The problems associated with open area test site (OATS) testing are well known and may, at times, seem insurmountable. Still, this is the only worldwide official technique for checking the EMC compliance of most products.
Most commercial standards for radiated emissions specify the use of an OATS and an antenna at a distance for making measurements. Standards requiring this technique include FCC (ANSI C63.4) and virtually all common EN standards except EN 55014 for household appliances and hand tools and EN 55015 for luminaries.
In many parts of the world, certification based on the results of self-testing is legally accepted and has become normal practice. Also, precompliance testing prior to visiting a test lab is used almost universally because it can offer significant time and cost savings.
According to figures published by the EMC Test Laboratories Association, 60% of the products offered for testing by a test lab fail to meet the required standards. The costs of modification and additional visits make the concept of self-testing worthy of serious consideration.
The key to any self-testing is the test site. A realistic assessment of the three major sources of uncertainties associated with OATS testing on typical self-test sites is summarized in Figure 1.
The numbers found in Figure 1 are the average of more than 150 sites and show that the problems related to the test site are, by far, the most important. A lack of appreciation of these factors is commonplace with most EMC test-equipment suppliers.
Test sites suffer from two major problems:
- Background (ambient) interference.
- Test-site distortion.
These two factors are entirely independent, and each requires a separate solution. Indeed, a solution for one may exacerbate the other. For example, the classic solution for ambient interference—a screened room—creates the worst possible situation for test-site distortion (see the sidebar Screened Room Nightmare).
Background interference causes several distinct difficulties:
- EUT emissions may be buried under the background interference if frequencies coincide.
- Strong background signals such as FM broadcast transmissions may cause compression and false indications in lower-cost spectrum analyzers.
- Background instability makes finding EUT emissions difficult.
Test-site distortion relates to the fact that the site attenuation factor (with the exception of test laboratory sites) largely is unknown. This is particularly true of sites used by manufacturers for self-testing/self-certifying or precompliance testing.
Although the measurement errors between the antenna and the instrumentation may be reasonably small, the great unknown is the relationship between the signal radiated at the EUT and the signal received by the antenna. Deviations of more than 20 dB between what actually is received and what should be received are common. To understand why there is such a high potential for error, see the sidebar Bad Reflections.
In spite of these difficulties, emissions measurements on homemade sites can be achieved with a reasonable level of accuracy as long as techniques to alleviate background interference and test-site distortion are applied. Significantly, reports from users have indicated that the correlation between these results and test-lab results has been within accepted measurement uncertainties.
This does not mean that the quality of the site is immaterial. The better the site, the better the results. However, for many companies, limited resources, time, and space often result in the EMC engineer making the best of an inadequate situation.
Background Interference
Several techniques are used to identify and measure EUT emissions in the presence of background signals. Virtually all rely on the measurement of the background first, then switching on the EUT and identifying additional signals.
Some conventional analyzers use a manual technique in which the operator marks the background peaks for elimination from the peak list. This technique can be tedious and sometimes ineffective, particularly if the EUT produces broadband noise rather than distinct peaks. Broadband noise is difficult to identify if it already is present in the background.
A second technique obtains the spectrum of the background, stores this as a background trace on the screen (the background signature), then scans a second trace, this time with the EUT on. New emissions can be seen as an increase of the second trace over the background signature.
Unfortunately, this technique does not work since typically 90% of the changes between the two spectra will be due to the changes in the background interference. Background interference usually is very unstable. Many signals are intermittent, such as navigation beacons, telemetry, ground mobile transmissions, machinery noise, and air-traffic control. For the technique to work, the background signature must be stable.
A refinement offers two techniques for obtaining a stable signature: average and peak hold. Each monitors the background for a period of time. During that time, intermittent signals will be caught and included in the signature, and the resultant trace becomes stable.
Any change to the spectrum that occurs when the EUT is switched on is almost certainly due to it rather than a background change. Of the two techniques, average is the most effective. Peak hold tends to be quicker and must be used if the EUT emissions are intermittent or discontinuous.
Having obtained the two traces (background only and background + EUT), the next problem is how to determine the true level of the signal strength from the EUT. An example illustrates the situation: If at one frequency the background noise level is 40 dBµV/m and rises to 46 dBµV/m when the EUT is switched on, what is the actual level of the signal from the EUT? The obvious answer of 6 dBµV/m is very wrong because we are dealing with a log (decibel) scale. If we subtract two log numbers, we are effectively dividing one by the other.
