Systematic Errors in Normalized Site Attenuation Testing

Zhong Chen and Mike Windler

Computing a correction factor for biconical antennas reduces the systematic errors for calculating normalized site attenuation and improves the accuracy of the theoretical model.

Accurate antenna factors are the foundation of ANSI C63.4 normalized site attenuation (NSA) measurement.¹ NSA is defined in ANSI C63.4 as the site attenuation divided by the antenna factors (AFs) of the radiating and receiving antennas (all in linear units). The AF is further defined as a quantity relating the strength of a field to the output voltage measured at a 50- load connected to the terminals of the antenna. Although NSA makes site validation tests possible with common EMC broadband antennas, the use of the term normalized should not be taken literally. The AFs of the transmitting and receiving antennas are normalized (divided) from the site attenuation; however, other factors that are antenna-specific, such as antenna pattern, mutual coupling between antennas, and antenna coupling with ground plane, are not completely normalized from NSA. The definitions of NSA and AF may lead one to assume that these two physical phenomena are mutually exclusive. That is, the antenna being used for site attenuation is assumed to be unaffected by the site itself or the measurement procedure. Unfortunately, this is absolutely not true.

In this article, the free-space AF and AF derived in the specific NSA geometries (e.g. height, polarity, and distance) are discussed in the context of NSA measurements. Broadband biconical antennas are emphasized because of their widespread use for EMC measurements. Various ground-plane effects and other factors that influence NSA measurements are described, and a numerical correction factor is introduced for broadband biconical antennas. This study demonstrates that numerical results can be correlated to experiments conducted in a highly controlled calibration process on a premium-quality (near-ideal) ground plane, which will be discussed later. To obtain the lowest possible measurement uncertainties, multiple measurements were made. This article also provides the measurement uncertainty evaluation and a description of the measurement procedure.

Geometry-Specific AFs and Correction Factors

The combination of measurement distance, antenna polarity, and transmit height is referred to as a measurement geometry. The NSA test described in ANSI C63.4 requires measurements to be made in the geometries shown in Table I. The receiving antenna is swept from a height of 1–4 m. Because the measured attenuation is affected by more than just the site and the antenna factors, the mutual impedance correction factor (AF) is defined in ANSI C63.4. The resulting NSA equation is

NSA(dB) = V_Direct(dB V) – V_Site(dB V) – AF_T(dB m¹) – AF_R(dB m_–1) – AF_TOT(dB) [1]

where

AF_T = antenna factor of transmitting antenna

AF_R = antenna factor of receiving antenna

AF_TOT = mutual impedance correction factor for resonant dipole antennas.

The AF_TOT is specified in ANSI C63.4 as the mutual impedance correction factor; however, it encompasses more than just the mutual impedance correction. The term AF_TOT also corrects the near-free-space AF to account for phenomena such as antenna mutual coupling, antenna pattern, nonuniform field illumination, antenna coupling with ground plane, and near-field effect. Berry et al² calculated AFs for Roberts' dipoles using a numerical method (method of moments), and their results were later adopted by ANSI C63.4.

 

Measurement Distance (m)		Antenna Polarity	Transmit Antenna Height (m)
10		Horizontal	1
10		Horizontal	2
10		Vertical	1
10		Vertical	1.5
3		Horizontal	1
3		Horizontal	2
3		Vertical	1
3		Vertical	1.5

Table I. NSA measurement geometries.

The term AF_TOT for broadband antennas has not been calculated or included in existing standards; however, this calculation is unnecessary if NSA measurements are performed in the same geometry as the antenna calibration--through the geometry-specific AF. Essentially, this is a site intercomparison method. In using Equation 1 for both theo- retical model and actual measurement, AFs are simply intermediate variables. Lacking the correction factor for broadband antennas, the 1988 edition of ANSI C63.5³ permitted geometry-specific calibrations for such specific purposes. The 1998 edition of ANSI C63.5 removed the provisions allowing geometry-specific antenna calibrations. However, the correction factors for broadband antennas, which will be shown to constitute a significant portion of the total site evaluation budget (normally ± 4 db), are not included.

