Medical Electronics Manufacturing Fall 1998
Common-mode currents on printed wiring boards can be controlled through simple design concepts.
Thurman J. (Bill) Ritenour
Electromagnetic compatibility is a dual phenomenon, involving both immunity to and suppression of electromagnetic interference. Immunity has typically received more attention within the medical electronics community, but suppression is equally important. Emissions testing is a required step in CE marking, and the international standard EN 60601 covers both immunity and suppression. Emissions requirements are detailed in EN 55022.
Perhaps the most difficult aspect of emission control concerns common-mode (one-way) currents generated on printed wiring boards (PWBs). Once generated, these currents are difficult to eliminate. Adding to the problem are medical equipment statutory limits on leakage currents mitigation that prohibit decoupling capacitors and the increasingly common practice of using plastic enclosures. Accordingly, designers need to minimize these common-mode currents at their source—the PWB—and in order to do so, they must understand their origin. This article describes the mechanism responsible for generating common-mode currents and recommends a method for minimizing them.
By the early 1980s, it was common knowledge that many emission problems were caused by common-mode cable currents. A number of current-reduction fixes were developed, but the underlying problem—the generation mechanism—was not known.1 Moreover, until a product was built and tested, the retrofits needed to make it meet emission standards also were not known. Confounding matters even more, the common-mode currents were frequently not even related to the signal frequencies in the culprit cables—and when they did match, emission magnitude far exceeded the values that could be related to the signals.
It wasn't until the 1980s that ground-plane noise was identified as the source of common-mode current generation.2 The ability of ground-plane noise to drive cables also was initially illustrated in 1988 and by the mid-1990s, the model began appearing regularly in print. Most articles not only presented the model, but also supported it with measurements. A very useful one was Dockey.3
Figure 1. Time history of typical logic gate currents and resulting board bounce (Vgp).
The model is useful because it explains the mystery of common-mode current generation—but best of all, it provides methods to minimize it during PWB design and system mechanical layout prior to test. Design engineers can reduce common-mode currents during PWB design by minimizing signal trace height above the ground plane. Likewise, enclosure designers can reduce the remaining common-mode currents by judiciously picking chassis-to-PWB bonding locations. What's more, the model enables test engineers to predict the likelihood of excessive emissions based on measured PWB ground-plane differential voltages.
Board Bounce. Figure 1 shows a schematic representation of how ground-plane noise is generated. Voltage spikes (Vgp) or board bounce are caused by logic transitions and are induced differentially across PWB ground planes, thanks to the ground planes' slight (but significant) inductive nature. The phrase board bounce is useful in describing ground-plane inductive spikes because of its similarity to the widely recognized term ground bounce, which describes microcircuit ground-pin inductive spikes that occur during logic transitions.
Figure 2. PWB resonance model.
Figure 2 shows a PWB driven into resonance as a consequence of harmonically rich Vgp spikes.3 Note that the PWB as drawn can resonate at multiples of one-half wavelength (l/2, l, 3/2, etc.). Figure 3 also illustrates the consequences of PWB resonance—radiated emissions.3 Clearly, it is possible not only to force PWBs into harmonically related resonance at frequencies related to their length, but also to create emissions that equal or exceed regulatory limits. Figure 4 shows an empirical relationship between PWB width and length.3 Note the calculated value of loop emissions on both figures that, at board resonance, falls short of measured emission levels by between 2 and 26 dB, depending on PWB width.
Figure 3. PWB length-to-width effects on resonance.
Figure 4. PWB emission model.
Referring again to Figure 1, one sees a time history of power (ic), signal (is), and ground (ig) currents that will flow as a consequence of circuit operation.4 The left-hand gate's output voltage (V0) is typical, and is a square wave with 4-nanosecond rise and fall times, as shown. Note that the resulting signal current (is) is composed of positive and negative 4-nanosecond wide pulses approximately shaped like rounded square waves. This is a consequence of V0's rise and fall times (current through a capacitor is expressed by the derivative of its driving voltage—in this case, the positive and negative 4-nanosecond ramp voltages driving the single line plus gate input capacitance). When is returns to its source, small-voltage spikes (Vgp) occur directly under the signal trace.5 The spikes are caused by ground-plane inductance (Lgp) and the fast rise and fall times (tr/2) of is. On Figure 1, Vgp is calculated to be 6 mV (peak) using an Lgp of 0.4 nH.6 Figure 11 (of Reference 6) also shows that Lgp can be made to vary by more than 10:1 as trace height is dropped. This is the most important aspect of common-mode current generation reduction.
