In this issue you will find two practical papers that should interest members of the EMC community. The first is entitled “When Correlation Lets You Down, Pre-Compliance Testing Made Usable” by Mathew Aschenberg, Derrell Gottwald, and Charles Grasso. This paper will provide assistance to the many of you who have struggled with using pre-compliance measurements to predict the results of compliance tests. The second, entitled “Reducing Emissions in the Buck Converter SMPS,” by Scott Mee and James Teune is a revised and expanded version of the paper they presented at the 2002 IEEE International Symposium on EMC in Minneapolis. It has been reprinted by permission. The paper is a very practical discussion of how to reduce emissions in these power supplies. I think you will find the way they relate before and after emission measurements to circuit measurements is an especially useful way to quantify the effect of design changes. The purpose of this section is to disseminate practical information to the EMC community. In some cases the material is entirely original. In others, the material is not new but has been made either more understandable or accessible to the community. In others, the material has been previously presented at a conference but has been deemed especially worthy of wider dissemination. Readers wishing to share such information with colleagues in the EMC community are encouraged to submit papers or application notes for this section of the Newsletter. See page 3 for my e-mail, FAX and real mail address. While all material will be reviewed prior to acceptance, the criteria are different from those of Transactions papers. Specifically, while it is not necessary that the paper be archival, it is necessary that the paper be useful and of interest to readers of the Newsletter. Comments from readers concerning these papers are welcome, either as a letter (or e-mail) to the Associate Editor or directly to the authors.

When Correlation Lets You Down, Pre-Compliance Testing Made Usable
Mathew Aschenberg, Derrell Gottwald, and Charles Grasso
Mat.Aschenberg@ieee.org, Derrell.Gottwald@EchoStar.com, Chasgrasso@ieee.org
EchoStar Technologies Corporation, Englewood, Colorado

This step in the process compares, point for point, the data taken at both sites. By examining this comparison we can generate two vectors. The first vector is an examination of the measurement variation between the two sites. The second vector is an evaluation of the tendency of one setup to measure higher or lower than the other.

1. Create Comparison Vector:

Equation 1
Where n is the number of samples, X_i is the i^th number sample, m is the estimate of the mean, and Q is the estimate of the standard deviation. The accuracy of this estimate is directly proportional to the number of measurements taken.

3. Calculate the measurement offset between the sites by calculating the mean of the Comparison Vector.
The mean is a theoretical value, which is best estimated by taking an average of the data. (Equation 2)

 

Equation 2
Where n is the number of samples, X_i is the i^th number sample,m_est and mest is the estimate of the mean. The accuracy of the estimate is directly proportional to the number of measurements taken.

The measurement variation and measurement offset calculations describe the past performance between the pre-compliance setup and the OATS. It is critical that these measurements are taken with care to accurately characterize the difference between the two sites. A problem occurs with this when maximized data is taken at one site and not another. It is important to compare apples to apples.

Also worth noting is that the 25 samples required above are just a starting point. The real power of this process is that it is self-reinforcing. With each unit tested, more data points can be generated to add to the Comparison Vector. After enough data is collected, the measurement variation and measurement offset calculations can be calculated as running quantities, giving frequency effects more visibility.

Predicting the Future:
Once the wort is completed, the prediction process is a simple matter of overlaying a normal distribution curve, defined by the standard deviation and mean estimated in the wort, onto the measured data of the unit under test.

1. Measuring the EUT.

m _corrected = m _emission + m_est

2. Calculate the probability of each individual emission passing the FCC limit.
The probability of each individual emission passing the regulatory limit is obtained by integrating the normal distribution up to the FCC limit using mcorrected as the mean and Qest as the standard deviation.

3. Calculate the probability that the EUT will pass the regulatory limit.

The total probability is calculated by taking the product of the individual probabilities.

The application of this process is much simpler when put into a spreadsheet.

This example uses the frequency dependant measurement offset and measurement variation, which can be done when enough comparison data is gathered.

