Radiated-emission test specifications set a limit below which the electric field (E-field) strength at a given distance from a piece of equipment under test (EUT) must lie. It is a requirement of ISO 17025 that the uncertainty of measurement be ca lculated ("General Requirements for the Competence of Testing and Calibration Laboratories," 1999). The uncertainty is an estimate of the deviation from the true value of the measurand.

Figure 1. UK national standard ground plane at NPL, measuring 60 x 30 m.

For most measurement quantities, the smallest uncertainty that can be achieved is determined by national measurement institutes (NMIs). The test laboratory will use this as a starting point on which to base its measurement uncertainties. Usually, the test laboratory employs equipment calibrated by calibration laboratories. The chain of measurements running from the test laboratory to the national laboratory is known as traceability. Each NMI pools its measurements with those of other national laboratories in a process of international intercomparison. The uncertainty calculated by the test laboratory will always be greater than the minimum uncertainty internationally agreed upon.

NMIs that have signed a mutual recognition agreement (MRA) commit to participation in international intercomparisons that form the basis of the key comparison data published by the Bureau International des Poids et Mesures (BIPM; http://www.bipm.fr). Under the MRA, the NMIs agree to recognize each other's certificates of measurement at the level of the uncertainties that is the basis of the key comparison data. Antenna factor is the responsibility of Working Group GT-RF of the Consultative Committee for Electricity and Magnetism, part of BIPM (http://www.bipm.fr/enus/2_Committees/CCEM.shtml). This article addresses the traceable use of antennas for measuring electric field strength.

The Calculable Dipole Antenna

An achievable target uncertainty for E-field strength for test laboratories is ±1 dB. This is equivalent to an uncertainty of +12.2/­10.9% in linear terms. It is a combination of the uncertainty of antenna factor and power. Antenna factor (AF) is the ratio of the strength of the E-field in which the antenna is immersed to the voltage at the output of the antenna transmission line of a given characteristic impedance, commonly 50 W. The output is connected to a receiver, which is effectively a power meter. National laboratories can at best measure E-field strength to an uncertainty of approximately ±0.2 dB, or 2%. This is achievable through use of a calculable dipole antenna and a national-standard power meter. Analytical formulation of AF for a resonant dipole antenna has been shown to be accurate to better than ±0.05 dB.

To be able to use a dipole at more frequencies than just the single frequency at half-wave resonance is desirable. Numerical computation has been used to show that AF can be calculated with an accuracy of better than ±0.2 dB over bandwidths exceeding 200%.^1,2 The calculable dipole antenna is made up of two parts, a dipole element and a balun. The dipole impedance and the transmission loss between a pair of elements are calculated numerically, and the balun S-parameters are measured by means of a vector network analyzer. The two parts are combined in an impedance matrix to give the total transmission loss of the antenna pair, from which AF is calculated.

Verifying AF. In order to establish a national standard, AF must be verified by measurement. The great difficulty in establishing free-space conditions for omnidirectional VHF dipole antennas can be overcome by creating a perfect mirror over which to measure the dipoles and then relying on Maxwell's equations to compensate for the presence of the mirror. A. A. Smith adapted the Friis formula for the three-antenna method for use above a ground plane.³ With a sheet metal ground plane at least 20 x 15 m in dimension and flat to within ±5 mm, AF can be verified to better than ±0.15 dB between 60 MHz and 1 GHz. It is necessary to verify that the effects of edge diffraction are a small component of this uncertainty; likewise, reflections from the antenna supports and feed cables. The national-standard ground plane at the National Physical Laboratory (NPL), shown in Figure 1, is made of continuously welded galvanized mild steel plate, measures 60 x 30 m, and is flat to within ±5 mm over 95% of the area. Its size ensures that edge effects on site attenuation are negligible for horizontal polarization and, for vertical polarization, less than ±0.15 dB about the major axis and less than ±0.05 dB about the minor axis. The large size also enables verification of calculated AF down to 20 MHz.

