PCB. Electromagnetic radiation due to high speed logic from different PCB layouts. (First Draft)

Transcription

1 EMC CONSULTING INC. P.O. Box 496, Merrickville, Ontario, K0G 1N0 Phone: (613) Fax: (613) Web Page: PCB. Electromagnetic radiation due to high speed logic from different PCB layouts. (First Draft) The data and information contained within this report was obtained from an independent R&D project funded by EMC Consulting Inc. The contents may be used and quoted but the source must be referenced in any publication D.A. Weston PCB2.rep.doc K. McDougall

2 Contents 1) Introduction ) Summary of the best and worst PCB configurations in the differential configuration. 5 3) Common mode current flow on the PCBs ) The PCB test set up ) The PCB layouts tested Basic PCB description PCB Dimensions and trace characteristic impedance as built Design goal for layouts Trace dimensions and impedances ) Summary of differential configuration results and conclusions at low frequency Trace Types ) Measured data and detailed PCB comparisons for the differential PCB configuration and Transmission Lines Microstrip Comparison Differential Trace Stripline Comparison Transmission Line Comparison Adding a Cable ) Summary of single sided signal PCB configurations at low frequency and high frequency ) Measured data and detailed PCB comparisons for single sided configurations Microstrip with and without attached cable Stripline comparisons at low frequency Single sided stripline radiation compared to microstrip ) Comparison of stripline PCBs with vias removed in the single sided configuration Predicted effectiveness of a localized stripline Measurements of striplines with vias removed in the three layer fully shielded configuration ) Driver configurations used with the ECL100 device ) Comparison of logic types

3 1) Introduction This report provides data on radiated emission measurements from fourteen different PCBs, with and without attached cables, and compares them to each other and to the FCC and EU radiated emission limits. These measurements were designed to be a continuation of measurements described in Reference 1. The measurements in 1 were made on PCBs with large traces and large distances between traces, with uncontrolled characteristic impedances, as well as with complex unmatched loads and with an upper frequency limit of 1GHz. The dimensions of the PCBs described in this report were chosen to be typical for four to twelve layer PCBs and were designed to have either 100Ω or 50Ω differential impedance and in some cases to have 50Ω characteristic impedance between one trace and the ground plane for single sided measurements. The fourteen PCBs include: ƒfour types of microstrip/ image plane layouts ƒeight types of stripline layouts ƒtwo types of transmission line layouts. The above PCBs were tested with single sided or differential drivers, and with and without a wire attached to the return for the transmission line, or in the case of the microstrip or striplines to the ground plane at the load end. The ground in the layouts with differential traces forms an image plane, although in this report the PCB is still referred to as a Microstrip. In the low-level radiation stripline layouts it was found that, although very short, the vias and traces connecting to the load and the load itself will be a significant source of radiation. Initially measurements were made with the original four layer PCBs connector shielded and the load and signal vias on the top of the board and the bottom of the board shielded as well as unshielded. The shield in these four layer boards was imperfect and the final tests (not yet performed) will be with three layer boards having bare copper ground on top and bottom. With the three layer board the copper shield can be soldered around its periphery. Due to the multiplicity of configurations a minimum of 47 tests were performed over the frequency range of 50MHz to 8GHz. All of these tests were initially performed using an ECLinPs differential driver as the source. After the 47 tests a number of the PCBs will be tested again (with and without attached cable) using different logic drivers. In some cases these drivers are single sided only. This preliminary report does not contain measurements on any but the ECLinPs driver as the source. All of the selected devices will have the capability of driving either 100Ω or 50Ω loads, and all of the PCB layouts are designed to have either 100Ω or 50Ω differential characteristic impedances or 50Ω single-sided characteristic impedances. The addition of a wire to the load in the transmission line PCB configuration, or to the underlying ground plane in microstrip or stripline PCBs simulates the connection of the signal return in a cable or the shield of a shielded pair cable and shows the effect of 3

4 common mode current flow on cables. The mechanism for this and further modeling and measurements are shown in Reference 2 and is summarized in section 3. After the initial tests, modifications were made to some of the PCBs, such as drilling out some of the vias down the length of the stripline and increasing the ground either side of one of the transmission line layouts. Photo 1.1 shows the enclosure, SMA connectors and PCB. One of the problems with drilling out vias on a stripline with an overall upper and lower ground plane is that holes are introduced into what was a full plane. This problem will be avoided with the three layer PCB which will be made with the distances between the vias varied from 3mm, 6mm and 12mm from board to board. Photo 1.1 Source enclosure, SMA connectors and a sample PCB. 4

5 2) Summary of the best and worst PCB configurations in the differential configuration The summary of the worst to best-case PCB layouts is for a real world configuration of the stripline configuration where the load resistor and upper signal pads are unshielded. This means that the results for the best stripline would be significantly better if the load had a very effective shield over it. This comparison is designed to help engineers and PCB layout personnel in making a decision on what type of PCB to use. The magnitude of the difference between PCB emissions is described in the following sections. All dimensions are in thousands of an inch (mils). Note that all PCBs are four layer. When the distance between the upper and lower ground planes of the stripline is 21mil then the upper ground plane is the second layer down and layer 1 (the top layer) is used to connect to the case of the connector and for the signal pads for the load resistor. 5

6 Differential TX2 J1/R Highest TX2 J3/R M2 J1/J S5 J3/J S3 J1/J S3 J3/J S2 J3/J M1 J3/J S4 J1/J Lowest Figure 2.1 Ranking of highest to lowest emitters among the PCB traces 6

