Site Validation Theory 101: Techniques and Methods

Metrology versus EMC

Qualifying a test site means different things to different people. It shouldn't be surprising that the type of testing performed determines the requirements for a test site. And although it would be nice to be able to perform any test with absolute accuracy and precision (zero uncertainty), it's important to realize that this is not always necessary. To illustrate this point, compare the requirements of a metrology lab with that of an EMC lab.

Metrologists are concerned with determining the absolute value of a measurable quantity. This can be as simple as comparing the thing being measured (the measurand) with a fixed reference object used as the standard of measure. A simple example of this would be using a standardized ruler to calibrate another ruler. This reference object is typically referred to as a "standard." This should not be confused with the other types of standards this article will refer to, such as the ANSI or CISPR standards, which are standardized procedures rather than standard objects.

In a more complicated situation, the measurand may be determined by using various physical laws to relate it to simpler measurable constants. For example, a mercury thermometer combines the thermal expansion properties of liquid mercury and a linear scale to allow temperature to be measured by measuring distance. The gauge on the thermometer doesn't measure temperature directly, but rather, the mercury serves as the transducer between temperature and length. Known physical laws that control how mercury expands with temperature make it possible to determine the temperature change from the change in the length of mercury in the tube. Note that this may not be the actual procedure typically used in designing and calibrating a mercury thermometer, but the principles applied here indicate one way of calibrating a thermometer.

This example serves not only to show how one physical property can be represented in terms of other simpler physical standards, but also to demonstrate that more than one method can be used to make a measurement that is just as "right" as another. For example, rather than the procedure above, one could create a standard thermometer by defining a degree to represent a certain change on the face of the thermometer and then use that thermometer to measure the temperature and transfer the appropriate marks to a completely different type of thermometer. The point is that often there can be many different ways to achieve the same measurement.

So what does all this have to do with test site validation? The test site becomes a factor in the generation of a known electric field and in the ability to measure that field precisely. Various quantifiable physical laws govern the relationship between a test site and the measurand as determined on that site. The test site actually becomes both a measurement standard (like a ruler) and a form of transducer (like the mercury). In order for metrologists to obtain valid measurements on a test site, they must be confident of both the physical constants the site represents and the physical relationship the site has to the measurand. It should be apparent that the quality of the measurement is directly related to the quality of the site, and any deviation from the physical ideal is unacceptable for metrology purposes (see Figure 1). (A standard ruler that is a bit shorter than the length it was defined to represent would not be a good standard.) This deviation contributes directly to the uncertainty of the final measured value.

Figure 1. Comparison of metrology and EMC measurement requirements. Metrology must attempt to determine absolute values with respect to a preset scale. EMC must verify that a level is below (or above) an acceptance criterion. The absolute value of that criterion does not necessarily matter if it was set based on a specific test method. Any errors in the test method are also included in the acceptance criterion.

By contrast, consider a test site used for emissions measurements. Consider first the reason for performing emissions measurements: an authoritative body, whether a government regulatory agency (FCC, etc.) or company management, determines that there is a need to ensure that the potential for interference from an equipment under test (EUT) is minimal. That's not to say that an EUT must have zero emissions, but rather that those emissions are kept within an acceptable level. It is apparent that just as it's impossible to expect zero emissions, it's also impossible to test every possible environment an EUT may be in once placed in the real world.

Therefore, the efforts of the standards bodies (ANSI, CISPR, etc.) have not always provided test methods that guaranteed a perfectly accurate measure of the electric field emitted from an EUT. Instead, the regulatory agencies are typically concerned primarily with ensuring that everyone uses the same technique and gets the same result. Think of it this way: If everyone uses a given technique to measure an emission, and that value is deemed to have a sufficiently low potential for interference, does it really matter what the value actually is? That is, if everyone uses a ruler that is 1 in. short to measure a piece of string, and that piece of string is the required length for the job, does it matter that it's really 11 in. instead of 12? The required length was determined to be the length of the ruler. It just so happens that the ruler was 11 in. long. What's important is that the measurements were consistent, not that the actual value was known.