The end result of the process is a number 0-100 that “grades” the unit as to its total likelihood of passing the FCC limit. In our experience, total probabilities 60 and lower aren’t worth taking to the OATS. While EUTs with a total probability above 60 have been consistently passing designs. Our comfort level has been around 60, but this level will inevitably change for different companies and different products

Conclusion:
While correlation is important for policing authoritative OATS sites, it is not necessary for pre-compliance measurements. Using some basic statistical calculations, on data already gathered during the standard design cycle, pre-compliance data becomes a very effective and valuable tool in determining a pass or a fail. The described approach enables the tester to examine pre-compliance data and accurately predict if the unit will pass or fail the upcoming compliance scan regardless of site to site correlation.

Biographies

Mathew Aschenberg earned his BSEE degree from Colorado State University, Fort Collins, Colorado in 1997. He is an Agency Engineer for EchoStar Technologies Corporation. He has been a member of the IEEE since 1998 and currently serves as Secretary and Vice-Chair Elect for the Rocky Mountain Chapter of the IEEE EMC Society.
Derrell Gottwald has a Bachelor’s degree in Mechanical Engineering from the University of Colorado. Since completing undergraduate studies, he’s taken a number of graduate classes in applied statistics at CU. Among his professors was Dr. Karen Kafadar, a well-known statistician who earned her doctorate under John Tukey at Princeton. Dr. Kafadar’s teaching philosophy, in part, stresses the importance of collaboration with experts outside the statistics realm. It is possible for a scientist to conduct an investigation without statistics, but not so for a statistician to contribute without understanding the discipline involved. Derrell thanks Matt and Charles for their time explaining the nuances of emission testing.
Charles Grasso earned his BSEE degree from Kingston Polytechnic, London, England in 1977. In 1977, he joined Burroughs Corporation as a fledgling engineer just as the EMC discipline was beginning to gather steam. His manager volunteered him as the assigned EMC Engineer and his life changed from then on. He has worked at StorageTek, Ansoft Corporation and is currently a Senior Compliance Engineer at EchoStar Technologies Corporation specializing in circuit/system design and verification, switching power supply noise and specifications as they pertain to EMC and Signal Integrity. He is a member of the IEEE, a member of the dB Society, Vice-Chair and Chair-elect of the Rocky Mountain Chapter of the IEEE EMC Society.

Errata Some of you may have noted that there were numerous errors in Jasper Goedbloed’s paper that appeared in the last issue of the Newsletter. Further, the figures were printed with unacceptable resolution. I would like to offer an apology for both of these. The manuscript supplied by Dr. Goedbloed was free of any errors and contained figures with fully acceptable resolution. I take full responsibility for the errors and the reduction of resolution that appeared in the final version that appeared the Newsletter. They were introduced in the editorial process and final checking was not adequate. We have decided to rectify this situation in two ways. First, the correct manuscript has been posted on the Web at https://www.ieee.org/organizations/pubs/newsletters/emcs/summer03/ Second, we want to call to the attention of our readers the three most significant errors that appeared in the paper. These are: 1) The equation on the line below Eq. (9), where 'Z-sub-2 =' should be 'Z-squared =', 2) The equation on the line below Eq. (20), where a superscript hyphen is printed, where a minus sign is needed between the two cross products, and 3) The correct version of Figure 6 and its caption, where two dashed line squares are clearly visible has been reprinted below: Fig. 6 The states (a) transmission and (b) reception. The dashed lines indicate parts that need not necessarily be accessible. Robert G. Olsen Technical Editor

Reducing Emissions in the Buck Converter SMPS
Scott W. Mee
Electronics Engineering, Johnson Controls, Inc. ASG.
Holland, Michigan USA
Email: scott.mee@jci.com
James E. Teune
Electronics Engineering, Johnson Controls, Inc. ASG.
Holland, Michigan USA
Email: james.e.teune@jci.com

Abstract
Switched Mode Power Supply demands are increasing, as the electronics industry requires more DC-DC conversion. In the past, linear regulators have been used to regulate power, but as the difference between supply voltages and desired output voltage increases, they become very inefficient. The BUCK power supply is efficient in converting higher voltages to lower voltages, but unfortunately in the process, both change in current (dI/dt) and change in voltage (dV/dt) are experienced. These changing parameters can cause excessive emissions in the RF spectrum, in conducted and radiated forms. We will examine the modes under which these emissions are allowed to propagate, and investigate techniques used to reduce them.

