Separate emissions by bandwidth
Set up your EMI receiver or spectrum analyzer correctly to insure valid emissions measurements.
David Bare and Tom Parker, Elliott Laboratories, Sunnyvale, CA -- Test & Measurement World, 2/1/2002
Electronic products emit energy from switching power supplies, radio outputs, and microprocessor clocks. You need to measure these emissions and their harmonics to verify compliance with EMC standards. How you set up the measurement system can affect your measurements.
A signal's sidebands, harmonic content, and pulse repetition frequency play a role in how you should set up your measurements. Improper settings can lead to measurement errors. Those errors often come from an overloaded instrument or from extraneous signals that infiltrate your measurement.
You can perform an EMI emissions scan with either an EMI receiver or a spectrum analyzer. Both instruments use internal bandpass filters to set the bandwidth that they use to scan across a frequency range. EMI receiver manufacturers refer to a filter's bandwidth as the instrument's intermediate frequency (IF) bandwidth. (Spectrum analyzer manufacturers refer to IF bandwidth as resolution bandwidth.) EMI standards such as CISPR 16 specify the IF bandwidth you need for a compliance test—200 Hz, 10 kHz, and 120 kHz (Ref. 1).
During an EMI scan, you'll get the best results if you use an IF filter with a bandwidth that passes energy from just one emissions signal or harmonic at a time. Unfortunately, that's not always possible because of the signal's spectrum of the emissions and the fact that a compliance test must use specified IF bandwidths. You may need to adjust the receiver's IF bandwidth and signal attenuation to get proper measurements based on the signal's spectrum—narrowband or broadband—and compensate for those adjustments in your test results. The terms "narrowband" and "broadband" refer to an emission's bandwidth relative to a receiver's IF bandwidth, not to a signal's fundamental frequency.
If, during an EMI scan, just one spectral line from an emission or harmonic falls within a given IF bandwidth, then you have a narrowband emission. If you get more than one spectral line within an IF bandwidth or if you can't distinguish any spectral lines, then you have a broadband emission.
 |
| Figure 1. A typical narrowband signal shows just one peak within an IF bandwidth (for EMI receivers) or a resolution bandwidth (for spectrum analyzers). |
 |
| Figure 2. A 120-kHz IF bandwidth may make a signal appear to be broadband. |
 |
| Figure 3. At a 10-kHz IF bandwidth, the signal in Figure 2 takes on narrowband characteristics. |
 |
| Figure 4. A filter with a large shape factor (wide plot) will let more energy enter a measuring system than a filter with a small shape factor (narrow plot). EMI receivers use filters with small shape factors and are therefore more selective than spectrum analyzers. |
Microprocessor clock signals, which typically range from a few megahertz to hundreds of megahertz, produce emissions at the clock's fundamental frequency plus higher-frequency harmonics. Assuming the clock fundamental is narrowband and because the harmonics appear at multiples of the clock's fundamental frequency, you'll never see emissions from more than one harmonic within a typical measurement IF bandwidth. Figure 1 shows a narrowband emission. In this spectral plot, most of the signal's energy falls within ±3 divisions (±60 kHz) of the signal's peak.
Broadband emissions occur whenever a signal's pulse repetition frequency is less than a receiver's IF bandwidth. Such a condition occurs with a 30-kHz power-supply switching frequency and a 120-kHz IF bandwidth. Emissions from three or four harmonics may fall within one IF bandwidth. The EMI receiver will display the sum of all energy within that IF bandwidth, which will produce a measurement that's higher than you'd get from just the fundamental signal or one harmonic. Figures 2 and 3 show the same emissions signal viewed differently. In Figure 2, the 120-kHz IF bandwidth makes emissions appear broadband. In Figure 3, though, a 10-kHz IF bandwidth reveals peaks 57 kHz apart. The narrower IF bandwith also reduces the measurement's noise floor, which brings the repetitive emissions into view.
You can use other methods to determine if you have a narrowband signal. For example, suppose you measure an amplitude that varies as you change the instrument's IF bandwidth. Changing the IF bandwidth affects the number of spectral lines within that IF bandwidth. Because the EMI receiver measures the total energy in a given IF bandwidth, the receiver will detect more energy as you widen the IF bandwidth. With a narrowband signal, however, the amplitude of the signal won't change because the IF bandwidth includes only one spectral component.
Use the center frequencyYou also can use an EMI receiver's or spectrum analyzer's center frequency to determine if an emission is broadband or narrowband. If you can tune the EMI receiver's center frequency plus or minus two IF bandwidths from the emission of interest (without changing the IF bandwidth), and the amplitude remains fairly constant (less than a 3-dB change), you have a broadband emission. Narrowband signals such as those from microprocessor clocks should reveal a noticeable reduction in amplitude as you tune your instrument's center frequency away from the peak of the signal.