With the increased accuracy in numerical modeling of complex antennas, determining numerical correction factors for broadband antennas becomes feasible. This numerical modeling technique is similar to the technique already in use for dipole antennas. The numerical approach, which has been experimentally validated, is preferred. In the numerical technique, NSA measurement is traceable to the theoretical results. The experimental method, on the other hand, is traceable only to the original antenna calibration, which, in turn, is restricted by the ground-plane quality and measurement quality in the original calibration process. However, for antennas that cannot be described by numerical or analytical methods, the measurement (geometry-specific AF) approach has the most potential for yielding the lowest uncertainties.

Broadband Versus Dipole Antennas

For many years, dipole antennas have been used successfully to conduct NSA tests, and the correction factors are currently provided in ANSI C63.4-1992. However, for both tramsmitting and receiving dipole antennas, the length must be changed for each of the 27 frequencies to be measured. Because using dipole antennas is very time-consuming, broadband antennas have become the most commonly used antennas for NSA testing.

The use of broadband antennas is required on alternative test sites such as weather-protected open area test sites and semianechoic chambers. Broadband antennas are necessary because frequency-swept measurements are needed to determine whether an alternative site produces unwanted reflections, which often do not coincide with the 27 frequencies of the tuned dipole. Furthermore, broadband antennas are commonly used for product measurements, and therefore these antennas are readily available for NSA testing without incurring additional costs.

Characterizing Antennas and Free-Space AFs

ANSI C63.5 provides several methods for calibrating broadband antennas, including biconical and log periodic dipole arrays. The most commonly used method is the standard site method (SSM), which requires three pairs of measurements on three antennas over a large, flat, metal ground plane. During each of the three measurements, the transmitting antenna is set at a fixed height, while the receiving antenna is scanned from 1 to 4 m for a 3- or 10-m separation distance. Site-to-site intercomparison is also a valid method for characterizing test sites, which is possible because ANSI C63.5-1988 specifies both horizontal and vertical polarizations for several transmit heights.

Free-space AF is defined as the AF measured from the bore-sight of the antenna in a free-space environment without a ground plane or any other scattering effects present. In addition, the incident electromagnetic wave must be a plane wave. The free-space AF is a theoretical definition that describes the intrinsic property of an antenna. It is not, however, a universal AF. It is well known that AFs can change significantly when the antennas are used in an environment that fails to satisfy the free-space condition. One argument that favors free-space AF over geometry-specific AF is that free-space AF provides an acceptable average for antennas used in different geometries, and thus eliminates the need to apply multiple AFs at different geometries. Free-space AFs are specified for product emissions measurements such as those described in ANSI C63.4 or CISPR 22.

Based on the theory that a free-space AF can provide an acceptable average AF for product measurements, ANSI C63.5 and other international standards have been moving toward requiring free-space calibrations. In fact, the ANSI C63.5-1998 version allows only a near-free-space calibration. Other geometry-specific methods of calibration are no longer permitted. This article demonstrates that NSA measurements made with free-space AFs without the appropriate AF_TOT correction term will include significant systematic errors. Using geometry-specific AFs has become industry standard to compensate for the errors in the SSM.

Because of concerns arising from the change in ANSI C63.5-1998, an ad hoc group that formed to address this issue was recognized by the ANSI Accredited Standard Committee as an official working group under C63 Subcommittee 1 in late 1998. The working group, which consists of members from National Institute of Standards and Technology, chamber and antenna manufacturers, EMC test labs, chamber and test-site users, and independent consultants, is developing a set of correction factors (AF_TOT) that can be used for broadband antennas. These correction factors will be similar to those for dipole antennas. In the interim, the ANSI C63 committee recommends the use of ANSI C63.5-1988 (instead of ANSI C63.4-1992) for NSA measurement.⁴

Antenna Calibrations in Different Geometries

Smith originally proposed SSM for calibrating dipole-like low-gain EMC antennas.⁵ SSM is achieved over a conducting ground plane, and the ground-plane reflection is mathematically removed in the calculation. The free-space calibration, as defined in ANSI C63.5-1998, is a specific SSM setup. It requires that the transmitting antenna be set at a 2-m height. The receiving antenna scans from 1 to 4 m, and the separation distance between the transmitting and receiving antennas is 10 m. For a common EMC broadband antenna in the frequency range of 30–1000 MHz, this method purportedly yields a near-free-space AF, which is generally within 0.5 dB of the true free-space AF. This claim is based on the fact that this specific antenna setup minimizes effects such as ground-plane coupling, antenna mutual coupling, nondipole pattern, and non–plane wave illumination. However, this method does not include these effects in other NSA geometries for which they are not considered negligible.