Introduction
One primary contributor to the low frequency emissions is the switching frequency of the converter, found typically in the 100’s of kHz range. Energy at the fundamental frequency along with several of its harmonics can find its way out onto the wire harness and radiate effectively. These emissions are derived from, among other things, the sudden changes in current flow (dI/dt) as a result of the regulator (SMPS IC in Figure 1) turning on and off during its periodic cycle.

Figure 1. Typical BUCK SMPS Circuit Diagram.
When the SMPS IC turns ON, the current flows through L1, SMPS IC, L2, and is delivered to the load (in parallel with C4). When the SMPS IC turns OFF, the current flowing through the SMPS IC stops. At this same moment, the energy stored in the inductor (L2) is released to the load, as the “free-wheeling” diode (CR1) begins conducting. It is this switching that creates current flow discontinuities at the input to the power supply. These current spikes in turn can drive the wire harness, attached to the product (Position 1 in Figure 1), like a transmitting antenna. Equation 1 can be used to calculate resonant frequencies (Hertz) of the cabling; substitute the length of the attached cable (meters) for l, and the speed of light (3x108 m/s) for C. Using 2, 4, and 20 times the length of the attached cable for l allows other resonant frequencies to be calculated. If any resonant frequencies of the cabling correspond with undesired RF frequencies coming from the Power Supply, the resonance can exaggerate the RF emissions problem.

l = C / f (1)

Equation 1. Wavelength.
Figure 2 shows discontinuities at the input to the power supply (measuring between VIN of the SMPS IC and ground) while the regulator is switching. Notice that the discontinuities at VIN correspond directly to sharp changes in the SMPS IC’s output voltage (measured between Figure 1 position 4, and ground).

Figure 2. Switching waveforms.
At the output of the regulator (position 4 in Figure 1), dur-ing switching states, two separate resonant RLC networks can be defined. These networks produce an under-damped response when “excited” by a step function (ie. switching!) and allow high frequency ringing to oscillate for several cycles. Broadband RF emissions commonly seen anywhere from 40 – 140 MHz are a direct result of this ringing. The R, L, and C components that make up the networks are defined by the path that the current flow takes during each switch state. RLC network #1 is formed when the SMPS IC output turns OFF. In this state the current flows from ground through CR1, L2, C4, and the LOAD. Each of these devices has impedance that is made up of R’s, L’s, and C’s (including the PCB layout traces and parasitics). When combined, these properties form the resonant network that gets “excited” by the step response of the switching (ON - OFF, or OFF-ON). RLC network #2 is formed when the SMPS IC output turns ON. In this state, the current flows from the OUT pin of the SMPS IC (Figure 1), through L2, C4, and the LOAD. Each of these two paths has a unique frequency response, and is tuned differently. An RC snubber circuit (R1 & C3) can be added in parallel to the output of the regulator to create a more “dampened” response in the two unique RLC networks. If values are chosen correctly, the snubber circuit will reduce the amplitude and number of cycles of the unwanted ringing. See waveforms in Figure 3 that illustrate this ringing. Note the frequency of the ringing is directly related to high frequency emissions seen during testing (see Figure 4 @ 60MHz).

Figure 3. Leading edge, ringing waveforms.
Ringing measured on the output of the regulator (position 4 in Figure 1) also affects the current flow through the output inductor, and causes a corresponding change in the magnetic field surrounding the device. This changing magnetic field can couple onto neighboring traces or planes allowing the RF energy to proliferate throughout the board.

