It’s no secret that a uniform, worldwide EMC test standard is long overdue.
Part 1 of this article, which appeared in the September 2002 issue, introduced the idea of creating a universal EMC testing standard for tests performed within the United States. Such a standard would eliminate many testing problems and significantly reduce manufacturing costs and time-to-market.
Part 1 addressed the different tests required by MIL-STD-461E, FCC, and EU standards; the testing environment; power-line impedance control; and measurement distance. Part 2 focuses on measurement errors, test setups, antenna characteristics, receiver detector function, and receiver impulse bandwidth. These areas also must be coordinated to create a universal test procedure.
Measurement Errors
Measurement errors are introduced by instrumentation errors, the test environment, the calibration sources, and the measurement procedure interpretation/implementation. Using NIST-traceable calibration facilities and NARTE-certified EMC engineers and technicians will eliminate most of the procedural errors. A universal testing standard would help eliminate most of the human errors.
Errors can be systematic or random, meaning that they bias the measurement positively or negatively each time or they are independent random occurrences. Errors are functions of many variables, and some of these are not easily defined. Additionally, some of the errors are skewed in strange ways so the error-probability distributions are not always normal Gaussian distributions. Overlaying these varying error distributions to arrive at a calculated error value is difficult.
MIL-STD-461E attempts to circumvent this problem by requiring that the entire measurement system be calibrated from the sensor to the display using standardized reference signals for each procedure. This does an excellent job of minimizing the errors for conducted measurements, but radiated measurements still are prone to errors resulting from interaction with the surroundings.
ISO/IEC 17025 General Requirements for the Competence of Testing and Calibration Laboratories (December 1999) describes measurement uncertainty requirements for both testing and calibration. For a universal standard, each procedure should be calibrated as required by MIL-STD-461E, and ISO/IEC 17025 should be used.
Sensor errors and receiver errors have the greatest effect on EMC measurements. In frequency ranges that overlap, the antennas and line impedance stabilization networks (LISNs) largely are the same for FCC, EU, and MIL-STD-461E specifications. However, in frequency ranges that are different, additional equipment is required to cover the range. MIL-STD-461E conducted measurements start at 30 Hz and radiated measurements at 10-kHz. Commercial conducted tests begin at 150 kHz and radiated at 27 MHz.
Radiated Measurements
The antenna factor (AF) relates the incident electromagnetic field (V/m) to the voltage developed at the output terminals of the antenna (Vo); that is, AF = (V/m)/Vo. By generating a known field from a calibrated source antenna and substituting the antenna to be calibrated, the antenna factors can be determined. By far, the best approach is to calibrate each antenna, although some manufacturers rely on batch calibration, and a few simply use the antenna-factor curves that are provided in the Society of Automotive Engineers antenna documentation.
Tuned dipoles typically serve as the calibration reference antennas. NIST data indicates that tuned dipole antenna factors can be determined to within £0.5 dB and biconical antenna factors £3 dB. Near-field errors add another »1.5 dB for frequencies below 100 MHz. Transmission-line effects on the electric field (EF) add another »1.0 dB.
These numbers are troublesome, especially since errors cannot exceed 2 dB per MIL-STD-461E. But the big problem is errors introduced by the voltage standing wave ratio (VSWR). The biconical antenna and its several extended-range derivatives have among the worst VSWR values, ranging to approximately 13:1.
The variation in peak-to-peak voltage (Vpk-pk) for an antenna with an available antenna voltage eg connected to a transmission line with a length ?/2 is:
eg/(1 + 1/VSWR)
Error pk-pk = which equals VSWR
eg/(1 + VSWR)
Expressed in decibels: Error pk-pk (dB) = 20 log VSWR
For an antenna with a VSWR of 13:1, the error is approximately 22 dB. Adding a 6-dB pad will lower the VSWR to approximately 1.5 and the pk-pk error to 3.5 dB. A 20-dB pad will reduce the VSWR to approximately 1.013 and the pk-pk error to about 0.11 dB. Always use the largest pad at the receiver front end that does not result in out-of-limit readings.
Receivers also have VSWR errors, but because receiver input impedances are more nearly 50 ? over their entire frequency range, the VSWR error generally is less than 0.5 dB. Other errors, such as those associated with the RF attenuator, impulse bandwidth, and gain stability, add up to less than about 1.5 dB to 2 dB.
The biggest problems are receiver saturation and mistaking spurious responses for equipment-under-test (EUT) signals. Most of the newer computer-controlled receivers check for these problems. Even if they must be done manually, such checks should be required by a universal specification.
Test Setup
MIL-STD-461E places the EUT on a stationary ground plane and moves the antennas to minimize the problems with beam width and gradient. Worst-case emissions areas are determined by inspection. In contrast, the FCC/EU tests are performed with the EUT placed on a turntable and rotated through 360° while simultaneously raising and lowering the antenna.
Regardless of which approach is used, the test RF environment must have conducted and radiated RF ambient levels below the specification. If the tests are performed inside a shielded enclosure, the enclosure must be large enough to handle the EUT and the necessary test antennas to minimize error caused by internal reflections and antenna loading.
Both the MIL-STD-461E and FCC/EU tests specify sufficient spacing to minimize these effects. Neither standard adequately addresses the problem of enclosure resonance although the addition of absorber material helps, and MIL-STD-461E does permit the use of mode-stirred chambers.
The radiated test time is determined by multiplying the number of modes (M) of operation × time (t) to complete each operating mode × number of frequencies (F) tested × test sample (S) face locations × antenna polarizations (P) × number of antenna locations (L). MIL-STD-461E specifies scan times, antenna locations, and bandwidth, but the EU test requirements are not quite as thorough.