7 3) Common mode current flow on the PCBs Many articles on theoretical PCB radiation discuss only the radiation due to differential mode signal current flow, specifically the signal current between the source and the load, returning on either a return trace of a signal ground. This type of analysis on a microstrip PCB often uses a quasi TEM mode, split into radiation from the horizontal current and the contribution from the vertical currents in the load. Transmission line equations or full wave computation based on the method of moments is often employed with the capability of analyzing 3D structures. Another approach is the Conjugate gradient FFT (CG-FFT) method. This type of analysis results in a 40dB/decade increase in radiation with frequency up to the frequency where the PCB length is a resonant length. Above resonance the increase is typically 20dB/decade. These results are very different than presented in this report, as well as in Reference 1, Reference 2, Reference 3 and Reference 4. The reason is that common mode (C/M) current sourced radiation is ignored in many analyses. In a practical PCB where signal traces are not in the exact center of the board common mode currents are generated due to lack of symmetry. The C/M current in a microstrip PCB is typically generated by displacement current which flows as the capacitance of the line is charged and discharged. Thus the displacement current is nonuniform with a maximum at the source end of the line and a zero at the load end. Unlike signal current which is uniform down the length of the board. Although not a typical or practical PCB layout, a symmetrical microstrip with the traces in the center of the board and the source generator buried behind the ground plane powered by a battery and with no attached cables, C/M currents are still generated due to the displacement current. If a single wire is attached to the ground plane of a microstrip PCB then the common mode current continues to flow on the wire, whereas the differential mode current returns in the ground plane. Another source of C/M current when a wire is attached to the microstrip ground plane is the voltage drop in the ground plane due to the returning signal current. If the Microstrip is close to a metal structure such as the enclosure in photo 1.1 then displacement current flows between the signal trace and the enclosure increasing C/M current flow even further. This is discussed and analyzed in Reference 4. The CG-FFT program does predict radiation from a microstrip with a high permittivity substrate, like a PCB, but assumes a symmetrical infinitely small source. The radiation as a result of C/M current is ignored and this results in a massive 60dB under-estimation of the fields from the structure!! It is clear how common mode current is generated from a microstrip but why does this happen with the stripline PCB. In practical stripline PCBs and in the four layer boards presented in this initial report, the signal vias come up to a load resistor on the top of the board. When the signal is differential i.e. two signal traces in the stripline then the signal and return are very close to being balanced around ground and the level of radiation is very 7

8 low, as is the common mode current. When the signal is single sided, one side of the load resistor and its trace is at the ground plane potential and the level of radiation from the load and signal trace is much higher, as is the C/M current. 4) The PCB test set up The frequency of the source was swept using voltage controlled oscillators controlled by a saw-tooth oscillator. The driver was contained in a small metal enclosure which also contained batteries, power supplies and the swept frequency source. The enclosure lid was sealed using finger stock material and also using dimpled adhesive copper EMI shielding tape. Before every test the PCB was replaced by two 50 Ohm terminations and the radiation from the enclosure and supply leads (ambient) was measured. In all the tests with four layer boards this ambient level was always lower than the PCB radiation. At the start of each new day of testing the copper tape was replaced to ensure shielding integrity. Each PCB was manufactured to contain two circuits (layouts) and have four SMA connectors mounted on it (2 per PCB layout). The signal from the source was brought out of the box using semi-rigid cable and two SMA connectors. The centre conductor of each coaxial connector is used for one signal line of a differential pair. Only one of the two connectors is used for testing single ended drivers and loads. Thus the signal source in its enclosure can be used with any of the different PCB layouts and is itself not a source of radiation. However, at high frequency a significant quantity of emissions was found to be leaking out of the SMA connectors on the board and thus, the connectors on the boards were shielded part way through the testing. Later it was found that the load resistor and pads on the top of the PCB and the signal pads on the rear of the PCB were also a source of radiation and these were then shielded. The level of shielding achieved with the four layer boards is limited, when the three layer boards are manufactured the shielding should be much more effective. The majority of the measurements were made on an Open Area Test site located in an area of relatively low ambient. The receiving antennas were located at a distance of 3m from the turntable and the site has met all the Normalized Site Attenuation requirements. At low frequency the emissions from some of the stripline PCBs were so low that measurements had to be made in an anechoic chamber with additional damping. This chamber is described in Reference 1. Photo 4.1 and Photo 4.2 show the enclosure and attached PCB. 8

9 Photo 4.1 Overall stripline PCB and enclosure with limited shielded connectors and loads 9

10 Photo 4.2 Localized stripline with localized upper ground plane PCB and enclosure For the differential measurements the PCB was broadside on to the measuring antenna and the elevation angle of the PCB was varied until a maximum level of emissions was measured. It was expected that varying the azimuth angle i.e. rotating the PCB relative to the antenna should not significantly change the emissions level for a PCB without attached wire. However the C/M and D/M current flow in the PCB varies in phase down the length of the PCB when it is electrically long and we found that at high frequencies it was important to rotate the turntable to achieve maximum emissions. As expected rotating the turntable did not change emissions at low frequency where the electrical length of the PCB and enclosure is short. For all of the single-sided measurements both the elevation angle and the azimuth angles of the PCB were adjusted for maximum emissions whereas with the differential measurements only the elevation angle was changed. For many of the stripline PCBs, especially with a single-sided signal, maximum radiation occurred with the PCB vertical and with the load facing the antenna. Probing the PCB showed that indeed the load radiated at a higher level than the edge of the stripline, even when ground plane vias where drilled out. 10

11 5) The PCB layouts tested 5.1 Basic PCB description Seven four layer PCBs were manufactured each containing two circuits per PCB. The basic layouts built on these PCBs were: a) Two different transmission lines on one PCB b) Four different microstrips on 2 PCBs c) One stripline with localized upper ground plane connected by vias to lower full ground plane d) Two stripline PCBs with full upper ground plane but with the vias connecting top and bottom ground planes close to the signal traces e) One stripline with full upper ground plane and with the vias connecting top and bottom located around the periphery of the PCB. Once trace 2cm from the edge and one 5.6mm from the edge (at edge). Photo 5.1 shows the signal and return traces of the transmission line on one side and the transmission line with its traces widened by conductive tape. Photo 5.2 shows the M1 J3/J4 microstrip with 100Ω differential traces and 50Ωsingle sided traces. M1 J1/J2 has different dimensions and a 50Ω differential impedance. Photo 5.3 shows the M2 J3/J4 microstrip with 100Ω differential traces and 50Ωsingle sided traces. M1 J1/J2 has different dimensions and a 100Ω differential impedance. Photo 5.4 shows the S2 J3/J4 stripline with 100Ω differential traces and 50Ωsingle sided traces. S2 J1/J2 has different dimensions and a 50Ω differential impedance. Photo 5.5 shows the S3 J3/J4 and S3 J1/J2 with 100Ω differential traces and 50Ωsingle sided traces. The only difference is that the traces in J3/J4 are 2cm from the edge of the PCB and J1/J2 are 5.6mm from the edge. Photo 5.6 shows the S4 J3/J4 with a 50Ω differential trace and S4 J1/J2 with a 100Ω differential trace. 11

