Examining the Use of Fully Anechoic Rooms for Full-Compliance EMC Testing
The adoption of standards for using 3-m fully anechoic rooms for full-compliance testing will save time and provide more flexibility than open-area test sites.
Using fully anechoic rooms for full-compliance EMC testing is not far away. Both CENELEC and CISPR are considering changes to enable such testing. Major changes include dispensing with the ground plane and setting emission limits that are independent of the current open-area test site (OATS) limits. A draft standard, prEN 50147-3, will enable full compliance testing.1
Current standards are based on radiated-field measurements at a 10-m distance from the product. Compliant semianechoic rooms are generally 20 m long. The adoption of EN 50147-3 will allow room lengths less than half of this. Measurements in a free-space environment will save time and provide flexibility, because it will be feasible to simply scale the measurement distance and limit level according to size of the product being tested and the frequency used. Many test houses are already using a single facility for both immunity testing and precompliance emission testing in the frequency range of 30 MHz to 18 GHz. The use of ferrite tile absorber has made it possible to reduce room size and meet the ±4 dB requirement.
This article presents the reasons for adopting a product-to-antenna separation of 3 m in a quasi-free-space environment. This environment enables the length of the room to be less than 9 m, which reduces wall area by a factor of about four compared with a conventional 10-m chamber. Since the cost of the room and absorber is related to the surface area, the cost of a 3-m chamber would be roughly 25% of a 10-m chamber.
The currently allowable method for measuring radiated emissions is to place the product above a large metal ground plane with a distance of 10 m to a receiving antenna.2 In practice, reflecting obstacles such as buildings must be at least 20 m away. This requires a fairly large land area, and such a test site is usually an outdoor OATS. Television transmissions and other radio traffic often far exceed the limit to which a product is being tested, which creates extra work to discriminate the product's emission from the ambient signals.
International standards committees are considering the use of alternative methods for EMC testing. To eliminate ambient RF interference, one solution involves building a large screened room with walls and ceiling lined with absorber. Such absorber is not perfect, but this method has become acceptable because existing standards tolerate a relatively high level of reflections. The insertion loss between two antennas is allowed to deviate by ±4 dB from the value for theoretical insertion loss on an ideal site. This is known as the CISPR NSA (normalized site attenuation) criterion.2 Such facilities are termed semianechoic rooms because the metal floor is not absorber lined.
Radiated-emission testing is required over the frequency range of 30 MHz to 1 GHz. Industry is pressing for this to be extended to 2.7 GHz. Demand for even higher frequencies is soon to follow. To guarantee that a 10-m chamber meets the CISPR criterion, the screened (metal) room must be about 20 m long. The pyramidal absorber (RAM) must be up to 2.5 m deep to be effective at frequencies as low as 30 MHz, where the wavelength (l) is 10 m. In the last decade, this absorber has been replaced by a combination of 6-mm-thick ferrite tiles and 1-m-long RAM, enabling some reduction in room size. Such chambers typically cost $1.5 to $3 million. It is estimated that more than 90% of tested products measure less than 1 m3, which means that accurate field measurements could be made at distances of less than 10 m.
It is becoming increasingly common for EMC test houses to save on costs by using 3-m ground-plane sites, either outdoors (OATS) or indoors (semianechoic rooms). Some CISPR standards, such as CISPR 11, do not allow for a relaxation of the limit, whereas others, such as CISPR 22, allow the 10-m limit to be raised by 10 dB to compensate for the increase in signal at the shorter 3-m distance. Measurements of radiated emissions have shown that the differences between the received signal at distances of 3 and 10 m can vary between 2 and 15 dB. NPL (Teddington, UK) has made measurements on specially designed equipment under test (EUT) that radiates from a slot (see Figure 1a) or from a wire (see Figure 1b), driven by a battery-powered comb generator inside the box. The box measures 48 x 48 x 12 cm.
Figure 1. A special equipment under test was used to measure differences between signals at distances of 3 and 10 m. The battery-driven comb generator radiates from a slot (a) or from a wire (b).
The radiated emissions were measured at NPL's national standard ground plane, which is a continuously welded, 60 x 30-m steel sheet. It is flat to ±5 mm over 95% of the area, and the NSA of this ground plane is within ±0.1 dB of the theoretical value over the frequency range of 20 to 1000 MHz for horizontal polarization. This constitutes an OATS, and measurements were made for antenna-to-EUT separations of 3 and 10 m. The emissions were also measured in NPL's 3-m fully anechoic room, which has dimensions of 7.1 x 4.8 x 4.9 m. The NSA performance of this room at 15 points in a volume was just better than ±4 dB.
