Modern embedded system designs present challenges for EMI testing and troubleshooting that didn’t exist years ago. These challenges include switching power supplies, high speed system clocks and data buses, bursty information transfers, transmission line and termination issues, and spread spectrum clocking, as well as the integration of wireless interfaces and connectivity. Adding to the challenge is that most of these technologies can lead to issues that are transient, load dependent, and highly variable over time.
With budgets and time pressures greater than ever, it’s critical that designers working on embedded systems test for potential EMI issues and head them off as early in the design process as possible. Traditional tools often are not adequate to identify the source of EMI problems in today’s electronics. Fortunately, the combination of a practical approach to testing and new tools like the mixed domain oscilloscope (MDO) help eliminate guesswork. An MDO incorporates a wideband spectrum analyzer along with a traditional mixed signal oscilloscope, helping to make troubleshooting transient RF and EMI signals faster and more efficient.
Catch EMI problems early
EMI compliance is a fact of life for virtually any embedded system, with considerable variability across countries and industries. Specifications define levels for unwanted conducted and radiated emissions, as well as susceptibility/immunity standards for the device. Conducted emissions span 9 kHz to 30 MHz while radiated emissions range from 30 MHz to 6 GHz.
Typically, compliance measurements are complex and therefore expensive and are conducted at a test house using a stepped EMI receiver in an anechoic chamber. There’s no debating the value of only having to going to the test house once, saving time and expense. The way to ensure a single trip – or at least greatly improve your odds of success – is by carefully conducting pre-compliance measurements. Although you will still need to go to the test house, you can catch problems early on throughout the design process. What’s more, scanning doesn’t have to take a long time.
For pre-compliance scanning, most designers opt to use a general purpose spectrum analyzer rather than invest in a specialized EMI receiver. The key is to understand the differences between the them. Factors that should be considered include resolution bandwidth (RBW), the number of trace points, dwell time, support for CISPR detectors, and antenna factors. On the physical set up, biconical and log-periodic antenna are both good choices, along with a tripod and possibly a pre-amp. In the absence of an anechoic chamber, you can often find an RF-quiet location such as a boardroom or underground parking garage. Although it is difficult to completely duplicate EMI lab conditions, it is possible to make an accurate approximation by paying close attention to as much detail as possible.
A common tool for EMI pre-compliance testing is a swept-tune spectrum analyzer . This traditional architecture offers good dynamic range and good sensitivity, but is limited to two measurements: frequency vs. amplitude and amplitude vs. time. This is important because when you are performing a peak scan and find a frequency of interest, the swept-tune spectrum analyzer can be placed into zero span mode to look at power vs. time of the signal in order to determine periodicity.
Since this measurement is seen through the eyes of the RBW filter you can only go as wide as the RBW allows. As shown in the architecture of a swept-tune spectrum analyzer in Figure 1 , the video bandwidth filter is a post-detection low-pass filter that smoothes out the trace. For some EMI specifications, you’ll need to pay attention to frequency range and the number of trace points the instrument provides. Some low-end spectrum analyzers provide about 500 trace points – not enough for gigahertz of spectrum. When selecting a swept-tune spectrum analyzer for EMI testing, that number ideally should be in the thousands for better frequency resolution.
Figure 1: A swept-tune spectrum analyzer offers good sensitivity and dynamic range for EMI scanning, but low-end model aren’t able to handle gigahertz of spectrum.
As shown in Figure 2 , a real-time spectrum analyzer (RTSA) uses an ADC to create time domain samples, which are digitally down-converted to create I-Q, which then are passed along to a real-time engine. It’s real time because no samples are dropped during processing. An RTSA offers wide capture bandwidth since it’s not RBW limited, along with comparable sensitivity and dynamic range. From a speed perspective, RTSAs are significantly faster than swept-tune spectrum analyzers for narrow resolution bandwidth. Since the signal can be looked at from an I-Q perspective, you can do multi-domain analysis in a correlated manner, including frequency, amplitude, and phase vs. time. For instance, if you put a marker in one domain it automatically tracks in another. This isn’t helpful for EMI scanning, but it is helpful for EMI diagnostics.
Figure 2: A real-time spectrum analyzer offers high-speed spectrum measurements and correlated measurement domains, helpful for EMI diagnostics.
Another instrument for debugging potential EMI issues and EMI pre-compliance is a mixed domain oscilloscope (MDO). An MDO combines a dedicated spectrum analyzer, analog oscilloscope channels, and digital logic analyzer channels with time correlation across at all inputs. Unlike a swept-tune spectrum analyzer that sweeps from left to right and looks at the spectrum through the eyes of the RBW filter, the MDO instantaneously digitizes up to a full 3.75 GHz span. This in turn enables global triggers across all channels with common acquisition control. The ability to trigger on the RF, digital or analog inputs and maintain precise correlation with the other inputs is useful for EMI diagnostics.
Debugging EMI issues
While unwanted emissions can come from any number of places in an embedded system, the first place to look is the power supplies, especially those with switch mode operation, which often leads to ringing. Depending on the amplitude of the ring, this a ripe source of RF energy. From there, the next most likely source of RF energy is the clock and data. If you are using spread spectrum clocking, it’s important to know how well it’s working.
Resonances are another prime source of EMI issues. These could be emanating from the board itself, wiring geometries, cabling and faulty shielding, improper shielding, and mechanical connection issues. Any location where signals are either coming in or going out are potential problem areas.