Layout Considerations
One of the most important layout considerations for buck converters is to minimize parasitic capacitance and inductance at the output of the regulator. Parasitics contribute significantly to the ringing and other distortion on the output waveform. Do not place a ground plane on any layer of the board stack-up directly below inductor L2; this creates the parasitic capacitance that should be avoided. An option to consider would be to place a power bus (+) beneath inductor L2 if using a multi-layer board.

Loop Areas
Loop areas need to be controlled and minimized (physical area) in the layout to reduce overall emissions. A loop formed between the output pin of the SMPS IC and the ground on the load (connected across C4) contains large amplitude dI/dt waveforms and must be controlled to the fullest extent possible. Another loop is formed between position 1 and position 3 in Figure 1. Finally, a loop area formed between the input to the supply (position 1) (usually at the main connector) and the ground to the SMPC IC, contains small discontinuities (dV/dt and dI/dt). Loop areas are kept small through proper floor planning and routing of traces. Initially our output stage (CR1, L2, C4) was not properly designed and the loop area formed there was approximately 3 in2. Redesigning the output stage allowed us to reduce the area to 1.5 in2. This change brought improvements in the amount of cycles and amplitude of the unwanted ringing on the output (Figure 1 position 4). Although this was not the only change to the design, reducing this loop area had a direct correlation to reducing broadband emissions in the frequency range of 40MHz – 140MHz (Figures 4 & 5).

Feedback Trace
The feedback trace is used by the SMPS IC to sense the status of its output (connected between C4 and the feedback pin of the SMPS IC). Therefore, it is critical to route it away from any noisy circuitry, in particular, the output inductor L2. If using a multi-layer PCB, the feedback trace of the regulator can be embedded in a layer below the top layer (further down the board stack-up) and shielded by ground from the layers above and below. A ferrite may also be placed in series with this trace to reduce RF energy that can enter the SMPS IC and can adversely influence the output waveform.

Component Considerations
Whenever possible, use surface mount devices to minimize lead inductance. L2 should be a closed core inductor. This type of component keeps most of the magnetic field confined within the core during the switching of the regulator. Industry suppliers have various options to choose from. The best approach is to ask vendors for samples, and try each type during testing. Choosing the right device is a key factor in reducing emissions, since the field surrounding the inductor is constantly changing (i.e. ringing, switching current) and can couple to neighboring components and traces, etc.

Snubber Circuit
A snubber circuit can be used in parallel with the output of the SMPS IC (Figure 1 position 4 to ground) to reduce distortion and ringing on the output waveform. Typically this circuit is located between the “free-wheeling diode CR1,” and the output inductor L2. This circuit works like a high frequency shunt to allow RF energy to return to ground (i.e. capacitor to ground). One important factor to consider when choosing to use a snubber circuit is that you must sacrifice some efficiency in the power conversion. Some of the power will be dissipated in the snubber circuit itself. The series resistor is used to limit the amount of RF current taking this path to ground and the capacitor is used to “tune” the frequency. The switching waveform at the location of the snubber circuit is primarily a square wave, and Fourier theory states that it contains high frequency content that is dependent upon the rise time of the waveform. Consequently, without a series resistor in the snubber circuit, there will be significant power dissipation in the shunt capacitor. The snubber circuit can be effective at reducing broadband RF emissions typically seen between 40MHz and 140MHz. Typical values are R = 20 ohms, and C = .01uf. Also, low parasitic inductance components should be used to avoid forming resonant tank circuits. Therefore, avoid using wire-wound resistors, or leaded capacitors.

Diode
The diode (CR1) should be placed on the same side of the PCB where the other power supply circuitry is placed. This is done in order to minimize inductance (by avoiding the use of vias) while the diode is in “conduction” mode and current is flowing through the diode (CR1) and out to the LOAD (C4). This proved to be very beneficial for reducing the emissions near 70 MHz. Diodes that are available in the industry can have various switching characteristics such as: soft start, ultra slow start and fast start. These terms refer to how fast or slow the diode switches from the reverse diode block mode (when the SMPS IC output is ON), to the forward conducting mode (when the SMPS IC is OFF). Much of this parameter has to do with the forward voltage required to “setup” the p-n junction. Tradeoffs must be considered when selecting the freewheeling diode and the output inductor. The longer the inductor is in a non-steady state mode, the more heat it will be required to dissipate.

