54 CHAPTER 3 SEPARATION OF CONDUCTED EMI The basic principle of noise separator is described in this chapter. The construction of the hardware and its actual performance are reported. This chapter proposes a device which is developed to decipher CM and DM noise from a Conducted EMI. A Noise Separator using power combiner is used as a diagnostic tool to investigate the Conducted EMI of a power converter. This approach is useful in the analysis of predicted spectrum of CM and DM noise voltages. From the noise diagnosis, valuable information can be obtained for improving the filter design. 3.1 INTRODUCTION The total Conducted EMI noise is caused by two systems such as CM and DM Noise. The CM noise is connected to a capacitive coupling of switching voltage into LISN and the DM noise is connected to switching current. CM and DM noises arise from different sources, so it is essential to have different filters for attenuating CM and DM noise. Therefore, the first step in designing the EMI filter is to separate the CM and DM noises. For this purpose, a noise separator is used. Figure 3.1 shows the setup for Conducted EMI measurement using noise separator.
55 LISN EMI Filter 50µH POWER SOURCE 50µH DM NOISE SOURCE (SPC) 0.1µF 0.1µF 50 50 V L V N NOISE SEPARATOR SPECTRUM ANALYZER CM Figure 3.1 Coupling Paths in Conducted EMI Test Setup An isolated half-bridge AC-DC converter is used as the noise source. The DM noise is caused by the unfiltered portion of the switching current flowing through the 50 resistors. The CM noise is caused by the displacement current flowing through the 50 resistors. The displacement current is coupled with parasitic capacitances between semiconductor devices and chassis, and the transformer inter-winding capacitance. The CM noise current flows through the two 50 resistors in parallel, i.e., through both line and neutral to ground, while the DM current flows through the two 50 resistors in series. It is therefore concluded that the EMI measured voltages on the line and the neutral are different. One is CM + DM, and the other is CM - DM.
56 In order to separate DM and CM noise, the noise separator should satisfy the following requirements. i. Input impedances are always real 50 and are independent from noise source impedances. ii. iii. Output is (V X V Y )/2 for DM noise measurement and (V X +V Y )/2 for CM noise measurement. Leakage between the CM and the DM at output should be small. 3.2 NOISE SEPARATOR The basic concept of noise separator is very simple. Figure 3.2 and Figure 3.3 show the diagram depicting this concept in which the two signals derived from the LISN consist of both CM and DM noises. One of the signals is the vector sum of the two modes of noise (CM + DM) and the other signal is the vector difference of the two modes of noise (CM - DM). L 0 C 1 LISN L 1 C X A To EUT CM + DM 0 Power Combiner CM N C 2 L 2 C X B CM - DM 50 SPECTRUM ANALYZER Figure 3.2 DM Rejecter
57 L 0 C 1 LISN L 1 C X A To EUT CM + DM 180 Power Combiner DM N C 2 L 2 C X B CM - DM 50 SPECTRUM ANALYZER Figure 3.3 CM Rejecter According to the description of standard measurement, a 50 resistance should be placed between point A and ground while measuring the line Conducted noise. The same comment applies to the neutral side also. When a combiner is used, the internal 50 resistors in the LISN are not connected. The 50 resistance between point A and ground (also point B and ground) is the reflected 50 input impedance of the spectrum analyzer. Both inputs are of equal amplitudes. It is assumed that the CM current is evenly divided between the two input terminals, which are often true except in extreme cases. The challenge of implementing the basic concept described above is to maintain the accuracy for the frequency ranging from 150 KHz to 30 MHz. A small error of phase or magnitude introduced in the process of summing or the phase subtracting leads to large percentage of error in the separator. The device that meets such accuracy requirements is the power combiner.
