This paper was originally presented at the ERA conference "EMC 94 - The Clock is Ticking - Countdown to Compliance", 24th February 1994, ERA Report 94-0092.
Revised March 2000
This paper will discuss the options for EMC testing for compliance with the EMC Directive from the point of view of a manufacturer who wishes to achieve as much progress as possible, in-house, on a limited budget. It is not addressed to test houses nor to those manufacturers who have the resources to emulate most or all of the facilities of an accredited test house in their own premises. There are many small-to-medium sized enterprises who are able to dedicate a modest budget of several thousands or tens of thousands of pounds to an in-house EMC test set-up and who wish to gain the maximum benefit from so doing. This paper is addressed to them.
It will discuss only radio frequency (RF) EMC testing, which is taken to include ESD and conducted transients; low frequency power disturbances and mains harmonic emissions are not covered, despite their undoubted importance in some sectors.
Firstly, it will look at the difference in requirements between the various types of test: diagnostic, pre-compliance, full compliance and production testing. The minimum requirements for an RF EMC test set-up will be described. The size and nature of the EUT (equipment under test) has a bearing on the test facility that is finally chosen. Lastly, some upgrade paths from the minimum requirements are described.
|Conducted RF immunity|
|Radiated RF immunity||?||?||?|
|- essential, - likely, ? - maybe|
While a product is in development, the EMC effects of alternative design decisions and of remedial measures will need to be checked. Different layouts, different filters and different screening regimes need to be compared and this should be done as development progresses rather than at its conclusion, as otherwise a design which is ineffective in EMC terms may have excessive development resources spent on it. Therefore there is a need for a diagnostic test facility - which need not be truly equivalent to the compliance test set-up - to be available throughout the development phase.
A hidden advantage of this facility is that it enables a database of knowledge of the effects of alternative design choices to be built up. This will prove invaluable later when the possible EMC effects of post-design changes need to be assessed.
Once you have a prototype that represents the model which will be subject to full compliance testing, it is advisable to carry out some preliminary tests to the same standards. This has several purposes:
A full compliance test set-up must be capable of carrying out all those tests that the manufacturer intends to do in-house, to the required standard setup. The choice of which tests to do in-house and which at an external test house is affected by several factors, the most prominent being
The EMC Directive is silent as regards production EMC testing, although the Commission's explanatory guidelines  say that the manufacturer should take "all measures necessary" to ensure that the manufacturing process ensures compliance of each product with the Directive's requirements. Typically, this will involve some form of sampling regime or "golden product" comparison . The type of test set-up needed for this purpose is different to that needed for compliance testing: the emphasis here is on simple, go/no go comparative tests that can be carried out by semi-skilled technicians. A repeatable test jig with instrumentation that will detect expected variations due to production tolerances may be all that is required .
For emission testing, the basic tool is a spectrum analyser. Most commercial emission standards cover the frequency range 150kHz to 1GHz; in a few cases, for lighting equipment in particular, the bottom end of the range is extended down to 9kHz. Extended low frequency coverage tends to increase the cost of a spectrum analyser, since it has an impact on the required local oscillator and IF chain quality. Hence, low-cost analysers tend to be restricted in their LF coverage and may not even be specified down to 150kHz.
They may also not reach an upper frequency of 1GHz. The need for measurements up to this frequency is partly dictated by the type of emissions expected. Low performance electronic and electromechanical products are unlikely to suffer serious emission problems above 300MHz and indeed some product-specific standards (notably EN55014) do not at present extend above this frequency. If you make only low frequency analogue products with no inherent emission sources then you could quite seriously justify no emission testing at all. On the other hand, microprocessor circuits should certainly be tested up to 1GHz. Higher frequency tests, up to the 5th harmonic of the operating clocks, are now on the horizon for some products.
Other important cost trade-offs with spectrum analysers include frequency stability and settability, available resolution bandwidths (RBW) and quality of RBW filters, noise floor and sensitivity (which is also a function of RBW), and the presence or lack of digital signal processing functions such as markers, trace storage, `max hold' and averaging, and stored settings. All of these factors differentiate between the ability to use an analyser for measurement and diagnostics, and the availability only of a "window" into the frequency domain without the ability to interpret the contents of that window.