To obtain the correct result, we must convert to linear numbers (µV/m) by antilogging the dBµV/m numbers, then subtracting, then converting back to the decibel scale by logging the result. This technique is defined in CISPR 11, Annex C. If we do the math, the answer is 40 dB.
In practice, this technique is far from foolproof. Errors occur because the background behaves in a perverse manner, sometimes appearing just at the wrong moment and wrong frequency to mask the EUT signals.
A strategy that significantly reduces ambiguities is to perform an initial pre-scan of the EUT with a near-field probe, using a peak-hold function to collect all potential emissions frequencies from inside and around the product. Save this result to file and compare it to the background scan. Any differences in peaks that do not correlate with the near-field peaks are almost certainly not due to the product.
Some engineers complicate the issue by attempting to involve the phase relationship between background signals and EUT emissions. This would only be significant if the two signals were phase-locked in some way or had an identical frequency. The chances of this occurring are extremely remote.
Even in the unlikely event that two significant signals are very close in frequency, such as 100,000,000 Hz and 100,000,001 Hz, the resulting phase vector would rotate through 360° at a 1-Hz rate. This would be effectively averaged out by the quasipeak detector.
Another relatively new technique requires a two-channel analyzer with synchronized sweep plus two antennas. One antenna is located close to the EUT and the other remote from the EUT. Both antennas will receive the background signals, but only the one close to the EUT will receive the EUT signal.
Consequently, if the difference between the two signals can be calculated in real time, this will show only the EUT emissions. The system claims to be very effective but obviously involves some special hardware and an additional antenna.
Test-Site Distortion
Unwanted reflections, ground-plane imperfections, lack of height scanning, and imperfect antenna calibration are typical factors that degrade even apparently good test sites.
It is an obvious folly to ignore these factors, and it is no help simply to insist that the test site comply to the standards. For all but the very few companies with large EMC test budgets, such sites are out of the question.
So, what can be done to reduce the test-site error to acceptable proportions?
A technique has been devised in which a source—the Emissions Reference Source (ERS)—is used to calibrate the site. Having calibrated the site, a correction factor can be applied to any EUT measurements on that site so that the measurements appear as though a near ideal test site was used.
Each ERS is measured and traceable to a national reference site (in this case, NPL in London), using a test technique that is strictly correct. This measurement gives the calibration data for the ERS.
If this ERS now is measured for emissions on a second perfect site, the results should be exactly the same as the calibration data. If, on the other hand, the ERS is measured on an imperfect site, the results will differ from the calibration data. The difference is due to the test site plus any other factors in the measurement system including the antenna and the analyzer. Essentially, this is a true end-to-end calibration process involving every step from the source of the signal to the result on the analyzer.
This process includes the following:
- Test-site attenuation error—reflections, height scanning (or lack of), and ground plane (or lack of).
- Antenna factor errors.
- Antenna cable attenuation.
- Calibration error in any preamplifier.
- Analyzer gain error.
- Detector error and any other factors that may influence the result.
Factors that cannot be included in this list, such as ERS frequency and amplitude stability as functions of time and temperature, must be taken into account. In practice, however, these measurement uncertainties are insignificant.
Tests have shown that the ERS exhibits a typical worst-case stability of less than 0.05 dB/°C and 0.1-dB/month drift. Fundamental frequency accuracy is 80 ppm, but the automated software that controls the site calibration process automatically measures and corrects for any ERS/analyzer frequency mismatch.
This technique can be implemented in two ways: manually and automatically. While these procedures are not specified in any standards, they can significantly improve measurement accuracy.
In the automatic implementation, the calibration data is loaded into the analyzer so that the system knows what the ERS should measure. Then, there are two stages in this process:
- Calibration—The ERS is located where the EUT would be placed and the emissions from it are measured. This result is correlated against the calibration data, and correction factors are calculated.
- Correction—This stage applies the correction factors to the results obtained from the EUT.
Although this sounds simple, there are many traps for the unwary. The most significant problem is the ambient environment. This must be measured and taken into account during both stages. Remember that this process produces the correction for a signal emitting from a certain location to the antenna.