The go-to tool for isolating sources of energy on a PCB is a near-field probe. An E-field or stub probe is useful for high voltage, low current sources and provides maximum sensitive when held perpendicular to the source. An H-field or loop probe is better suited to low-voltage, high-current sources and provides maximum sensitivity when held parallel to the source. In general, near field probes can’t be calibrated, but this isn’t a problem from a diagnostic perspective because you are more concerned about relative changes. For instance, if you have a spot frequency that you know is a problem and want to mitigate with shielding or a change in the design, you can measure before and after to find out how much you’ve affected the signal.
Multiple time-correlated domains speed troubleshooting
Amixed domain oscilloscope combines a dedicated spectrum analyzer, analogoscilloscope channels, and digital channels with time correlationacross all inputs. Anyone who has used a spectrum analyzer is familiarwith time domain and frequency domain views, but how are the two linkedtogether? The solution used in the MDO is spectrum time. The orange barin Figure 3 indicates when the spectrum on the display hashappened. Spectrum time allows the user to step spectrum time throughanalog time.
Atraditional spectrum analyzer takes one snapshot of the RF signal atone point in time. An MDO takes many snapshots, each time aligned withthe analog and digital signals. This enables a design engineer to seehow their RF signal is changing over time. It’s important to note thatthis is not just FFT on a scope, but an independent channel that has arecord length that is time correlated to the analog and the digital anddigital channels.
Example: Switch Class D Amp
In the set up shown in Figure 4 involving a switch class D amplifier, the MDO’s spectrum analyzer islooking at a center frequency of about 250 MHz with a span of 500 MHz.This capture was taken with an H-field probe directly over the switchingIC itself. There are two band cursors set up between 12 MHz and 108MHz. This is not a zero span like on a regular swept-tune spectrumanalyzer, which would be able to look only at power over time throughthe eyes of the RBW filter. The power over time trace on the MDO isactually looking at span power, or span power over time. Consistentspikes in amplitude vs. time are apparent. In this case, there is a lotof spectrum content between 80 and 108 MHz. This isn’t good because thisspectrum content correlates with the FM broadcast band.
In Figure 5 ,we are probing the control frequency on the switcher itself. Note thatwhen we do a single shot acquisition, the amplitude peaks directlycorrelate to the rise and the fall of that control location. This is ahuge tip off as to the source of unwanted emission.
Thisproblem was a challenge to diagnose because while a regular standalonespectrum analyzer showed a peak in the 80-100 MHz range it was unable toshow when the peak was occurring or how often it occurred. As such, itwas very difficult to correlate the peak to the switch controlfrequency. Using the MDO, we were able to isolate the problem in under15 minutes.
Example: USB Interference
In this example,channel 1 is on a USB line together with an H-field probe input intothe MDO’s spectrum analyzer. In Figure 6, the MDO is looking atamplitude vs. time as well as the RF spectrum at about 390 MHz. A directcorrelation between a USB message and a spectral peak on top of a broadnoise floor is apparent. This is where having 1 GHz of instantaneousbandwidth is helpful. When looking at the changes to the noise floor, tothe eye the spectral peak doesn’t actually look that big. But whenlooking at it from an amplitude vs. time perspective, it is easy to seewhere peaking occurs in that energy.
In this case, furtherinvestigation into this problem revealed that it wasn’t actually the USBline that was the issue. Instead the energy came from themicroprocessor when it generated a particular USB message. This would bea difficult problem to solve without the MDO’s ability to correlate ananalog signal with amplitude vs. time, and look at the spectrum overtime to determine the actual coincidence.
Advanced EMI diagnostics
Formore advanced EMI diagnostics, real-time spectrum analyzers withdigital phosphor or persistence displays are helpful tools, particularlyfor transient or bursted signals. An example is trying to figure outthe second harmonic content from a fundamental that we know is bursted.In the past, with gated sweeping it would be necessary to ensure thesweep was occurring at the time that the source was actually on.
In Figure 7 ,an RTSA is looking at an 8 GHz span of real-time data. This is possiblebecause it has the ability to step the real-time engine across thecomplete tuning range of the spectrum analyzer. With an RTSA, the usercan not only define the step size and the start frequency and stopfrequencies, but can specify a dwell time as well. The probability ofintercept is extremely high on a step-by-step basis which means that theprobability of intercept over the complete span is much greater thanwith a normal swept-tune spectrum analyzer.
Figure 7: A real-time spectrum analyzer provides a high probability of intercept of transient signals, even over an 8 GHz span.
EMIdiagnostics and troubleshooting of modern designs present uniquechallenges. Pre-compliance scanning will save both time and resources.To do this effectively, it’s important to pay careful attention to thetools you select and recognize their limitations. For EMI diagnostics,near-field scanning lets you quickly pinpoint trouble spots. Coincidenceis key to identifying sources of unwanted EMI emission. The ability toconduct analysis across multiple, time-correlated domains, includinganalog, digital, and RF, both speeds troubleshooting and reduces debugtime.
Faride Akretch is a technical marketing manager for Tektronix .In nearly 20 years in the industry he has held a variety of positions,including application engineer, product marketing, and business andmarket development, in Germany, Japan, and the United States. He holds amaster's degree in electrical engineering/electronics from theTechnical University in Berlin, Germany.