Most analysers intended for EMC use include CISPR-specific RBWs (9kHz and 120kHz) and the quasi-peak detector function. These are irrelevant for diagnostics. They are essential for full compliance tests, but not really necessary for pre-compliance work, where the error introduced by using the normally-available 10kHz and 100kHz bandwidths and a peak detector is swamped by other errors in the measurement setup. (Quasi-peak detection is important if the predominant emissions are pulsed broadband, such as from motors and switching operations.)
A particularly useful "accessory" for the spectrum analyser is a tracking generator. In many cases, this has to be ordered when the original instrument is purchased, and cannot be added on later. Including a tracking generator with the spectrum analyser greatly expands its measuring capability. The tracking generator is a signal generator which is locked to the analyser's measurement frequency and swept at the same rate. If it feeds a network whose output is connected to the analyser's input, the frequency response of the network is instantly seen on the analyser.
The tracking generator/spectrum analyser combination can be used for several EMC-related tests:
To sum up with regard to spectrum analysers: the analyser is the basic tool for EMC testing, and as with all tools, the best available for the job should be chosen. A good performance analyser is worth far more pro rata than a low-budget one: i.e., an analyser costing £10k gives much more than five times the usability of one costing £2k. Proper diagnostic work needs the ability to identify the frequencies and amplitudes of signals, determine their nature, and compare changes over the full frequency range of concern. Low-budget analysers do not offer this ability.
For a budget of around £10k, there are several analysers available. Alternatives from this level up to £25-30k include PC receiver adaptors and full specification measuring receivers (Figure 1). If the intention is to carry out full compliance emission testing with a high degree of equipment utilization, then these should be considered seriously. They are less suitable for diagnostic and pre-compliance work because of their lack of speed and flexibility.
The instrumentation needed for immunity testing is less clear-cut. As a minimum, an electrostatic discharge (ESD) simulator is advisable; ESD immunity tests are a requirement for virtually all electronic products that come under the product and generic immunity standards. There are several sources of ESD generators which are capable of contact discharge to the IEC 61000-4-2:1995 specification , available for around £3.5-4k. A contact discharge capability (as distinct from the older air discharge method) is a necessity since, as well as being a mandatory requirement under most product standards, it is far superior in terms of repeatability and hence its suitability for diagnostic purposes. A self-discharge loop is a recommended accessory, which provides an intense local H-field that can aid in diagnosing susceptible areas; the broadband nature of the discharge field allows it to partially emulate RF immunity testing in some circumstances.
The immunity standards also specify conducted fast transients to be applied in common mode on signal and power ports - in some cases, only to signal ports with connected cables greater than 3m in length. These are particularly upsetting for microprocessor circuits and the tests can easily be applied during development. As with the ESD test, the chosen test set-up and generator should be capable of a good degree of repeatability and close control of output level in order to facilitate diagnostics. Such generators are available for around the £5k mark. An alternative near-zero cost set-up is the "chattering relay": a high power contactor is wired such that its contacts interrupt its coil current to give a continuous on-off oscillation, and the EUT is positioned next to the wiring harness. This test is severely stressful for the EUT, but it is non-standard and hardly repeatable.
Surge and RF immunity testing are more problematical and are best considered as an extension to the minimum test instrumentation requirements.
If the intention is to leave RF immunity testing to an external test house, and ESD/transient generators do not need extra transducers, then the major transducers required for a minimum set-up are for emissions purposes: these include a LISN, a set of near field probes, and, perhaps, a set of antennas.
