The EMC Continuum
By Martin Rowe, Senior Technical Editor- May 1, 2010
Rolla, MO—EMC is one of those engineering disciplines that require a great deal of practical insight, which is why many EMC engineers have spent decades in the field. But those engineers will eventually retire. A new crop of engineers well-versed in EMC will need to step in.
That's where EMC labs such as the one at the Missouri University of Science and Technology (Missouri S&T) and other universities play an important role. In this college town about 100 miles from St. Louis, the lab's faculty is training future EMC engineers to understand why EMC problems occur and how to mitigate them.
The Missouri S&T EMC lab has six faculty members and about 25 graduate students who work closely with industry. Some of the companies working with the EMC lab through its EMC Consortium have included Altera, IBM, Cisco, Apple, Sony, Samsung, Freescale Semiconductor, NVIDIA, and LG. Recently, the EMC lab has started the Center for Electromagnetic Compatibility (Ref. 1), an Industry & University Cooperative Research Program funded in part through the National Science Foundation (Ref. 2). Three other universities—Clemson University, the University of Houston, and the University of Oklahoma—are also part of the Center.
Graduate students and faculty at Missouri S&T, which until January 2008 was known as the University of Missouri—Rolla, work closely with industry through the lab's EMC Consortium. Companies pay $60,000 a year for membership, which subsidizes students and faculty who investigate EMC problems, often to a fundamental level. "Industry engineers often need to find fast, effective, and inexpensive solutions to specific EMC problems and don't have the time to fully explore the fundamental causes of the problem," said Professor Richard DuBroff.
The knowledge gained through research is often incorporated into EMC software tools. "Consortium members are interested in how we develop software tools to help them in designs," said DuBroff. "We develop diagnostic techniques and theories on how circuits interact. Companies in the Consortium come to us to get a better understanding of EMI [electromagnetic interference] and how to prevent the problem from recurring. We sometimes can't get into the details of a design, but we try to understand why an EMC fix might work."
Associate professor David Pommerenke emphasized that "We try to study the most difficult EMC problems such as intrasystem coupling. In cellphones, for example, we've seen RF signals interfere with audio."
Pommerenke advises students on numerous other projects, many of which revolve around ESD (electrostatic discharge) and radiated immunity. "We try to understand and make projections through models for circuit behavior," he said. "We look for coupling paths and try to learn where current flows."
Figure 1. To understand the effects of ESD, students constructed two boards connected by a shielded ribbon cable. The second flip-flop on the chip is not used.
Under Pommerenke's direction, MSEE candidate Argha Nandy studied how ESD can travel through a flat flexible cable and couple into a circuit. Nandy developed a test set (Figure 1) consisting of two PCBs (printed-circuit boards). One board lets him inject an ESD pulse whose current travels through a flexible circuit to a second board that contains a D flip-flop IC. Nandy has two versions of the test board. One contains a typical TTL flip-flop, while the other contains a high-speed (8-GHz) device.
To characterize the flex cable, Nandy injected pulses into the blank board (left side of Figure 1). Using an oscilloscope, he measured how the pulse propagated through the flexible circuit to the flip-flop, looking for pulses on the flip-flop's clock line that would cause the device to change state. He also developed a full-wave model of the circuit in both the time domain and frequency domain. Figure 2 compares the measured and predicted voltage at the flip-flop's clock input. The peak is enough to produce a change of state in the flip-flop.
Nandy and Pommerenke presented their findings at an EMC Consortium meeting in 2009. You can download a copy of their presentation, "Numerical prediction of ESD upset level."
Pommerenke teaches students to use a systematic approach to locating sources and resonances in circuits. "Suppose that noise in a laptop's memory couples into another part of the system. We want to understand the coupling paths. We run EMI scans on a product to find how electric and magnetic fields radiate in a circuit." Using mathematical processing of scan results, students find the coupling paths.
"We are developing software in Matlab to do the analysis," noted Pommerenke. "When we're satisfied with the models, we'll implement them in C++."
Software tools predict EMC behavior
Pommerenke's students aren't alone in developing software tools. Associate professor Daryl Beetner spends much of his time working with students to develop computer models of circuits that Consortium members use to predict EMC behavior. Beetner, who is a professor of computer engineering, works with a Consortium member company to model how ICs respond to EFT (electrical fast transient) events. His current goal: Develop a model of how energy couples into ICs so the company can design parts with greater immunity.
To develop the models, Beetner and PhD student Ji Zhang try to acquire as much information about an IC as possible. In some cases, IC manufacturers will provide equivalent circuits, but rarely will they provide complete schematics of an IC's protection circuits.