When selecting diodes from vendors, look for the symbol “Vf” that indicates the forward voltage required to turn the diode on. A smaller voltage rating means that the diode will turn on faster. Each diode has its own characteristic impedance that can affect the nature of the high frequency emission (40MHz – 140MHz). The process of selecting the right diode can be trial and error. This is due to the parasitic inductance and capacitance inherent and unique to each individual layout. If you have test equipment available, one of the best ways to approach this is to obtain samples from vendors and place each diode on the PCB while observing the RF emissions performance.

Front End Filtration
Clean input power is critical for “quiet” SMPS IC operation. In order to reduce the RF emissions within the low frequency range (i.e. switching frequency up to several Megahertz), front-end filtration must be carefully selected. Given that there are two types of discontinuities seen at the input to the SMPS IC (i.e. voltage and current), two types of filtering need to be addressed.

Series Inductance
The series inductor (L1) stores energy and releases it as necessary to reduce current spikes. Caution should be used when selecting this inductor. First, the device should be properly rated for steady state current flow. Second, a balance should be considered between having enough inductance to smooth larger current spikes, but also keeping a low series DC resistance. Large voltage drops can occur depending upon current draw, subsequently reducing the overall voltage available to other circuitry including the SMPS IC.

Bulk Capacitance
Bulk parallel capacitance is required at the input of the SMPS IC to remove voltage discontinuities, and should be placed as closely as possible to the input pins of the SMPS IC. However, if the bulk capacitor is too large, the charging period will be long and may cause large current spikes (i.e. higher emissions). If using Electrolytic capacitors, it is best to choose the lowest ESR (Equivalent Series Resistance) possible. This ensures that stored charge is delivered, with the lowest impedance, to the SMPS IC. In some cases it may be necessary to add a small ceramic capacitor in parallel with the input between the bulk capacitor and the SMPS IC. Ceramic capacitors have very low ESR and can a provide charge at faster rates to reduce unwanted swings in the voltage.

Common Mode Choke
Common mode chokes are effective input filters. From our experience, ferrite core CMC’s were effective at reducing emissions in the ringing frequency range (40 MHz – 140 MHz) but did not help at the switching frequency (260 kHz). Using an iron core transformer at the input did the opposite where it adequately reduced switching frequency noise but did not reduce the ringing frequency emissions.

Design Methodology
One of the best ways to approach a power supply design is to focus only on the power supply section initially. Re-move any circuitry from the PCB that is not related to the power supply. This allows the designer to implement a layout that is as close to ideal as possible, given the shape and size of the PCB. An artificial load can be made to draw the same expected current, as the load will draw on the board in the power supply’s final application. The artificial load can be connected during initial testing to obtain results. We recommend using a conducted emissions test as a way to benchmark and monitor improvements as the design is optimized. One primary advantage to this method is that efforts are focused on the power supply design. Not having the other application specific circuitry placed on the PCB makes it easier to manipulate the power supply design. It also requires less time to make changes and release another prototype revision. After the optimal power supply design has been determined, the remaining circuitry can be added. Changing as little of the power supply section as possible will maintain the improved EMC performance that has been achieved.

Radiated Emissions Comparison
Broadband RF emissions can be seen in Figure 4, in the frequency range between 40MHz and 140MHz. This broadband “hump” is attributable to the high frequency ringing on the output of the SMPS IC. Removing the parasitic capacitance and inductance found at the output, choosing the appropriate switching diode, and selecting proper snubber circuit values reduced these emissions. Figure 5 illustrates improvements in RF emissions gained by optimizing these design parameters.
Improvements were obtained by the following changes:

Figure 4. Radiated Emissions (25MHz-200MHz) – Before.