It is only necessary to verify AF at two or three frequencies, because once the principle of calculability is established it applies at all frequencies. For increased confidence, measurements are compared with calculations for a variety of combinations of antenna heights and separations, including near-field separations. Particular care is required at the higher frequencies. At 1 GHz a ground plane flatness of ±5 mm is required, and the antenna heights have to be set within the same tolerance relative to the mean ground plane level. Also, because commonly used antenna support structures have dimensions on the order of one wavelength at 1 GHz, special measures have to be taken to reduce reflections, including reflections from the antenna feed cable. One approach is to support the antennas on expanded-polystyrene blocks and to replace the coaxial cables with optical fibers via RF/optical links on the antenna output connectors.

A "perfect" ground plane is necessary for validating the calculable antenna, but once the computed AF is proven by measurement to be within a given uncertainty, such perfection can be dispensed with. The calculability of the AF now depends only on the mechanical reproducibility of that antenna design. The NPL design yields free-space antenna factor (AFfs) at an uncertainty of less than ±0.2 dB. When the antenna is used above a ground plane, that plane, in order to retain this low uncertainty, must be of such quality as to ensure that a sufficient image of the antenna is formed. For this to happen, the ground plane must extend far enough in all directions around the antenna. The NPL ground plane ensures that the uncertainty of ±0.2 dB also applies to the concept of antenna factor in the presence of a ground plane (AFgp). In normal antenna usage, AFgp can differ from AFfs by up to ±2.5 dB, so accurate knowledge of AFgp can be important.

Minimizing Antenna-Related Uncertainties. If a test house wants low antenna-related uncertainties, it can use a calculable dipole antenna and follow the guidelines for ground plane construction given in CISPR 16-1:1999 (Specification for Radio Disturbance and Immunity Measuring Apparatus and Methods, Part 1:1999, Apparatus). A ground plane size of 20 x 15 m should suffice for antennas placed centrally on the site and separated by less than 10 m, for frequencies greater than 30 MHz.

The uncertainty of AF will also depend on the flatness and electrical conductivity of the ground plane, an edge termination sufficient to reduce edge effects, and the magnitude of unwanted reflections from the antenna feed cable and the antenna support and surrounding objects. These factors are less demanding than the specification needed for the "golden" ground plane that was used to establish the principle of the calculable dipole once and for all. In addition, the uncertainty of AF will depend on the measured S-parameters of the balun.

Figure 2. Typical antennas used for EMC testing: biconical, log-periodic, tunable dipole, calculable dipole, and monopole.

Design of a Calculable Dipole Antenna. The following details of design for the calculable dipole are recommended: The antenna elements should be light so that their dielectric support at the feed point can be small and therefore less electromagnetically intrusive. The gap at the dipole feed point should be minimized. It should be possible to disconnect the elements from the balun so that the set of nine S-parameters of the three-port can be measured. This willenable regular checks on balun performance and allow the antenna to be used without matching pads, thereby maximizing the available antenna gain. It should also be possible to mount the elements some distance in front of the antenna support structure in order to minimize the reflections from it. The calculable antenna is ideal for site evaluations, for which the balun should be capable of handling sufficient power to enable ambient interference to be overcome. The dipole could also be used as a reference for EMC field probes calibrated in anechoic chambers; for example, to achieve a field strength of 3 V/m at a distance of 3 m, the balun power handling should be rated at least 2 W.

CISPR 16-1. A description of the calculable resonant dipole appears in CISPR 16-1:1999. The construction of the antenna is outlined and the method of verifying the performance of the antenna by measurement is described in detail. The source code is given in an annex of the publication, so that coupling between a pair of resonant antennas above a ground plane can be calculated. This 41-page addition of text to CISPR 16-1 is the result of work done since 1994 by an ad hoc working group of CISPR subcommittee A. It represents the first of three units of work on the measurement of E-field strength for radiated-emission testing, specifically, the establishment of a calibration test site (CALTS) on which to measure antennas. The second part concerns acceptable methods of calibrating antennas, and the third part, the verification of open-area test sites (OATS) on which emission measurements are made. The second part should be in final-draft form after the June 2000 CISPR annual meeting.

