A common approach to dealing with EMI still seems to be ignore it as long as possible, and hope it won't be a problem. As you know, this approach rarely works, which becomes painfully evident in the eleventh hour of a new product development cycle.
There are numerous sources of EMI in switchmode power converters, but they all boil down to changing voltages and changing currents, which are unfortunately intrinsic to switchmode power conversion.
Many of the most troublesome EMI sources to locate and resolve are due to dynamic magnetic (or “H”) fields generated by rapidly changing currents. Once generated, these fields radiate EMI and induce noise-voltage spikes and transients in nearby unshielded circuits and magnetic devices. Noise appears to have no clear source in sight. A typical switchmode converter will have several H-field noise sources, making the problem seem insoluble. An H-field probe allows you to individually locate, identify, and remedy the EMI sources.
Unlike voltage or “E” fields, high-frequency H-fields are very difficult to shield. Most conventional magnetic shielding materials become more conductive than magnetic above a few tens of kHz, but they are not particularly good conductive materials. Conductive materials can shield magnetic fields, by generating eddy currents that “cancel” or block the H-field, but with a few special exceptions this requires nearly solid and completed conductive enclosures. Any small cracks, gaps, or holes in the path of the eddy currents allows the magnetic field to “leak out” to an astonishing degree, with disappointing results. Shielding magnetic fields to reduce EMI is often not an economic option in cost-sensitive applications.
The best solution is simply not to generate significant spurious high frequency magnetic fields in the first place. With sufficient experience a designer can learn to minimize most sources of EMI by design. Even then, a few can slip through.
H-Field Probe
During my formative years designing multi-kW converters, I found that a small H-field probe can be useful in “sniffing out” (isolating, locating, and identifying) magnetic field sources of EMI. It seems, the smaller the probe the more useful it becomes — despite a decreasing sensitivity.
The H-field probe is simple in principle, consisting of nothing more than a loop or coil of wire with a connecting cable for a scope or spectrum analyzer. Such a probe produces a voltage proportional to the rate of change of flux inside the coil, and could more accurately be called a dФ/dt probe; however, “H-field probe” is a more common term. The probe's “derivative” response is preferable to a “flat” response, as induced noise voltages are also proportional to the rate of H-field change. Thus, the probe waveforms are often a recognizable component of the switching noise transients seen when a scope replaces the spectrum analyzer connected to the line impedance stabilization network (LISN) or antenna used to measure EMI.
A larger pickup coil, around 1 cm (⅜-in.) in diameter or more, can tell you there is a noise problem here somewhere, but it will seldom be able to identify the conductor(s) carrying the offending current(s) — particularly when several nearby conductors also carry high dI/dt current. For most applications the sense coil should be no more than about 1.5 mm (1/16-in.) in diameter. A 10-turn coil is also a reasonable compromise between coil length, sensitivity, and bandwidth. The induced EMF increases with more turns, but inductance increases even faster, reducing output in the UHF band when resistively terminated. It's also desirable to use a Faraday shield for the H-field sense coil. This prevents dV/dt induced “capacitive” pickup, which only leads to confusion in interpreting the probe output.
A “Miniature EMI Sniffer™ Probe” has been in the industry for about five years, but it's not adequate for all tasks. Tiny, lower power converters in portable electronics need an even smaller probe size, while larger, higher power converters need a probe that can reach a further distance in between closely spaced components. In other cases it would have been desirable to reach around corners, or to probe fields near traces on a p. c. board closely mounted to a chassis or another board. Finally, it's sometimes necessary to have a probe with finer discrimination, or the ability to resolve individual fields when several H-field noise sources are close together. A range of H-field probes now meets these various needs (see Photo on page 24).
Construction
Fig. 1 shows the construction of Model E101, a general-purpose H-field “EMI Sniffer™ Probe,” with an axial 10-turn coil at the probe tip. The probe is Faraday-shielded and insulated with Kynar and high temperature epoxy.
Fig. 2 illustrates the probe response to a uniform H-field (when terminated in 50Ω). The voltage is essentially proportional to frequency of about 200 MHz to 300 MHz, a range that includes virtually all EMI produced by switchmode supplies. Although characterization has only been done to 500 MHz, the useful response should extend to a few GHz.
As shown in Fig. 3a, on page 30, the spatial response varies near an “isolated” conductor (such as a p. c. board trace) carrying high frequency current. With the probe held normal to the board and moved towards the trace, the voltage increases inversely with the distance to the conductor center, reaching a maximum near the conductor edge. The voltage then drops rapidly to zero over the conductor, inverts phase, and builds to a similar maximum again at the opposite edge, and then decays as the probe moves away from the conductor. The sharp null and phase inversion at the center of the conductor provides a high resolution of the path of the high frequency current.
When the current has a return path on an adjacent (or nearby) conductor, the response is somewhat different, as shown in Fig. 3b, on page 30. As before, the voltage builds upon approaching the conductor pair, reaching a first maximum near the first conductor edge. The first response null and phase inversion moves somewhat outside the center of the first trace, and reaches a stronger second maximum midway between the conductors. The response then has a second null past the center of the second conductor, and reaches a third maximum (similar to the first) at the outer edge before decaying with distance.
