GTEM Radiated Emissions Correlation Testing and FDTD Modeling

Transcription

1 GTEM Radiated Emissions Correlation Testing and FDTD Modeling Timothy E. Harrington, Zhong Chen, Michael D. Foegelle EMC Test Systems, L. P Kramer Lane Austin, TX, USA Abstract: A three-dimensional finite-difference time-domain (FDTD) model is used to simulate basic GTEM to OATS and GTEM to free space radiated emissions correlation algorithms. The GTEM models are terminated both with the typical resistor/foam absorber load, or alternately with a numerical absorbing boundary condition to simulate a perfectly-matched load. Radiation from dipole and loop antennas and a 19 dummy EUT is simulated in FDTD GTEM, OATS, and freespace models. Comparison is shown of simulated and measured data for these same antennas and EUT fed by a comb generator. Very good agreement is shown between the measured and computed results. Several considerations regarding correlation algorithm application in one-port TEM waveguides are discussed. INTRODUCTION Beginning in the late 1970 s, personnel at the US National Bureau of Standards performed extensive theoretical development and practical application of the simulation of OATS emissions testing based on voltage readings in twoport TEM cells [e.g. 11. These methods were later reduced by Wilson et. al. for use in one-port TEM cells, the most common of which is the GTEM (Gigahertz TEM) cell [2]. Since then, numerous variations on the basic Wilson threeposition algorithm have appeared [3-51, and many successful GTEM to OATS correlation results have been published. While much effort has and is being devoted to TEM waveguide immunity modeling and characterization, virtually no work has been done on TEM waveguide emissions test modeling. Several questions remain on the validity of using TEM waveguides for emissions testing. Although progress is being made on GTEM analytical descriptions, for example using Schelkunoff s generalized transmission line method [6], numerical modeling typically can be more versatile and useful for investigating emissions correlation algorithm issues. In this paper, a previouslyvalidated finite-difference time-domain (FDTD) GTEM model [7] is used to simulate the correlation of GTEM receive voltages to OATS and free-space electric fields. A goal of this work is to aid in predicting the response of various EUT classes in a GTEM, and to be able to later use the method to develop recommended EUT size limits and cable layouts. Although this study considers only the GTEM, many of the observed emissions test deviations are also expected to occur in other types of nonuniform TEM waveguides [S-lo]. Most of the GTEM to OATS measurement results published to date consider only discrete frequency data. Usually no rigorous explanation has been postulated for any frequencyby-frequency anomalies and discrepancies between OATSmeasured and GTEM-simulated data. Better insights into the mechanics of correlation can be gained by using sweptfrequency radiators or dummy EUTs, such as dipole antennas or equipment cases excited by a battery-powered broadband comb generator source. In addition, as GTEM use above 1 GHz is considered [l 1, 121, it is mandatory to fist have a better understanding of most aspects of correlation physics below 1 GHz. The present study focuses on modeling and testing dipole and loop radiators and a 19 metal box dummy EUT without cables from 30 to 500 and 1000 MHz. Frequently there are questions about the efficiency of the GTEM termination or load, and the influence of any load imperfections on correlation discrepancies. To examine the influence of termination quality, FDTD-modeled correlation results are shown for a GTEM terminated with the standard resistor board and pyramidal foam absorbers, compared to a GTEM model terminated with the Liao numerical absorbing boundary condition (ABC). The absorbing boundary simulates the behavior of an ideal, reflection-free GTEM [7]. This gives some idea of what the best achievable GTEM is for any load modifications. The results below show that at higher frequencies, many deviations are independent of the load, which means the deviations may be due to unavoidable non-uniform waveguide effects of the flared geometry [ 131. CANONICALEUTDESCRIPTIONS To lay the groundwork for later study of the testing of actual complex EUTs, it is best to begin by gaining a thorough understanding of the behavior of canonical dipole and loop radiators in GTEM cells. Measured and simulated data using a common comb generator-type source (RefRad = Reference Radiator [14]) feeding a 30 cm square loop and a 40 cm dipole antenna is considered, with a 5 MHz spacing radiated 770