The primary goal of validating an emissions test site should be to ensure that all test sites are the same, rather than that all test sites are perfect. Please note that this is in no way meant to imply that the methods presented by the standards bodies aren't also intended to try to produce the correct results. However, often financial and other considerations require that the methods written into standards provide the best results available with current technology and without excessive cost. It is usually expected that the method will have some uncertainty associated with it, although sometimes that uncertainty is larger than the authors of the standard might have been able to determine at the time.

Bearing in mind that requirements are really determined by the application, here are some of the techniques currently in use and under study for site validation.

Normalized Site Attenuation

The first method to consider is the ANSI C63.4 normalized site attenuation (NSA) method.⁴ As others have discussed, there is considerable dispute about what this method really is and what it's supposed to be. The ANSI NSA concept was introduced by Smith, German, and Pate in 1982 and was adopted into the C63.4 standard and, indirectly, into the C63.5 standard.^5–7 It presents a normalized representation of the site attenuation (path loss) for a pair of electrically small (with respect to the wavelength and with respect to the dimensions of the system in question; in this case, the antenna heights and separation distance) dipole elements placed over an infinite ground plane.

The method includes standardized separation distances, transmit antenna heights, and height scan ranges for the receive antenna. The formulas are provided for both horizontal and vertical polarizations of the antennas. The formulas contain the antenna pattern corrections (sin) for the vertical point dipoles and assume a circular pattern for the horizontal ones. The combination of position and polarization for a given measurement is often referred to as a geometry. The underlying assumption of both of these standards is that the site attenuation (SA), which is the loss measured between two antennas, is equal to the theoretical NSA for the given geometry, plus the antenna factors of the two antennas, when represented in decibels.

The ANSI C63.4 standard states that antennas used to perform an NSA measurement may be calibrated per C63.5. Since both methods use the same mechanism to determine two different unknowns—antenna factors or measured (versus theoretical) NSA—this becomes a circular problem. If one can assume a perfect site, then it is possible to determine the antenna factor per the C63.5 definition. On the other hand, if one can assume good antenna factors per C63.5, then one can validate a site per C63.4. But, using this method, how can one verify the quality of a site without starting from a perfect site? That was the problem that faced the ANSI committee, and they thought they'd solved it. In the end, it was just hidden, and the result was that every site validation per ANSI C63.4 was really just site intercomparison. Chambers and weather-protected open-area test sites (OATS) were compared to good OATS under the guise of comparing measured NSAs to the theoretical values. To understand how this happened, it is important to look at some of the limitations of the ANSI theory.

ANSI Method. The first thing to note is that the antennas used for qualifying the site are never electrically small. In fact, in case of dispute, FCC falls back to tuned dipoles, which are the largest antenna element that will work properly for a given wavelength. So, one might expect that sites measured using the ANSI method with large antennas might not exactly match the theory for a point dipole. This is, of course, the case.

It was evident almost immediately that the ANSI NSA model failed to account for the mutual coupling between large antennas like tuned dipoles.⁸ To compensate for this, a set of tables was included with correction factors for tuned dipoles, which can be calculated analytically. Unfortunately, other EMC antennas cannot be corrected as easily. In 1994, Gavenda pointed out that the ANSI NSA formula was missing the near-field correction term, which causes a significant difference at 30 MHz.⁹ The ANSI model had assumed the far-field radiation pattern of a point dipole, but then allowed it to be used in instances where the assumption was invalid. In recent years, more-significant errors have been shown to exist.^10,11

With all of these discrepancies, how did anyone ever validate a chamber using the ANSI NSA method? Remember that C63.4 specified a set of geometries for the NSA measurements and stated that the antennas could be calibrated per C63.5. The original wording in C63.5–1988 states that the 10-m horizontally polarized calibration is the preferred method, but provides information for calibrating in special geometries with the statement, "NOTE: If measurements are needed in special geometries such as 3-m separation or vertical polarization, see Appendix A." This statement describes exactly what's done when measuring NSA, so it was an obvious step to calibrate antennas in the geometry in which they'd be used. This immediately solves the problem of the errors listed above. Because the magnitude of each error is a function of the geometry, the resulting antenna factor from C63.5 becomes specific to that particular geometry (and thus the term geometry-specific antenna factor [GSAF], which has come into use recently). When that antenna factor is used to validate a site in the same geometry, the same errors exist in the site attenuation of the site to be validated, and they cancel out the errors in the antenna factors.