58 3.3 POWER COMBINERS The noise separator is built using the principle of noise rejection accomplished by the use of power combiners. A power splitter is a commonly used RF device for splitting an input signal into two signals with equal amplitudes and a specified phase angle. When used in reverse, a splitter becomes a combiner. Two types of combiners, 0 power combiner and 180 power combiner are used in the noise separator. The block 0 power combiner is a device that cancels the DM component and lets through the CM component, as shown in Figure 3.4. The block 180 power combiner cancels the CM component and allows the DM component, as shown in Figure 3.5. The output of a 0 power combiner is the sum of the two input signals, and the output of an 180 combiner is the difference of the two input signals. The design and fabrication technique of a combiner is closely related to those of a wide-band transformer. In Figure 3.4 wide band transformer act as a summer and it produce the CM noise in the secondary side. + CM+DM R int + CM-DM Ro CM + - Figure 3.4 0 Power Combiner
59 + CM+DM R int CM-DM + R o + DM - Figure 3.5 180 Power Combiner In Figure 3.5, two wide band transformers are used. First transformer act as a adder, second transformer act as a subtractor. Here added result of CM noise is grounded. Difference of two noises is carried out by the second transformer and the resultant noise of DM is appeared in the secondary side. The accuracy of combining must be maintained over a larger frequency range. In addition to the accuracy, a combiner should also provide proper input impedance to avoid disturbance during the measurement. Power combiners meeting the above requirements are commercially available. 3.4 PERFORMANCE OF THE NOISE SEPARATOR A two-way 0 power combiner (ZFSC-2-6-75) and a two-way 180 power combiner (ZFSCJ-2-1) manufactured by Mini-Circuits are used. The noise separator is used to measure the CM, the DM, or the total noise by means of a three-way built-in switch. 0 combiner is used to reject the DM noise and let the CM noise go through intact. This portion of the noise
60 separator network is called Differential Mode Rejecter (DMR). A 180 combiner is used to reject the CM and permit the DM and is called Common Mode Rejecter (CMR). In a measurement, the CM and the DM noise can be selected by a switch in the noise separator. To measure the total noise, the input signal bypasses both DMR and CMR. The performance of the noise separator has been evaluated by measuring the rejection attenuation of both modes of noise going through the CMR or the DMR. Figure 3.6 and Figure 3.7 show the diagrams of the experimental setup to conduct such evaluation. An impedance analyzer is used to generate input voltage, Vi, of sweeping frequency and to plot the output to input ratio Vo/Vi. In Figure 3.6, CM signal V i is generated as the input of CMR or DMR. In Figure 3.7, input signal Vi is split into DM signal by the 180 power splitter, which serves as the input to CMR or DMR. The four parameters defined to evaluate a noise separating network are the Common Mode Insertion Loss (CMIL), Differential Mode Insertion Loss (DMIL), Common Mode Rejection Ratio (CMRR) and Differential Mode Rejection Ratio (DMRR). The transmission coefficient of noise separating network S 21 is defined in Equation (3.1). S 21 = 20 log (Vo/Vi) decibels (db) (3.1) where, Vi and Vo are the input and output voltage of the network. As the transmission coefficient S 21 is CMIL/DMIL, the Vi and Vo represent the same mode signals, and as S 21 is CMRR/ DMRR, Vi and Vo represent different mode signals. It should be mentioned that the attenuation should not be less than -5 db and the rejection ratio should not be more than -40 db.