Unless your products to be tested are exclusively battery-powered, then a LISN/AMN (Line Impedance Stabilising Network or Artificial Mains Network) is essential. All CISPR-based standards require a conducted emissions measurement on the mains port. Making diagnostic tests on this port is also convenient and repeatable. The purpose of the LISN is to offer a defined RF impedance to the mains terminal of the EUT and to provide a defined coupling route from the EUT to the measuring instrument. The main concern in choosing a LISN is its current-carrying capacity - all devices will have a standard RF impedance as defined in CISPR 16-1 . (Other measurement standards, such as automotive or aerospace, call for different values of impedance characteristic.) Coupling from the LISN to the spectrum analyser should be via a transient limiter to protect the analyser's input. Best practice use of LISNs is described in , available in summary form from the Elmac Services web site.
Near field probes are essential for diagnostic purposes but irrelevant for pre- or full compliance, since they cannot be used to predict or measure far field performance. H-field (magnetic) probes are the more useful, but because of Faraday's law (which states that the coupling is linearly dependent on frequency) at least two units are needed to cover the frequency range from 9kHz to 1GHz with adequate sensitivity. VHF probes are easy to build in the laboratory (see Figure 2), or a commercially available set can be used. An H-field probe is used to detect regions of high di/dt flow on pcbs, or to detect discontinuities in shielded enclosures. It can also be used in reverse , to inject RF fields locally and hence pinpoint regions of high susceptibility.
Conventionally, radiated emissions measurements extend from 30MHz to 1GHz and are made on an open area test site (OATS). Not all products have to be tested for radiated emissions: the generic emissions standard  makes it clear that this is only necessary if the product uses processing devices operating at frequencies above 9kHz. Thus investment in radiated test equipment may be avoided if only certain kinds of product are to be tested.
Historically, the wide frequency range made it necessary to use at least two broadband antennas for full coverage; the universal standard types are the biconical for 30-300MHz and the log periodic for 300-1000MHz (some suppliers offer a breakpoint at 200 rather than 300MHz). For many types of product the investigation can be curtailed at 300MHz and thus only a biconical is required. These antennas are supplied with a table of calibration factors to convert from voltage at the antenna terminals to field strength. Accurate use of these factors is contingent on the antenna being used in an open area - proximity to other conducting objects, including the EUT, distorts the response of the antenna and invalidates its calibration. This problem is most acute for the biconical at low frequencies. A critical requirement for any antenna is that it should maintain balance, and should be linearly polarized. There are some circularly polarized antennas (notably the conical log spiral) which are used in some military tests, which are invalid for tests to CISPR-based standards.
The BiLog antenna (and its derivatives) is now available which covers the whole frequency range in one sweep. The principal advantage of this is a significant reduction in test time for a pre- or full-compliance test. Since diagnostic work is very often restricted to a subset of the full range, this is less of an advantage if regular compliance testing is not envisaged.
Sensitivity of the system (analyser plus antenna) can be a serious problem, especially if the more severe (Class B) emission limits are to be applied at a distance of 10m and above about 600MHz. Under these conditions the noise floor of a typical analyser, corrected by the antenna factor of a typical log periodic, is actually higher than the limit level of 37dB(mV/m). Testing at 3m gives a few dB margin but this is hardly adequate, and a better solution (in the absence of a test receiver) is a wideband pre-amplifier between the antenna and the analyser, to lower the effective noise floor by 20-30dB.
The minimum requirement for in-house diagnostics and pre-compliance testing is a dedicated test area with a ground plane. This test area may be in the development laboratory, provided that precautions are taken that adjacent equipment does not affect the testing and is unaffected by it. For instance, laboratory computers and test instruments may emit significant RF noise or may be disrupted by nearby ESD or transient events. They should be powered off during testing, or should be sufficiently distant from the test area. The test area itself should be supplied from a separate mains circuit, preferably but not essentially via an RF filter.
The ground plane is an essential part of all conducted EMC tests as it provides the RF ground reference against which measurements are made and stimulus signals are generated. It can consist of sheets of copper, aluminium or steel, laid on the floor of the facility and bonded to each other at close and regular intervals. The sheet thickness is not critical unless measurements are being made on equipment which radiates high magnetic fields at low frequencies (for example, high power switch-mode power supplies). The ground plane should extend by at least 10cm and preferably more than 0.5m around the boundary of any expected test setup, i.e. including the EUT, the test instrumentation, the interconnecting cables and any ancillary support equipment. For RF purposes it need not be connected to any external ground, but for safety it should be bonded to the mains safety earth.