For a MIL-STD-461E or FCC/EU test, acquiring a complete and exhaustive data set can take a long time. As an example, assume a device has two modes of operation and each mode takes 3 s to test. The turntable requires 30 s to complete one revolution, the antenna needs 10 s to change from minimum to maximum elevation, and there are two polarizations. The frequency range is 30 MHz to 1,000 MHz using a 100-kHz or 120-kHz bandwidth.
A 970-MHz frequency range requires 9,700 100-kHz bandwidths. Time is 2 × 3 × 30 × 10 × 2 × 9,700 = 34,920,000 s or approximately 280 years. Most labs will perform this test in three or four days and stake their reputation on the validity of the data.
Antenna Polarization
Both military and commercial tests originally called out manually tuned dipole antennas. Later, shortened broadband dipoles were introduced into EMC measurements, and the specifications began requiring that both vertical and horizontal polarizations be measured. The measured signals aren’t necessarily polarized in the same way as the antenna so the arriving signal at the required measurement distance may be cross-polarized with the antenna. Polarization error in decibels resulting from polarization mismatch is:
Error (dB) = 20 log (cos ?)
where: ? = angle between polarizations
The largest angular difference is 45°, which equates to a -3-dB systematic error.
Field Gradient and Beam Width
MIL-STD-461E radiated tests may begin at 10 kHz, depending on which branch of the military service, which emissions or susceptibility tests, and which conducted susceptibility tests have been applied. At 10 kHz, the wavelength is 30,000 m, and the near-field/far-field crossover occurs at 4,774 m. At the lower frequencies, regardless of the EUT’s size and the antenna location, signals arriving from all parts of the EUT will be near field.
But for a test condition where the antenna is placed close to a large EUT, as the frequency increases, distant signals arriving from the EUT’s farthest point now may be plane wave, and signals arriving from the nearest parts of the EUT closest to the antenna still may be in the near field. Finally, as the frequency continues to increase, both the distant signals and the signals arriving from the nearest part will be plane wave.
This problem and the antenna beam-width problem are illustrated in Figure 1. The error in the measured field from the distant signal paths compared with the nearest path is computed using the following equation:
Error(dB) = 20 log (Rn/Rf)n
where: Rn = near distance in meters
Rf = far distance in meters
n = 1 for all signals in the far field (Rn and Rf >> ?/2?)
n = 2 for Rn in the near field, Rf in far field
n = 3 for all signals in the near field (Rn and Rf << l/2?)
The errors can be substantial when both Rn and Rf are near field. For example, when Rf = 1.5 Rn, the error is approximately 10 dB. The beam-width problem occurs because the antenna coverage in azmuth or elevation does not include the entire test sample. These errors can be minimized by using multiple antenna locations or a turntable.
Receiver Detector Function
Probably the greatest problem to overcome in creating a universal testing standard is the receiver usage. This includes both the detector function and bandwidth. Most receivers can measure the signal in a number of different ways, and they have several detector functions available. The three most popular are true peak, average, and the CISPR quasipeak.
Digital devices tend to respond to the true peak values of a transient or pulse signal. Analog devices more closely respond to the average value of the interfering signal. Peak and average values can be easily related to one another mathematically. This certainly is true for the less complex waveforms. MIL-STD-461E limits use true peak.
The FCC/EU specifications rely on the CISPR quasipeak detector function. It was developed during the mid 1930s to emulate the feelings of listeners to the interference occurring on AM broadcast band radios. The quasipeak detector’s weighting circuits with long charge/discharge times preclude making automated measurements.
Measurements are performed by first making measurements using a true peak detector. If an unwanted signal that exceeds the specification limit is discovered, then the receiver is switched to quasipeak and that signal is remeasured to see if it passes or fails.
Because the quasipeak detector discriminates against low-repetition-rate pulses and the true peak detector does not, a device can create significant transient interference to surrounding digital devices and still be within specification limits when measured using quasipeak detectors. Measurements made using simultaneous true peak and average detection are better suited to a universal EMC test standard.
Impulse Bandwidth
Signals come in two varieties: broadband and narrowband. The distinction is based on the bandwidth of the receiving circuit. It could be the bandwidth of the device being interfered with, but in this case we are referring to the measurement receiver. As shown in Figure 2, the receiver passband characteristics do not greatly affect the measurement of narrowband signals. That is not true for broadband signals, and most signals today originate from digital devices and are broadband.
If all receivers had a rectangular passband, the bandwidth could be specified, and measurements made with one receiver would be the same as those made with another having the same passband. Unfortunately, that is not the case. Even if all receivers were nominally identical, each one still would have some minor variation in its functional characteristics resulting from component tolerances.
The characteristic that permits one receiver to be compared with another when measuring broadband signals is its impulse bandwidth. Unfortunately, not one of the current standards requires receiver impulse bandwidth measurements, partly because receivers with wide-open front ends, such as spectrum analyzers, have problems with overloads.
Of course, they may have the same problem during a test. For additional information, refer to IEEE 376-1975 Standard for the Measurement of Impulse Strength and Impulse Bandwidth.
There are other errors to minimize and other measurement details to tweak, but taking care of the problems discussed in this article would make it easier to create a universal EMC test standard—and we certainly could use one.
About the Author
Ronald W. Brewer is an internationally recognized EMC authority named a Distinguished Lecturer by the IEEE EMC Society. The NARTE-certified EMC/ESD engineer has worked full time in the field for more than 30 years, specializing in EMC systems design, integration, and shielding. Mr. Brewer completed undergraduate and graduate work in engineering at the University of Michigan, serves on the IEEE EMC Society Board of Directors, and recently was featured in the CBS History Channel Special “Three Air Crashes: Common Links.” P.O. Box 1221, Leesburg, VA 20177, 703-727-4150, e-mail: [email protected]
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2002 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
September 2002