Mutual Coupling between the Antennas and the Ground Plane. Input impedance and the effective antenna length change with orientation, which causes the AF to oscillate from its free-space value as the antenna height changes. Generally, the closer the antenna is to the ground plane, the stronger the coupling effect is. Horizontally polarized antennas couple more tightly than vertically polarized ones. Moreover, different antennas also produce different reactions.

Mutual Coupling between Antennas. When performing the NSA test, the transmitting and receiving antennas interact with each other, which also influences antenna calibration. The behavior of each antenna is modified by the presence of the other.

Dipole Pattern. In both SSM and NSA formulations, a dipole antenna pattern is assumed for all antennas under calibration, which mathematically removes the ground-plane reflection. This assumption is generally valid for biconical antennas below 200 MHz and for tuned dipole antennas. It can, however, cause significant errors for typical biconical antennas used above 200 MHz or a log periodic dipole antenna where the radiation pattern deviates from that of a short dipole.

Nonuniform Wave Illumination. To entirely satisfy a free-space calibration condition, not only must the antennas be situated in a free-space environment, but also, the impinging electromagnetic wave must be a plane wave. In an SSM calibration or an NSA measurement, especially for a 3-m separation distance, the antennas are in the Fresnel zone. Greater distances better simulate the plane wave condition. More importantly, in an SSM, the field at the receiving antenna is a vector sum of the direct wave and ground-plane-reflected wave from the transmitting antenna. The vector sum can change rapidly with small variations in geometry, therefore changing the uniformity of the electromagnetic fields. The nonuniformity of the incident wave also causes differences between a geometry-specific AF and a free-space AF.

Near-Field Effects. Smith's SSM formulation assumes that the receiving antenna is in the far field of the transmitting antenna. If this is not the case, the E-field falls off as an inverse function of R (distance). This study shows that this is not always true, especially at lower frequencies. It is possible to have an error of up to 1.5 dB at 30 MHz by disregarding 1/R² and higher terms.⁶

In an SSM calibration, all of these errors are combined. Even if it were possible to quantify each error source separately, the simple sum of the errors does not represent the total error in the method. Numerical methods, however, can simulate the full calibration process and provide a correction factor that is the vector sum of all errors.

Deterministic Modeling of Biconical Antennas

The biconical antenna chosen for the numerical study was approximately 1.35 m in length (from tip to tip). Construction of the cage elements is specified by MIL-STD 461A-1968. One radial strut is used in each biconical element to suppress the element resonance at about 280 MHz. This construction represents almost all biconical antennas currently used in the EMC industry. Numerical electromagnetic code (NEC) 2, a wire-based moment method, was chosen for the numerical calculation.⁷ There are several reasons for using NEC. It is freely available on the Internet (http://www.qsl.net/wb6tpu/swindex.html), and it is the least demanding on computer hardware for this application compared with other techniques, such as finite difference time domain, transmission line, or finite element methods. A thin-wire model can sufficiently and accurately represent biconical antennas.

 

Case	R (m)	Horizontal Polarization			Vertical Polarization
		h1 (m)	Max AFror f200 MHz	Max AFror f200 MHz	h1 (m)	Max AFror f200 MHz	Max AFror f200 MHz
1	3	1	2.0	2.8	1	1.8	4.6
2	3	2	1.7	2.8	1.5	2.2	2.5
3	10	1	1.2	1.2	1	1.4	1.4
4	10	2	1.0	1.0	1.5	1.1	2.1

Table II. Calculated maximum AF_TOT in dB for geometries listed in ANSI C63.4.

