EMC legislation continues to evolve

The problem hit makers of domestic appliances that contain small, electrically noisy motors that could be tested to EN55014; an older, less stringent standard. This problem became apparent in the UK when hairdryer maker Helen of Troy UK (HoT) was taken to court because one of its appliances interfered with tv signals in the home: a red flag when it comes to the EMC Directive. Article 4 of the original EMC Directive specified that equipment needing certification should be built so 'the electromagnetic disturbance that it generates does not exceed a level allowing radio and telecommunications and other apparatus to operate as intended'. Had this gone to court, it might have shown the use of EN55014 was not in the spirit of the EMC Directive. But HoT pleaded guilty and the issues around the case were never tested. Revisions have attempted to reduce gaps such as these between the generic Directive and product-class specific standards. In the case of the hairdryer, the problem was the version of EN55014 in force at the time assumed the maximum frequency of intense emissions from domestic appliances would fall below 300MHz. The hairdryer motors were producing interference somewhat higher in frequency. The latest version of the standard, which came into force a couple of years ago, extended this to 1GHz. Revised in 2004, and finally coming into force before the end of the last decade, the EMC Directive has been slightly simplified in terms of language and clarifies situations where there are, potentially, different options open for certification. Under the old regime, to avoid problems such as those encountered by HoT, consultants were advising manufacturers to consider working to the generic Directive if there was a gap between it and a product oriented 'harmonised' standard. Now, the Directive makes it clearer that the preferred way of determining whether a product conforms is to use the relevant harmonised standard. The manufacturer still has the choice of deviating from the harmonised standard, but has to go through the job of preparing an EMC assessment, as with the generic Directive, and provide documents to support the assertion that the apparatus meets the Directive's requirements. Originally, to complete an assessment, manufacturers had to use third party test houses. One of the big changes in the 2004/108/EC Directive is to remove this need – although many companies will take that path as it will help remove doubt in the engineers' minds about how well the assessment has been put together, should a Trading Standards office repeat the spot checks that caught out HoT. A further change that may provide some scope for legal argument comes when the Directive considers the issue of electromagnetic immunity. For equipment to pass, it needs to have 'a level of immunity to the electromagnetic disturbance to be expected in its intended use which allows it operate without unacceptable degradation of its intended use'. Defining what 'unacceptable' means could prove troublesome. However, harmonised standards will often cover just what kind of behaviour is 'unacceptable' if they apply to the product. The phrasing is somewhat different to the original, which read 'The apparatus has an adequate level of intrinsic immunity of electromagnetic disturbance to enable it to operate as intended'. Europe is far from being alone in operating EMC emissions and immunity standards. Some 50 countries now have regulations; the good news being that, for the most part, they are the same standard. For example, the main core standard for determining appropriate levels of conducted and radiated emissions from IT equipment is contained in CISPR 22 – or EN55022 in its Europeanised form. The tricky part is that each country may not be using the same version of CISPR 22. One example is for radiated emissions above 1GHz all the way up to 18GHz – a response to the way that peak emission frequencies have crept up in many processor based systems. Taiwan introduced these last year and China brought them in earlier this year, but Europe will not enforce these regulations until the autumn. Japan decided to defer its introduction to match Europe. Within a few years, CISPR 22 will hit the end of the road as several new core standards are being developed in response to the way that electronic systems have evolved. CISPR 22 reflected the world of the 1980s, when the only devices with high processor clock frequencies were office bound computers. The new standards, such as CISPR 32 for radiated emissions, are designed for a much broader category of systems that come under the banner of multimedia devices. The intention is to get around problems that might arise in testing; for example, a digital camera that might have to connect to both PCs and TVs or, indeed TVs that contain computing equipment. Under the existing regime, the appropriate standards are different, depending on whether the device is classed as IT or broadcast equipment. CISPR 35 for immunity takes a different approach to previous standards. Instead of trying to shoehorn equipment into product types, the proposed standard will determine which tests a device will have to pass by the functions it performs, such as displaying images or playing music. There are some highly specific tests such as 'broadband impulsive conductive disturbances' that internet routers and similar devices have to deal with. As CISPR 32 has much in common with its predecessors, it is closer to final publication than CISPR 35, which saw a couple of false starts before IEC members decided on the idea of testing based on functions, rather than product type. In principle, CISPR 35 is due to be completed in 2013. Whether it will hit its expected deadline remains an open question. For the most part, filtering and shielding remain the weapons of choice for the EMC engineer. Even metamaterials have been invoked as possible candidates for EMI shielding; not so much to cut general radiated emissions, rather to bend and warp them around a potential trouble spot – such as a mobile phone user's head. The unusual refractive properties of metamaterials make it theoretically possible to reduce the level of a phone's radio signal that passes into a user's head when it is held close to the ear. Instead, the signal is bent so that stray emissions tend to pass out the side of the phone or through the hand. Other techniques attempt to go to the source of the problem in digital systems. Peak clock frequencies in computers have topped out for the moment because of concerns about power consumption. But they have been creeping up in consumer and industrial systems as their compute workloads have increased. Because clock signals tend to generate a lot of harmonics, manufacturers have encountered problems with shielding because short wavelengths leak more easily through tiny openings in cages and cans. In 1997, in the company's technical journal, Cornelis Hoekstra of Hewlett-Packard documented a technique that some of the design groups had put into action to cut EMI and which greater impact on high harmonics. The technique was to modulate the main system clock frequency to turn an interference spike into a smear with a smaller peak. This paper has become the standard reference for what has become known as spread-spectrum clocking (see fig 1). In the paper, Hoekstra explained why a greater effect was seen with higher harmonics. "The absolute value of frequency deviation increases linearly with harmonic number, so that spectral energy is spread over a larger range at higher harmonics, while the width of the filter over which spectral energy is measured is fixed." Hoekstra noted that the system clock is generally the best target. It is the biggest source of EMI because, in synchronous systems practically all activity is driven by this clock. Spreading this clock over a range of frequencies can therefore lead to a big reduction in EMI at the peak frequency. The first schemes to be used at HP were based on very simple square-wave modulation. The result was not a direct shift from one frequency to another – the signal feeding voltage-controlled oscillator would lag and overshoot. So a plot of frequencies looked more like a ringing square wave. Despite the simplicity of the scheme, it successfully spread the peaks so that equipment could pass FCC radiated emissions tests. However, the scheme could lead to jitter that was unacceptable for some circuits. So, some HP design teams decided to add sigma-delta modulation to smooth the frequency transitions; the resulting frequency plot looked more like a stepped square wave. This method has become the core of most dithered or spread-spectrum clock generators used today. Ideally, design tools would show design engineers what works and what does not, but this remains more an area of research than of development. It is computationally expensive and difficult to estimate accurately the level of interference that a given design will produce – or be affected by – during simulation. Testing tends to provide a more reliable answer, although EMC simulation is becoming more common in the area of SoC design to prevent noisy digital circuits from upsetting more sensitive analogue subsystems in the same package. Simulation technology will gradually improve, which should make it easier to assess which circuit design techniques will have the greatest impact on EMC. At the same time, standards around the world are being aggregated into more comprehensive documents, such as CISPR 32 and 35. This should simplify the job of working out which standards apply to a product. But it means that, as the second decade of the EMC Directive Europe draws to a close, the rule will be more legislative change, rather than less.