Either broadband or narrowband emissions can overload a receiver. To test for an overload condition, tune the receiver's center frequency to match the signal you want to measure. Adjust the instrument's input attenuators until the signal appears well above the noise floor. Don't let the input amplifier clip the signal, because it will produce gain compression or harmonic distortion.
Insert an external attenuator in front of your receiver and "switch in" a fixed amount of attenuation. If you use a spectrum analyzer, you'll often need to insert an external preamplifier between the EMI antenna and the instrument. Without attenuation, you could overload the preamp, which will send a distorted signal to the spectrum analyzer. If you overload the external preamplifier, attenuators in the spectrum analyzer won't remove the distortion that the overloaded preamplifier produces—the analyzer's attenuators will simply reduce the amplitude of the distorted signal. Initially, don't concern yourself with the absolute signal level; just look at the differences in amplitude when you change increments of external attenuation.
If the signal amplitude doesn't decrease by the amount of external attenuation you add, then you may have gain compression or harmonic distortion. Add more external attenuation until switching in and out an increment of attenuation produces the same amount of change in the measured amplitude. For example, a 3-dB increase in external attenuation should produce a 3-dB reduction in the signal's amplitude. You must, though, keep the signal at least 6 dB above the instrument's noise floor.
If a change in external attenuation never matches the difference in measured amplitude and you run the signal into the instrument's noise floor, you must narrow the IF bandwidth. A smaller IF bandwidth lowers the instrument's noise floor because it lets less total power, including the receiver's own noise power, enter the measurement. The smaller IF bandwidth lets you add more external attenuation to reduce the overload.
Compensate for differencesWhen you're finally ready to make the measurement, be sure to add the amount of external attenuation you've inserted to the measured value. You need that attenuation value to get the true amplitude of the signal. If you add 12 dB of external attenuation, for example, then remember to add 12 dB to the amplitude you measure. If you make a broadband measurement at an IF bandwidth other than that specified by your test standard, you must apply a bandwidth correction factor to your measurement. The correction factor will depend on the type of broadband signal—coherent or incoherent—that you measure.
Coherent broadband emissions come from individual impulses that originate in repetitive signals such as those from switching power supplies, commutators, ignition systems, and fluorescent lamps. They typically have a fundamental frequency plus harmonics. When there are random neighboring spectral components, however, the broadband emission is incoherent. Incoherent signals often come from nonrepetitive sources such as DC powered gas lamps or high-voltage corona discharges. These sources produce emissions over a wide frequency range (perhaps into the gigahertz range) from just one pulse.
For coherent broadband signals, apply the following correction factor:
CF = 20 * Log (Reference IF Bandwidth / Actual IF Bandwidth).
For incoherent broadband signals, use:
CF = 10 * Log (Reference IF Bandwidth / Actual IF Bandwidth).
Because it's often difficult to discriminate between coherent and incoherent signals, it's best to use the larger correction factor. In both cases, use the reference IF bandwidths that the emissions standard dictates. Even emissions that fall just outside your IF bandwidth can cause incorrect measurements, because the IF filter may let through some energy from frequencies close to its edges. In general, EMI receivers use filters with sharper cutoffs than spectrum analyzers. The sharper filter's shape factor makes an EMI receiver more selective and less prone to overload than a spectrum analyzer.
Know those shapesA filter's shape factor shows how sharply an IF filter attenuates signals from a center frequency. The ratio of frequencies for the filter's 60 dB bandwidth to its specified bandwidth defines the filter's shape factor. A 10-kHz IF filter will attenuate frequencies by 3 dB at ±5 kHz from the filter's center frequency. A filter with a shape factor of 10 will attenuate signals ±50 kHz from the center frequency by 60 dB.
Assume you try to measure an emission with a 10-kHz IF filter with a shape factor of 10. Suppose, though, you have a signal 25 kHz from your IF filter's center frequency. Too much of the unwanted signal will likely get through the IF filter. The lack of attenuation may cause you to falsely believe your EUT will fail a compliance EMI scan.
Replace the IF filter with one that has a shape factor of two (typical of an EMI receiver), and you will sufficiently attenuate the unwanted signal. Figure 4 contrasts the difference in shape factors between a spectrum analyzer (wide plot) and an EMI receiver (narrow plot). Clearly, the EMI receiver uses a sharper filter.
You can use the simple tests we've described to eliminate many common measurement errors. While spectrum analyzers require more external components than EMI receivers do, you can still use them effectively in precompliance EMI scans. For making final measurements (particularly in a crowded RF environment), however, we prefer EMI receivers.
| Author Information |
| David Bare received his BSEE from the University of the Pacific and his MBA from Santa Clara University. He has over 25 years experience in EMC and electrical product safety. |
| Tom Parker received his BSEE degree from San Jose State University in 1970 and founded Elliott Laboratories in 1979. He served as Elliott's president until 1994, and he is currently chairman of the lab and editor of Elliott's compliance advisory service. |
| Reference |
|