Based on the ubiquitous nature of ground-plane inductance, one would expect ground bounce to exist on all logic PWBs—and it does! It is not uncommon to measure ground-plane differential-voltage peaks that are in excess of 200 mV. Ott refers to 150–200-mV ground-plane differential noise (board bounce) as a good design goal.3
Ground-Plane Inductance. The idea that ground-plane inductance is significant enough to generate small-voltage spikes across PWBs will surprise many readers. But what is most surprising about ground-plane inductance is that the small-voltage spikes comprise the common-mode current generation mechanisms in and of themselves. One would reason, then, that if Lgp is reduced, emissions would be reduced, too. Dockey experimentally verified this deduction by dropping his signal trace height from 0.063 to 0.008 in., resulting in a measured emission reduction of 16 dB.3 Additionally, Holloway predicts a resultant net inductance drop of 4.8 to 1 or 13.6 dB for a matching change (see Figure 11).6 Since emission intensity is proportional to Vgp, Holloway's prediction matches Dockey's measured emissions reduction to within 2.4 dB!
Skeptics should bear in mind that although these ground planes are just thin sheets of copper (sometimes full of holes), they are metal and can be excited into oscillating at frequencies proportional to their significant dimensions. The ground plane in Figure 3 resonates at 375 MHz when excited by the fixed-amplitude 2-mA sine waves. In a similar fashion, the board-bounce spikes shown in Figures 1 and 2 contain harmonics that serve as sources to resonate the ground planes and, as will be discussed later, drive attached cables.
PWB resonance often occurs when a PWB's physical dimensions are much shorter than its clock's wavelength. As previously noted, Vgp, or board bounce, is rich in higher-frequency clock harmonics that can excite the PWB. But suppose that the PWB's dimensions are such that the harmonics skip over the PWB's resonant frequency? In practice, the likelihood of exciting PWBs at frequencies near their physical resonance is much higher than might be expected. In contrast to the simple model shown in Figure 2, reality is more complex. Many extra voltage spikes induced across PWB ground planes as a consequence of the use of either a single clock or several clocks. Also, other spikes will occur at clock subharmonics, depending on the design of the logic's divide-by-n's. This composite effect produces a dense spectrum, and frequencies are usually produced near one of the PWB's resonance frequencies (l/2, l, 3/2, etc.). An observant reader might comment that Figure 3's example board is resonant at 375 MHz, and if clock rates are kept well below 100 MHz and edge rates (logic rise time) curtailed, resonance problems should be minimal. Unfortunately, this observation is valid only if cables are not attached to the PWB—a rare event, because power, at the very least, will be delivered to the PWB via wires. And as Figure 4 demonstrates, wires become part of the resonating structure, causing resonance at much lower frequencies.
Figure 5. Effect of cable on resonance and resulting common mode current generated emissions.
Figure 6. Cable effects on PWB emissions.
Figure 7. Model of cable driven by board bounce voltage (Vg). Adapted from reference 7, equation 8.21 on page 417.
Attaching Cables to PWBs. Figures 5, 6, 7, 8, and 9 illustrate that when wires are attached to PWBs, the resonant frequency of the total structure drops, giving the attached cables a reasonably low input impedance (ZIN) (less than 1000 ).7 This allows reasonably large cable currents (5 to 50 µA) to flow into the cables—assuming that reasonably large values of board bounce occur (10 mV or more)—and possess spectra within a decade of resonance. The board bounce serves to drive the cables as if they were antennas, and they emit RF. Figure 5 shows the progression in detail. Figure 6 shows the measured results of adding a 0.72-m wire to the PWB in Figure 3 and demonstrates that a PWB/cable combination resonates at approximately the half-wavelength of their combined length.3 Note that in Dockey's article, wires are not connected to the signal source; board bounce alone provides excitation voltage sufficient to drive them as antennas.
Figure 8. Plot of input impedance of resonant wire versus resonant frequency harmonics and subharmonics. Extracted from Reference 8.
Figure 9. Effect of adding structure to PWB model.
