EMC is all relative. For a given set of RF environmental conditions, at a given location, and at a specific relative instant in time, the equipment has EMC or the absence of failure. Change any of these conditions with respect to the equipment’s orientation and internal operation and an EMC failure may not occur because both the electromagnetic environmental (EME) disturbances and the equipment’s response to these disturbances are statistical in nature.
We can make the probability smaller, but there is no way by testing to duplicate all the possible combinations of frequencies, amplitudes, modulation waveforms, spatial distributions, and relative timing of the many simultaneous interfering signals that an operating system may encounter. As a result, it’sgoing to fail. So, does this mean we shouldn’t do EMC tests? Of course not.
The Need for an EMC Safety Margin
The approach used for systems and equipment that must function in critical applications is to establish an EMC safety margin (EMCSM), which in effect, increases the severity of the EMC test. Normally, an EMCSM is established to double the worst-case environments to set a test level in which a unit should operate when exposed to any assumed combination. Otherwise, there is no way to even guess what combinations of EMI frequencies and levels the equipment will withstand.
If it is known from the beginning that EMC problems might exist in the equipment being used in critical applications, and there is great risk of failure, then a safety margin is required. If various military standards are reviewed or if we refer to CISPR’s 80/80 rule, a 6-dB EMC safety margin is specified to allow for variations in manufacturing.
This doesn’t cover measurement equipment errors, errors that result from the interactions of EMI with the surroundings, or the standard’s misrepresentation of the operational environment. Consequently, an EMCSM much higher than 6 dB may be required. Some studies suggest 12 dB may be necessary. Others indicate 20 dB, and 40 dB also has been cited.
The value is directly related to the cost of the system and the difficulty of fixing problems after the system has been deployed. Even though these EMCSM values may be correct, they can seem too conservative for equipment manufacturers that insist that the equipment emissions or susceptibility levels need only be equal to or less than the spec limit.
Keep in mind this simple rule: A signal or response at the specification limit has an equal chance of being above or below the limit. Murphy’s Law, however, states that with a 50-50 chance for a given outcome, there is a 100% chance of failure.
Any system that contains intentional RF emitters or circuits that may be susceptible to levels within 20 dB of the RF ambient levels at the system’s operational locations should have EMC requirements imposed. And these requirements need to be verified by testing the system.
Complete and Exhaustive Data Set
Testing at higher RF levels helps to identify the problem areas, but that is not enough. We must adopt a test philosophy like NASA’s. I’m sure most of you have heard the motto that expresses the complete NASA testing philosophy: Test like you fly…Fly like you test. In other words, perform a test so that it represents the conditions that the EUT will experience when it’s being used, commonly referred to as a complete and exhaustive data set.
As an idealistic example for a spacecraft to establish the EMC susceptibility limits, a long-term EME survey should be made starting at the manufacturing location, continuing through the transportation phase to the storage location, and then finally adding the on-orbit usage environment. A composite worst case would then be developed from this information and the device designed and tested accordingly.
Keep in mind that at each of the steps along the way, the EME is continually changing. These changes can evolve slowly as new equipment is developed that uses RF spectrum, or the environment can change extremely rapidly from being a clear day to one filled with thunderstorms and associated lightning transients or from sitting quietly on the ground to being launched on a planetary mission.
The environment does not change as rapidly for equipment used in fixed locations as it does for mobile equipment. For equipment carried on aircraft or launch vehicles/spacecraft, the EME includes the flight path. For non-geosynchronous spacecraft, the EME is the entire Earth’s surface. For ground vehicles, the EME is not quite so vast but still a rather large geographic area.
Because of the logistical difficulty of maintaining an up-to-date composite database covering such a large geographic area, military and commercial standards are used at a cost of either over- or under-design. Even then, it is prudent to use an up-to-date specification.
Why Things Are Getting Worse
During the past 20 years, there have been significant technological advances, especially with semiconductor devices. Unfortunately, these advances have made the EMC problem worse. The quest for faster and cheaper data processing has resulted in the following configuration changes to large-scale integrated circuit designs:
- Smaller geometry to reduce self-inductance.
- Greater packaging density.
- Shorter substrate conductors
- Lower 0/1 state change voltage swings.
- Lower power requirements.
- No internal protection diodes.
- Limited shunt protection capacitance.
- Plastic packages.
- Higher device speed directly related to bandwidth.
Just as Moore predicted 30 years ago, component density doubles about every two years. Likewise, data processing is increasing at an exponential rate. Each of these advances has increased computing speeds while reducing costs. However, they also have resulted in increased emissions and greater susceptibility.
Fortunately, test and design standards have evolved in an attempt to keep up with these technological innovations. New equipment should be designed and tested using the current standards that, at this time in the military sector, are MIL-STD-461E for equipment and MIL-STD-464A for systems.