In addition to learning about ICs, students learn how energy gets into the devices. For example, current could enter an IC directly through an I/O line. Electromagnetic fields caused by interfering signals can also couple to I/O lines, producing unwanted current. Students run tests by firing an EFT generator into traces connected to the IC and tracking how current travels through a circuit.
Beetner and Zhang are currently measuring and modeling the I/O circuits in a microcontroller, and they plan to publish their results in a paper at the 2010 IEEE EMC Symposium in July (Ref. 3). Figure 3 shows the test setup they use to measure the microcontroller's V-I characteristics at any two pins. An Agilent Technologies function generator creates the test signals, and an Agilent oscilloscope measures the response. One oscilloscope probe measures the voltage across the IC pins; the other measures the voltage across the 47-Ω resistor to find the input current. "We know there's some kind of diode structure in the microcontroller," said Beetner. "We look for breakdown voltage, parasitic resistance, and threshold voltage to build our models."
From the measurements, Beetner and his students developed a model of the IC that's far simpler than the actual device because of the difficulty in modeling active components. They also developed a model for a test board. Using the models, they apply a simulated EFT signal to the board and then predict performance of a powered device. "We have a good SPICE model of the IC that we want to use to predict the IC's susceptibility to electromagnetic fields produced by an ESD gun," Beetner said. IC designers can now use that model as a design tool.
Assistant professor Jun Fan has also been involved in the development of software tools that model circuits for signal integrity and intrasystem interference. A former EMC laboratory student, Fan worked in industry after completing his PhD and before joining the EMC lab faculty. He and his students study signal integrity to look for degradation of desired signals that can lead to false triggering in a digital circuit. Very often, a degradation in signal integrity is associated with some parasitic electromagnetic effects, similar to those effects associated with EMI.
Fan has updated the university's course on signal integrity and has continued the development of existing courses in computational electromagnetics. The signal integrity course is available on videotape and online though Missouri S&T's continuing-education program.
Fan is currently investigating how to use acoustic mechanical vibration as a means of detecting EMI at frequencies much higher than audio. "We're looking for traces of interfering electrical sources in other parts of a system that manifest as localized vibrations. For example, we might modulate an RF system's digital circuits with a 2.4-GHz signal and look for traces in the RF section."
Finding solutions to EMC problems often requires an understanding of the materials used in electronic products. Research associate professor Marina Koledintseva studies how a material's composition and geometry affect its ability to absorb unwanted signals. From her research, she can develop mathematical models of a material's frequency behavior.
Koledintseva currently works with eight students on four projects. Working with equipment-maker John Deere, her students are investigating how to design ferrite chokes for use in automotive applications. Common-mode currents in cables can cause interference with circuits and systems, and Koledintseva's students look to mitigate common-mode currents with ferrite chokes.
Koledintseva is also working with Cisco to characterize PCB dielectric materials, research that is important for signal integrity and high-speed digital design. The roughness of a PCB trace's surface can lead to signal loss in transmission lines, and this loss will affect the quality of the extracted dielectric parameters. This lack of accuracy becomes a problem as frequencies exceed 1 GHz. Koledintseva said, "We developed a method called DERM [differential extrapolation-redistribution method] that lets us separate the loss of conductors and dielectric materials on PCBs that takes into account the roughness of conducting surfaces in our analysis."
Custom probes simplify calculations
Measurements that students make to find coupling paths often require custom probes. DuBroff and Pommerenke advise students in the design and construction of near-field probes.
"Near-field probes can give a good representation of a field without too much mathematical processing," said DuBroff, whose background is in antenna theory. Probes respond to the angle of incidence of a field to a probe and to the field's activity. The probe will average the field over its loop area.
Electric fields, however, can affect a probe pointing in any direction. They can couple on the metal part of the probe and induce currents. Thus, electromagnetic fields aren't completely decoupled from a circuit. To compensate for that, Dubroff and others developed a process that involves measuring a known source with a probe in different orientations and then trying to characterize the probe's response. From that response, students can use software to compensate for deficiencies in the probe.
There's a tradeoff between a probe's size and its sensitivity. Because a loop probe responds to the time-changing magnetic fields that pass through the loop, the smaller the probe, the finer the resolution. Finer resolution makes isolating an emissions source easier, but it also makes the probe less sensitive. Pommerenke is pushing the limits of probe manufacturing by working with students to develop a probe that will measure the current in a single ball of a BGA (ball-grid array) IC package.