Antennas Commonly Used for EMC Testing

Current standards for radiated-emission testing set limit levels for E-field strength over the frequency range of 30 MHz to 1 GHz. Biconical and log-periodic dipole array antennas (called log antennas for short) are commonly used for this testing. Some typical antennas employed in EMC testing are shown in Figure 2. Draft standards are currently extending the range for limit levels to 18 GHz, but the scope covers up to 400 GHz. The overlap frequency range of 200 MHz to 18 GHz can be covered by commercial log and horn antennas. (Generally, horn antennas are used above 1 GHz, but they are not discussed at length here because they are only used in a free-space setup and their calibration is relatively uncomplicated.) At 200 MHz the horn antenna is rather large; above 3 GHz the log antenna can be rather delicate and its cross-polar performance degraded.

Antennas used to set up high field strengths for immunity measurements do not have to be calibrated; however, a calibration does provide the necessary functional checks. The important design principle for immunity antennas is that their return loss be sufficiently high to eliminate the need for very high power amplifiers.

The application of antennas in EMC testing has been described generally elsewhere,^4,5 so this article focuses on three intrinsic problems and three preventable problems involved in their use. The frequency range under discussion is 30­1000 MHz, and the figures are worst-case deviations for both biconical and log antennas designed for this range. The biconical antenna typically has a dipole tip-to-tip length of 1.4 m and is designed to operate at 30­300 MHz. The log antenna, made up of half-wave-resonant elements, has a typical length of 0.6 m in the direction of propagation and is designed to operate from 200 MHz to 1 GHz. Uncertainties quoted in this article are based on a coverage factor k = 2, which provides a level of confidence of approximately 95%.

Problems Intrinsic to Use of Antennas for EMC Testing. Intrinsic to the use of an antenna with a ground plane is the interaction of the antenna with its image. The AF of a horizontally polarized biconical antenna can change by 1.8 dB when its height is changed from 1 to 4 m. Another problem is that the antenna is never receiving the signal only in its boresight direction (the maximum of the polar diagram). The signal reflected off the ground plane is incident at a greater angle from boresight than is the direct signal from the EUT. When the antenna is vertically polarized, the intended signal is suppressed typically by 0.5 dB on a 10-m site and by 3 dB on a 3-m site, depending on frequency and antenna type. The third intrinsic problem relates to log antennas. Their phase center moves with frequency, so the separation distance of the antenna from the EUT is not the specified 3 or 10 m at all frequencies. For a log antenna operating above 200 MHz whose reference point is at the element that resonates at 300 MHz, this causes uncertainties of 0.2 dB for a 10-m separation and 0.8 dB for a 3-m separation. These are guideline values and apply to antennas operating in free-space conditions. Over a ground plane, assuming height-scanning for maximum signal, the values could be slightly greater.

These three intrinsic problems are factors only at ground plane sites (and the second and third problems exhibit their greatest uncertainties at the minimum site distance of 3 m). They do not apply to free-space sites. A correction can be applied for the phase center in a free-space environment. An NPL calibration certificate gives phase-center information specific to each model of log antenna. Since 1991, CENELEC Working Group SC210/WG4 has been evaluating fully anechoic rooms (FARs) as a means of making emission measurements in free-space conditions. The group produced a draft standard, prEN 50147-3, which was circulated to national committees for comment in December 1996. The comments were acted on, and a second version of the standard was circulated in February 2000 (prEN 50147-3:January 2000, Basic Emission Standard, Part 3: Emission Measurements in Fully Anechoic Rooms). NPL research has found that whereas emission results from an EUT on a 10-m OATS and in a 3-m FAR were in general agreement within ±3 dB, the results on a 3-m OATS differed by up to ±8 dB. The differences will generally be larger in 3-m semianechoic rooms (SARs) because these are designed, for economy, to just achieve the ±4-dB normalized site attenuation (NSA) criterion (found in Annex L, CISPR 16-1:1999), whereas most OATS will be within ±1 dB.