The balun impedance also affects the AF. Two types of baluns are currently available: 50- baluns (1:1 impedance transformation ratio) and 200- baluns (4:1 impedance transformation ratio). A 50- balun was selected for this study. Obtaining the results for a 200-‡ balun is a straightforward calculation. The calculated free-space AF of the 50- balun biconical antenna is shown in Figure 1. Figure 2 shows the calculated correction factor for the NSA (AF_TOT) for a pair of horizontally polarized biconical antennas, and Figure 3 shows the correction factor for a pair of vertically polarized antennas. Table II summarizes the results. If the biconical antenna is used only from 30 to 200 MHz, using the free-space AF causes a maximum deviation of 2.2 dB. If the biconical antenna is used from 30 to 300 MHz, the maximum deviation is 4.6 dB, which occurs for vertical polarization at R = 3 m and h1 = 1 m. In this case (AF_TOT = 4.6 dB), it is impossible to assess compliance of a theoretical perfect site (infinitely large and perfectly conducting ground plane) with the theoretical NSA model without the AF_TOT correction factor (Figure 4).

 

Figure 1. Numerically modeled free-space antenna factor of a biconical antenna (plane wave illumination).

Figure 2. Numerically modeled AFTOT for horizontally polarized biconical antennas. R is the separation distance, h1 is the transmit height, and h2 is the receive height, all in meters.

 

Figure 3. Numerically modeled AFTOT for vertically polarized biconical antennas. R is the separation distance, h1 is the transmit height, and h2 is the receive height, all in meters.

 

Figure 4. Numerical and measured AFTOT if applying R=10m, h1=2m, horizontal AF for 3-m distance vertically polarized biconical antennas (R=3m, h1=1m). The measured data were obtained at the National Institute of Standards and technology (NIST) ground plane in Boulder, CO, in 1997.

It is interesting to note that for the R = 3 m, h1 = 1.5 m vertical polarization case (Figure 3), the AF_TOT curve shows two discontinuities with a sudden ascent from about 115 to 140 MHz. The ANSI model predicts that there is a sudden change in height for E^d_max at 115 MHz during the 1~4-m height scan, whereas it actually occurs at 140 MHz for biconical antennas.

Empirical Validation of the Model

Validation of the numerical model raises several technical issues. The numerical model is based on a perfect test site. To simulate a perfect site, a premium site was selected that employed 50 x 80-m ground plane with a rated flatness of ±4 mm. The ground plane is made of a 30 * 60-m welded solid steel–copper composite surrounded by 10 m of mesh on all sides. To minimize the effects of site imperfections, a statistical approach to the geometry-specific calibrations was used.

Five sets of height-swept measurements were taken for each geometry at a different location on the ground plane. The locations for the transmitting and receiving antennas formed a 3-m-diam circle. The standard deviation of the five measurements at each frequency step was used to evaluate the site quality. For an ideal site and ideal equipment, the standard deviation would be zero. Figure 5 illustrates the equipment configuration. Figures 6 and 7 illustrate the worst-case standard deviation of the data acquired. The standard deviation of these five values is a worst-case contribution to uncertainty that can be associated with the site imperfections, cable lay variations, amplifier gain stability, distance measurement errors, antenna height start/stop errors, antenna travel smoothness (directivity), and height versus frequency variations.

 

Figure 5. Equipment configuration.

Figure 6. Standard deviation of five geometry-specific (horizontal) antenna calibrations taken at different locations on the ground plane. Five locations with horizontal polarity, R=10m, h1=2m.

Figure 7. Standard deviation of five geometry-specific (vertical) antenna calibrations taken at different locations on the ground plane. Five locations with vertical polarity, R=10m, h1=1m.

A pair of identical (e.g. same make and model) biconical antennas was used. All nonsite uncertainty contributors were minimized. For example, the measurement distance, transmit height, and starting receive height were verified in each geometry for each of the five positions. As the small value of the standard deviation data in Figures 6 and 7 illustrates, this site is close to ideal.

The evaluation of measurement uncertainty associated with this model validation included receiver frequency span error, receiver amplitude error, mismatch, cable coupling, and site imperfections. Uncertainties were calculated for each geometry in each frequency span. Table III shows a typical example of the uncertainty calculations.

f = 72.5–115 MHz			Uncertainty
Source	Distribution	K	(dB)
Receiver frequency error	Rectangular	1.7	0.41
Receiver amplitude error	Rectangular	1.7	0.09
Mismatch	U-shaped	1.4	0.05
Site and system uncertainty	T-distribution	2.8	0.04
Combined uncertainty	Normal	1	0.24
Expanded uncertainty		2	0.49

Table III. Example: uncertainty calculation.