So why not just keep doing things the way we've always done them? What's wrong with using GSAF and letting the errors cancel? First, the recently approved ANSI C63.5–1998 removed the note mentioned above so that the loophole no longer exists.¹² The real result of using GSAFs for site validation is that the site under test is actually being compared with the site on which the antennas were calibrated (the reference site). For calibration, SA_ref = GSAF₁ + GSAF₂ + NSA_theory, whereas for site validation, SA_site = GSAF₁ + GSAF₂ + NSA_site. Therefore, there was no difference between comparing the measured NSA, NSA_site, with the theoretical value, NSA_theory, and direct comparison of the site attenuation of the test site, SA_site, with that of the reference site, SA_ref. This is exactly what ANSI was trying to avoid from the start.

One additional wrinkle is the fact that the uncertainties are cumulative. The antenna factor contains the uncertainty due to the site it was calibrated on, but the quality of the site under test can only be known to the uncertainty of the antenna calibration. Using the ANSI NSA method, it becomes impossible to verify that the site the antenna is calibrated on is good enough to calibrate an antenna with a low enough uncertainty to verify the quality of a site. The uncertainties become circular too (see Figure 2).

Figure 2. Because the antenna calibration and NSA measurement form a circular pattern, the ANSI C63.4 NSA method results in site intercomparison. The site being verified is compared to the site on which the antennas were calibrated.

The goal of the ANSI NSA method was to develop a way to verify a site from basic principles rather than requiring a golden site as a standard that everyone compares to. Such a golden standard becomes complicated when talking about a test site. The golden-standard test site would be expensive to construct and maintain, and that expense would be transmitted to everyone, who would then be required to have antennas calibrated on that site in order to validate their own test site. Additionally, with no independent verification of the quality of a site, the golden site actually becomes nothing more than the standard reference site, rather than the perfect site. The ANSI committee has always tried to define a low-cost alternative that removes the restrictions imposed by having one or two golden sites.

The Arguments

The change to the wording in C63.5 has opened the Pandora's box of site validation and has prompted many disagreements over what the standard intended, as well as what the right method is. Although there is an overall desire to find a better solution to the problem, there is also a need to determine what to do in the meantime. There are two main camps in the debate, the free-space antenna factor (FSAF) proponents and the GSAF proponents.

Free-Space Antenna Factors. On one hand, proponents of FSAFs argue that their use was the original intention of the ANSI standards. In terms of removing the dependency on site intercomparison, this is certainly true. If the antenna factors are made independent of the calibration method, they become a suitable reference standard for determining the behavior of a site. The recent changes to C63.5 are further evidence of this. The changes moved the standard closer to a pure antenna calibration standard. However, in turn, the changes lost sight of the manner in which these antennas are typically used.

Unfortunately, the arguments for using FSAF are weak because each geometry results in errors that are specific to that geometry. These errors are related to the type of antennas used, their polarization, and their proximity to the ground plane and each other. These errors are not predicted by the ANSI NSA theory, and thus, when comparing the measured NSA to the theoretical value, the errors still exist. These errors can be greater than the ±4 dB allowed by C63.4 at some frequencies for some antennas. A perfect site, therefore, might fail an FSAF NSA while, conversely, a truly defective site might pass.