Originally, measurements at a distance of 3 m were countenanced as an additional measure only if it was impossible to measure EUT signals in the presence of ambient interference, and only at frequencies where there was interference. Now it seems to be common practice to cover the whole frequency range on a 3-m site. In fact, in some countries the 3-m ground plane site predominates over the 10-m site, and the regulatory authorities appear to have lost sight of the reasons why a 3-m separation was to be used only as a last resort. A consortium of eight organizations, sponsored by the European Commission, conducted a two-year study on the viability of testing at a distance of 3 m in FARs, which showed that the elimination of the intrinsic errors ensured good agreement between 10-m OATS and 3-m FARs.^6,7 The study also revealed the large variations in results caused by the treatment of EUT cables. The variations were just as large on the OATS as in the FAR. (The issue of reproducibility of measurement has not yet been completely resolved.) Prescribing and fixing cable layouts in order to improve reproducibility is gradually becoming acceptable. For example, see Figure 9 in ANSI C63.4:1992, Methods of Measurement of Radio Noise Emissions from Low-Voltage Electrical and Electronic Equipment in the Range 9 kHz to 40 GHz.

Bilog Antennas. Bilog antennas, combining aspects of biconical and log-periodic dipole array antennas, were introduced by Chase EMC Ltd. in collaboration with York University in 1994. The advantage of the bilog is that a swept frequency measurement from 30 MHz to 2 GHz can be performed without changing antennas. The disadvantage is that the antenna is about 1.4 m long, which creates a potential for increasing measurement uncertainties, especially at a 3-m distance. These increases will certainly be less in a 3-m FAR than on a 3-m OATS or in a 3-m SAR, but a full study to quantify them has yet to be done. In the meantime, it is advisable to use separate biconical and log antennas to achieve lowest uncertainties, especially for NSA measurements in the evaluation of sites.

Preventable Problems Connected with Antenna Use. The three problems with the use of antennas alluded to previously as preventable are balun imbalance, cross-polar impact, and erratic resonances in log antennas.

Balun imbalance. A problem that had serious ramifications for EMC testing in the past is imbalance in the baluns of biconical antennas. Many antennas in current use have poor balance. One popular low-cost model was supposed to be improved by reducing the AF at 20 MHz, but measurement reproducibility suffered greatly. A connection to ground was removed and this caused changes in readings of as much as ±15 dB when the antenna was vertically polarized. This led to reports of problems with reproducibility of emission testing and various theories about how to use and calibrate antennas, nearly all of which became redundant once imbalance was found to be the cause. Notably, the use of a pair of these "improved" antennas for the NSA site validation measurement made otherwise-good sites look unusable. The model was modified in about 1995, but some original ones are still in use.

NPL identified this problem in 1990 and devised a balance test that has been a routine part of antenna calibration ever since. It involves recording the signal received by a vertically polarized antenna, then inverting the antenna and noting the change in signal. Any change exceeding ±0.5 dB is caused by common-mode current on the portion of feed cable that is aligned parallel to the antenna elements, whose radiation field interacts with the antenna field.

Other balun designs, particularly the more-complex high-power baluns, exhibit imbalances up to ±5 dB, mostly below 150 MHz. It is possible to reduce the effect of imbalance by placing ferrite clamps on the antenna feed cable, with one next to the antenna output connector and others spaced at intervals of about 20 cm. A balun test is to be described in a future issue of CISPR 16-1.

Cross-polar level. Log-periodic dipole array antennas have asymmetrically mounted dipole elements. This construction leads to cross-polar radiation. A cross-polar level of only ­20 dB can result in an uncertainty of ±0.9 dB in the measurement of the signal with the intended polarization. In most radiated-emission tests, this component of uncertainty is usually less than ±0.9 dB because the higher of the horizontal and vertical readings is taken. (One reading being larger than the other implies that the lesser signal will be having less cross-polar impact.) However, the worst case--when the magnitudes of the horizontal and vertical components are equal--is possible, and it is desirable to have a cross-polar rejection of 20 dB or higher. Many log antennas now in use in test laboratories have cross-polar rejection only as low as 14 dB at 1 GHz, and much worse than that at higher frequencies. A cross-polar test is to be described in a future issue of CISPR 16-1.