Using the Probe
Most switchmode converters generate noise largely below 30 MHz, so it's measured as conducted EMI with an LISN, with the conducted noise on each power line usually fed to a spectrum analyzer for measurement. This typical situation illustrates the use of H-field Sniffer™ Probes.
In switchmode circuits, nearly all noise is generated during (or near) a switching transition, when a power semiconductor turns on or off. However, the various potential noise sources don't occur simultaneously, but actually occur one after another with a short time delay. For example, there may be noise generated by the gate drive signal to a FET; then, the actual drain voltage falls followed by the reverse recovery of an output rectifier. Next, there's the forward recovery of the voltage clamp diode, or later reverse recovery of the same clamp diode, and so forth.
A time domain analysis is a powerful tool for locating EMI sources in SMPS, so an oscilloscope can replace the spectrum analyzer because analog scopes with a clean, smooth trace are best for diagnosing EMI. The quantization steps, trace “sparkle” and other characteristics of digital scopes make EMI harder to diagnose, particularly at low levels.
To investigate the EMI, you must feed the LISN output to one scope channel, with the probe output supplied to a second channel. In doing so, equalize propagation delays along the two signal paths using similar cable and wiring lengths along each path. Preferably, trigger the scope sweep by a third input connected to a switching waveform that just precedes the noise transient of interest, such as a transistor drive signal. With the gain on the probe channel initially set to maximum, “sniff around” the converter to find something happening in precise synchronism with the noise transient. The probe waveform will not be identical to the noise transient, but will usually have a strong resemblance to at least one component of the transient.
Next, move the probe closer to the suspected source while reducing scope gain. You can locate the conductor carrying the responsible current by its sharp response null on top of the conductor (even if in a buried layer or on the opposite side of a p. c. board), with inverted polarity on either side. Trace the noise current path as much as possible, and identify the current path on a circuit schematic. The source of the noise is usually evident from the current path and the timing information. (For more information on locating EMI sources in SMPS and typical fixes, see references [1], [2].
Probe Types and Applications
The H-field probes typically have an axial 10-turn coil of AWG No. 34 wire in the probe tip, with a coil diameter of 1.5 mm (0.060 in.). A female BNC connector mates to standard 50Ω coaxial cable, and the probe body is about 3.0 mm (0.12 in.) in diameter over shield and insulation. There is an internal 50Ω series-terminating resistor, but because of probe inductance it also requires a 50Ω load termination to eliminate noticeable transmission line reflections. Probe body length, or “reach,” is 2 in. (5 cm) nominal unless noted. Kynar tubing and high temperature epoxy insulate all probe bodies.
Model E201 is a “Micro Sniffer™ Probe,” similar in construction to the E101 but “necked down” to less than 1.7 mm (0.070 in.) in diameter at the tip, with a 0.075 mm (0.030 in.) diameter 10-turn coil of No. 40 wire. The smaller size and higher resolution of these probes allow the detection of EMI in very small converters used in portable and handheld equipment (cell phones, PDAs, etc.). The smaller coil's output is about 23% of that of the standard probe coils.
You can use the model E301 probe, made with semirigid coax, for “long reach” and more inaccessible applications. Standard nominal body lengths are 6 in., 9 in., and 12 in., but custom lengths of up to 48 in. can be produced. You can bend these probes to reach around corners and fit into tight and awkward spaces. Depending upon the tightness of the bends, the probe can be reshaped a number of times before failure.
The model E401 Right Angle Probe (Fig. 4) has a sense coil at right angles to the probe axis. Although a standard or long reach probe could reach between a p. c. board and a chassis (or two boards), the axial sense coil isn't suitably oriented to accurately locate and resolve EMI generating currents in the board traces, a problem resolved with this probe. The square probe body also has standard nominal lengths of 6 in., 9 in., and 12 in., and custom lengths of up to 24 in. are available on special order. A red reference mark embedded in the BNC connector indicates the coil orientation.
The model E501 high discrimination probe can resolve closely spaced H-field EMI sources. This probe has a 10-turn coil wound as a “figure 8,” effectively two “Siamese” or binocular coils wound in opposite directions, as shown in Fig. 5, on page 34. Individual coil diameters are 1.0 mm (0.040 in.), wound with No. 40 wire. This construction has minimal response to uniform ambient fields, as the voltages produced in the two “coils” are in opposition and effectively cancel. However, when placing the probe over a current carrying conductor, the fields in the two coils are in opposition, and the voltages now add. Thus, the probe has a maximal response when directly over a conductor, with a quickly diminishing response to either side, as shown in Fig. 6. Opposing red and black reference marks on the BNC connector indicate the coil orientation and sense polarity.
References
B. Carsten, “Sniffer™ probe locates sources of EMI,” EDN magazine, June 4, 1998, pp.155-164.
http://www.bcarsten.com for a condensed EMI Sniffer™ Probe application note. E-mail: [email protected].
For more information on this article, CIRCLE 332 on Reader Service Card