Figure 5. Radiated Emissions (25MHz-200MHz) – After.

Conducted Emissions Comparison
Figures 6 and 7 illustrate improvements made in the low frequency range (150kHz – 2MHz) for a Switched Mode Power Supply operating at 150kHz. In this case the design was using a bulk capacitor for input filtration, and most of the discontinuities in the voltage were removed. However, current spikes were still present, and a series inductor was added to remove current spikes (L1). A value of 33uH was chosen in order to smooth supply current to the SMPS IC during switching.

Figure 6. Conducted Emissions (150kHz-2MHz) – Before.

Figure 7. Conducted Emissions (150kHz-2MHz) – After.

Figures 8 and 9 also illustrate improvements made in the low frequency range (150kHz – 2MHz) for a Power Supply operating at 70kHz. The following changes were made to this power supply design to obtain the improvements:

Figure 8. Conducted Emissions (150kHz-2MHz) – Before.

Figure 9. Conducted Emissions (150kHz-2MHz) - After

Conclusion
Improving the EMC performance of a Switched Mode Power Supply design requires attention to details in the following key areas:

References
[1] Hnatek Eugene, Johnson Alan, “Designing Electromagnetic Compatibility Into DC/DC Converters and Switching Regulators,” IEEE EMC Symposium Paper, (1955-1995)
[2] Jung, Walt, McDonald, John, “Noise Reduction and Filtering for Switching Power Supplies,” Design Segment, www.sss-mag.com/pdf/ad9.pdf (2002).
[3] Keith Hardin, Greg McClure, Robert Menke, “Methods for Identifying Causes of EMI Emissions from Switched Mode Power Applications,” IEEE EMC Symposium Paper, (2001).
[4] National Semiconductor, “LM2676 SIMPLE SWITCHER‚ High Efficiency 3A Step-Down Voltage Regulator,” National Semiconductor, www.national.com, (1998).
[5] National Semiconductor, “LM267X 3A, 5A Evaluation Board,” Application Note 1135, www.national.com (1999).
[6] Paul, C.R., “Introduction to Electromagnetic Compatibility,” John Wiley & Sons, Inc., New York, NY (1992)
[7] Ridley Engineering, “Snubber Circuit Design,” Design Tips, www.ridleyengineering.com/tips.html (2001).
[8] SMPS Technology Knowledge Base, “Power Supply Question and Answers,” Web Based Forum, www.smpstech.com/qa/qa0005.htm (2000). EMC

 

Biographies

Scott Mee received the BSEE degree from Michigan Technological University in 1998 with specialties in RF Communications and Discrete Electronics. Since 1998 he has been working at Johnson Controls Inc. as a Sr. EMC Engineer. He has over four years of experience in the automotive EMC industry. His responsibilities include: EMC product compliance for telematics, multimedia, and other automotive electronics, product debugging and EMC failure analysis. He has published and presented technical papers at past IEEE EMC Symposia. He has also co-authored and published a paper in ITEM magazine titled: EMC Simulation for the Reduction of Emissions in Telematics Designs – 2002. His professional interests included EMC software simulation, product EMC failure analysis, PCB design for EMC, and EMC test methodology.

 

Jim Teune is a Lead Engineer for EMC Test and Design at Johnson Controls ASG. His responsibilities include directing the EMC design support activities for product development, technical operations of the EMC test laboratory (including an Open Area Test Site), and maintaining Johnson Controls’ AEMCLRP requirements for automotive laboratory accreditation. He has published and presented technical papers at past IEEE EMC symposia. He has 15 years experience working in the area of automotive EMC design techniques and test methodology. He previously worked with Intel Corporation. He holds a BSEET from DeVry Institute of Technology and is a NARTE Certified EMC Engineer. His professional interests include diagnosis of product EMC issues, and pursuit of methods to recreate field failures through combined test exposure.

 

If you would like to contact the IEEE Webmaster
© Copyright 2003, IEEE. Terms & Conditions. Privacy & Security

return to contents