12 Photo 5.7 shows the S5 J3/J4 with 100Ω differential and 50Ω single sided traces. S5 J1/J2 has a 50Ω differential trace. The dimensions for these PCBs are shown in Section All of the striplines have through hole plated vias tying the upper and lower ground planes together at 3mm centers. As it is uncommon in a practical PCB to make a perfectly localized stripline with the vias connecting the bottom and top ground planes at very close proximity (for example 3mm) other tests were performed on the striplines with differential traces with the ground plane vias selectively drilled out to increase the spacing or to create a slot and these are described for differential and single sided PCB configurations in section 7.2. Some measurements were made with the load resistor and upper signal pads for the resistor shielded as well as the SMA connector shielded. However the shielding was very limited and it was found that emissions from under the load shield were higher than the top or edge of the stripline, even with vias drilled out. For this reason another set of 3 layer stripline PCBs were manufactured with bare copper on the upper and lower surfaces. This enables the shields to be terminated as effectively as possible and comparative measurements of the three singlesided striplines will be made. Instead of the vias drilled out, these three layer striplines were made with vias at 3mm, 6, and 12mm spacing. These measurements have not yet been performed. However the measurements presented here with the load unshielded is perfectly valid and provides useful data as in a real PCB, traces and ICs will exist outside of the stripline. Thus the best microstrip may be almost as good as the worst stripline when the load is unshielded. 12

13 Photo 5.1 Transmission line PCB TX2 13

14 Photo 5.2 Microstrip PCB M1 14

15 Photo 5.3 Microstrip PCB M2 15

16 Photo 5.4 Stripline PCB S2 connectors and load with limited shielding 16

17 Photo 5.5 Stripline PCB S3 connectors and load with limited shielding 17

18 Photo 5.6 Stripline PCB S4. One side load and connector shielded the other side unmodified 18

19 Photo 5.7 Stripline PCB S5 One side load and connector with limited shielding At present additional three layer stripline PCBs are being produced. 5.2 PCB Dimensions and trace characteristic impedance as built Design goal for layouts The goal was to design PCB configurations with differential characteristic impedances of 50Ω and 100Ω and where possible also single sided (trace to ground plane) impedances of 50Ω. The dimensions were dominated by the impedance requirements but were also chosen to be manufacturable using the dimensions of typical multi-layer boards. Another factor in the choice of layout was to have as much variation in the different dimensions as possible to see the effect of changes in height, trace spacings and trace width. We have assumed board thickness of 10 mil for two layer boards and 31 to a maximum 125 mil for twelve layer boards. The minimum layer thickness of 5.8 mil is used, although boards with a smaller layer thickness is now possible. By changing the width of traces and the separation between traces the layer thickness may be adjusted to values other than used in the investigation. The typical type of PCB configurations are: 19

20 Microstrips Striplines Transmission lines Microstrips Microstrip PCB 1 has a height between the lower surface of the trace and the upper surface (closest to the trace) of the ground plane of 5.8 mil and so this might be contained in anything from two layer to 12 layer boards. Microstrip PCB 2 has 31 mil between the ground plane and the trace and so could be accommodated in a two layer board or a four layer board with 10.3 mil between layers. Using 6.2 mil as the layer thickness, the Microstrip 2 could be incorporated into anything from six to twelve layer boards. The earlier investigation described in Reference 1 shows that placing an image plane between the trace and ground plane of a microstrip degrades the performance considerably and is discouraged. This image plane in a typical PCB might be a VCC plane connected to the ground plane with decoupling capacitors. Thus the layers directly between the signal traces and the ground plane forming the microstrip must be kept clear of copper. In practical PCB layouts it has been found empirically that the inclusion of an image plane between the upper signal trace and ground plane (which carries the signal return current) in a microstrip invariably degrades the performance. For this reason the inclusion of an image plane imbedded in a microstrip was not repeated in the data described in this report. Although for consistency we refer to these PCB layouts as microstrips when measuring with the source connected across the two traces and the load between the two traces at the far end and with no connection of the signal to the ground plane, the configuration is really differential traces above an image plane Striplines Three of the four stripline PCBs tested (S2, S3, and S5) were a four layer design, so that the upper ground plane is the second layer down and the top layer is used for the connector and load resistor connections only. Although the 4 th (bottom) layer is ground plane the upper and lower surfaces of the PCBs are covered in solder mask and so no bare copper can be used for the shielding. The striplines were closed i.e. through hole plated vias are used down the length of the PCB to stitch together the upper and lower ground planes. In the case of the stripline in which the vias were selectively drilled out eventually all of the vias are removed and the stripline is what is sometimes referred to as an open stripline. Stripline 2 is manufactured using local striplines, with the vias forming the stripline close to the traces and with full upper and lower ground planes. This might be typical of boards for which two ground planes have been envisaged but in which local striplines are required. In a 20

21 practical PCB other traces and power planes contained on different layers must be kept away from the layers covered by the stripline. Obviously these layers are available outside of the local stripline. In stripline 2 the traces are located symmetrically between the two ground planes and the height between the ground planes is 21 mil. Using a 10.5 mil thickness the board could be 4 to 6 layer and with a layer thickness of 6.8 mil, the board could be eight to twelve layer. Stripline 3 has the traces placed symmetrically between the ground planes and the height between upper and lower ground planes is 21 mils. The upper ground plane is stitched around its periphery forming an overall stripline. Ideally this would be the upper and lower layers of the PCB and using a 5 mil thickness a five layer board could be manufactured, or a four layer using 6.2 mil thickness. If the upper and lower layers are just used for IC pads and any low level control or analog signal traces, then a six layer board is feasible. Adding a third ground plane, two striplines may be formed in an eight or ten layer board. Stripline 4 has local striplines (vias close to the signal traces) with full upper and lower ground planes. Here the traces are asymmetrical (often referred to as a dual stripline) and 12.3 mil from the upper ground plane and 41.5 mil from the lower ground plane. This means that with 6.2 mil layers that the board could be from 4 layers to as high as twelve using 6.2 mil thickness with the ground planes at layer two and eleven. Stripline 5 has the same height between upper and lower ground planes as stripline 2 (21mil) and the same stack up may be used. However the upper ground plane is local covering the stripline traces only and so this layer may be used for other traces outside of the upper ground area Transmission lines Transmission lines are simply traces side by side or one on top of the other and so may be located on a single layer, without ground planes and so they can be accommodated on a PCB with any number of layers. However as full ground planes are highly recommended, and this report shows why, the transmission lines are typically confined to two layer PCBs where ground planes are not feasible Trace dimensions and impedances The POLAR Instruments SI6000 Controlled Impedance Field Solver, which uses advanced field solving methods, was used to calculate PCB trace impedance for both the single ended and differential designs. Using the SI6000 C program the trace width and separation were varied until the correct impedance was found. Another approach was to input the height and the required impedance and see what values of trace width and spacing were required to achieve the impedance. The POLAR Instruments CITS500S Controlled Impedance Test System uses Time Domain Reflectometry to measure the impedance of the finished four layer PCBs. Test coupons were manufactured with each PCB and the impedance of these coupons was measured. The measured impedance was dependent on the position of the measurement point down the trace and so average values were taken. The test coupons showed variations in the test for 21