Erratic resonances in log antennas. A good log antenna design achieves a smooth, monotonically increasing antenna factor with rising frequency. Most commercial antennas qualify as good in this respect. However, after a period of use, some models exhibit erratic narrow-band resonances that are caused by the breakdown of RF contact between the dipole element and the body (transmission line) of the antenna. These erratic resonances can cause underestimation of the signal by, typically, 10 dB. The breakdown of contact is exacerbated when the antenna is used for high-power transmissions, as in immunity testing. Then, arcing in the tiny gap between the element and the body occurs, along with a buildup of oxide.

The EMC tester cannot easily know when the log antenna is being affected in this way; such knowledge would necessitate a transmission test before each use of the antenna. A return loss measurement usually reveals whether one or more resonances exist, but a full transmission test is needed to measure their magnitude. This can be an awkward problem for calibration laboratories because it can hold up the calibration for a customer while the repair is negotiated. One solution is to design antennas with elements welded to the body. These are available from some manufacturers.

Table I displays components of a typical uncertainty budget for radiated-emission testing attributable to well-designed antennas, free of the three preventable problems just described.

Uncertainty Component				Uncertainty (dB) with a Ground Plane (3-m OATS or SAR)					Uncertainty (dB) in a Fully Anechoic Room (3-m FAR)
				Bicone		LPDA1			Bicone		LPDA
AF calibration				±1.0		±1.0			±1.0		±1.0
Antenna directivity, antennas vertically				+0.5 ­0.0 polarized		+2.0 ­0.0			±0.0 0.0		+0.2 ­0.0
AF variation				±1.8 with height		±0.5			0.0		0.0
Antenna phase-center variation, assuming height-scanning fo rsignal maximum on OATS, SAR.Phase center can be corrected for FAR.				0.0		±0.5			0.0		±0.2
AF frequency interpolation when measurement interval is >10 MHz, <50 MHz)				±0.3		±0.3			±0.3		±0.38
Antenna balun imbalance coaxial output cable parallel to dipole elements				±1.0		±0.0			±1.0		±0.0
Cross-polarization, assuming cross-polar suppression of 20 dB				±0.0		±0.9			±0.0		±0.9
Measurement distanceerror ±2 cm				±0.1		±0.3			±0.0		±0.3
Height of antenna above ground plane, height error ±2  cm				±0.1		+1.02			­0.0		No ground plane
Mismatch: antenna-receiver				+0.9/­1.0		±0.3			+0.9/­1.0		±0.3
1 LPDA = log-periodic dipole array. 2 The uncertainty in field strength of an LPDA above a ground plane at fixed height is greater than that achieved with height-scanning, because a height scan ensures in-phase components of direct and ground-reflected signals.

Table I. Typical achievable uncertainty components of commonly used antennas vertically polarized over a ground plane and in a fully anechoic environment, for an EUT-to-antenna separation of 3 m. All are rectangular distributions except mismatch, which is U-shaped, and antenna factor (AF), which is normal with k = 2.

The CISPR Reference Antenna

The CISPR reference antenna is the half-wave tuned dipole. The use of broadband antennas is allowed only if their antenna factors are referenced to the free-space AF of the tuned dipole and if their directivity is within certain bounds (CISPR 16-1, Paragraph 5.5.5.2). This situation was acceptable in the days when uncertainties with broadband antennas were in excess of ±2 dB and the free-space AF of a resonant dipole could potentially be calculated to an uncertainty of less than ±0.5 dB. However, the way CISPR 16-1 mandates the dipole to be used contains a serious technical flaw. It is assumed that emission testing involves varying the height of the antenna between 1 and 4 m over a conducting ground plane. The AF of a dipole tuned to 30 MHz differs from the free-space value by up to 4 dB in this height range, a fact not considered by the standard. In practice, the difference is closer to 2 dB at the height at which the signal maximum is measured in emission testing. But this is still a long way from the ±0.5-dB uncertainty that the user believes the reference antenna is meant to have. Interestingly, the AF of the biconical antenna at 30 MHz varies by less than 0.2 dB over the height range 1­4 m because at this frequency it behaves like a short dipole.

