Modeling of EM1 Emissions from Microstrip Structures with Imperfect Reference Planes Bruce Archambeault IBM Raleigh, N.C. Introduction The EM1 radiated emissions from most all commercial electronic products must be controlled to meet both US, European, and other regulatory requirements. Reducing product development and manufacturing costs are considered very important in order to remain competitive. Most engineers try to exercise some level of EM1 control at the printed circuit board (PCB) level when ever possible to help reduce the costs associated with shielding etc. Since most high speed signals have better signal integrity and better EM1 control with microstrip or stripline PCB structures, they are used extensively. Proper design using a microstrip or stripline is fairly straightforward, as long as some well published guidelines are followed. Unfortunately, other design considerations often force a deviation from these typical design guidelines. In these cases, engineers are often at a loss to predict the severity of the effects caused by these design guideline deviations. This paper briefly describes an effort to use numerical modeling tools to simulate a PCB with a microstrip and understand the effect of improper microstrip design. Measurement data is used to first validate the model, then various microstrip designs are implemented, and their effects noted. This effort was part of an effort to develop a set of internal microstrip design guidelines which went far beyond the normal guidelines. Initial Model The initial model was created using the Method of Moments (MOM). A wire frame ground reference plane was used with a single short microstrip. The ground reference plane mesh size was 5 mm near the microstrip, and 1 mm away for the microstrip. The microstrip itself was selected to be one inch long (centered on the reference plane), while the ground reference plane was 28 mm long by 12 mm wide. These dimensions were selected to allow comparisons to measured data by Dockey and German [ 11. The microstrip was driven at one end, and terminated in a 5 ohm load at the far end. The entire structure was placed flat.8 meters above a ground plane. Figure 1 shows the electric field levels at a distance of 3 meters over a ground plane for both the measured data and the results of the initial model. The receive location height was scanned from one to four meters, and rotated 36 degrees around the PC board. As can be seen in Figure 1, the modeled results agreed within 3-4 db over most of the frequency range. This agreement was considered good. Measured data also existed in the same reference for the same microstrip board structure, but with a 75 cm wire attached to either end of the board. This configuration was modeled, and the electric fields levels are shown in Figure 2. The modeled results agreed well except for predicting the resonant frequencies. The first resonant frequency for a 28 mm long structure with two 75 cm wires would normally be expected at about 85 MHz. The model predicts the first resonance at about 9 MHz, while the measured results show the first resonance at about 7 MHz. While the overall shape of O-783-414-6/97/$1. 456
Figure 1 Model and Measured Results E-Field Comparison for Initial Model 6 Initial Microstrip Model Comparison to Measured Results Horizontal Polarized Electric Field at 3 meters 5. g 4. 5 % 1 > 9 3. z ii.p z g 2 -Measured ------Model 1. 1.OE+OZ Frequency (MHz) l.oe+o3 Figure 2 Model and Measured Results E-Field Comparison for Model w/wires 6. Initial Microstrip Model with 75cm Wires Comparison Between Modeled and Measured Results Horizontal Polarized Electric Field at 3 meters -Measured /.,I..Mode,. Frequency 1 (MHz) 457
the modeled and measured responses agree, the measured results seem to consistently show a lower than expected resonant frequency. Since the main goal of this work is to compare the difference between different microstrip configurations, the difference in measured and modeled resonance frequencies are not important. So, again, the model was considered valid. RF Current along the PCB Reference Plane Edge One definite advantage of using modeling tools is there is extra information available that would be impossible to obtain through measurements. For example, since the MOM technique finds the currents every where on the structure being modeled in order to find the radiated fields, those currents are available for analysis. In this example, the current level in the ground reference plane along the edge of the PC board at 1 MHz is shown in Figure 3. The current distribution along the edge clearly resembles a dipole s current distribution. (Note the length of the PC board is shorter than % wavelength at 1 MHz.) Figure 3 Distribution of RF Current Along Reference Plane Edge RF Current Along Edge of Reference Plane for Short Microstrip @IO MHz 15 Position Along Edge (mm) RF Current through the PC Board on the Reference Plane The currents in the ground reference plane can be analyzed for all locations, of course. Figure 4 shows the current levels on the reference plane at 1 MHz for cuts through the PC board. Position #I is through the center of the PC board, and includes the ground reference plane directly under the microstrip. Position #2 is about one-third of the way towards the end of the PC Board, and does not include the area under the microstrip.. Position #3 is near the end of the PC Board. While it is expected that the currents under the microstrip would be highest, the currents 6 mm away at the edge of the PC Board are 458