Figure 2 shows the signal of the slot mode measured at a distance of 3 m in a fully anechoic room, which is taken to be the reference plot. Only one face of the EUT was measured. The 3- and 10-m OATS measurements have been corrected by vector subtraction of the ground-reflected signal according to Smith's formula.3 Although the measurements show general agreement within 3 dB for the 10-m OATS and for the 3-m fully anechoic room, they show differences of up to 10 dB for the 3-m OATS. The measurements were also made in a 1750 GTEM cell, and these were also within 3 dB of the fully anechoic room measurements.
Figure 2. Measurements of the equipment-under-test slot mode. This graph plots the correlation of fully anechoic room measurements to open-area test sites at 3 and 10 m.
Figure 3 shows the same measurements taken using the wire mode. Height scanning was used on the OATS, but only one fixed height was used in the fully anechoic room. This could explain the discrepancy of up to 7 dB between the fully anechoic room and the 10-m OATS; however, the difference is up to 15 dB for the 3-m OATS.
Figure 3. Measurements of the equipment-under-test wire mode. This graph plots the correlation of fully anechoic room measurements to open-area test sites at 3 and 10 m.
These measurements demonstrate that the correlation between measurements made on a 10-m OATS and in a 3-m fully anechoic room is significantly better than their correlation with a 3-m OATS. It is anticipated that the correlation with a 3-m semianechoic room would compare even worse because of wall reflections. It is clear that measurements at 3 m should not be taken on a ground plane site. However, when performed in a fully anechoic roomeven with an NSA of ±4 dBthe result agrees favorably with the 10-m OATS data. Note that in Figure 2, strong ambient signals appear on the OATS plots but not on the plots for the fully anechoic room.
The advent of the 3-m fully anechoic room heralds a major advance in radiated emission testing, the simulation of a free-space site. Had low-cost absorber technology been available at the inception of EMC standards, measuring apparatus in free-space conditions likely would have been the chosen method. The method that has evolved effectively doubles the measured radiation because the signal reflected from the ground plane adds to the signal transmitted directly from the EUT to the antenna. Because the wavelength is relatively large compared to the size of the EUT, the EUT radiates in all directionsmaking the reflected signal almost as strong as the direct signal.
Some have justified this anomaly on the grounds that apparatus in everyday use will entail reflections from surfaces, such as a floor with a reinforcing wire mesh. It is better, however, to know the intrinsic radiation from an EUT in free space and, if necessary, estimate the effect of reflections for a given location by calculation. One might argue that if the radiation level is doubled, it is simply a matter of doubling the limit level. However existing standards overlook the fact that for horizontal polarization at frequencies below 100 MHz, the direct and reflected signals destructively interfere, yet the limit line assumes in-phase addition of the two signals. This means that a product can radiate excessively below 100 MHz, and the radiation will not be detected in the test. The introduction of a new standard creates an opportunity to remedy such anomalies in existing standards.
In addition, above 200 MHz the NSA test is performed with a log-periodic dipole array (LPDA) antenna in the measurement volume. However, it is also more stringent. EUTs are unlikely to have the 6-dB directivity of LPDAs, and their radiation must be assumed to occur in any direction. To compensate for this anomaly, the new standard uses a dipole antenna, which makes for a more realistic NSA test.
Small chambers, sometimes known as compact chambers, have been in use for many years. Such rooms, either fully or partially lined with absorber, may be as little as 6 m long. Compact chambers are used for precompliance testing and are not to be confused with fully anechoic rooms, which are intended for full compliance testing against a standard. A radiated-emission test is a fairly elaborate procedure, involving scanning the receive antenna in height from 1 to 4 m (to avoid nulls caused by the ground reflection) and rotation of the EUT in a horizontal plane through 360º.
Furthermore, since most of the emission comes from the power leads and communication cables, the cable layout must be varied until the maximum emission is detected. This procedure must be done in small steps over the frequency range of 30 to 1000 MHz. In a precompliance test, the nonconforming frequencies are identified by rapid frequency scanning so that full compliance testing can be done for a reduced set of frequencies. Recently, it has become common to line the floor with absorber to avoid time-consuming height scanning.