22 the 100Ω impedance of from 97.8 to 108.9Ω and the 50Ω impedance from 47.2 to 54.6Ω. The PCBs chosen for the test had the closest impedances to the specified. The cross section through each of the PCBs with dimensions and the characteristic impedances are shown in Figure 5.1 to Figure 5.7. Photo 5.1 to Photo 5.7 show the PCBs. All dimensions are in mils. The photos show the two SMA connectors which are used to connect the differential signals to the PCB. When only one trace is used in a single sided connection then only one of the SMA connectors are used only one trace is used and the load resistor is between this trace and the ground plane =GLII Ω 7 =GLII Ω = Ω Figure 5.1 Microstrip 1 dimensions and impedances 22

23 =GLII Ω =GLII Ω = Ω Figure 5.2 Microstrip 2 dimensions and impedances =GLII Ω =GLII Ω = Ω Figure 5.3 Stripline 2 dimensions and impedances 23

24 mm cm =GLII Ω = Ω =GLII Ω = Ω Figure 5.4 Stripline 3 dimensions and impedances =GLII Ω =GLII Ω Figure 5.5 Stripline 4 dimensions and impedances 24

25 =GLII Ω =GLII Ω = Ω Figure 5.6 Stripline 5 dimensions and impedances 7; RWDJURXQGSODQH 7.5 = Ω = Ω Figure 5.7 Transmission line 2 dimensions and impedances 25

26 6) Summary of differential configuration results and conclusions at low frequency 6.1 Trace Types As expected, the stripline traces produced the lowest emissions of the three trace types tested, and transmission line traces produced the highest emissions. On the OATS the best stripline under test produced emissions 60dB lower than one of the transmission lines at its worst case. In that test the stripline had very imperfect shield over it s load and connectors and therefore these components still contributed to the emissions and the effectiveness of the stripline layout at reducing emissions in a perfect world (fully shielded stripline) is underestimated. However in the real world traces will run on the upper layer outside of the stripline and ICs will be sources of radiation. Microstrip traces generally produced higher emissions than the striplines, but far less than the transmission lines. In some cases a microstrip PCB performed as well as a stripline with an unshielded load. See Figure 6.1 and Figure 6.3 for comparisons of the best trace from each type on the OATS and Figure 6.4 for the worst. Figure 6.2 and Figure 6.5 show best and worst traces based on 100Ω differential tests performed in the anechoic chamber. As shown later, the radiation from S2 J3/J4 is initially lower than S4 J1/J2 when measured on the Open Area Test Site (OATS). However it was found that the most significant source of radiation is from the 50Ω load on S4 J1/J2. When this was covered in an imperfect a copper shield the emissions dropped below those from S2 J3/J4. Traces M1 J1/J2, S2 J1/J2, S4 J3/J4 and S5 J1/J2 had 50Ω differential configuration capability, however they have yet to be tested. The results from 50Ω differential testing will be presented in the next report. 26

27 Figure 6.1 Comparison of the best transmission line, differential microstrip (image plane) and differential stripline traces over the mid frequency range for the 100Ω differential impedance tests on the OATS. N.B. The load on S4 J1/J2 is unshielded as in the real world. 27

28 Figure 6.2 Comparison of the best transmission line, differential microstrip (image plane) and differential stripline traces at low frequency for the 100Ω differential impedance tests in the anechoic chamber. The load on S2 J3/J4 was unshielded, as in the real world. 28

29 Figure 6.3 Comparison of the best transmission line, differential microstrip (image plane) and differential stripline above 1GHz for the 100Ω differential impedance tests on the OATS with emissions normalized to M2 J3/J4. The S4 J1/J2 PCB had very imperfect shields over both its load and connectors. 29

30 Figure 6.4 Comparison of the worst transmission line, differential microstrip (image plane) and differential stripline traces at low frequency for the 100Ω differential impedance tests on the OATS with emissions normalized to M2 J3/J4. Stripline load was unshielded, as in the real world. 30

31 Figure 6.5 Comparison of the worst transmission line, differential microstrip (image plane) and differential stripline traces at low frequency for the 100Ω differential impedance tests in the anechoic chamber. The stripline load was unshielded, as in the real world. Based on the low frequency data collected in this investigation, a stripline PCB has the best likelihood of meeting the most stringent radiated emission requirements. Of the striplines tested, the ones with a localized stripline and a full upper ground plane produced lower emissions than other combinations in the 100Ω differential tests, especially the localized stripline with localized upper ground plane. However the predominant source of radiated emissions for the best stripline is almost certainly the load resistor and signal pads and not the stripline. Where radiated emission requirements are less stringent, a microstrip PCB may be sufficient, given that in practical applications components will not be shielded and thus load radiation will dominate on a stripline layout. 31