only about 45 db less than the current directly under the microstrip. Since these It is also interesting to note the shape of the currents on wide buss structures can be very current away from the microstrip area. The significant (many amps of current all current was maximum in the center of the together), then this level along a PC Board board, but was also very high along the edge can also be significant. edges. Figure 4 RF Current Across Reference Plane at Different Positions 8 7 RF Current Across Ground Refernce Plane for Short Microstrip @IO MHz.. 6 -... 5 -. l *. a 4-. l s B,I l z 3. q Position #2 2 4 6 8 1 12 Position Across Board (mm) Voids in Ground Reference Planes Although running a microstrip over a void in a ground reference plane is completely against all good EM1 design practices, this practice occurs occasionally due to other design constraints that have an over-riding priority. When these reference plane voids occur, the currents can not return to their source directly under the microstrip, and must find other (longer) paths. Since it has already been shown that the currents along the edge of the reference plane are important and available from modeling tools, then such voids can be simulated easily by forcing a break in the wire mesh at the appropriate location, and analyzing the currents. Figure 5 shows the difference in the RF current along the edge of the ground reference plane at 1 MHz between the case of a normal reference plane and a reference plane with a split directly under the microstrip and extending an inch on either side. The amount of current increase was significant, and the shape of the current distribution changed from a single peak sine to a three peak distribution. This effectively places much more current in the corners of the board than might have been expected. Non-Centered Microstrips A series of models were used to compare the effect on the radiated fields when the microstrip was moved closer to the edge of the PC Board. Figure 6 shows the change in electric field levels for a case with two different microstrip positions. Position #l 459
Figure 5 RF Current Increase Along Reference Plane Edge @loo MHz FromVoid in Reference Plane 9 Increase in RF Current Along Reference Plane Edge Due to Void in Reference Plane @IO MHz a 7 6 is P E f!5 ; C ; I 4 t 4 3 2 1 15 2 25 3 Position Along Edge (mm) Figure 6 Increase in E-Field as Microstrip is Moved Closer to Board Edge Increase in Radiated Field Strength Due To Offset Microstrip Positions Horizontal Polarized Electric Field at 3 meters Frequency (MHz) was with the microstrip 35 mm from the edge board. Since the microstrip was only 2 mm of the board, and position #2 was with the wide, a separation of 15 mm from the edge microstrip 15 mm from the edge of the seemed to be suffkient. However, as Figure 46
5 shows, the radiated fields increased significantly with this close proximity position. Summary The use of modeling to simulate the performance of a microstrip printed circuit board structure has been shown to be accurate when compared to measured data. A number of different features, including location near an edge or void in the reference plane, were modeled and the results compared to the normal case. These types of analysis are much more convenient to perform than building a number of different PC boards and measuring all the different possible configurations. The test time costs to build a good series of cases would be extremely expensive, and the uncertainties associated with typical EM1 testing could confuse the issue. Once the general model is validated, the possible combinations of different cases are endless. Additional information present from the simulations (such as RF current distribution across a ground plane) provide the ability to better analyze the causes of the RF emissions. Once the causes are truly understood, then the appropriate design countermeasures can be implemented much more efficiently and accurately. References [l] New Techniques for Reducing Printed Circuit Board Common-Mode Radiation, Robert W. Dockey and Robert F. German, 1993 IEEE International EMC Symposium, pp 334-339.