The FAR Project
Because small chambers cost less, the EMC community has sought to use them for full compliance testing for many years. In 1991, a CENELEC working group (SC210A/WG4) began probing the feasibility of introducing a standard with the important distinction that the limit was to be related to free-space emission levels rather than levels obtained above a ground plane. It started work on the standard for fully anechoic rooms, which presently is in final draft form as prEN 50147-3.
Gaining acceptance by the standards committees was a major hurdle to overcome. The challenge was to prove that the probability of an EUT passing the test in a 3-m fully anechoic room against the new free-space limit was the same as the probability of passing on a 10-m OATS against the existing ground plane limit. An example of a radiated-emission limit from CISPR 22 is 40 dBmV/m for Class A equipment in the frequency range of 30 to 230 MHz.
A consortium of eight European laboratories took up the challange of proving such equivalency. Sponsored by the European Commission over a two-year period, the FAR (fully anechoic room) project, which began in February 1997, was piloted by NPL.4 The project's aim was to determine the appropriate limit level and to demonstrate the equal probability of passing either test. The project partners were also members of SC210A/WG4, and ideas put forward by the working group were tested and honed by the project team. This resulted in a proposal that the existing 10-m OATS limit be raised by 5 dB for comparison with 3-m fully anechoic room results.5
Another finding of the FAR project was that it was overoptimistic to construct a chamber as small as a compact chamber. One goal of the project was to develop relatively inexpensive rooms. Using ferrite tiles, a room length of 6.5 m was conceivable, allowing the antenna and EUT to be more than 1 m from the end walls. Since height scanning was not required, reducing the room height to 3.5 m enabled the chamber to occupy one level of an existing building, a further cost savings. Existing anechoic materials, however, do not have a sufficiently low reflection coefficient to meet the ±4-dB NSA criterion in such small rooms. Moreover, the NSA test for FARs is more stringent than the NSA test for SARs (semianechoic rooms), because the chamber-wall reflections are relatively weaker in an SAR. This is the case because the undesirable wall reflections are compared with the direct signal between the two antennas, which, in an SAR, is almost doubled by the addition of the desired floor reflection. Taking this into consideration, a more realistic size of 9 x 6 x 6 m can result in a savings of 75%. The use of smaller rooms awaits improvements in absorber performance or the acceptance of chamber correction factors6 at frequencies below 120 MHz, where, coincidentally, it is most difficult to comply with the ±4-dB margin.
Suppliers of fully anechoic rooms should note that measurement uncertainty is associated with all measurements, including site evaluation. Suppliers cannot design a room that uses the entire ±4-dB NSA allowance. More realistically, it must be within ±3 dB, allowing ±1 dB for measurement uncertainty.
The current method of doing site reference measurements has proven inadequate. Free-space conditions cannot be achieved at a height of 5 m above ground for a biconical antenna separation of 3 m (see Annex 1 of Reference 1). Also in question is whether the NSA at the back of the volume should be weighted the same as the NSA at the front, because, after all, the EUT is rotated 360º in azimuth. For a small chamber, the NSA always worsens as the distance increases. Perhaps the back NSA measurement should not be mandatory.
Some uncertainty surrounds the validation of a 3-m fully anechoic room using a bilog antenna, which itself may be 1.4 m long. NSA measurements are less likely to be disputed if they are done with a biconical antenna from 30 to 200 MHz and a log periodic antenna from 200 MHz to 1 GHz, especially for chambers that border the ±4 dB limit.
A further finding was that radiation from cables made comparisons between fully anechoic rooms and OATS difficult because of the complexity of reproducing the cable layout in the two facilities. It is well known that 6-dB variations are common for full compliance tests conducted at two different accredited OATS, and that the cause of this variation lies mainly with the cable layout. Most manufacturers, therefore, attempt to achieve results 6 dB below the limit to compensate for this potential variation. The penalty is the increased cost of design and production measures to reduce emission levels.
Because of the strong incentive to make tests more reproducibleto within ±2 dBWorking Group 4 agreed that the cable layout would be fixed and that any cable that had to go outside the measurement volume (e.g., the power lead) would have its emissions suppressed by a ferrite clamp. This was another example of the working group's attempt to produce a new standard that improved on the old methods. As it turned out, and this should have been foreseen, the clamp altered the termination condition of the cable, shifting the emission maxima in frequency. Reproducibility is still a primary goal, and the working group continues to address cable layout and termination.