Figure 5 combined with Figure 8 essentially illustrates the common-mode generation mechanism. The combination shows how common-mode currents drive system cables, which then radiate at frequencies not necessarily associated with resonance but with board bounce Vgp, spectra (Fi), and cable input impedance (ZIN), as shown in Figure 8. Common-mode current is simply the spectra voltage at Fi divided by the attached wire's impedance of ZIN. Figures 7 and 9 show more complex structures based on the same basic mechanism. Board bounce drives attached cables with common-mode current—and they radiate.
Figure 8 plots the absolute value of cable input impedance (ZIN) as a function of of cable resonance.8 This graph's simplicity is deceiving. The figure allows first-order estimates of ZIN in terms of the cable plus PWB resonance (Fr), which in turn enables an estimation of the cable input current; however, accurate estimates of cable input impedance (ZIN) are difficult because of the complexity of the PWB and cable combination.9 ZIN can vary greatly from the /2 resonance shape pictured. A comfortable approach is to treat the emission estimates based on Figure 8 as no closer than ±6 dB. It's a good idea to modify ZIN, too; assume that values greater than 2000 remain at 2000 and values less than 100 remain at 100 .
Figure 3 measures emissions produced by a four-member family of fixed-length PWBs of various widths. The emissions are the result of driving each of the PWBs with a fixed-amplitude 30–1000 MHz 2-mA sine wave current. The current was injected into identical 1-in.-long by 0.06-in.-high 50- Z0 loops centered in the middle of each PWB. All emission measurements were taken 0.8 m above the ground plane and parallel to it with the PWBs face up. The measuring antenna was 3 m away from the PWBs. During the measurements, each PWB was rotated 360° in eight equal increments. Measurements were taken at each increment while the antenna height was varied from 1 to 4 m. The maximum emissions at each increment were added to obtain the final emission plots. As can be seen from Figure 3, it is possible to force an 11-in.-long (28 cm) PWB into electrical resonance using low-level signals and a short 1-in.-long loop centered in the middle of the PWB. One can conclude, then, that the PWB electrical resonance is a function of PWB length—regardless of width and not the signal loop dimensions. Also, for high trace-to-ground-plane height, very low values of signal current will cause emissions in excess of regulations. Lastly, PWBs become better radiators as they grow narrower along the axis perpendicular to their axis of excitation.
Estimating Radiated-Field Intensity
Using the equation in Figure 5 and modeling the length of the cable plus PWB as 1 m, the amount of common-mode current necessary to cause emissions in excess of the EN 55022 Class B limits at 10 m is about 2 µA (see Table I). It is very likely that this could be provided by the 6-mV board bounce and 4-nanosecond rise/fall times on Figure 1. At 100 MHz, spectra amplitude of slightly less than 0.6 mV (20 dB down from 6 mV) is all that is necessary to provide 6.0 µA (recall that at resonance, ZIN is at a minimum). Considering that board bounce is usually 15–30 times larger than the 6-mV peak calculated,2 the likelihood of board bounce containing 0.6-mV spectra near 100 MHz is pretty high.
|1. E = (2 10-7 x F x L x l)/D where:|
E—absolute value of field strength in V/m maximized with respect to antenna polarity and cable aspect
F—frequency in Hz
L—length of cable in meters
l—common-mode current flowing along cable in amperes
D—distance from cable to antenna in meters
2. EN 55022 emission limits at 10 m are 30 dBµV/m from 30 MHz to 230 MHz and increase by 7 dB for frequencies above 230 MHz
Table I. EN 55022 Class B maximum allowable common mode cable currents for information technology equipment. The table shows the maximum common-mode current that, when it flows along the various lengths of cable shown and at the stated frequencies, will cause far-field emissions1 equal to EN 55022 Class B limits.2
A simple estimate of the emissions resulting from the 6-mV board bounce in Figure 1 acting on the PWB-plus-cable structure in Figure 5 illustrates the usefulness of the board-bounce model (note: in the example, the clock FO is assumed to be 5 MHz and its nearest harmonic to 112.5 MHz to be within ±11 MHz). The estimation exercise will reveal how close the estimated emissions are to EU emission limits (40 dBµV/m at 100 MHz and 3 m distance):
a. The first order approximation of the PWB plus cable's resonant frequency is
b. The first-order approximation of Vgp-generated spectra amplitude at 112.5 MHz with a clock FO (assumed to be operating at a subharmonic of 112.5 ±11 MHz) is:
1. VFi 6 mV maximum, where VFi is approximated by the expression VP times duty cycle4 (assumes a maximum variation of 10 to 1 in amplitude for multiples of FO near Fr)