Today’s Equipment and Systems Standards
The limits provided in these standards reflect current experience, EMC environments, and electronic designs. Figure 1 indicates how the military EMC test requirements for radiated emissions and susceptibility have changed since MIL-STD-461 was introduced in 1967.
Although this is a reasonably comprehensive test procedure, it is difficult to cover everything. Accordingly, companies working with spacecraft should obtain the new MIL-STD-1541B as a guide for tailoring the limits and setting appropriate EMC safety margins for mission-critical circuits.
MIL-STD-461 has always encouraged contractors to make tailoring decisions provided the rationale for such tailoring is included in the EMC control plan. Tailoring is performed before testing, and the levels and procedures are based on a priori knowledge and understanding of the operation of the equipment and the EMC environment in which it is used. The process of simply drawing a new limit line a few decibels above the measured failure values after the qualification test is completed and calling that tailoring is unacceptable.
Even when a system meets the most up-to-date EMC requirements, it still will fail. In some cases, the specification simply cannot be applied.
For example, receivers and transmitters have special considerations. Both are exempt from meeting the radiated emissions and susceptibility requirements within their operational pass bands. A receiver capable of detecting a 0.5-µV signal certainly cannot withstand a 200-V/m susceptibility level within its pass band. Likewise, a 100-W transmitter won’t meet a 25-µV/m emissions requirement within its pass band.
So even though an up-to-date standard combined with intelligent tailoring may have been used, receivers and transmitters still must be protected from outside influences. Additionally, EM field coupling is extremely complex, making it nearly impossible to predict system-level performance based on subsystem/equipment test results.
Consequently, it is important to perform systems-level/operational EMC tests. These tests should mimic the actual operational environment as closely as possible.
That is the basis of NASA’s systems-level testing philosophy of test like you fly. Unfortunately, this may be extremely expensive, difficult if not impossible, and time-consuming.
For example, testing the on-orbit radiated susceptibility of a spacecraft and mimicing its operational environment requires the EMC test facility to generate a complex EME during a cryogenic-thermal vacuum test being performed on a high-level shock and vibration table under weightless conditions. It would almost be cheaper to launch it and then test the device while in orbit.
In an attempt to provide EMC evaluation close to the actual operational environment, the NASA Glenn EMC facility combines EMC measurement with cryogenic thermal temperature and a near-space vacuum. The EMC chamber is 100 feet in diameter and 122 feet high.
Shortcomings of the Standards
EMC tests were developed to make them easier to perform, provide better repeatability from one lab to another, and reduce testing costs. But even though the EUT may pass its EMC tests, it still may fail when it’s put into service. Laboratory tests performed in accordance with the standards are not adequate to guarantee that an EMC failure will not occur during actual operation because the tests do not represent the operational EME. However, they can provide a good understanding of the potential EMC problems.
Figure 2 defines an approach to determine the potential effects from electromagnetic energy on systems. The worst shortcomings of EMC standards testing tend to be the radiated tests. That will be covered later, but while on the subject of test shortcomings, we have to question whether the equipment has been tested.
Has the Unit Been Tested?
Government organizations are divided on the need for equipment to meet both emissions and susceptibility requirements. The FCC, for example, only places requirements on the emissions from the EUT and does not consider the device’s susceptibility.
This approach covers less than half of the problem. It’s true that some level of reciprocity exists between emissions and susceptibility, and what is done in the design to solve one also helps the EMC design for the other issue. That is why the military and EU specifications emphasize both emissions and susceptibility.
Often, new equipment isn’t tested because it’s approved by similarity to equipment that has been tested. However, a single EUT tested 10, 20, or 40 years ago cannot possibly be used to represent similar equipment being built today.
The input-output transfer functions may be similar, but that does not mean that the new device has the same or even remotely similar EMC characteristics as the 20-year-old product. In many cases, it is not even possible to procure the semiconductor devices that were being built 10 years ago, and today’s fast low-level semiconductor devices generate greater emissions and are more susceptible than yesterday’s devices.
It’s often not known until in-the-field failures occur whether a given EUT passed its previous EMC tests because it was well designed or poorly tested. Plus, the only way to determine continuing EMC compliance is by performing periodic production sample testing during the life cycle of the equipment. All that would be known is that when the EUT was tested in that environment in accordance with the procedures used at that specific instant in time, the device is defined as having EMC.
Even doing this with current production devices, the component, measurement-equipment, and test-setup/facility variations often result in large differences in EMC measured data. This measurement uncertainty is being addressed by today’s EMC test procedures.
A good reference to learn more about the statistical nature of EMC measurement uncertainty is IEC CISPR 16-4. This uncertainty deals only with the random and skewed errors associated with making measurements to a standard and does not address how well the standard’s test environment mimics the actual operational environment.