61 In CM noise rejection test, a CM signal is applied to both inputs of the rejecter. Ideally, the rejection is infinity for CMR and 0 db for DMR. The CMIL result of the high-performance noise separating network is shown in Figure 3.8. As the frequency goes up, the CMIL declines slightly but remains above -2dB. Impedance Analyzer Single output Ref Test ~ Vi CMR or DMR ~ Vo Figure 3.6 Measurement of Attenuation of CM Signals for CMR Impedance Analyzer Single output Ref Test ~ Vi 180 Power Splitter CMR or DMR ~ Vo Figure 3.7 Measurement of Attenuation of CM Signals for DMR For DM noise rejection test, a DM signal is created using 180 splitter. Ideally, the rejection is infinity for DMR and 0 db for CMR. The
62 DMRR result of the high-performance noise separating network is shown with a good performance in Figure 3.9. Figure 3.8 CMIL of the Noise Separating Network Figure 3.9 DMRR of the Noise Separating Network
63 The DMRR rises with the increasing frequency and at 30MHz, the maximum frequency for Conducted EMI noise measurement, the DMRR remains below -40dB. The measurement results prove that the noise separating network can separate the Conducted EMI noise efficiently. The wiring of noise separator hardware should be kept as balanced as possible to maintain better performance. 3.5 USE OF NOISE SEPARATOR IN POWER CONVERTER EMI MEASUREMENT An isolated half bridge AC-DC converter is used in EMI measurement. All the test results are displayed in three parts: Total noise, DM noise and CM noise. Test frequency range covers 150 KHz to 30 MHz (FCC Class B). All the horizontal axes are in linear scales. Figure 3.10 shows the hardware result of the Total, DM and CM noise voltages without filters using noise separator by spectrum Analyzer Agilent E4411B (9 KHz 1.5GHz). (a) Total Noise
64 (b) DM Noise (c) CM Noise Figure 3.10 Hardware Results of Noise Separator without Filter All EMI spectrums displayed on the spectrum analyzer are employed by a peak detector and a 10 KHz resolution bandwidth. At low frequency, DM noise is dominant. The pulse current of the switching frequency generates the
65 EMI noise at low frequency harmonics. At high frequency, CM noise is dominant. Discharging and charging current of the parasitic capacitance produce EMI during the fast and high voltage transition. The test EMI filter diagram is shown in Figure 3.11, where L CM is a CM choke, L DM is a DM choke, C X1 and C X2 are DM capacitors which are called X capacitors, C Y is a CM capacitor which is called Y capacitor. The CM choke L CM has two identical windings wound on the same core. Ideally, a CM choke has no effect on the DM noise due to the cancellation effect of the two identical windings. However, its leakage inductance due to coupling imperfection provides filtering effect to the DM noise. Usually the leakage inductance is about 0.5% to 2% of CM inductance. L DM L CM1 º º L CM2 AC Input 18µH C x1 0.22µF 18µH C Y C Y 4.7nF 4.7nF Output C x2 0.22µF Leakage Inductance º 0.9mH º 0.9mH Figure 3.11 Test EMI Filter Figure 3.12 shows the hardware result of the Total, DM and CM noise voltages when the test EMI filter is used. From the results obtained, it is clear that, to improve the performance, the dominant-mode noise needs to be suppressed. There is no point in adding additional filter components to suppress the less-dominant mode because it will not affect the total noise.
66 (a) Total Noise (b) DM Noise
67 (c) CM Noise Figure 3.12 Hardware Results of Noise Separator with EMI Filter From the experimental results and measurement analysis, the proposed noise separator has the following advantages. 1) Input impedances are always real 50 and are independent from source impedances. 2) Outputs are exact DM and CM noise voltages. 3) DM and CM are simultaneously measured using the same circuit. 3.6 CONCLUSION Noise separator is a valuable tool to identify the dominant mode of EMI noise. This device consists of only passive components and requires no power supply. The information obtained provides the designers for noise diagnosis with guidance for improving the filter design. Table 3.1 shows the hardware results for various frequency ranges using the noise separator. The results obtained satisfy the FCC class B regulations.
68 Table 3.1 Comparison of Noise Separator Results Frequency in MHz Without Filter (dbµv) With Filter (dbµv) Total DM CM Total DM CM 10 83 76 80 36 33 32 20 80 78 77 28 24 23 30 76 74 78 24 20 20 The results shown in this chapter demonstrate the efficacy of a power combiner as a diagnostic tool for Conducted EMI problems. Experimental results verify that the proposed noise separator using power combiner satisfies all the requirements. Results have shown that the proposed method is very effective and accurate in identifying and capturing EMI features in SPC. The method presented is not only limited to half-bridge converters, but it can be applied to any different converter topologies.