2 spectrum. The original reference radiator concept [ 151 was to use short and long dipoles and a 30 cm square loop for site evaluation. Above about 30 MHz, the loop was found in this study to behave more like an electric dipole source with two strong dipole moments. While not really a pure magnetic source, this loop may still simulate radiation from some types of EUTs. A 48 cm x 40 cm x 16 cm rack-size dummy EUT was proposed in [16] for site assessment, and it has been used in the CENELEC/EU emissions testing in fully-anechoic rooms (FAR) project [17, 181. The dimensions of this EUT are such that it can be numerically modeled with a grid size of 4 cm or less. Figure 1 shows a schematic of the dummy EUT and dipole and loop. The comb generator is installed inside the box. As used here, the dummy EUT has two operational modes. In the fat dipole radiating mode, the lower box is completely closed, with an electrically-isolated pin coming through the top center and connected to the center of the top plate spaced 4 cm above. In the slot mode, a plate with a 12 cm x 4 cm slot is fastened to the side of the box and the offset top plate is removed. The outer conductor of a coaxial cable is connected to the bottom of the slot, with the exposed center conductor traversing and connected to the top of the slot to form a voltage source across the gap. Figure (48 cm) rack-size dummy EUT with separate slot or offset-top-plate radiating modes, and comb generator with dipole and loop antennas. FDTDMODELINGCONSIDERATIONS A general description of the GTEM FDTD model and its use for field uniformity behavior has appeared previously [7], and most of the details will not be repeated here. Modeled results for GTEMs 750 (4 m length) and 1750 (7.7 m length) are shown below. A staircase grid approximation is used for the GTEM flared walls. The septum is aligned with one of the mesh planes to minimize current flow perturbation. Resistors are included in the model by assigning appropriate cell-edge conductivity values to give an aggregate 50 Q total impedance. Pyramidal absorber shapes are also gridded up by staircasing, and material-parameter averaging is used at the surface cells to minimize air/dielectric mismatch. In the GTEM 750 model the pyramidal absorbers are placed parallel to the backwall, while in the GTEM 1750 model the absorbers are spaced off from the backwall and follow a spherical placement scheme. The code includes a dispersive material subroutine to account for a simple frequencydependence of the absorber material parameters. The dispersive FDTD method was tried only with the dipole in the GTEM 1750 model, and was found to give only a slight reduction in the low-frequency ripple. Thus all results included here used constant absorber permittivity and conductivity. Alternatively, the resistor/absorber load in the model can be replaced by a Liao numerical absorbing boundary condition, to simulate the performance of a reflection-free GTEM. The Liao ABC GTEM is a mathematical ideal. Such a GTEM has yet to be realized in practice. The same boundary condition is used for the OATS model, with a one-cell thick ground plane located 5 cells from the lower boundary, and a minimum 1 m separation from the receive point and EUT wall to the boundary. The OATS model calculates only the field at the observation point, independent of any receive antenna. While useful results are possible with some FDTD geometries at a resolution of 10 cells per wavelength, best comparison to measured results for the GTEM model is with 30 cells/,% minimum. A GTEM 750 model was run with both 1 and 2 cm cell sizes up to 1 GHz. The 2 cm grid gave the same response as the 1 cm up to 500 MHz, so 2 cm was used for most geometries here to allow reasonable GTEM 1750 and OATS model run times. Results for the 2 cm grid OATS model should be valid up to the 10 cells//2 frequency