Geometry-Specific Antenna Factors. GSAF proponents, on the other hand, largely point to the fact that the method does provide the correct result because the geometry-specific errors cancel. A certain amount of uncertainty is added, because it is not possible to directly verify the quality of the reference site. This can be minimized, however, by selecting a reference site with specific physical qualities (size, flatness, etc.). FSAF proponents may argue that this method could be easily abused by selecting a bad site as the reference to allow a bad test site to pass. However, one could argue for leaving a known error in the FSAF method, so it is evident that any method is vulnerable to abuse.

An added advantage of the GSAF method is that FCC has accepted this method since its inception. This reinforces the concept that the accuracy of the method matters less than that everyone follows the same method approved by the regulatory body. Finally, the ANSI C63 committee has issued an interim statement that C63.5–1988 (which allows GSAFs) shall be used for determining antenna factors for site validation until the C63.5 working group determines a better solution.

Nonetheless, the FSAF method is the better method from a metrology standpoint. Because the FSAF is a physical property of the antenna—independent of the geometry in which it was used or how it was calibrated—the FSAF can be used as a reference for determining site behavior. The problem is that the ANSI NSA formula does not predict the behavior of typical EMC antennas with FSAFs. In addition, the behavior is dependent on the type of antenna used. So the problem becomes one of predicting the behavior of a real antenna when used over a ground plane.

CALTS

In the mid-1990s, CISPR introduced the calibration test site (CALTS) method.¹³ The term actually refers to the type of site being qualified. A second type of test site, which has different test criteria, is referred to as a compliance test site (COMTS). This method uses a standardized tuned dipole set for which the behavior can be calculated using an analytical model. The antenna design includes a detachable balun, so that the portion of the antenna performance that cannot be calculated can be easily calibrated out.

The antennas are placed at a variety of heights and spacings, and the results are compared to theory. Because the theory predicts the performance of the antenna used at each position, this method does not suffer the geometry-specific limitations of the ANSI NSA method. CISPR proposes different qualification levels determined by this method: one for general-purpose emission test sites (COMTS) and another for antenna calibration–quality test sites (CALTS).

From a test lab standpoint, however, this method has several limitations and disadvantages. First, the antennas are extremely expensive, and the measurements are long and tedious (costing more each time they're performed). Second, the vertical calibration method and requirements have yet to be defined. Finally, the test method doesn't necessarily reflect the use of the site. Although this may not seem important, an emissions site may not need to be tested to the same level as a precision calibration site. This may become more evident at higher frequencies where directional antennas such as horns are typically used. At higher frequencies, the behavior of the test site outside the beam of the antenna will have less effect on the measurement. This can't be said for a site used for a measurement where the antenna may be focused toward different parts of the site at different times. Narrowing the site validation method to a specific application may not be the best solution for everyone, but it may provide some time and cost savings to the EMC industry.

The Future

Where do we go from here? The ANSI C63.5 working group is currently studying several possibilities. Recent work has shown that it is possible to accurately model the behavior of typical EMC antennas such as biconical antennas.⁹ The likely result will be a set of ANSI NSA correction tables for some common biconical designs similar to the tables for tuned dipoles. Another possibility is to use a calculable bicon, which has a similar design to the calculable dipoles used for the CALTS method.¹⁴ Although more expensive than a traditional biconical antenna, the calculable antenna wouldn't require calibration (beyond the balun calibration) to determine its antenna factors or its behavior on a test site. This would enable determination of the NSA of a site within about 0.5 dB (better than typical antenna calibrations) without the use of a reference site. Additional work remains to be done for higher frequencies and other types of antennas. A possible solution for log periodic antennas is forthcoming.¹⁵

Other methods suggested include using a round-robin test of many OATS to determine a virtual golden site to which all sites could be compared. This method ensures that everyone involved in the round-robin is testing on an equivalent site (within the statistical variation of the sites), but tends to make the test artifact or antennas being used the golden standards, with many of the same problems that are inherent in a golden site.

Another generalized method would use a combination of strict dimensional criteria and repeated statistical measurements of a site to determine that site's uncertainty. The site could then be used to calibrate antennas for site intercomparison with the known uncertainty applied to the site validation measurement. This method has yet to be proven to determine the quality of results it could produce.