32 7) Measured data and detailed PCB comparisons for the differential PCB configuration and Transmission Lines 7.1 Microstrip Comparison Figure 7.1 shows a comparison of the emissions from the three differential microstrip traces tested on the OATS with 100Ω differential impedance. We refer to this configuration as a differential microstrip but in reality it is differential traces above an image plane, as none of the signal current returns in the ground plane. Figure 7.2 gives the same comparison, but with traces M1 J3/J4 and M2 J1/J1 normalized to M2 J3/J4 and then smoothed out. Embedded figures on the plot show the width of the traces, their separation and their height above the ground plane. From both plots it is clear that M2 J1/J2 produces higher emissions than M1 J3/J4, by approximately 4 to 10dB. The differences between these traces are the width of the traces and their position. M1 J3/J4 has a slightly wider separation between its traces, which are thinner and both traces are closer to the ground plane than those of M2 J1/J2. Figure 7.1 and Figure 7.2 present the same data in different manners. Smoothing data is done by connecting the peaks and removing the dips from a given dataset. When smoothing is carried out on original data we end up with an overall trend of the highest emissions for each trace. Trace subtraction data looks much different than original data because dips in trace data become peaks when subtracted from another trace that is relatively flat. If we then smooth the resultant trace the separation between the original traces becomes emphasized. Refer to Figure 7.3 for a visual explanation. In comparing M1 J3/J4 to M2 J1/J2 we predict a reduction of 14.5dB due to height and an increase of 6.5dB due to an increase in the separation. Therefore the delta between the two layouts is expected to be 8dB and the measured is 4dB to 8dB. Unfortunately we cannot use these relationships to predict the level of reduction for all configurations. For example the measured reduction between M2 J3/J4 and M1 J3/J4 is 10 to 15dB whereas we would predict a reduction of 30dB due to height and separation. However we see that, at least from 220MHz to 310MHz, the higher the height above the ground plane and the larger the distance between traces the higher the level of emissions. Also M1 J3/J4, which has the lowest height, is uniformly lower than the other two microstrips at almost every frequency. In the older [as in those from reference 1] measurements on microstrips with single sided signals we saw a direct relationship between emissions and height and we see a measured 12-14dB difference between M1 and M2, as described in section 7, and a predicted 14dB difference due to height and this confirms our earlier measurements on single sided microstrips. The resonances in M2 J3/J4 seem to occur at different frequencies than the other microstrips, however if we were to compare the emissions at the corresponding resonances, M2 J3/J4 would fall between M1 J3/J4 and M2 J1/J2 in terms of emissions. The traces of M2 J3/J4 are as far from the ground plane as those of M2 J1/J2, but they are wider and more widely spaced to ensure the same characteristic impedance. 32

33 Measurements taken in the anechoic chamber suggest M2 J3/J4 produces the highest emissions. However, emissions from M2 J1/J2 were still found to be higher than those of M1 J3/J4, as was found on the OATS. See Figure 7.7 for a plot comparing the microstrip emissions measured in the anechoic chamber. Above 1GHz only M2 J3/J4 was used for the 100Ω differential impedance tests. At high frequency this trace will represent an average microstrip PCB. Figure 7.1 Direct comparison of the microstrip PCB trace emissions taken on the OATS at low frequency for the 100Ω differential tests. 33

34 Figure 7.2 Comparison of the differential microstrip PCB trace emissions, normalized to M2 J3/J4 emissions taken on the OATS at low frequency for the 100Ω differential tests. The relative plot on M1 J3/J4 in Figure 7.2 is much higher at 400MHz than a comparison between the PCBs shown in Figure 7.1. This is because the traces were smoothed at different points in the process and so in one case the separations are emphasized whereas in the other the proximities are emphasized. In both plots M1 J3/J4 is overall lower than the other two microstrips. 34

35 Figure 7.3 Smoothing original trace data compared to smoothing trace subtracted data. From the measurements we find that attaching a 1m cable to the ground plane of the differential trace with image plane changes the emissions. Theoretically, for a differential trace the current in the ground plane is due to the electromagnetic image in the ground plane. This current smears over a larger area than for the single sided microstrip, where all of the signal return current flows in the ground plane. In the symmetrical differential trace over an image plane configuration the total voltage drop in the ground plane is theoretically zero, and no common mode current flows. However the two traces are not in the center of the PCB (non uniform), also differential current results in radiation, especially from the unshielded load resistor, and this couples to an attached cable, and so the composite radiation from the differential traces and the cable is different than the same PCB without cables. When the cable and PCB are 1m long the frequency at which maximum radiation is predicted is by 1m = 0.5λ. This corresponds to a frequency of 150MHz. (N.B. When a wire is attached to a large conductive enclosure connected to a ground plane the maximum emission often occur when the cable is 0.25λ long). Figure 5.3 shows the radiation from the M1 J3/J4 PCB with an attached 1m long cable. The radiation from the PCB with attached wire is lower than without the wire at some frequencies which seems counter intuitive. However as described in the Reference 3, Radiated emissions, predicted and measured from a voltage source in a 0.15m section of wire (simulating a common mode voltage in a PCB ground) and the effect of an attached wire this is true as long as the wire is in the same plane as the measuring antenna. When the wire is mounted on a turntable 35

36 and rotated with reference to the measuring antenna then the level of emissions with and without the wire are virtually the same as explained in reference 3. This leveling off in radiated emissions above the resonant frequency is often seen in measurements of a product with attached cables tested on a turntable. It is contrary to the predictions based only on the differential mode current, which predicts an increase of 20dB/decade with increasing frequency after the resonant frequency. The emissions from the M1 J3/J4 PCB with and without the wire attached to the ground plane are shown in Figure 7.4 from MHz with the wire broadside on to the antenna. In these tests the turntable was not rotated. The length of the wire and PCB in wavelength is provided in Figure 7.5 and for the PCB alone in Figure 7.6. From 200MHz to 550MHz maximum emissions from the PCB, enclosure and wire occur when the total length equals 0.5λ, 1.5λ and 2.5λ and for the PCB alone when its length is equal to 0.5λ. 55 Microstrip PCB With Attached Cable Vs Without, OATS Connectors Unshielded E - Field (dbuv/m) Frequency (MHz) M1 J3/J4 No Cable (s=17, h=5.8, w=7) EN Class A FCC Class A M1 J3/J4 With Cable (s=17, h=5.8, w=7) EN Class B FCC Class B Figure 7.4 Emissions from the M1 J3/J4 microstrip with and without a 1m wire attached to the ground plane. Turntable was not rotated. 36

37 E-Fields From M1 J3/J4 and Attached Cable with Respect to Frequency and Total Wavelengths (data labels represent total wavelengths) 50 Total PCB and Cable Length in Wavelengths (λ) E-Field (db µv/m) MHz MHz 354MHz MHz MHz Frequency (MHz) Figure 7.5 Emissions from M1 J3/J4 with attached wire versus PCB, enclosure and wire length in wavelengths. 37