Missouri S&T students and faculty also use field probes to induce voltages and currents into devices as they research device failures. They often mount the probes on a scanner from Amber Precision Instruments (which leased space in the same building as the EMC lab) and look for emissions or inject interference into a board under test (Figure 4). Giorgi Muchaidze, a senior engineer at Amber and a former student at Missouri S&T, contributed to the scanner's design and uses it to create maps of PCBs or ICs that show the sensitive areas. He uses a TLP (transmission line pulser) in the scanner to produce pulses with controlled rise times of around 200 ps. (The TLP produces pulses with more controlled shapes than an ESD simulator.) The electromagnetic field that a pulse generates can couple into PCB traces and produce current in the circuit. "We look for functional failures," he said. "We want to see how the current flows in the EUT [equipment under test]."
The scanner uses probes developed at Missouri S&T. Muchaidze demonstrated a magnetic-field probe that has a 5-mm × 5-mm loop. He uses the probe to inject magnetic fields generated by a TLP pulse into a circuit and runs scans in both horizontal and vertical orientations to find fields generated by the resulting current.
Other equipment includes a mechanical arm that can hold an ESD gun and move it to a device under test. Students use the arm because they can program it to achieve repeatable distances from the device. Students often use it to test handheld devices such as cellphones and MP3 players.
The lab also includes one semianechoic chamber, one shielded room, several network analyzers and spectrum analyzers, a direct-contact probe station for working with very fine-pitch components, and some specialized equipment assembled in the lab that measures ESD and near-field radiated emissions.
The laboratory has equipment for making swept-frequency measurements of two-port parameters, impedance and materials characterization, radiated emissions, and ESD. These facilities cover the range of 100 Hz to 20 GHz. A four-port network analyzer and test set also lets students and faculty make balanced differential S-parameter measurements. Time-domain measurement equipment includes high-speed digital oscilloscopes for measuring signals with subnanosecond rise times and for time-domain reflectometry characterization of signal paths.
Over the years, test-equipment manufacturers and EMC Consortium members have come through with donations and substantial discounts to equip the lab. For example, ETS-Lindgren donated a shielded room, and Advanced Micro Devices donated an ETS-Lindgren semianechoic chamber. Each chamber is large enough to permit a 1-m distance between the EUT and a transmitting or receiving antenna. "We don't use the chambers to make compliance measurements," said Tom Van Doren, professor emeritus. "We use them to make baseline measurements, then make comparisons after design changes to verify their effectiveness."
Both chambers had to be disassembled at the donor facilities and reassembled in Rolla. Research technician Kenny Allison noted that "disassembling a chamber is more difficult than assembling one." The semianechoic chamber, which had ferrite tiles glued onto its walls, wasn't built to be moved, so Allison had to cut through the tiles to reach the joints between the walls, then reinstall the tiles during reconstruction.
Professor James Drewniak, who has been a self-proclaimed "loud voice" for the lab, has taken on the role of fundraising and equipment acquisitions. He came to the University of Missouri—Rolla in 1991. Drewniak, along with Van Doren and Todd Hubing (now at Clemson University), was looking for a niche because "everyone was doing computational magnetics and we needed to do something else."
To get the lab equipped, Drewniak approached several companies. Bill Curran at Lindgren (now ETS-Lindgren) donated a shielded room worth $30,000. Mike Hart of EMCO (also now ETS-Lindgren) donated $20,000 worth of antennas. Tektronix sold oscilloscopes, spectrum analyzers, and time-domain reflectometers to the lab at cost. Hewlett-Packard and later Agilent Technologies have also provided oscilloscopes, signal sources, and network analyzers at cost. "The EMC community has been very good to us," said Drewniak with a smile.
Van Doren added "We can always use a few more 1-GHz oscilloscopes. They're fast enough for measuring noise in power supplies and for many other student projects. A lot of equipment has 20-MHz clocks, and we can use 1-GHz oscilloscopes to measure clock harmonics."
Besides bringing in funds, Drewniak brings in students from all over the world. "Missouri S&T students come to the lab when they hear about us as undergraduates. Each year, we put out a call for student projects."
Students from China learn about the EMC program through professors at their undergraduate universities. "We have colleagues at several universities in China who send us students," said Drewniak. Numerous other students come from India and Europe. The ability to work with commercial companies in the Center for Electromagnetic Compatibility is another reason that students travel to Rolla to study. Those company members often hire the lab's students after graduation.
The students usually take classes in the morning, then work in the lab during afternoons and evenings. Drewniak noted that much of the lab's funding is used to pay for student stipends and tuition.
Programs such as the one at Missouri S&T produce engineers who will solve EMC problems in ICs, boards, and systems. Those problems will only get more difficult to solve as clock speeds increase and circuits shrink. Missouri S&T EMC lab graduates will solve many of those problems, which will result in safer and more reliable electronic products.
(Editor's note: The original version of this article omitted Tom Van Doren's first name and title. )