether or not they have a problem before the prototype is built, although the fundamentals of EMI are well understood.
"The simulation techniques are well known and established: 3D Maxwell equations have been in use for more than 20 years," says Barchanski.
Brim claims Bell Northern Research developed a simulator that resembled the Spice engine used widely for analogue circuit simulation, but for EMC calculations. "It worked out the voltages and currents that a circuit would see so that they could be back annotated into a simulation of their design. Engineers could ask: 'If I have all these trace currents going around, what will the average look like and what will be the emissions from that?'
"They found at the time they got a pretty good representation," says Brim, but the technique was not widely adopted.
The problem is a combination of getting good data into the tool in the first place, finding the compute power to execute accurate, system-level simulations and then making sense of the data they produce. Barchanski says the models quickly become large and unwieldy but: "There are ways to tackle this complexity, such as segmentation. I can divide my systems into subparts and simulate them as standalone entities, then bring them together to simulate the whole system. But you need to know where to divide it.
"The other approach is to use brute-force computation and adding more processing hardware," Barchanski says, noting that engineers in Asia tend to favour the hardware-centric approach whereas those in Europe are more likely to try segmenting the design.
"They are both valid approaches but if you segment you need to put in some knowledge beforehand. With brute-force approaches, you get a complete picture in one step but it takes longer to simulate or needs more hardware."
A recent development in computing to use graphics processors (GPUs) as computation engines has made the brute-force approach more appealing. "GPU computing provided a step change in the computing power you can get."
Perfecting the design
There is a limit to how accurately simulation can represent actual system behaviour, Brim says. "It's probably not going to be feasible for the whole system where the system includes cables and enclosures. Subtle changes to the design can lead to big differences in performance. For example, how does cable get soldered to the board. How is it mounted? Connectors have huge geometric complexity. Do you approximate that to make the simulation simpler? You need so much detail for success that I don't think it will happen.
"Although some say 'look at how computer resources have proliferated' and they can compute huge things real fast. But there is so much you have to specify," Brim adds, noting that EMI shielding itself adds additional complexity because of the 3D shapes that are involved.
Brim says one approach is to take account of those differences in small tests. He cites one example of an engineer working on consumer electronics designs who has been testing eight different cable types under various circumstances, such as if they touch the board or not, to see how those changes affect performance. "The cables don't look very different but you can get compliance on four and failures on the other four."
Although simulation is becoming more common, design-rule checking (DRC) can play a big role in reducing the risk of a design flaw being exposed during testing.
Brim says: "We call it physical DRC to find potentially bad things. The rules say don't route across something that's devoid of return paths. You combine that with subsequent simulation but using rules you have a powerful way of determining that your boards are quiet, although it might not help with cables.
"Even the rules checks are not performed by an EDA tool, it's being done by a simple script or just by an EMC engineer looking over the shoulder of the design engineer, saying 'don't do that'. However, I think that will change in the direction of greater automation if the rules get more strict and people are not able to meet compliance."
For the time being, EMC analysis is likely to remain a mixture of rules, testing and increasing amounts of simulation.