38 Microstrip PCB Without Cable, OATS Connectors Unshielded E - Field (dbuv/m) λ (400MHz) 0.45λ (500MHz) 0.47λ (525MHz) 0.5λ (555MHz) Frequency (MHz) M1 J3/J4 No Cable (s=17, h=5.8, w=7) EN Class B FCC Class B EN Class A FCC Class A Figure 7.6 Emissions from M1 J3/J4 PCB alone (no attached cable) with PCB length in wavelengths. A comparison between the three microstrips measured in the anechoic chamber with the antenna located at a distance of 1m is shown in Figure 7.7. The emissions from the same microstrip without attached cable and measured on the OATS with the antenna at a distance of 3m, is shown in Figure 7.8. We would expect the relative level of radiation to be a function of the height of the differential traces above the image in the ground plane as well as the separation between the traces. As the height increases the emissions increase and as the separation between the traces increases so do the emissions. The source driver is low amplitude LVDS and even with this low level signal all of the microstrip PCBs are above the Class B limit. Only M2 J1/J2 approaches and exceeds the Class A limit. M1 J3/J4 is the best with emissions above the Class B limit at only 710MHz and MHz. 38

39 Figure 7.7 Comparison of differential microstrip PCB trace emissions taken in the anechoic chamber at low frequency for the 100Ω differential tests. 39

40 Differential Microstrip PCBs Without Attached Cable, OATS Connectors Unshielded E - Field (dbuv/m) Frequency (MHz) M1 J3/J4 (s=17, h=5.8, w=7) M2 J3/J4 (s=170, h=31, w=51) Figure 7.8 Comparison between the microstrip emissions measured on the OATS. 7.2 Differential Trace Stripline Comparison Figure 7.9 shows a comparison of the emissions from each of the five stripline traces tested with 100Ω differential impedance at low frequency. In none of these measurements was the 50Ohm/100Ohm loads shielded. Below 200MHz none of the stripline traces radiated above the noise floor on the OATS. For this reason data below 200MHz id not presented. Although all stripline traces had very low emissions, S4 J1/J2 was the best of the striplines, especially from MHz and above 600MHz. The overall stripline with vias around the outside of the board was expected to be the best performer. In this configuration having the traces 2cm in from the edge of the PCB reduced emissions compared to 6.4mm from the edge, which is to be expected. However the difference was only about 3dB. In the differential mode configuration with both an unshielded load and with a poorly shielded load the localized striplines with overall upper ground plane were better. The emissions from PCB S2 J1/J2 with a height of 21thou was lower than the emissions from S4 J1/J2 which has a height of 55 thou. For S5 J3/J4, with a localized stripline and localized upper ground plane, from MHz has emissions were substantially higher than the other traces and this was seen in the early measurements, reported in Reference 1. In the single sided configuration the signal return current flows in the upper and lower ground planes and down the length of the vias in the side of the stripline and so the magnetic field shielding effectiveness of the stripline probably dominates. For the differential stripline the return current is confined to one of the traces and the magnetic and electric fields incident 40

41 on the inside surface of the stripline tend to cancel. The shielding effectiveness is likely to be a combination of E field and H field. In the original radiated emissions measurements on singe sided striplines with a poor shield the radiation from the load predominated, despite having a shield of limited effectiveness over the load. As expected when comparing the emissions on the same PCB for the differential trace configuration versus the single sided, the differential emissions are much lower. In other words the emissions from the unshielded load connected to a differential signal is much lower than when the load is connected single-sided. Figure 7.9 Comparison of stripline PCB trace emissions for the 100Ω differential tests at low frequency on the OATS. Loads unshielded, as per the real world. On the OATS S2 J3/J4 produced the next least emissions up to 400MHz. Above 400MHz S4 J1/J2 was overall lower except from 900MHz to 1GHz. The comparisons up to 400MHz are close to the measurements made in the anechoic chamber up to 400MHz. As shown in Figure 7.10, when tested in the anechoic chamber, S2 J3/J4 performed slightly better than S4 J1/J2. Both of these traces have local striplines and full upper ground planes. 41

42 Anechoic chamber measurements confirm that S5 J3/J4 produces the worst emission profile of all the striplines below 400MHz, as was also observed on the OATS. Like S2 and S4, S5 J3/J4 also has a local stripline, as shown in Figure 7.9, however its upper ground plane only extends to the edge of the localized vias. All of the other striplines have full upper ground planes, which is almost certainly the reason that S5 J3/J4 is not as effective at reducing emissions. Figure 7.10 Comparison of stripline PCB trace emissions for the 100Ω differential tests at low frequency in the anechoic chamber. Loads unshielded, as per the real world. 42

43 Figure 7.11 Comparison of S3 PCB trace emissions for the 100Ω differential tests at low frequency. In comparing the levels, the measurements on S3 J1/J2 and S3 J3/J4 in figure 5.11, it can be seen that at high frequency the layout with the traces 5.8mm from the edge of the PCB has higher emissions but only by 2 7dB. It was found that the sides and rear of the SMA connectors mounted on the board contributed to emissions from the stripline and so these were shielded for some of the differential signal measurements (as denoted by CS). To demonstrate that the load made a significant contribution to PCB trace emissions, a duplicate of PCB S4 J1/J2 was made and a copper shield soldered over its load and over the vias connected to the load on the opposite side of the board. The shield was found to be imperfect, however the 100Ω differential impedance low frequency measurements were taken from this modified board and compared to the data for S4 J1/J2. LS denoes Load Shielded. Figure 7.12 through Figure 7.14 confirm that indeed the load does contribute to the radiated emissions. Figure 5.14 shows S4 J1/J2 produced even lower emissions with both its load and connector shielded, no matter how poorly, than the other PCBs. 43

44 Emissions from Differential PCB S4 J1J2 With and Without Shielding over the Load, After One Trace Smoothing (C = Connector, L = Load, S = Shielded) S4 J1J2 S4 J1J2 LS E-Field (dbµv/m) Frequency (MHz) Figure 7.12 The effect of shielding (poor shield) the load on differential PCB trace S4 J1/J2 when measuring the radiated emissions on the OATS. This result has an impact on PCB board design and has been seen many times in practical, stripline PCB layouts. In many of these boards the emissions are predominantly from the ICs, connectors external (to the stripline) traces and oscillators, with emissions from the stripline at a very low level. This means that radiated emission limits may be exceeded even using the best stripline layout. As expected the radiation from traces outside of the stripline (unshielded) increases with increasing frequency as illustrated very well in Figure 7.14 for S4 J1/J2 versus S4 J1/J2 LS. 44

