Predicting Vehicle-Level EMC Performance Utilizing On-Bench Component Characterization Results

Transcription

1 Predicting Vehicle-Level EMC Performance Utilizing On-Bench Component Characterization Results CI-IINGCHI CHEN IEEE MEMBER Ford Research Laboratories Rm.3010sRL/MD# Rotunda Drive Dearborn, MI Tel : , Fax : cchen4@ford.com Abstract -This paper proposes a simple, but powerful technique to predict the system-level electromagnetic performance ~Forn the results of onbench component characterization. Using a network analyzer, the components of a vehicle ignition system were characterized on-bench up to IGHz, and then the results were manipulated numerically to estimate fhe system-level performance. Results have shown that the prediction matches the open-area-test-site (OATS) test results pretty well, which shows that this can be a very handy tool for predicting possible system-level problems from componentlevel features, or the problems can be traced down to subsystem- / component-level to pinpoint the troublesome bottle neck(s). I. INTRODUCTION Ignition systems are known to be the dominant broad-band noise source on spark-ignited gasoline vehicles. Through decades of efforts, many cost-effective EMI mitigation techniques have been developed, like the adoption of resistive spark plugs, resistive high-tension wires, by-pass capacitors, etc. These strategies were mostly evolved based on experts profound knowledge/experience, plus extensive experiments and full-blown tests to confirm the effectiveness. Modeling approaches are known to be tough and imprecise because of complicated component geometry/system structure, plus the fact that the wave-length of interest (centimeters for GIIz noise ) may not be shorter than the system or even the component dimensions; therefore, no one has ever been confident to predict the overall performance from well-known subsystem/component characteristics. Very often, full-blown whole-vehicle tests have had to be carried out to investigate the impacts of a minor modification, a fairly timeconsuming and costly exercise. Worst of all, if the outcome. is satisfactory, it is difficult to provide solid/quantifiable explanation, and if it fails, experts know-how was the only thing to count on. To address this deficiency, effort has been taken recently. For example, methodologies have been proposed to evaluate the radiated EMI pattern from a complicated wire harness [l]; however, the noise distribution mechanisms for the component embedded inside the engine block have not been well-studied. To supplement, this paper presents an empirical strategy to estimate the system level performance from the component features measured on-bench. Simple algorithms are then derived to integrate the component characteristics to anticipate the system-level features. Evidences have shown that the prediction matched the vehicle-level test results pretty well, showing it a very handy tool for debugging/developing next generation products. II. SYSTEM DESCRIPTION Figure 1 COP ignition system configurations. The circuitry investigated in this paper, as shown in Figure 1, is the state-of-the-art coil-on-plug (COP) ignition system [2], which has allured a lot of attention due to the potential benefits of lower overall cost, increased engine packaging flexibility, and supposedly reduced EMI radiation. The operation principle of this system is virtually the same with traditional ones, where the coil current is 1 765

2 regulated by the electronic switch in the controller, located in the passenger compartment. A long wire harness is needed to connect the electronic switch and the COP. When the electronic switch is turned on, magnetic energy is stored into the coil, and when it s off a high voltage is generated on the secondary of the coil. This elevated voltage is passed through a coiled spring down to the spark plug to produce the desired electric spark. A boot resistor can be added between the coil and the plug to farther reduce the EMI emission. In order to mimic the on-vehicle arrangement, components inside the engine well ( spark plug, boot resistor, & coil ) were characterized on-bench by surrounding them by a piece of rubber boot (cut from the actual boot used in the system ) and sheltered by a copper tube with similar diameter to the engine well, and with both side completely terminated, as shown in Figure 4. Figure 5 shows the S-parameters of a resistive spark plug measured under this setup. III. NOISE DISTRIBUTION MECHANISMS & COMPONENT CECARACTERIZATION The electric spark has long been suspected the dominant wide-band ( KHz -GEIz) noise source. To confirm this assumption, a simple test was arranged, where the spark plug was removed while the coil excited under the same ignition cycle. Under this condition, the coil generates the highest output voltage yet no electric spark occurs. Interestingly, no MHz-GHz noise was detected, proving that the electric spark is the OiVZY high-frequency noise source. With this in mind, the noise distribution mechanism is clear : it propagates back through the spark plug, boot resistor ( if exists ), ignition coil, wire harnesses, then radiates to the ambient, as shown in Figure 2, and the mechanism can be described by the block diagram in Figure 3(a), where the components/subsystems are modeled as two-port networks as shown in Figure 3 (h). To characterize these components up to GHz range, S-parameters are adopted [3], and which can be measured by network analyzers, Figure 3. (a) Block diagram for ignition noise distribution mechanism. (b) Component/subsystem represented by a two-port network. network analyzer Figure 4. On-bench setup for component characterization Figure 2. Spark noise distribution and radiationmechanisms