of 1.5 GHz. A typical problem space size is 132 x 385 x 202 or 10.3 million cells. This geometry fits into less than 0.3 of 1 GB available RAM on a 500 MHz Alpha-based workstation. With FDTD, the field and voltage responses from the timedomain Gaussian or modulated-gaussian pulse source are Fourier-transformed to get the wideband frequency response for each EUT simulated. For an accurate comparison of the model and test data it is necessary to know the output voltage of the comb generator. To estimate a scale factor, the comb generator output voltage was measured through a known attenuator into a 50 Q receiver. The difference between this forward power and the power from the 1 volt FDTD source was used as a normalization factor. Since the septum is aligned with a mesh plane in the model and the floor is staircased, in both the modeling and testing described below the EUT axes are aligned parallel to the septum. In contrast, most GTEM users align the EUT axes parallel to the floor for convenience. From results not shown here for the plate-mode dummy EUT, the difference between these two orientations was found to be negligible. In the 2 cm grid the dipole was modeled as 18 cell-edge 771

3 wires with a one-cell gap feed. Similarly, a one-cell voltage gap and a one-cell edge wire spanned the two-cell gap between the top plate and box and the slot of the dummy EUT. CORRELATION ALGORITHM CONSIDERATIONS All testing in the GTEMs, OATS, and FAR was done with an HP8546A EM1 receiver. In the OATS data, about of the 195 comb test frequencies were omitted due to ambient interference. The receive antennas were an EMCO 31 IOB biconical for MHz and an EMCO 3147 log for MHz. The antennas were calibrated in a 10 m distance, 2 m transmit height, 1 to 4 m receive height geometry (ANSI C63.5), which has been seen to yield antenna factors that are within 0.5 to 1 db of the free-space value. No additional antenna factor corrections were used. The OATS tests were done on a 26 m x 24 m ground plane. The FAR results were obtained on-axis in a 4.9 m x 3.7 m x 3.1 m chamber lined with highly-loaded 0.61 m foam pyramid absorbers, which do not begin effective absorption until above 80 MHz. This FAR was very small, with the biconical antenna tips only about 30 cm from the foam. septum, the z-axis is longitudinal from feed to load parallel to the septum, and the x-axis is transverse. In several of the graphs below, to spread out the curves for easier viewing, a constant 10, 20, or 30 db is subtracted to offset the curves from each other. For example, in Figure 2 GTEM ABC-10 db means 10 db has been subtracted from the actual response to prevent the curves from overlapping, and similarly for all other graphs. Details of the three-position correlation algorithm are given elsewhere [2, 31, and will not be repeated here. The correlation routine is based on the assumption that the EUT Dipole and Loop Antennas Model and Test may be reasonably represented by its initial multipoleexpansion moments, namely the electric and magnetic dipole moments. Three voltage measurements are then made in the GTEM from which the EUT total radiated power may be Figure 2 shows receive voltages for the three positions of the 40 cm dipole modeled and measured in a GTEM 750. The model cell size was 1 cm. Here the dipole length is a little calculated. The individual dipole moments are not determined more than half of the septum height. Similar voltage separately. The total radiated power is then used to simulate the maximum EUT fields over a ground plane or in free space, based on a model of co-polarized dipoles, either horizontal or vertical, radiating that same total power. In most cases, a three-position GTEM measurement will overestimate the emissions from a product [ 191. However, in some cases it may be necessary to do a total of 6, 9, 12, or 15 positions to responses occur for the dipole modeled in GTEMs 500 and It is believed that staircasing effects in the model are negligible, because if not the longitudinal component would be expected to get larger at higher frequencies. Voltage is underpredicted by the model from about 500 to 900 MHz, where calculated values then go above the test levels. This is possibly due to an imprecise source normalization as get the