The project allayed the objections of those who feared that a distance of 3 m would be accompanied by near-field errors. Designers of dimension D microwave antennas use a distance criterion of 2D2/l for the minimum separation to avoid near-field effects. This implies a distance of about 7 m at 1 GHz. If the measurement is done at 3 m instead, the uncertainty is no greater than 1 dB, which is minimal compared to the overall uncertainty of EMC measurements. In fact, near-field errors become significant for distances of less than l/2p, when the electric and magnetic components diverge rapidly in magnitude. The near-field effects at 30 MHz on a 3-m range are less than 1 dB, becoming negligible above 80 MHz. The FAR project set a maximum diameter of 1.2 m for the EUT volume. In order to keep the uncertainties in check, the 3-m distance to the receiving antenna was measured from the front of the EUT rather than from the center as is the practice for semianechoic rooms.7
Transmission Line Modeling
It is possible to accurately compute the NSA performance of a fully tiled chamber using the transmission line modeling (TLM) code.8 The ferrite tiles are represented by a filter on the boundary of the TLM grid.9 NPL has compared TLM calculations with measured NSA for its fully tiled screened room with dimensions of 7.1 x 4.8 x 4.9 m. The agreement is generally better than ±1 dB. NSA can vary by several decibels depending on the room's dimensions and the room's length-to-width-to-height aspect ratios. TLM is a powerful tool for predicting whether a proposed room size has any chance of meeting the proposed NSA performance.
A Versatile Facility
Besides being compliant for radiated emissions, some apparatus must be tested for immunity to outside RF disturbances. Another raft of standards addresses immunity testing, which involves irradiating the product with a field of 3 or 10 V/m and noting whether it fails to function. A combined facility could be used for both emission and immunity testing. Furthermore, such a facility could be used for testing in the frequency ranges of 30 to 1000 MHz and 1 to 18 GHz.
New standards will require testing above 1 GHz to be done in a free-space environment. These standards should be formalized soon. An advantage of drafting a new standard, such as EN 50147-3, is the opportunity to rectify the limitations of existing standards. Unlike OATS, fully anechoic rooms are not affected by inclement weather. Improvements so far have included removing the ground plane, using a small dipole for NSA testing up to 1 GHz, and setting the separation distance from the front of the EUT volume instead of the center. Moreover, fully anechoic rooms use free-space antenna factors and therefore render reduced antenna-related uncertainties.
Testing efficiency can be improved by avoiding the need to height-scan the antenna in a fully anechoic room. However, above 400 MHz, more than one fixed height may be required for EUTs whose vertical radiation patterns have nulls. If measurements are to be made with uncertainties at less than an agreed amount, the separation distance must be appropriate to the size of the EUT. The draft standard prEN 50147-3 allows a 3-m range for equipment with a diameter of less than 1.2 m (including cable layout). One advantage of a quasi-free-space site is that the signal strength decreases inversely with distance from the EUT. This enables the use of any distance defined by the EUT size and frequency. According to guidelines in EN 50147-3, the distance, d, should be greater than 0.3 l, and the EUT dimension should be less than 0.4 d. There should be no objections to changing the emission limit accordingly. For example, if the distance were increased from 3 to 4 m, the limit level should be decreased by 20 log (4/3) dB.
CISPR is very cautious about changing limits and must be certain that the fully anechoic room would truly enable this flexibility. The scaling of limits for different ranges in free-space conditions will provide more-comparable results than scaling of measurements made above a ground plane. Free-space antenna factors can be used, resulting in lower uncertainties. The use of emission measurements in a fully anechoic room as an acceptable test method was introduced to CISPR/A at its annual meeting in June 1999.
Early EMC measurements were made in parking lots, fields, or unlined screened rooms. The unlined rooms acted like a cavity, producing uncertainties of more than ±20 dB. In the 1980s, test sites were standardized using a metal ground plane outdoors, and ±4 dB became regarded as an acceptable NSA deviation.
The increase in ambient RF interference is a strong driver for the use of anechoic chambers. However, 3-m semianechoic rooms can produce unacceptably large errors. Work by the FAR project has proven that a 3-m distance can be used as low as 30 MHz. In a fully anechoic room with an absorber-covered floor, products were shown to have an equal probability of passing the emission test as on a 10-m OATS.