Initial Test Conditions
During typical EMC tests, power is supplied to the EUT from shielded enclosure filters via either 10-µF feed-through capacitors or 50-? line impedance stabilization networks (LISNs). This results in an artificial source impedance that does not represent reality.
Operational power-line impedance varies from 0.001 ? to approximately 600 ?, and it does this as a complex function of frequency. Source transmission characteristics depend on whether real power is a battery, a small 1- to 20-kW portable generator, or a giant 10-MW generator at a nearby power plant and whether the power-line is buried, aerial, run in plastic, aluminum, or steel conduit or simply lying on the ground.
Wire and cable are carefully laid out on the test stand or bench to provide controlled common impedance loop conditions with each box or sub-assembly diligently arranged in a straight line facing the test antenna. Even though the standards require the side with the greatest emissions or the greatest susceptibility to be facing the antenna, both of which vary with frequency, most of the time this is determined by visual inspection. How I don’t know.
This might be acceptable if the tests are performed in a reverberation chamber or a simple metal-box enclosure. But as soon as an attempt is made to make the enclosure environment appear as a free-field environment through the addition of absorber materials, all those stray reflected signals that might have been captured are lost. The linear layout, as opposed to a real layout, may make testing easier but also introduces errors.
To reduce the overall test time in a large system, modes of operation are chosen based on the type of test. The highest emission mode is used while performing radiated and conducted emissions measurements, and likewise the highest susceptibility mode is used during radiated and conducted susceptibility.
It’s hard to define these mysterious modes since they are not really distinct modes of operation. In the good old days when systems had only a single mode of operation, such as either on or off, making a determination of what mode to use was simple. Now, however, systems are so complex we just have to be certain that the configuration used assures that all circuits are operational during the tests.
Each circuit produces interference independent of other circuits in the system that may or may not be correlated. If the sources are correlated, the levels add directly, but the correlation factor can be either positive or negative. This degree of correlation is unlikely, so from an EMC perspective circuits are assumed to be uncorrelated. With this assumption, the emissions from uncorrelated circuits add together as the sum of their individual noise power
Pt = P1 + P2 + . . . Pn
This gives
E2/Zt = E12/Z1 + E22/Z2 + . . . En2/Zn
If we simplify the expression by assuming both the voltages and the impedances are equal, the expression becomes
Et = (n En2)0.5
which is true for plane waves. Accordingly, all circuits must be operational during the tests or the measured levels will be incorrect.
Single-Frequency Tests
One of the biggest departures from reality is the single-frequency susceptibility test. Both MIL-STD-461 and the EU susceptibility tests are performed by sweeping a single constant-amplitude narrowband signal through the test-frequency range. This doesn’t represent the ambient.
The actual RF environment has many simultaneous signals. Some are narrowband, some broadband. Plus, the broadband may be random or coherent.
The signals use an assortment of modulation schemes. Originally, the modulating frequency and waveforms were selected based on those that would cause the greatest stress on the EUT.
Although the specification still calls for the test signal to be modulated, the characteristic of the modulating signal has changed over the years. That requirement has been simplified until now all that is required is a 1-kHz 50% amplitude-modulated square wave. This minimum is accompanied by a tailoring statement suggesting that using a waveform similar to that encountered in the actual environment may be a good idea.
In the actual environment, there are large numbers of narrowband signals occurring simultaneously. These signals have random/pseudo-random variations in frequency, amplitude, spatial characteristics, and timing.
When they are sufficiently close in frequency that multiple narrowband signals fall within the pass band of the receptor circuit, they appear to the EUT as a broadband signal with a complex modulation and randomly varying amplitude within the pass band. There also are broad bandwidth signals occurring in the mix at the same time
The illumination of the EUT by these signals is taking place from all directions simultaneously, not just from a test antenna located 1 meter or possibly 3 meters away as required by unit test standards. The effect is to submerge the EUT in a susceptibility signal whose coupling influence covers the entire range of frequencies from all directions, with randomly varying amplitudes, and all at once.
A number of studies have been performed by NASA and others that indicate that one susceptibility signal can bias a circuit and additional susceptibility signals added to it can result in failure. An approach to more closely simulate the environment without increasing the test complexity a great amount would combine the single narrowband modulated signal with a broadband signal from a high-voltage impulse generator.
Susceptibility Criteria
Although possible, especially with highly sensitive receivers, an EMC susceptibility test typically does not cause permanent damage to the equipment, only upset, malfunctions, and temporary degradation of performance while the susceptibility signal is present. As a result, the performance of the EUT must be monitored continuously during the test with pass or fail determined in real time.
It’s generally not possible to determine EMC susceptibility by measuring and comparing the performance of the EUT from before the test with its operation after the test. Yet, some manufacturers attempt to do this.