2. VFi 0.6 mV minimum up to 6 mV maximum (per the above assumptions).
c. The calculated common-mode current (ICM) amplitude at 112.5 MHz caused by VFi (based on Figure 8) is
(Remember, input impedance to cables can vary by up to 50% or more from the /4 resonance shape pictured in Figure 8. Never use ZIN values greater than 1000 or less than 100 .)
d. The calculated far-field emission |E|µV/m that would result from the common-mode current estimate shown above and in the equation in Figure 5 is
|E|µV/m = (0.628 • ICM, µA • Fi, MHz • L, meters)/d, meters
= (0.628 • 6 (up to 60) • 112.5 • 1.0)/3.0
= 142 µV/m up to 1416 µV/m, or
= between 43 dBµV/m up to 63 dBµV/m, which is between 3 and 23 dB in excess of the EU limit.
This example clearly illustrates why so many products experience excess emissions. Millivolt low board bounce can cause attached cables to emit at levels well above the regulatory limits. Recalling that real-life board-bounce amplitudes are usually between 100 and 200 mV peak, the general problem is much worse than the example; however, if trace height were reduced from 0.06 to 0.005 in., this example would show an emission reduction of 25 dB (20 log [0.005/0.09] –21 dB), which would bring it back into compliance. For further emission reduction, PWB's and cable shields can be RF grounded.
Grounding Techniques to Reduce CM Current
Figures 10–12 show several applications of two concepts of reducing the common-mode currents in cables as they leave enclosures. Applying these concepts usually drops emissions by 10 to 20 dB. The first approach involves reducing the ability of PWBs to drive their attached cables by referencing the cable/PWB interface to the surrounding enclosure such that the interface becomes the counterpoise (base) of a quarter-wave (/4) antenna. The second approach entails providing a low-impedance path for common-mode current to return to its PWB source at the cable-enclosure interface, which also reduces common-mode cable current.
Figure 10. Minimizing common-mode current exiting a product by shunting it to chassis.
Figure 11. More minimization techniques for shunting ICM to chassis.
Figure 12. Moats are a baited trap. They are an alternative quiet zone.
Figure 13. Beware of chopped ground plane greater than unchopped plane. Detours in ground plane cause inductive hits on Lgp and high-frequency emissions.
Quarter-Wave Counterpoise (The Quiet Zone). Figure 10 illustrates the first technique, also known as the quiet zone method. In this setup, one end of the PWB is electrically bonded to the enclosure (which must be metal) using a low-impedance RF bond to establish a zero reference point, as sketched. The chassis reference point acts as an antenna counterpoise (chassis space capacitance is higher than cable space capacitance). As a consequence, the PWB resonates as a quarter-wave antenna referenced to the cable/PWB/chassis connection, as shown. Since the PWB-to-chassis RF potential is extremely low, no common-mode current can be generated. Often, electrical design constraints do not permit direct bonds between the PWB and chassis. The metal-to-metal bonds are replaced with series capacitors called RF shunts that tie the PWB ground planes to chassis above a few MHz. To be effective, the shunts must have low series inductance, be placed within l/4 in. of each side of each cable connector, and connect to the chassis through a bond with low RF impedance. Short and wide is good. Any version of steel—unless plated with zinc—will not be adequate because high-frequency skin effects create high surface resistance at frequencies greater than 10 to 20 MHz. This arrangement usually drops emissions below 300 MHz by more than 10 dB.
Low-Impedance Shunt. The concept of providing a low-impedance return path from the cable/PWB interface back to the PWB is illustrated in Figure 11. In Figure 11a, a low-impedance bypass capacitor acts as the shunt for all wires leaving the enclosure. The capacitor return path is through impedance Zshunt back to the chassis and must be lower than ZIN. Emission reduction is proportional to Zshunt/ZIN. Commercially available ribbon and dB connectors with built-in decouplers are most effective—but beware! Decoupling of high-frequency noise also shunts the high-frequency portion of the signals and disrupts system operation (most of today's interfaces operate above 10 MHz). Bypassed connectors are not recommended for high-frequency (>5 MHz) applications; however, signal suppression can be avoided by shunting only the common to the chassis if the cable is shielded. An example of this configuration is the coaxial LAN connection shown in Figure 11b. Since only the signal return is shunted, signal integrity is preserved. The arrangement works as follows: The signal wire is enclosed within the coax cable's shield, which reduces its radiation effectiveness.
