an md&di july 1998 column
emi field notes
radiation patterns offer insights that can help engineers find solutions to emi problems and therefore speed a device's time to market.
radiation patterns produced by electromagnetic interference (emi) emissions are a useful diagnostic tool for solving emi problems in electronic medical devices. emi radiated from computing products often has antenna-like radiation patterns, and one of the most common is the figure-eight pattern produced by dipole-like structures. this article discusses why emi sometimes forms these patterns and how they can be measured and interpreted.
attempting to identify these patterns has some limitations, however. not all emi produces well-recognized patterns, and identified patterns may still lead only to educated guesses about the radiating structures.
one way to view electronic devices is as a cluster of many transmitters (sources) and antennas (radiators). cables, printed circuit boards (pcbs), enclosure parts, and other components can act as antennas when they conduct emi currents from clocks and other sources. these currents may exist because a structure is part of a ground return path or because the structure has received and reradiated emi from somewhere else.
antenna-like structures need not be resonant to be emi problems. however, when they are resonant, they are more likely to produce stronger emissions. higher clock speeds, and consequently higher harmonics, increase the energy available at higher frequencies. relatively small components and structures that were emi-quiet with slower clocks can become efficient radiators at higher speeds because energy is more likely to exist that will excite their resonant modes.
figure 1. a simple dipole less than a wavelength long produces a distinctive pattern.
structures in electronic devices often behave like dipoles. in its simplest form a dipole is any long, thin conductor with a radio-frequency current flowing in it. if its length is less than a wavelength at the frequency of a current flowing in it from a clock harmonic, it will radiate energy in a figure-eight pattern, as shown in figure 1. the polarization of the emissions can also provide important information. for example, if a dipolelike structure is oriented horizontallythat is, parallel to the floor in normal operationthen its radiation will be horizontally polarized.
beam width and multiple lobing
the specific shape of emi patterns and the number of pattern lobes can provide further information about the radiating structure. for example, as a dipole gets longer (in terms of a wavelength) the figure-eight pattern will first become narrower (prolate), as shown in figure 2. then, dipoles longer than about one wavelength will have several major and minor lobes. when this occurs, it is possible to estimate the physical length of the structure acting as a dipole.
figure 2. dipole patterns versus length in wavelengths.
polarization can be used to differentiate dipolelike structures from other structures. for example, a figure-eight pattern and horizontal polarization implies a dipolelike structure. a figure-eight pattern and vertical polarization implies a more complex radiator. kraus explains many common antenna types and their patterns.1
regulations require that computing products be rotated during emi tests to find the angle of maximum signal strength. this technique can be used to measure emi patterns too. emi strength should be measured as a function of angle, and then these data should be plotted (figure 3). automated measurements are easier, and certainly faster, than manual methods, but either produces usable results. the trade-off is time versus the expense of automation.
figure 3. this typical polar plot of emi from real hardware shows a somewhat distorted dipolelike pattern.
angular resolutions of 5°10° should be adequate for showing figure-eight patterns. one way to estimate the level of resolution needed is to look at the dimensions of the device in terms of a wavelength. for example, a 1/2 wavelength at 300 mhz is 0.5 m. if most of the structures within the product are less than this, and if it's unlikely that the device's clock harmonics have much energy above this frequency, then multiple-lobed patterns characteristic of electrically large antennas are unlikely. of course, this is only a starting point. patterns can be remeasured with finer resolution if the initial results are inadequate.
there are two ways to address the emi of a structure that is acting as an antenna: reduce the emi current flowing in the structure or reduce the structure's efficiency to act as an antenna. source suppression is usually the least expensive way to reduce radiated emi, as long as it is caught early in product development when changes are relatively easy to implement.2
source suppression means turning down the transmitter, thereby reducing the energy available to excite structures and radiate. one of the most effective ways to achieve suppression is to slow down the rise and fall times of the clock signals to only what is absolutely necessary. another option is to use a slower clock. sometimes the only change required is to install a series resistor in the clock line. this resistor forms a simple resistor-capacitor low-pass filter in conjunction with the inherent capacitance between the clock trace and the ground plane on a pcb.
resonant structures, which are efficient radiators, often occur when a cable or other structure is nearly a 1/2 wavelength long. consequently, it is often sufficient to change the physical dimensions of a resonant structure or change the clock frequency so that the structure is not resonant at clock harmonics. in some cases plastic parts can be substituted for metal ones.
when it is impossible to change the dimensions of a radiating structure, it may be possible to dampen its resonance or attenuate induced emi currents. for example, i once found a structural supporting bar that was inadvertently resonant.3 nearby circuitry induced emi currents into the bar, and it radiated strong, horizontally polarized emi with a dipolelike figure-eight pattern. replacing the bar with a reinforced, plastic part would have been too expensive. although the bar was fastened at both ends, the support was grounded at only one end. placing a toroidal ferrite around this support reduced the emi to tolerable levels and was far less expensive than other solutions.
improving the grounding of a radiating structure may reduce the emi currents on the structure. however, this may backfire if it changes the structure's resonant frequency to a lower, and potentially stronger, clock harmonic. consequently, reducing emi at one frequency may increase it at another. when evaluating emi fixes, it is imperative to evaluate the effect of a change at several frequencies.
the size, shape, and orientation of a structure acting as an antenna can be inferred by carefully examining the patterns of its emi signals. for example, let's say a device has a horizontally polarized emission with a figure-eight pattern at some frequency, f1 = 156 mhz, from a computing device with a 12-mhz clock.
1. immediately note that:
a. the emi is at the 13th clock harmonic: 156/12 = 13
b. a dipolelike structure can be inferred as oriented horizontally.
c. its length is likely to be at least 1/2 wavelength:
2. is the figure-eight pattern prolate?
a. yes: estimate that the length equals roughly 1/21 wavelength.
b. no, lobes look more like circles: estimate that the length is 1/10 wavelength:
3. gain additional insight by looking at patterns at progressively higher harmonics. find a frequency where the pattern begins to show a six-lobed pattern of a dipole about 11/4 wavelengths long. let's say this occurs at 276 mhz (the 23rd harmonic), then:
so in this example, the engineer would look for a horizontally oriented structure at least 0.9 m long, oriented perpendicular to the direction of the main or strongest pattern lobes.
several factors can make the pattern identification technique difficult to use. when designing antennas, strong, steady signals are desirable for pattern measurements. but when it comes to emi, weak emissions are preferred. emi signals with low signal-to-noise ratios can make pattern identification difficult. this can be further exacerbated if the emi fluctuates rapidly.
devices are also different from intentional antennas because they usually contain multiple random radiators or incidental antennas. almost any conductive structure or part can act as an antenna. they often all do to some extent. each antenna-like structure has its own pattern, so emi patterns are actually many patterns combined.
finally, the location of cables and other components may vary enough to cause unit-to-unit emi variation. for a resonant structure, the exact frequency may vary greatly depending on the position or even the tightness of fasteners.
nevertheless, looking for emi emission patterns can be a useful technique for diagnosing emi problems. once found, these patterns are one more piece to help solve what at first may seem like a confusing puzzle.
1. kraus jd, antennas, 2nd ed, new york, mcgraw-hill, 1988.
2. kimmel wd, and gerke dd, electromagnetic compatibility in medical equipment, piscataway, nj, ieee press, 1995.
3. roleson s, "finding emi resonances in structures," emc test des, 3(1):2528, 1992.
scott roleson is an emc and telecommunications engineer for hewlett-packard (san diego).