maximum reading [3-5, 19, 201. The cubical FDTD discussed in the previous section. In general a good mesh used here is most easily amenable to studying the agreement is shown between the vertical component model three-position algorithm. The advantage of GTEM over and test, and this is most important component since signals some other types of TEM waveguides is that the cell factor, or quasi-static cross-sectional field distribution, is calculable. This gives predicted OATS field strengths very about 5 db or more below this have little influence on correlated field strength in the three-position method. The main peaks in the longitudinal response are predicted. close to actual measured levels without the need for any additional correction factors. The OATS-equivalent field strengths for dipole model and test in a GTEM 750 are shown in Figure 3. Below 500 MHz GTEM CORRELATION TEST AND MODEL RESULTS the GTEM modeled results for 1 cm and 2 cm grid sizes are about the same. The ABC-termination is seen to have the most effect below about 200 MHz, with little improvement above this. As desired, the GTEM-predicted levels generally exceed the OATS values. The GTEMs used were EMCO 5407 and In all GTEM results, the actual EUT location and septum height were used as input parameters in the correlation, although variation of the effective electrical positions will influence the results. In the GTEM, the y-axis is vertical or perpendicular to the Figure 2. Calculated (talc) and measured (meas) GTEM 750 receive voltage for x-, y-, z-positions of the dipole, 75 cm GTEM septum height, 1 cm FDTD cell size. 772

4 overall response is near the 390 MHz dipole half-wavelength frequency. lot,,,,i,,,,i,,,,',,,,i,,',i,,,,',,,,i,,,,i,,,,i,,,,i Figure 3. Calculated and measured dipole maximum field strength on 3 m OATS (80 cm transmit height, l-4 m receive height, vert. pol.), and field strength calculated from the measured and modeled GTEM voltages of Figure 2. Field strengths predicted from a 2 cm grid GTEM 750 model with standard and ABC load are also shown. GTEM ABC-1OdB refers to 10 db subtracted from the actual response to keep the curves from overlapping. Figure 4 shows similar results for the dipole modeled and tested in a GTEM 1750 at a 1.5 m septum height. The longitudinal response for the dipole in the GTEM 1750 (not shown here) displays a 130 MHz peak typical of the field structure in this size GTEM, but a comparison of the ABC GTEM (which does not exhibit a longitudinal anomaly) and standard GTEM 1750 results in Figure 4 shows that this does 90 Figures 5 and 6 show horizontal and vertical polarization test results respectively for the 30 cm square loop on the OATS and in a GTEM 750, and in 2 cm grid GTEM and OATS models. Note the particularly strong overestimate from the GTEM readings compared to the OATS nulls. This is because the loop has strong electric-field pickup when aligned perpendicular to the septum in both the vertical and transverse planes. The OATS test only captures the strongest of these two dipole moments. As discussed in [19], the field overestimate may be caused by the neglect of phase differences in the three-position method. Height scanning the receive antenna will smooth the OATS nulls somewhat, but the GTEM will still in general overestimate the OATS field strengths. 80 I I 10 ~.~..I~~..I.~..I..~.I~...I...I~~,.I.,.,I...I, ll 10 Figure 5. Calculated and measured loop antenna maximum field strength on 3 m OATS (80 cm transmit height, 2 m receive height, horiz. pol.), and field strength calculated from measured and modeled GTEM 750 voltages. Dummy EUT Modeling and Measurement Figure 4. Calculated and measured dipole maximum field strength on 3 m OATS (80 cm transmit height, l-4 m receive height, vert. pol.), from a GTEM 1750 test, and from a GTEM 1750 modeled with the usual resistor/absorber load, and with the numerical ABC. not have a significant effect on the correlation response. The modeled response of the standard GTEM below about 200 MHz shows a sawtooth behavior, which is mostly bounded as an envelope by the measured GTEM data. The peak in the Figure 7 shows the receive voltages for three orthogonal positions of the plate-mode dummy EUT modeled (2 cm) and measured in a GTEM Here EUT maximum size is on the average one-third of the septum height. Excellent agreement is shown for the vertical and longitudinal components, with the vertical modeled sawtooth below 200 MHz falling within the envelope of the measured data. A strong longitudinal pickup is again seen near the 130 MHz anomaly. However, the OATS-predicted field strengths in Figure 8 show no apparent deviation from this effect. The OATS-equivalent field strengths for the plate-mode EUT model and test in GTEMs 750 and 1750 are shown in Figure 8. In two of the three positions, the EUT size is about two-thirds of the septum height in the GTEM 750, 773