A second technical flaw raises another problem. The tuned dipole is 4.8 m long at 30 MHz, so the standard states that the antenna's minimum usable height when vertically polarized is 2.75 m. However, the signal maximum occurs at a lower height. Thus, although a biconical antenna would be able to measure the signal maximum at the correct height, the reading of the reference antenna, which is at the "wrong" height, takes precedence. And this is not the only problem. At the higher reference height, the radiation pattern of the antenna suppresses the ground-reflected signal, and so the reading made by the reference antenna is doubly in error.

Another irony relating to the dipole length of nearly 5 m is that it was one reason for the original recommended separation of 30 m for EMC testing. When it had become clear that the specified emission limits were too low to be measured at 30 m, a compromise 10-m separation was established. By that time, the biconical antenna had been accepted as an alternative to the resonant dipole. And now a separation of 3 m is gradually winning acceptance by some national regulatory authorities, with the consequence of higher uncertainties on a ground plane site.

The biconical antenna would be far better suited to serve as the reference antenna under the conditions that CISPR 16-1 states must be applied to the tuned dipole. In an emission test conducted above a ground plane, the AF of the biconical antenna differs by less than ±1 dB from its free-space value and by less than ±0.5 dB over most of the frequency range. A vertically polarized antenna couples even less to the ground plane, and the AF differs less than ±0.5 dB over its entire frequency range. For products characterized by vertically hanging cables, the maximum reading will in most cases be the vertically polarized signal, so the uncertainty of measurements attributed to AF variation will be less than ±0.5 dB. Some laboratories nowadays offer uncertainties of less than ±0.5 dB for broadband antennas, and a test house that uses the tuned dipole for EMC testing will be regarded as very odd. Why? Because, first, the dipole has to be adjusted in length every time the frequency is changed, and, second, tuned dipoles are the antennas most sensitive to change in antenna factor over a ground plane.

CISPR subcommittee A has produced a document on the calculation of uncertainties for EMC tests that bears the draft number CISPR/A/256/CD. This draft draws on original work published by UKAS in document NIS81.⁸ Until the CISPR dipole reference is changed, uncertainties have to be related to the result that would be obtained by the free-space AF of the CISPR dipole. This has the undesirable effect of making the "true" value of the E-field strength that which is measured by the CISPR dipole rather than the value that actually exists at that point in space.

A better reference antenna would be the broadband calculable dipole. With it, antenna factor is calculated at the relevant height above the ground plane. Also, it can be used over a wide frequency band. NPL uses four dipole lengths to cover the frequency range 30 MHz to 1 GHz, which ensures that the AF at the band edges is not too high. The principles of this antenna are described in CISPR 16-1:1999, but the source code works accurately only for resonant antennas. A code that gives very accurate answers away from the resonant frequency is NEC2 (Numerical Electromagnetic Code, downloadable from ftp://ftp.netcom.com/pub/ra/rander/NEC). A user-friendly version for the PC is NEC-WIN Professional 1.1 (Nittany Scientific Inc., Hollister, CA). The broadband calculable reference antenna with customized software is available commercially from Schaffner-Chase Ltd. (Capel, Surrey, UK).

Antenna Calibration

Calibration Methods. Of the three methods of antenna calibration, the fundamental one is the three-antenna method. The unknown antenna factors of any three antennas can be determined after the insertion loss between each of the three possible pairs of antennas is measured. Three simultaneous equations based on the Friis formula can be written to calculate the unknown AFs from the three insertion loss readings. Many conditions that must be met in order to obtain the correct gain or antenna factor are described in textbooks on antennas. A key one is that unwanted reflections from the surroundings should be negligible.