45 Figure 7.13 The effect of even a poor shield over the load on differential PCB trace S4 J1/J2 when measuring radiated emissions in the anechoic chamber. Taking into account the modified S4 J1/J2 traces, the best stripline turns out to be the S4 J1/J2 PCB with the poor load shield. S4 J1/J2 has a full upper ground plane with the ground planes tacked together either side of the traces (localized stripline). However S2 J3/J4 may exhibit lower levels of radiation with the load shielded. Figure 7.14 is a plot produced by subtracting the S4 J1/J2 LS CS 4 centre vias emission data from the other stripline trace emissions measured in the anechoic chamber. The plot is an excellent illustration that the striplines produce similar emission profiles, but with varied field levels. In summary the PCBs with a local stripline but full upper ground plane performed the best. The next best was the full upper ground plane with the vias at the edge of the board, (the Full stripline). The next best was the local stripline with local upper ground plane. Although the performance of the better striplines may be limited by the unshielded load and S4 may be better with a more effective shield. 45

46 Figure 7.14 Differential stripline emissions relative to S4 J1/J2 LS CS measurements taken in the anechoic chamber. Figure 7.14 seems to show that S4 J1/J2 with the load shielded but the connector not shielded has emissions lower than the same configuration with both CS and LS. However this is not the case as the CS LS S4 J1/J2 had 4 center vias drilled out and this is why emissions are higher. Above 1GHz three stripline PCBs were tested, S2 J3/J4, S4 J1/J2 LS with a 4 vias wide gap in the centre of the stripline and S5 J3/J4 with 50% of its vias drilled out. All connectors were shielded on these boards before testing began in order to further isolate the emissions coming from the traces. The lowest emission profile of the striplines from 1GHz to 8GHz continued to be the S4 J1/J2 trace with load and connector shielded, even with 4 of its center vias drilled out creating a slot. At 3.55GHz emissions from S4 J1/J2 with 4 centre vias drilled out, fell below the noise floor. One reason may be that the signal is degraded i.e. rise and fall time increased or amplitude reduced with the center vias drilled out. 46

47 Figure 7.15 Stripline trace emissions from 1 8GHz in the 100Ω differential impedance tests. With the exception of the resonance from 2.3 to 2.8GHz, the differential stripline PCBs continued to produce lower emissions than the differential microstrip (image plane) PCB from 1 to 4 GHz. Above 4GHz the striplines S2 J3/J4 and S5 J3/J4 with 50% of its vias removed have little superiority over the best differential microstrip PCB (which is in reality differential traces over an image plane). S4 J1/J2 LS with 4 of its centre vias drilled continues to produce the least emissions up to 4GHz but this PCB had both load and connector shielded. This effect and the results with all of the vias removed is covered to some extent in section 10) and will be the topic in a separate study. Figure 7.16 gives a comparison of the differential striplines to the microstrip trace above 1GHz. One conclusion that may be made from these results is that differential signals above a ground radiate at about the same level as differential signals in a stripline probably because the radiation from the load in the striplines predominates at these frequencies. 47

48 Figure 7.16 Differential stripline emissions versus the differential microstrip (image plane) emissions from 1 8GHz in the 100Ω differential impedance tests. Normalized to M2 J3/J Transmission Line Comparison Two transmission lines were tested with 100Ω characteristic impedances, TX2 J1/R1 and TX2 J3/R2. Figure 7.17 shows a comparison of the emissions from the transmission line traces at low frequency on the OATS and Figure 7.18 shows the emissions measured in the anechoic chamber. N.B. The abrupt change in emissions at 200MHz was merely due to a change in antenna. The trace with the highest emissions was TX2 J1/R1, which is made up of three thin, widely separated traces. TX2 J3/R2 only has two thicker traces spaced much closer together. 48

49 Figure 7.17 Transmission line emissions at low frequency for the 100Ω differential tests on the OATS. 49

50 Figure 7.18 Transmission line emissions at low frequency for the 100Ω differential tests in the anechoic chamber. It would have been possible simply to increase the distance between the two traces to see the importance of keeping the traces close together but this would have meant that keeping the characteristic impedance of the transmission line at 100Ω would have been difficult if not impossible. Instead a single central line was used for the signal with the return current flowing on two traces either side of the center trace. This meant that the trace separation could be increased from 17.3 thou to 70 thou with the impedance kept at 100Ω. Using the Numerical Electromagnetic Code (NEC) to model the two structures at 550MHz and 400MHz we saw that the currents were exactly 180 out of phase in the two traces of TX2 J3/R2 and 180 between the center trace and the two outer traces of TX2 J1/R1 exactly as predicted. If we calculate the radiation from the two transmission lines based on the currents and distance of each trace from the antenna, we see TX2 J1/R1 should have a 21dB higher level of emissions than TX2 J1/R1. The measurements in Figure 7.17 show a difference of 26.6dB at 550MHz and 18.7dB at 400MHz. The major difference between our analysis and the measurements is that the two real test configurations have a 7cm wide, 6cm high and 9.5cm long metal enclosure located 4cm from the end of the transmission line and 5mm with the transmission line 5mm from one edge of the enclosure. Due to this proximity and because the traces are offset from the center of the enclosure common mode currents will flow due to displacement current between the traces and enclosure. Reference 4 describes this in detail, as well as Reference 1. 50

51 Both transmission line PCBs were tested above 1GHz. Also tested was TX2 J1/R1 with its ground traces widened and moved in closer to the signal traces. From 1GHz to 4GHz TX2 J1/R1 again produced the highest emissions. TX2 J1R1 GMC, the trace with the ground widened and moved in, produced somewhat lower emissions at most frequencies than the unmodified board, but TX2 J3/R2 still performed the best up to 4 GHz. A secondary effect of moving the traces in is that the characteristic impedance of the structure is no longer 100Ω. The peak emission from TX2 J3/R2 at 2.92GHz, however was 4dB higher than the peak emission from J1/R1 at 2.67MHz. See Figure 7.16 for the 1GHz 4GHz transmission line comparison with the emissions normalized to trace M2 J3/J4 emissions. At 4.1GHz TX2 J3/R2 took over as the worst transmission line PCB trace and continued to produce the highest emissions of all the boards up to 8GHz. At 4.58GHz TX2 J3/R2 produced emissions 29dB higher than TX2 J1/R1 and 21dB higher than TX 2J1/R1 GMC. This demonstrates that, at least for this layout, above 4GHz moving the ground in closer to the signal, has no significant effect on the emissions Comparison of Radiated E-Fields from Transmission Line PCBs without Attached Cable (OATS) GMC = Ground Moved Closer E - Field (db µv/m) Frequency (GHz) TX2 J3/R2 Connector NOT shielded(s=17.3, w=7.5) TX2 J1/R1 GMC Connector NOT shielded (s=70, w=5) Noise Floor Data Figure 7.19 Transmission line PCB comparison 1 8GHz 51