To keep total test time to a minimum, the monitoring criteria used for EMC testing generally are short and simple. This simplified approach to determining susceptibility may not reveal subtle susceptibilities that would occur in a more complex electromagnetic environment. In addition, the monitoring equipment and its application may influence the EUT’s susceptibility, causing it to be more or less susceptible as a direct result of the monitoring equipment.
Antenna and Enclosure Errors
In the past, EMC tests were performed in ordinary metal-box shielded enclosures that had the advantage of more closely representing the installed environment, especially with military equipment. Now, the tests are performed in enclosures loaded with anechoic materials that significantly reduce the amplitudes of simultaneously arriving reflected signals from other directions and further remove the testing from simulating the actual environment. This approach does improve repeatability from lab to lab, but it tends to reduce the detected emission levels and increase the levels required to produce equipment susceptibility.
Although the internal enclosure ambient is low and permits detection of radiated emission signals from the EUT, the enclosure supports reflections and standing waves. Because of constructive or destructive addition, there may be increased or decreased localized field strengths depending on the phase relationships of the arriving wave fronts.
During use, antennas are closer to the walls and ceilings in smaller enclosures, which results in antenna detuning. In other words, the antenna factors that were so carefully determined at an open area test site (OATS) no longer are correct.
Additionally, antenna calibration generally is performed in the far field. But measurements made at a 1- or 3-meter separation distance position the antenna in the near field for lower frequencies. The transition distance used by the EMC community is R = 300/2? F. The frequency corresponding to a 1-meter transition is 47.7 MHz; for 3 meters, it is about 15.9 MHz.
Antenna beamwidth and antenna gradient introduce more errors as a result of the linear layout of the EUT. This is illustrated in Figure 3. Since the antennas are not omni-directional, if the beamwidth angle does not cover the width of the EUT, then additional antenna positions must be used.
The same requirement for multiple antenna positions holds for antenna gradient, but the level of measurement error is much greater. This is especially true for MIL-STD-461, which uses a 1-meter antenna distance. In this case, the RF signals reaching the antenna from the ends of the linear EUT layout can travel substantially greater than 1 meter. At some frequencies, the fields from the center and end are in the near field, sometimes one is in the near field and one is in the far field, and at some frequencies, the distance places them both in the far field.
Summary
EMC testing for components and subsystems has evolved greatly over the last 40 years since MIL-STD-461/461A was released. This design and test-procedures document was a credible attempt at standardizing EMC tests and has formed the basis for many other government and industrial procedures for equipment and subsystem tests. Unfortunately, it was not designed for system-level testing, and any attempt to use the procedure for in situ operational tests is futile.
Systems-level tests are a step up from MIL-STD-461 or the equivalent commercial tests and determine if each piece of equipment and subsystem that composes the system will work together. But these tests do not include the operational EME. In test-like-you-fly configurations, the EUT is placed in the operational EME or a synthesized equivalent to ensure that the EUT performance will not be degraded by the environment and that the EUT does not contribute to the EME.
Unfortunately, duplicating the operational EME is impossible. Even coming close is very expensive because of the continually worsening adverse RF and physical environment such as shock and vibration, ESD, lightning, salt fog, or rain. These changing environmental conditions directly affect the performance of the EUT device.
Since the EME is continually evolving, it is not possible to guarantee future performance of a new EUT based on the performance of the same or similar devices that may have been fielded in the past. This is especially true if an EMCSM was not established. Without previous EMCSM data or current EMC test data that provides the susceptibility profile, it can only be assumed that the equipment is at the threshold of failure with respect to the original EMC test environment. If the EMC test standard used was MIL-STD-461C, that standard defines a 20-year-old environment that is not adequate to protect today’s equipment.
Equipment often fails in its operational environment, and manufacturers chalk it up to unknown causes when, in many cases, the problem is an EMC susceptibility problem. All such failures should be investigated and a post-mortem reporting procedure developed to determine the cause of the failure so future systems won’t have the same problem.
When it’s not possible to test like you fly and provide reasonable assurance that the system will work in an adverse environment, use a modern standard tailored to the most flight-like conditions possible and determine an EMCSM during the performance of the test. The recommended minimum EMCSM is 6 dB, but 12 dB would be better.
About the Author
Ron Brewer currently is a senior EMC/RF engineering analyst with Analex at the NASA Kennedy Space Center. The NARTE-certified EMC/ESD engineer has worked full-time in the EMC field for more than 30 years. Mr. Brewer was named Distinguished Lecturer by the IEEE EMC Society and has taught more than 385 EMC technical short-courses in 29 countries and published numerous papers on EMC/ESD and shielding design. He completed undergraduate and graduate work in engineering science and physics at the University of Michigan. e-mail: [email protected]
December 2007