This set-up together with the appropriate instrumentation and transducers will allow conducted RF emissions, ESD and conducted transients tests to be carried out to pre-compliance and perhaps to full compliance level. Radiated emissions tests require a test range as free as possible from extraneous reflecting objects and ambient signals. This is harder to arrange in an indoor development environment, but useful diagnostic information can be had from a set-up which positions the antenna at 1.5m high 3m from the EUT, without a ground plane between the two and with as much space as possible around the range. The checks should be carried out in horizontal polarization, with EUT cables oriented as far as possible in the horizontal rather than vertical plane, parallel to the antenna's plane of polarization. This will maximise coupling with the cables, which are very often the major source of radiated emissions. It will also give a rather more pessimistic result in the 30-200MHz range than would be found with a proper ground plane, since horizontally polarised ground plane reflections tend to reduce the measured value in this frequency range.
Full compliance radiated emission testing requires a well-specified OATS, while radiated immunity testing cannot legitimately be carried out without a screened enclosure or an enclosed cell, because of the need to prevent interference to other services. These further investments are discussed in section 5.
Both generic standards  include a passage which says "it may be determined from consideration of the electrical characteristics and usage of a particular apparatus that some of the tests are inappropriate and therefore unnecessary." Microprocessor-based products will normally require the full suite of tests, but many other types may not; for example, low bandwidth circuits such as analogue-only amplifiers may be regarded by inspection as being free of RF emissions, and apparatus containing only electromechanical devices may be said to be inherently immune at the prescribed levels of RF, ESD and transients. If your product range can be thus characterized as not needing some tests, then investment in that part of the test set-up is unnecessary. Product-specific standards will state more specifically which tests are required and what test setups will be needed, but often the same principle applies; for instance EN 55014-2 requires no immunity testing at all for non-electronic products within its scope.
All emission standards lay great stress on measurements of emissions on the mains port. If the apparatus to be tested does not have a mains port - if it is battery-portable or DC powered - then this test is "inappropriate and therefore unnecessary". A further factor is that the distinction between the environment classes of "light industrial" and "heavy industrial", and hence which part of the generic standards to apply, is made on the basis of how the apparatus is connected to the power network. Manufacturers of apparatus which is not mains-connected may have a greater freedom to choose which standard(s) to use.
Small units can be effectively tested for RF immunity within a TEM or GTEM cell, or may be subjected to conducted RF injection up to 230MHz. Their size therefore substantially affects the choice of, and investment needed for, an RF immunity test facility. On the other hand, large systems might only be subject realistically to in-situ testing, requiring perhaps a different set of instrumentation or diminishing the need for an in-house compliance facility.
A spectrum analyser alone is quite adequate for accurate measurements of continuous interference as generated, for example, by digital electronic clock sources. For pulsed broadband noise on the other hand, such as from motors or switching-type operations, it is susceptible to overload and non-linearity and should either be combined with a preselector, or substituted by a full specification test receiver, if proper compliance or pre-compliance measurements are needed. Generally, a spectrum analyser should be used with caution and a high degree of awareness on the part of the operator.
For full compliance radiated emission testing, a true open area test site (OATS) as discussed in section 16 of CISPR 16-1  is needed. The calibration of an OATS is described in terms of its normalized site attenuation (NSA), which is the overall loss between a transmitting antenna sited at the EUT location and the receiving antenna, corrected to remove the effect of the antenna factors. An acceptable test site for full compliance measurements must show a deviation in both horizontal and vertical polarization of less than 4dB from the ideal.
Achieving an adequate test site in the above sense requires considerable investment. First, there is the site itself; this should be level and free of obstructions, but close to a suitable building for the test and ancillary equipment. It must have a ground plane which extends at least 1m beyond the periphery of the EUT and the largest measuring antenna and which covers the area between the two. The EUT-antenna separation will normally be 10m (although 3m may be acceptable).