52 7.4 Adding a Cable Adding a cable to the load end of a PCB trace had different effects on the emissions, depending on the trace type. When attached to differential microstrip (image plane) or differential stripline PCBs the cable had little effect on the emissions. This is expected for differential traces over an image plane or differential traces in a stripline. This is because theoretically the common mode C/M current flow in the ground plane/planes should be zero. However some C/M current will almost certainly flow in a differential traces over an image plane due to the proximity of the PCB to the metal enclosure and the capacitance between traces and enclosure. This results in a displacement current flow between the traces and enclosure. In a differential stripline without shielded load some C/M current may flow due to capacitance between the unshielded load traces and the enclosure. Also, as described in section 7.1 Microstrip Comparison when the cable is electrically long the radiation may be equal to or lower than the PCB without cable when the PCB is not rotated! See Figure 7.20 for with cable / without cable comparisons of microstrip emissions and S4 J1/J2 LS CS emission. For the microstrips emissions were higher with attached cable from 60MHz to 170MHz where the total cable, PCB and enclosure length were electrically short. When testing transmission lines, there was a clear increase in the radiation from the PCB when a cable was attached. The addition of a cable to a transmission line sets up a path along which the common mode current in the transmission line can flow, creating an electric field. In the case of microstrips and striplines, adding cable should theoretically have no effect on the emissions when the traces are configured differentially above a ground plane. This is because the current flowing in the traces sets up an image in the ground plane with symmetrical lines of current flow below the balanced s. Current flowing in the image should therefore cancel and produce a voltage drop of 0V thereby producing no common mode current flow. In contrast, transmission lines do not have a ground plane and so the cable is attached to the signal trace at the load end of the PCB. Due to the voltage drop down the length of these traces, common mode currents are set up and flow out on the cable, as the results of this experiment indicate. Figure 7.21 provides for a comparison of transmission line emissions with and without an attached cable and for TX2 J3/R2 emissions which are much higher with attached cable. It will be seen in section 9) that with a true microstrip, where the signal return current flows under the trace in the ground plane, emissions increase significantly with attached cable before the cable, PCB and enclosure length becomes electrically long. 52

53 Figure 7.20 Emissions with an attached cable versus without cable for a differential stripline and differential microstrip traces. 53

54 Figure 7.21 Emissions with an attached cable versus without cable for the transmission line traces. 54

55 8) Summary of single sided signal PCB configurations at low frequency and high frequency Microstrip and stripline radiated emissions compared to the limits for EN55022 and FCC Part 15 are shown in Figure 8.1 without an attached cable and Figure 8.2 with a cable. Figure 8.1 Single-sided microstrip and stripline PCB radiation compared to the EN55022 and FCC Part 15 limits. 55

56 Figure 8.2 Single-sided microstrip and stripline PCB radiation with attached cable compared to the EN55022 and FCC Part 15 limits. 56

57 Figure 8.3 Frequency range where stripline PCBs radiate more than a microstrip PCB in the singlesided configuration. 9) Measured data and detailed PCB comparisons for single sided configurations. 9.1 Microstrip with and without attached cable For the differential mode connection of the microstrip where the ground plane forms an image plane the voltage drop in the underlying ground plane is very low and the radiation with and without attached cable is virtually the same (Figure 7.20). However when the microstrip is driven single sided the signal current flows on a single signal trace and returns in the ground plane. A voltage drop is generated in the ground plane inductance and the resultant common mode current flows out on the attached cable, resulting in increased radiation. The maximum radiation occurs at the frequency where the combined cable, enclosure and PCB ground plane length is 0.5 wavelength long. Figure 9.1 shows the relative E fields from 50MHz to 210MHz for the M1 J3/J4 PCB and Figure 9.3 for the M2 J3/J4 PCB trace. The cable is 1m long and the PCB ground plane and the enclosure are 0.165m long. The measured E field from M2 J3/J4 is maximum at 128MHz, close to the predicted 127.5MHz, whereas the maximum from M1 J3/J4 is at 85MHz. The relative levels of radiation from 200MHz to 530MHz are shown in Figure 9.4 for M1 J3/J4. The measured levels with and without the cable are practically the same. The cable is 57

58 electrically long above 200MHz which means that the current change phase down the length of the cable. The composite field at some distance from the PCB is due to the fields generated by the incremental currents in the cable. In the initial measurements the PCB was broadside on to the antenna and the turntable was not rotated. The results for M2 J1/J2 were also taken with the table rotated and with and without cable. The PCB was also rotated on the table in the full radiation measurements. The full radiation measurement was the maximum radiation for all possible angles. In Reference 3 the E field from a source and a cable were measured and also predicted using the Numerical Electromagnetic Code (NEC) with the wire broadside on to the measuring antenna/ field analysis point. In this configuration the E field reduced with increasing frequency with the wire electrically long. However when the wire was rotated horizontally around its axis both the predicted and measured field from the electrically long wire increased slightly with increasing frequency. It is this slight increase, compared to a PCB with no attached wire, which we would expect if the PCB with attached wire had been rotated. In the plots rotating the PCB with reference to the antenna is designated Full radiation pattern and achieves the maximum radiation from all possible angles. The following plots are with connector poorly shielded in some cases, as noted. 58

59 Radiation on the OATS from Single-Sided Microstrip PCBs With and Without an Attached Cable (CU = Connector Unshielded) Single-sided M1 J3/J4 CU Single-sided M1 J3/J4 CU w ith Cable Relative E-Field (dbuv/m) Frequency (MHz) Figure 9.1 LF Radiation from M1 J3/J4 (connector unshielded) With and without cable attached. 59

60 Figure 9.2 LF Radiation from m1 J3/J4 (connector shielded on top) With and without cable attached. 60