Alternative Test Sites

Other familiar proposed emissions test devices include TEM cells, GTEM cells, reverberation chambers, and a wide variety of prequalification test cells. Many who have used one of these devices would probably like to be able to perform full compliance tests in them. Currently, FCC only accepts GTEM for compliance measurements below 1 GHz if specific correlation conditions can be met. The same arguments hold here that were presented regarding determining whether a site is suitable. Clearly none of these sites or methods would be identical to that used by everyone else to measure an EUT. The argument of everyone using the same technique fails here. This leaves two other options.

Correlation to OATS. The first option, which has been the traditional approach, has been to show that the proposed test method can be correlated to the OATS method accepted by FCC. Although it was accepted for the GTEM, this method has had dubious success at best. In general, these devices behave nothing like an OATS, and the EUTs tested in them behave nothing like the simple devices used to model correlation.

Why try to determine some magical method of correlation at all? If one can prove (sufficiently to satisfy the regulatory body) that for a given device type the difference between the OATS measurement and an alternate method is always the same, then why not just apply that difference to future measurements and be done with it? In essence, that's what the GTEM correlation does. The relationship between the emitted power determined by the GTEM and the predicted field on an OATS is hidden behind elaborate equations, which happen to be based on the ANSI NSA formula.

The one obvious problem with this approach is that it opens up the possibility of using questionable or unproven methods. The task of validating a multitude of methods would not be an enviable one, and is certainly not something any regulatory body would be likely to do. Many of the precompliance methods would never be acceptable for regulated emissions tests. On the other hand, some alternative test methods are at least as good as the OATS method. Borrowing an argument from proponents of reverberation chambers, the reverb environment is often a more realistic test environment than an OATS or a fully anechoic room. This can be seen by testing a cell phone or laptop computer, items that may be used more often in a car or on an airplane than on a big, flat ground plane. Although it's impractical to compare results between different test methods, that doesn't necessarily mean that they're bad methods.

Acceptable Emissions Levels. The second option would require that a regulatory body define acceptable emissions levels for each method. Rather than trying to determine what level is equivalent to that found on an OATS, why not determine what level is deemed acceptable and leave it at that? After all, there is already a discrepancy between the acceptance levels for 3- and 10-m tests. Both tests have flat lines for the limit lines, but even in the simple approximation of the ANSI NSA, the relationship between those two distances is not flat. Therefore, if one were flat, the other would be an odd-shaped curve. FCC chose numbers that made life easier. If the 3-m test is a little tougher than the 10 m, that's the cost of doing emissions testing at 3 instead of 10 m. Given the precedent of 3- versus 10-m sites, why not consider other methods? If it meant that they could use alternate methods for emissions measurements, many proponents of these methods would probably accept tighter requirements and the use of the 10-m OATS as the final arbiter in case of dispute.

Whatever the outcome in terms of regulation, it really comes down to the users of these devices to gain their acceptance as alternatives. Given the arguments presented here, a combination of political pressure, quality measurements, and a little additional work could go a long way toward making alternate devices common fixtures in compliant emissions measurements. The formation of users groups for these products would be a simple way to start this process. Also, regardless of correlation, regulatory agencies will typically require a standardized test method before allowing it to become a qualifying method. With the magnitude of the problems facing the various standards committees, waiting around, expecting your favorite test method to be written into a standard, would be a mistake.

Conclusion

While the current status of EMC site validation seems quite confusing, there are some simple points to keep in mind. At this time, the GSAF NSA method is still the correct method to use to achieve FCC-compliant site validation. Other methods, which may or may not change the reference value that a chamber is compared to, are likely to be introduced. However, this shouldn't be a cause for alarm, since unless a decision is made to change tolerances (i.e., ±2 dB instead of ±4 dB), a site that has been determined to be good under current methods will still be a good site with a new method. The change in the definition of the verification method won't change what makes the test site good. Inequity between methods doesn't matter as long as all of the results are deemed acceptable.

References