The standard-antenna method involves comparing an antenna whose AF is unknown to a standard antenna whose AF is known. The unknown antenna is set up in a plane wave of arbitrary field strength, and then its place is taken by the standard antenna. The difference between the two readings is added to the standard AF to give the AF of the unknown antenna. Field uniformity does not have to be as good with this method as with the three-antenna method, and errors can be kept small if the antenna under test and the standard antenna have similar dimensions. Care has to be taken when this method is used with antennas such as log antennas, whose elements are distributed along the axis of propagation. A variant of this method is the standard-field method, in which a known field is set up using a standard antenna with a known input power.

The third method relies on the amenity of the antenna to calculation of its antenna factor. The resonant dipole antenna described above is an excellent candidate. The AF depends on the measured dimensions of the dipole element, the measured insertion loss of the balun, and the calculated impedance matrix of the coupling between a pair of identical elements. The AF of the dipole element can be known--with the aid of an accurate vector network analyzer to measure the balun S-parameters--to an uncertainty of less than ±0.2 dB over a broad bandwidth. Factors of more-complex antennas such as the biconical type can be calculated to uncertainties of less than ±0.4 dB.^9,10

Volumetric Measurements for Site Validation. Test sites for EMC testing are qualified by means of a comparison of the coupling between a pair of antennas on the site to be qualified with an expected value. Pre-1999 versions of ANSI C63.4 and CISPR 16-1 define this expected value as one calculated via a simple formula. Unfortunately, this formula does not take into account either the coupling of the antenna to its ground plane image or the radiation pattern of the antenna, both of which can have effects amounting to as much as 4 dB (see Chen and Windler). Table G.4 of CISPR 16-1 contains theoretical correction factors for this, but they apply only to tuned dipoles. Moreover, the site validation method relies on antenna factors that have been measured on another test site whose qualifications may be no better than the site being validated.

On the other hand, the site-reference method requires the use of a reference site, and it is a very simple method. The coupling between a pair of antennas is first measured on the reference site. Then the same antennas are set up in the same way on the site to be validated and the coupling is measured again. The difference between the two results is a measure of the quality of the site being validated. The ANSI method in C63.4 effectively does the same thing, but in a roundabout way and with higher uncertainties. Site validation has been driven by the ANSI method for many years, including in Europe, where a disguised ANSI method has long been used.¹¹ This is the dual-antenna-factor method in which, rather than individual AFs being calculated, the product of the AFs of one pair of antennas is measured. With this technique there is no need to measure three pairs of antennas, but only the pair being used for the site validation. There are fewer uncertainty components and, consequently, the total uncertainty is less. Improvements in antenna metrology over the past decade have led to much-reduced uncertainties for antenna calibration and site validation.

ANSI has argued that its method does not rely on the existence of "golden" sites. It may be that uncertainties can be reduced by taking the average of several measurements from different parts of an imperfect site, but this would be a laborious procedure not widely adopted. There are probably many high-quality sites that would be suitable to use as reference sites, but no method accurate enough to prove this was available until recently. The CALTS method described in CISPR 16-1:1999 will make possible accurate assessments.

Improvements in Antenna Metrology

The biconical antenna was originally designed to have a maximum operating frequency of 200 MHz. But with balun design improvements it became common to use the antenna to 300 MHz. A problem with the MIL-STD-461 design of biconical elements is that antenna performance is affected by a strong resonance above 260 MHz. This problem can be reduced significantly with the use of open-structure biconical elements,¹² which also makes possible a collapsible-element design that enhances antenna portability.

To measure the strength of a local field, a small antenna should be used because a large antenna integrates the field over the volume of the antenna. Another advantage of electrically small antennas is that they interact very little with their environment; consequently, antenna factor does not change with proximity to surfaces. The disadvantage of associated higher antenna factors can be overcome by the use of amplifiers. Small antennas are particularly useful for the validation of small anechoic chambers.

The coaxial feed cable can cause measurement uncertainties of as much as ±1 dB, and much higher ones if the antenna is unbalanced. But as affordable RF/optical links become available commercially, coaxial cable can be replaced by optical fiber. This also is particularly useful in the validation of anechoic chambers where there is a need to know the RF performance to fractions of a decibel.

References

Martin Alexander is principal research scientist at National Physical Laboratory (Teddington, UK).