Further, for full compliance purposes a turntable large enough to accommodate expected EUTs, and a mast to allow a receiving antenna height scan from 1 to 4m, are necessary. For the fastest measuring time these would be automated under computer control.
Finally, there is the problem of ambient signals and weather. Nowhere in the UK (except in very sheltered valleys) are the broadcast ambients likely to be below the Class B emission measurement limits; on industrial estates, or close to RF-intensive locations such as transmitting stations, electrified railways or airports, other ambients are likely to be significant. The presence of ambients compromises the ability to carry out full-frequency tests and, especially if the ambients are transient, can markedly increase the time spent per test. British weather is also a problem and if the test site is to be available in all weathers then a weatherproof shed or tent around the EUT is needed. This should preferably be of fibreglass or plastic. Moisture absorption must be kept to a minimum, and ice or water should not be allowed to build up on any surfaces.
Because of these various factors, any in-house OATS is likely to be less than ideal. Many companies settle for nothing more than the use of the car park on a dry weekend with the antenna on a tripod and the EUT on a wooden table. This will normally exceed the 4dB criterion by several dB but is useful and usually acceptable for pre-compliance purposes. Dispensing with the ground plane is necessary because there is no height scan, and this means that a 5dB reduction of the applied limit level is essential. Whether compliance statements can be made from measurements taken on such a set-up depends on the margin below the limits exhibited by the measured emission levels. If this margin is greater than the NSA variability then it may well be realistic to do so, but this approach runs the risk of penalizing unit cost by requiring excessive EMC control measures on the EUT.
Investment in a screened chamber will be determined by the following factors:
The specification of the desired screened enclosure will be determined by which of these factors is dominant. The size of the chamber will be dictated by the anticipated EUT size and whether or not radiated tests are desired. If the only purpose for the chamber is (b) above, then a small conductive tent or a small room with conductive wallcoverings - paint, foil or wallpaper - is sufficient. This will typically give 40-60dB attenuation at HF and above, the actual attenuation depending on the care that is taken to seal all potential apertures and seams. The cost of such a room need be no more than £3-4k. The same technique can be used for (c) but the enclosure will need to be larger to allow separation of the EUT and antenna from the walls - a minimum length dimension being 5m plus the EUT size, which allows a 3m measuring distance. The presence of conductive walls will drastically distort both the antenna and EUT radiation patterns and the path attenuation, so that such tests cannot be used for compliance purposes. They do, though, offer EUT characterization and diagnostics without the distraction of ambients. EN50147-2:1996, Anechoic chambers Part 2: Alternative test site suitability with respect to site attenuation gives criteria for variability of site attenuation and "grey factor" which, if met, allow alternatives to open area sites (such as semi-anechoic chambers) to be used for radiated emissions tests.
Radiated RF immunity testing requires a screened enclosure to attenuate the test fields such that they do not interfere with other radio services. Properly speaking, this calls for an attenuation of 80-100dB which is beyond the capability of the DIY methods outlined above, but can be achieved by a fully specified and purpose-built enclosure. This is approaching an order of magnitude more expensive and less flexible in installation. As with c) above, the size must allow for separation of the EUT and antenna from each other and from the enclosure walls.
To reduce uncertainty of applied field strength due to wall reflections, which can easily exceed 20dB, it is preferable for the surfaces to be lined with anechoic absorber material. Full anechoic lining is expensive and the carbon-loaded foam type takes up extra space, requiring a larger room. This is not needed by ferrite tiles, but these are heavy, needing structural reinforcement. Various methods have been proposed as a "halfway house" to reduce the peaks and nulls of the field distribution, including strategically located blocks of carbon foam absorber around the EUT , partial rather than complete lining of the walls with absorber, and using a non-regular enclosure geometry, as well as deliberately offsetting the EUT-antenna axis from the chamber centreline.