5 which is greater than the usual recommended one-third septum height EUT size. Very good agreement is obvious between model and test, with the GTEM in general overpredicting the OATS field. The periodicity is greater in the GTEM 1750 response due to the longer effective transmission line length. A dip is evident in the OATS and GTEM 1750 results, which may been an intermittent EUT instability, or it is necessary to look for other reasons why this effect is not present in the OATS model or GTEM 750 test. The low-frequency ripple is mostly removed in the ABC GTEM 1750 response, but above 200 MHz the response shape is about the same as for the standard GTEM. Figure 9 is the equivalent free-space field strength calculated from measured GTEM 1750 voltages. The GTEM correlation mostly tracks the FAR test results above 200 MHz. The case resonance in the 700 to 800 MHz range is shifted down in frequency in the free-space model compared to the test, possibly due to a small piece of absorber that was included inside one corner of the box in the model to accelerate convergence. 85 f OATS c-ale.-25 db 10~~.~~'~~~~'~~~~1~~~~I~~~~I~~~~I~~~~I~~~~I~~~~I~... Figure 6. Calculated and measured loop antenna maximum field strength on 3 m OATS (80 cm transmit height, 2 m receive height, vert. pol.), and field strength calculated from measured and modeled GTEM 750 voltages. GTEM 750 mew.40 db 5 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Figure 8. Calculated and measured plate-mode dummy EUT maximum field strength on 3 m OATS (80 cm EUT height, l-4 m receive height, vert. pol.), and field strength calculated from measured and modeled (2 cm grid) GTEM 750 and 1750 voltages. L Figure 7. Calculated and measured GTEM 1750 receive voltage for x-, p, z-positions of plate mode dummy EUT, 1.5 m GTEM septum height, 2 cm FDTD cell size. Finally, Figure 9 shows the slot-mode EUT free-space copolarized electric field strength measured on-axis in a 3 m FAR, and modeled in free space (2 cm grid). Below 200 MHz the FAR performance is deficient. Also shown in 2ot~~~~l~~~~1~~~~l~~~~I~~~~1~~~,I,,,,I,,,,I,,,,I,,,~ lf Figure 9. Measured slot mode dummy EUT maximum field strength in 3 m fully-anechoic room (on-axis, co-polarized), and free-space tield strength calculated from measured and modeled (2 cm grid) GTEM voltages. SUMMARY FDTD modeling has been used to provide useful qualitative and quantitative insight into the basic three-position correlation algorithm used with GTEM cells. Excellent agreement was shown between measured and modeled

6 responses from self-contained dipole, loop, and simple box EUTs inside the GTEM, and between actual measured and GTEM-predicted OATS and FAR field strengths. One significant result shows is that the influence of any unwanted longitudinal component may be negligible in final product compliance assessment. More work needs to be done on modeling and testing of cable placement effects. The absorbing boundary condition results show that there is room for low-frequency improvements in standard GTEM designs. However, improved GTEM terminations are not expected to remove all standing waves effects caused by the nonuniform-waveguide flared geometry. GTEM cells continue to be a useful facility for both pre- and full-compliance emission and immunity testing. The present study is a step towards the better understanding and application of GTEMs. All considerations of essential and non-essential modes aside, the proof of GTEM validity is in its use. While analyzing and predicting higher order modes is useful, what really matters is if field uniformity is within some agreed-upon tolerance, if a single TEM field component