For biconical antennas, frequency span error is the dominant contributor to uncertainty because the slope of the AF curve for a biconical antenna ranges from 30 to 70 MHz. The biconical antenna frequency range of 30–200 MHz was divided into four smaller spans of 42.5 MHz each to reduce this error.

Receiver amplitude error was evaluated in each frequency span. The measurement values recorded during the antenna calibrations are relative values that represent the difference between direct and indirect measurements. The relative amplitude accuracy of this measurement is a function of the difference between these values. The relative amplitudes of the measurements ranged from 6 to 28 dB. Linearity was evaluated for each frequency range with the applicable relative amplitude. This test employed two 0–11-dB stepped attenuators to reduce the connect/disconnect uncertainties.

Conclusions

The numerical correction factor for biconical antennas predicts a value for AF_TOT as high as 2.2 dB from 30 to 200 MHz and 4.6 dB from 30 to 300 MHz. Empirical testing validated this model. Figure 8 shows the value of AF_TOT for a pair of biconical antennas measured in the 10-m geometries. Note that the nomenclature is v – vertical polarity, 1 – 1-m transmitting height. These data demonstrate that NSA measurements made with near-free-space AF can have a systematic error as high as 2.2 dB from 30 to 200 MHz.

The measurement uncertainties for these tests are relatively small (<0.5 dB). These empirical tests strongly correlate with the numerically calculated values shown in Figure 9. The difference between the measured values for AF_TOT and numerically calculated values are well within the uncertainty of the measurements.

Figure 8. Measured values for AF for a pair of biconical antennas, R=10m. Difference between insertion loss measurement (AF) for a pair of biconical antennas used for NSA testing, R=10m, (average of five locations for each geometry).

Figure 9. Comparison of measured and modeled AF for a pair of biconical antennas, horizontal polarity, R=10m. Differences between insertion loss (AF) for a pair of biconical antennas horizontal (h1=2) - horizontal (h111).

The term AF_TOT was not calculated in the standard, which led to a significant error margin of ±4 dB. This article examined this measurement uncertainty and demonstrated that this error can be greatly reduced. Consequently, the existing NSA model in ANSI C63.4-1992 should be revised to include the numerically calculated values of AF_TOT. Furthermore, for antennas that do not have numerically calculated correction factors such as log/biconical hybrid antennas, consideration must be given to allowing antenna calibrations to be done in specific geometries with appropriate restrictions set for the site and for the methods used. Conceptually, it is preferable to perform site validation by using SSM, in which the site attenuation on a reference site is compared directly to that on the site under test, rather than to measure geometry-specific AF for an NSA calculation in a roundabout way.

References

1. "American National Standard for Methods of Measurement of Radio-Noise from Low-Voltage Electrical and Electronic Equipment in the Range of 9 kHz to 40 GHz," ANSI C63.4-1992, American National Standards Institute (1992), New York.

2. J Berry, B Pate, and A Knight, "Variations in Mutual Coupling Correction Factors for Resonant Dipoles Used in Site Attenuation Measurements," IEEE International Symposium on EMC (1990), Washington, DC.

3. "American National Standard for Electromagnetic Compatibility—Radiated Emission Measurements in Electromagnetic Interference (EMI) Control—Calibration of Antennas (9 kHz to 40 GHz)," ANSI C63.5-1988, American National Standards Institute (1988), New York.

4. ANSI C63 meeting minutes, December 1998, Baltimore.

5. AA Smith Jr, "Standard-Site Method for Determining Antenna Factors," IEEE Transactions on Electromagnetic Compatibility 24, no. 3 (1982): 316–322.

6. JD Gavenda, "Near-Field Corrections to Site Attenuation," IEEE Transactions on EMC 36, no. 3 (1994).

7. G Burke and A Poggio, "Numerical Electromagnetics Code (NEC)—Method of Moments: A User-Oriented Computer Code for Analysis of the Electromagnetic Response of Antennas and Other Structures," Lawrence Livermore Laboratory (1981).

Zhong Chen is senior electromagnetics engineer for EMC Test Systems LP (Austin, TX), and Mike Windler is associate managing engineer for Underwriters Laboratories Inc. (Northbrook, IL).

Back to January/February Table of Contents