An alternative to anechoic lining is to deliberately design a resonant (or "reverberating") chamber and use mechanical means such as a rotating paddle to vary the field distribution in real time. The result is known as a mode-stirred chamber. Its effectiveness increases with increasing frequency. With a suitable paddle and chamber design, a room size of 24 x 16 x 10 feet has demonstrated a field uniformity of better than 2dB above 600MHz, falling to 4dB at 400MHz, over one paddle rotation . A further considerable advantage of this method is that because the room is made deliberately resonant, much less radiated power is needed to achieve a given field strength. The disadvantages of mode-stirred chambers are their limited use at low-to-medium frequencies and the fact that the field is effectively modulated at the paddle rotation rate. Further developments may well reduce these problems; work is underway to draft a new standard (IEC 61000-4-21) especially for reverberating chamber techniques.
Small EUTs can be tested for RF susceptibility within a device known as a TEM cell. This consists of an enclosed transmission line of constant characteristic impedance Z0 within which a uniform field is developed (Figure 3). The EUT is placed between the plates in this field. The TEM cell's advantage is that because it is enclosed, high fields can be developed within it without the need for a surrounding screened enclosure, and so it can be used in the laboratory environment; also, it needs relatively little power to develop an adequate field strength. Its disadvantage is limited frequency range, since overmoding of the transmission line and consequent variation of the field structure occurs above a cut-off frequency dependent on the cross-sectional dimensions. The EUT should not occupy more than one-third of the space between the plates, and is not accessible for visual monitoring during the test. As an example, for an EUT height of 20cm the maximum frequency would be 200MHz.
A development of the TEM cell, known as the GTEM cell , removes the inherent upper frequency limit whilst retaining its other advantages. The GTEM is flared in contrast to the TEM cell's parallel plates and it is terminated in a wideband resistive absorber combination (Figure 4). It is considerably more expensive (prices vary from £31k to £100k depending on size) and therefore represents a sizeable part of the total EMC investment, but it is capable of taking a great deal of the EMC test load (including emissions diagnostics and possibly full compliance measurements), allows full frequency coverage in a single sweep, and may remove the need for high power amplifiers and antennas.
A screened room, TEM cell or GTEM is only part of the investment needed for radiated RF immunity tests, albeit a large part. The extra instrumentation needed for these tests includes:
It is easy to underestimate the power requirements, which are affected by the maximum required field strength at nulls in the chamber response, the directivity and gain of the antenna, the distance between antenna and EUT, and an allowance for modulation, losses and amplifier imperfections. Typically, at least 100W is needed up to 300MHz for 10V/m.
Because of the expense involved in all aspects of radiated testing, there is considerable interest in the alternative of conducted RF injection. IEC 61000-4-6:1996, Testing and measurement techniques - Immunity to conducted disturbances, induced by radio-frequency fields) has been published  for this purpose, and it offers both voltage and current injection methods. Also, there is a widespread body of expertise which has been developed around the use of bulk current injection for automotive and aerospace applications. This method is specified up to 400MHz, while the tests in IEC 61000-4-6 call for a frequency range of 150kHz to 80MHz with a possible extension to 230MHz depending on the dimensions of the EUT. The instrumentation needed for RF injection includes the swept or stepped signal generator as before, a power amplifier of substantially less power than for the radiated method, a set of current probes and a calibration jig, and a collection of coupling/decoupling networks for voltage injection. RF injection should be carried out within a screened room, but it can be of the low specification or DIY variety. It is eminently suited to diagnostic work by virtue of its repeatability and relative cheapness .
For pre- or full compliance testing a software control package for either or both emissions or susceptibility testing is worthwhile. Various packages resident on PCs written specially for EMC testing are available, costing around £2-4k. Apart from the host PC and GPIB controller card, all that is necessary to use this software is to ensure that all the instruments and accessories to be controlled are accessible via a GPIB interface. The benefits of software control far outweigh its modest extra cost and include:
Software will not remove uncertainties due to interpretation of test requirements and physical layout of the test setup, but it will reduce the routine operational duties of the test engineer and allow him or her to concentrate on these more important factors.