The shield's common-mode current is shunted to the chassis, and the reference point forces the cable origin to be the base of the coaxial cable–derived antenna. The cable radiation is reduced by the amount the shunt reduces the common-mode current. As before, the shunt impedance to the chassis is critical. The filter arrangement in Figure 11b is conceptually correct; however, discrete capacitors usually resonate below 20 MHz, and as a consequence are not useful for higher-frequency noise suppression. But the decoupling style is very popular, and commercial connectors with multiple high-frequency capacitors integrated into the connector to provide wider-band noise suppression are available. Filtered connectors usually drop emissions in the filter shunt's passband by between 10 and 20 dB.
The Moat and the Chopped Ground Plane. A special kind of noise-suppression technique, the moat, is shown in Figure 12. It combines the quiet zone concept with deliberate enhancement of PWB ground-plane inductance. A cut or moat is made in the ground plane from the connector to keep all digital noise away. Only the signals that drive the connector use the integral ground-plane. The jury is out on the effectiveness of the moat; however, its use leads to an undesirable arrangement known as the chopped ground plane, shown in Figure 13. This unfortunate design configuration is responsible for innumerable systems emission problems. Using it guarantees excessive emissions. It often results from a misunderstanding of the moat application. At its worst, the technique can induce designers to completely remove the ground plane from around the connector to keep all the digital noise away. The fix is always the same—redesign the board. Nothing else works. Even when cut ground planes do not lose ground integrity, as in Figure 13, they always raise PWB inductance by orders of magnitude, and the resulting common-mode current cannot be contained.
RF interference is a common occurrence, arising in every category of electronic equipment imaginable, from personal computers to defibrillators to satellite receiver stations, to name a few. Reliable and inexpensive fixes to control emissions are clearly needed. The ground-plane reduction technique described above is inexpensive, effective, and easy to apply—but strangely enough, not widely used. Implementation of this design concept can drop system emissions by 10 to 25 dB. An additional emission reduction of more than 10 dB is possible if the PWB signal commons can be bonded to the RF enclosure at the point where cables connect to them.
1. Parker JW, "A Useful Method of Evaluating Cable Radiation," in Proceedings from the IEEE International Symposium on EMC, pp 371–374, San Antonio, TX, 1984.
2. Ott HW, Noise Reduction Techniques in Electronic Systems, 2d ed, New York, Wiley, pp 313–320, 1988.
3. Dockey RW, "Asymmetrical Mode Radiation from Multilayer Printed Circuit Boards," in Proceedings of the EMC/ESD International Symposium, Denver, Cardiff Publishing, 1992.
4. Mardiguian M, Controlling Radiated Emissions by Design, Van Nostrand Reinhold, 1992.
5. Ott HW, "PCB Design Techniques: Ground-Plane Inductance," presented at EMC '97: A Colloquium and Exhibition on Precompliance EMC Testing Problems and Solutions, Portland, OR, April 1997.
6. Holloway CL, and Kuester EF, "Net and Partial Inductance of a Microstrip Ground Plane," in IEEE Transactions on Electromagnetic Compatibility, vol. 36, IEEE, pp 33–35, 1996.
7. Paul CR, Introduction to Electromagnetic Compatibility, New York, Wiley, 1992.
8. Leferink FBJ, "Reduced Radiated Emission by Application of the SNT [Signal to Noise Transformation]," in Proceedings from the IEEE International Symposium on EMC, Chicago, IEEE, pp 468–473, 1994.
9. Hockanson, Drewniak, Hubing et al., "FDTD Modeling of Common-Mode Radiation from Cables," from IEEE Transactions on Electromagnetic Compatibility, vol 36, IEEE, pp 376–387, 1996.
Thurman J. (Bill) Ritenour is an EMC compliance consultant based in Boulder, CO.