This correction cannot be applied to any other signal, such as an FM broadcast signal being transmitted from a mast 10 miles away. As a result, these other signals must be eliminated from the results before the corrections are applied.
Software supplied with the Laplace SA1000 Analyzer now has operating modes in which these steps are applied automatically, making the process transparent to the user. This technique provides a complete overview of the characteristics of a test site. The result is a quantitative comparison between the user’s site and an ideal site (NPL).
In the case of screened rooms (nonanechoic), the locations of the ERS and the antenna are critical. In small rooms, slight movement of either could have significant effects on the results.
In an OATS, ground conditions may affect the site calibration so it should be done each time the site is used. This technique can show good correlation with test-house results for products similar in size to the ERS.
With larger products, if the ERS can be accurately positioned, a good correlation is maintained. Even in cluttered sites where corrections can approach 20 dB, measurement errors are reduced to generally less than 6 dB, a similar level reported by many test laboratories.
The manual procedure includes a specific routine to ensure that the ERS is accurately located at the true source of emissions from the EUT:
- Prepare the test site and install the EUT.
- Use the measurement system in the normal configuration to detect any emissions from the EUT over the entire frequency range.
- Note the frequency of any abnormal emissions peak.
- Locate the source of the emissions peak as accurately as possible using a near-field probe. This location may not correspond to the location of the highest readings from the near-field probe.
- Remove the EUT and place the ERS as close as possible to the emissions source. Turn on the ERS.
- Measure the amplitude of the ERS emissions nearest the frequency under investigation.
- Compare the measured amplitude with the ERS calibration data. The discrepancy is the correction factor to be applied to the original measurement.
- By applying this factor, the measurement is normalized to correlate with an ideal-site measurement.
Conclusion
To summarize, when measuring radiated emissions, be aware that the test site is by far the greatest unknown. Reducing this unknown involves preparing a site to conform to the standards for an OATS, calibrating for site attenuation, or using a precalibrated chamber, ideally a full anechoic chamber.
Remember, using a screened room does not help, other than in locations that suffer from extremely high ambient levels. Using an ERS with a suitable analyzer and software can provide a cost-effective and simple means of reducing test-site error.
Ambient signals can be handled, but there is a limit. Particularly noisy sites make radiated emissions testing impossible, and the only solutions are to move to a different site or use a test cell or anechoic chamber.
Screened Room Nightmare
The belief that a screened room eases the problems of radiated emissions measurements is a common one. Unfortunately, screened rooms are not advisable and should be used only in special circumstances.
By definition, screened rooms are metal-clad spaces. Each wall, the roof, and the floor are reflectors. These reflections cause resonances or standing waves to be established between opposing walls and between the floor and roof.
Imagine that the room is 6 meters long. That means there are two walls 6 meters apart. A signal with a wavelength of 6 meters (50 MHz) will be in resonance across this space because any reflection from the walls will reinforce the standing wave.
This effect also applies to all harmonics of 50 MHz. For instance, at 300 MHz, a standing wave will exist, establishing a series of peaks and nulls along this axis. These will be separated by a quarter wavelength, for example, 25 cm. If an antenna is located on a peak, it always will read high at 300 MHz. But move it just 25 cm, and the signal will be reduced to virtually zero.
Similar effects will occur in the other axes, and all this will be further complicated by the simple reflections off the metallic surfaces of the room. This is not the best way to obtain realistic measurements.
Screened rooms do have their uses for radiated emissions work. The absence of background interference makes them ideal for establishing the frequencies of any emissions. If the testing is moved to an OATS, just the frequencies identified in the room can be spotted and measured. This avoids trying to locate potential emissions frequencies in a noisy environment.
Secondly, a room can help if simple comparative rather than absolute measurements are required.
| For more information, visit “Bad Reflections.” |
About the Author
David Mawdsley is the managing director and founder of Laplace Instruments. Mr. Mawdsley originally trained with Rolls Royce (Aero Engines) as a mechanical engineer and later was a manufacturing manager for Data Acquisition. He received a B.Sc (Hons) London in electronic engineering. Laplace Instruments, Tudor House, Grammar School Rd., North Walsham, Norfolk NR28 9JH U.K., 011 44 1692 500 777.
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2000 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
September 2000