The FAR project has recommended that the limit for a 3-m fully anechoic room be 5 dB higher than for a 10-m OATS. This will allow for the increased power at the lesser distance, but the doubling of the signal due to the ground-plane reflection no longer applies. It is physically more sound to measure the radiated field in a free-space environment than above an artificial metal plate. Measurement distances and emission limits in a fully anechoic room could be scaleable according to EUT size. Although emission measurements of a special EUT in a 3-m FAR agree with those made on a 10-m OATS, agreement is poor for a 3-m OATS and is predicted to be worse for 3-m semianechoic rooms.
The working group improved other aspects of the measurement method, and these improvements are reflected in prEN 50147-3. Improved methods of NSA evaluation have been introduced. Instead of using a directional antenna (above 200 MHz) as in current standards, the room is evaluated using a broadband dipole antenna to simulate the possible radiation in all directions from a product. The distance of 3 m is measured to the front of the EUT volume.
The greatest contributor to poor reproducibility is radiation from power and communications cables attached to a product. This is also true of existing OATS and semianechoic room measurements. EN 50147-3 attempts to improve reproducibility by requiring a fixed cable layout. Another issue still to be resolved is the treatment of floor-standing products. A proposed solution is to place them on a conducting strip that is grounded to the screened room.
A 3-m fully anechoic room is estimated to cost less than one-third the cost of a 10-m semianechoic room. Although certain large products will require a larger chamber, it has been estimated that more than 90% of products can be contained within a diameter of 1.2 m and can be tested at a 3-m range. While the cost of an FAR may not be as low as some manufacturers and users had anticipated, there is the potential for a large cost savings in the operation of an FAR as a combined emission and immunity facility up to 40 GHz. A possible future step is to further reduce the cost of rooms by using fewer ferrite tiles. Although reducing the number of tiles will worsen performance below 120 MHz, the wavelength below 120 MHz is much larger than the EUT, so chamber correction factors could be valid. The fully anechoic room is usually a better environment than an OATS for achieving acceptable reproducibility of measurements.
1. EMC Basic Emission Standard, prEN 50147-3, "Emission Measurements in Fully Anechoic Chambers," CENELEC SC210A/WG4, Brussels.
2. CISPR Publication 16, "Specification for Radio Disturbance and Immunity Measuring Apparatus and Methods, Part 1:1999, Apparatus" (for NSA criterion, see Annex L), central office of the IEC, Geneva, Switzerland.
3. AA Smith, RF German, and JB Pate, "Calculation of Site Attenuation from Antenna Factors," IEEE Transactions on EMC 24, 1982: 301316.
4. "Development of New Measurement Methods of the EMC Characteristics in Smaller Relatively Inexpensive Fully Anechoic Rooms," EU 4th Framework DG XII, SMT-CT96-2133, available from M. Alexander (NPL; Teddington, UK).
5. MJ Alexander, "The Use of Fully Anechoic Rooms for Full Compliance EMC Testing," IEE EMC York '99 Conference on Electromagnetic Compatibility, July 1213, 1999, Savoy Place, London.
6. CENELEC Report R110-003, "Guidelines on How to Use Anechoic Enclosures That Do Not Fulfill the Requirements Regarding Normalised Site Attenuation for Pre-Compliance Tests of Products," CENELEC, Brussels, November 3, 1995.
7. EN 50147-2, "Anechoic Chambers: Alternative Test Site Suitability with Respect to Site Attenuation," CENELEC, Brussels.
8. J Clegg, M Alexander, JF Dawson, J Jee, AC Marvin, B Loader, SJ Porter, "A Method of Reducing the Number of Ferrite Tiles in an Anechoic Chamber, IEE York '99 EMC Conference, July 1213, 1999: 5964.
9. JF Dawson, "Representing Ferrite Absorbing Tiles as Frequency Dependent Boundaries in TLM," Electronic Letters 29, no. 9, 1993: 791192.
Martin Alexander is principal research scientist for the RF & Microwave Group at National Physical Laboratory (Teddington, Middlesex, England). He recently joined the editorial advisory board for Compliance Engineering. He can be reached via e-mail at firstname.lastname@example.org.
Back to May/June
Table of Contents