is dominant within some tolerance, and if a reasonable amount of correlation is shown to the preferred emissions test methods. The results here show that is possible to draw useful heuristic conclusions from cubicalgrid FDTD GTEM modeling about design improvements and testing of complex EUTs, in support of the current international TEM waveguide test standardization efforts. ACKNOWLEDGEMENTS Many thanks to M. Alexander of NPL UK and W. Mtillner of ARCS Austria for use of the dummy EUT and for many useful discussions. PI REFERENCES Ma, M. T., and G. H. Koepke, A method to quantify the radiation characteristics of an unknown interference [I71 source, NBS Tech. Note 1059, Oct Wilson, P., D. Hansen, and D. Koenigstein, Simulating open area test site emission measurements based on data obtained in a novel broadband TEM cell, IEEE Natl. Symp. Electromag. Compat., Denver, CO, pp , l-181 r31 Wilson, P., On correlating TEM cell and OATS emission measurements, IEEE Trans. on Electromag. Compat., vol. 37, no. 1, pp. 1-16, Feb [41 Osburn, J. D. M., and E. L. Bronaugh, Advances in GTEM to OATS correlation models, IEEE Intl. Symp. Electromag.,[191 Compat., Dallas, TX, pp , [51 Lee, A.-K., An advanced correlation algorithm between GTEM and OATS for radiated emission tests, ETRI Journal, vol. 17, no. 3, pp , Oct PO1 161 Koch, M., C. Groh, and H. Garbe, Exact determination of resonant frequencies in TEM cells, Zntl. Zurich Symp. Electromag. Compat., Zurich, Switzerland, pp , r71 Harrington, T. E., GTEM fields FDTD modeling, IEEE Intl. Symp. Electromag. Compat., Austin, TX, pp , PI Hansen, D., J. Funck, D. Ristau, and S. Moessler, Comparing the field quality of the new EUROTEM to GTEM and fully absorber lined chambers, IEEE Intl. Symp. Electromag. Compat., Denver, CO, pp , PI Podgorski, A. S., and J. Baran, Radiated emission measurements conducted in the AST semi-anechoic chamber dual polarization broadband gigahertz field simulator and their verification with the measurements done on open area test site., IEEE Intl. Symp. Electromag. Compat., Denver, CO, pp , r101 Giipel, K. -D., S-LINE: a newly developed TEM waveguide for EMC measurements, Intl. Conf: Electromag. Interf and Compat. 97 (INCEMIC-97), Hyderabad, India, pp , El11 Thelberg, M. J., E. L. Bronaugh, and J. D. M Osburn, GTEM to OATS radiated emissions correlation from 1-5 GHz, IEEE Intl. Symp. Electromag. Compat., Chicago, IL, pp. 387, 390, 389, 388, 391, 392 (printed page sequence wrong, Figures 2 and 4 numbers interchanged), WI Clay, S., Improving the correlation between OATS, RF anechoic room and GTEM radiated emissions measurements for directional radiators at frequencies between approximately 150 MHz and 10 GHz, IEEE Intl. Symp. Electromag. Compat., Denver, CO, pp I 124, H31 Koch, M., and H. Garbe, Investigation of field distortion in a TEM-waveguide, Intl. Zurich Symp. Electromag. Compat., Zurich, Switzerland, pp , EMCO 4630 RefRad Reference Radiator System, EMC Test Systems, Austin, TX. [I51 Garn, H. F., and W. Mtillner, A fully automatic, accurate method for the performance evaluation and calibration of absorber-lined chambers for radiated-emission tests, IEEE Intl. Symp. Electromag. Compat., Cherry Hill, NJ, pp , [I61 Marvin, A. C., and S. M. Mann, Experimental study of the calibration of antenna-screened enclosure combinations for radiated emission measurements in the frequency range from 1 to 30 MHz, IEE Proc. -Sci. Meas. Technol., vol. 141, no. 4, pp , July Alexander, M. J., Development of new measurement methods and limits for EMC compliance testing in smaller relatively inexpensive fully anechoic rooms, Intl. Zurich Symp. Electromag. Compat., Zurich, Switzerland, pp , Welinder, J., J. Carlsson, L. Hasselgren, E. Rodriguez, Simulated behavior of fully anechoic rooms for measurement of radiated emissions, Intl. Zurich Symp. Electromag. Compat., Zurich, Switzerland, pp , Turnbull, L., and A. C. Marvin, A treatment of the phase properties of GTEM to open-area test-site correlation techniques, IEEE Trans. On Electromag. Compat., vol. 40, no. 1, pp , Feb Klingler, M., J. Rioult, J-P. Ghys, and S. Ficheux, Wideband total radiated power measurements of electronic equipment in TEM and GTEM cells, Intl. Zurich Symp. Electromag. Compat., Zurich, Switzerland, pp , 775