3 IV. SYSTEM FEATURES PREDICTION OATS,TEST RESULTS & To estimate the features of a cascaded system, the measured component S-parameters were first transformed to ABCD parameters by the following equations [3] : (1) B = z (1+~,1)(1+~22)-~,2~*, 0 2s21 c 1 (1-.~,*)(1-~22)-~,2~2, = , and 20 = characteristic impedance ( 50R ) With AEXD parameters on hands, the equivalent characteristics of a cascaded system can be easily predicted by multiplying the matrices of these components/subsystems I 10 Figure 5. The S-parameters of a 2.7 KG? resistive spark plug Note that the sequence of these matrices multiplied must match the order of the actual physical setup, and then if desired, these calculated ABCD parameters can be converted back to the equivalent system-level S-parameters. Figure 6 compares the measured and the calculated ABCD parameters of a combo, including the spark-plug, the coiled spring, and the coil. Although with minor discrepancies, the main profiles in the figure track each other pretty well,,, showing this simple reasoning process works very nicely in describing the key features of the system This is so because these components are embedded deep inside the engine well, where direct radiation is virtually blocked, and the mutual coupling between non-adjacent components is limited by the long, narrow engine well b

4 Among the measured/calculated S-parameters for the combo, the parameter 821, gauging the energy transmission efficiency from Port 1 ( tip of the spark plug ) to Port 2 ( low voltage side of the coil ), is a direct measure to how easy the spark noise could propagate to the wire harness. Figure 7 shows the equivalent $1 for two combos,., the first is the one mentioned above, and the second is with the addition of a boot resistor. The figure shows that the boot resistor could reduce the noise propagation efficiency by quite a few orders, except below loombz, and this merit has been confirmed by the vehicle-level test at an OATS ( tested according to the SAE-J551 staudard [4] ). The results are shown in Figure 8, where the predicted improvement ( amount of db deviation ) in Figure 7 has been dutifully reflected to the final qualitication results in Figure 6. V. CONCLUSIONS A simple and powerful technique for debugging/development a system/subsystem is proposed in this paper, which has successfully foreseen the performance improvement over the vehicle-level from on-bench component characterization. Through A-B comparison, potential system-level improvements/deficiencies can be captured in the very early component development stage, or complicated system-level issues can be traced down to the specific point(s) for the most effective adjustments. REFERENCES [1] W.T. Smith and RR. Frazier, Prediction of Anechoic Chamber Radiated Emissions Measurements Through Use of Empirically-Derived Transfer Functions and Laboratory Common-Mode Current Measurements, 1998 IEEE EMC Symposium Record, pp [2] Automotive Handbook, 3rd. Edition, pp , Bosch, Stuttgart, Germany, [3] D.M. Pozar, Microwave Engineering, pp. 235, Addison-Wesley Publishing Co., N.Y., [4] SAE-J551, Surface Vehicle Electromagnetic Compatibility ( EMC ) Standards, Society of Automotive Engineers Inc., PA, USa, Frequency MHz Figure 6. Calculated & measured parameters ofthe plug+boot-re&tor+coil combo

5 . I,\, 1-I I,, \I l/y IIll. +.m?!!v--!-v- f 2odb L t i i iiiii ~llipttnwlt!._ 30 Y *~cyu~) Figura 7. &I for a system with a boot rezistor and one without. 40 1Odi-l L 30 Figure 8. Radiated qualification test results from an OATS for a vehicle with COP ignition systems, wifh boot resistors and without