Speakers Franki Poon, Bryan M.H. Pong. Power e Lab. The Power Electronics Lab., Hong Kong University. copyright 2002

Transcription

1 Explanation of Electromagnetic Interference (EMI) in Switching Power Supply Speakers Franki Poon, Bryan M.H. Pong Power e Lab The Power Electronics Lab., Hong Kong University copyright A brief introduction to the Power e Lab Our Mission : To advance switching power converter technologies To develop excellence in power electronics technologies To apply the technologies to products and foster cooperation with industries Contact Person : Dr. Bryan M.H. Pong Power Electronics Lab., Department of Electrical & Electronic Engineering, Hong Kong University, Pokfulam Road, Hong Kong Tel : (852) Fax : (852) or mhp@eee.hku.hk Web site : 2

2 Our work in the Power e Lab High Power 120W AC/DCadapter Platform Half brick DCDC Converter Platform High current Converter Platform 1V5 200A ACDC Converter Platform 1V5 40A Battery charger Platform 12V 65W 4.5kVA Power Amplifier 3 More in the Power e Lab website What you are going to learn today.... Easy EMI Basics. What and how people test your product. Not so easy General attitude of an EMI engineer. Difficult How EMI is produced. More difficult How EMI comes and goes. 4

3 Basic Blocks Every EMI issue has the following 3 basic blocks Noise Source Coupling Path Receiver 5 Noise Sources Switching voltage and current waveforms in a switching power supply 6

4 100V 50V 100V 0V 0s 4us 8us 12us 16us 20us V(R1:2) Time 50V 10A 0V 0s 4us 8us 12us 16us 20us V(R2:2) Time 100V 0V -100V 0s 4us 8us 12us 16us 20us V(R3:2) Time 5A 0A 0s 4us 8us 12us 16us 20us - I(R5) Time 100 Waveform produces frequency spectrum Rectangular 10kHz 100V 500MHz 1.0V Forward 100uV 1.0uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V(R1:2) Frequency 100V 1.0V 100uV Bridge 1.0uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V(R2:2) Frequency 100V 1.0V uV 1.0uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V(R3:2) Frequency Waveform produces frequency spectrum Continuous Flyback 10kHz 1.0A 500MHz 100uA 0 1.0uA 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz - I(R5) Frequency Discontinuous Flyback 10 10A 1.0A 5A 100uA 0 0A 0s 4us 8us 12us 16us 20us - I(R6) Time 1.0uA 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz - I(R6) Frequency 8

5 100V 50V 100V 0V 0s 4us 8us 12us 16us 20us V(R1:2) Time 50V 0V 0s 4us 8us 12us 16us 20us V(R7:2) Time Effect of duty cycle Duty cycle = kHz 500MHz 100V V 1.0V 50V 0 0V 0s 4us 8us 12us 16us 20us V2(R8) Time 100uV Duty cycle = uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V2(R8) Frequency 100V V 1.0V 50V 0 0V 0s 4us 8us 12us 16us 20us V2(R9) Time 100uV 1.0uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V2(R9) Frequency 9 Effect of waveform slope (dv/dt( dv/dt) ns rise and fall time 10kHz 100V 500MHz 1.0V 0 100uV 1.0uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V(R1:2) Frequency 300ns rise and fall time V 1.0V 0 100uV 1.0uV 10KHz 100KHz 1.0MHz 10MHz 100MHz 500MHz V(R7:2) Frequency 10

6 Snubber circuit reduces dv/dt 11 Notes on Noise Sources Voltage and current waveforms Noise sources in switching power supply has little effect beyond 200MHz Waveshape affect base band frequencies Rising and falling edges affect high frequency spectrum Duty cycle has little effect on the spectrum 12

7 Basic Blocks Every EMI issue has the following 3 basic blocks Noise Source Coupling Path Receiver 13 Coupling Paths Coupling Mechanisms Contact Coupling Non-Contact Coupling Circuits visible on Schematic diagram Near field Far field Electric Field Magnetic Field 14

8 Receiver Standard test equipment set up Conducted Measurement set up Radiated Measurement set up 15 EMI standards Electromagnetic Emission (Our Interest) Electromagnetic Immunity 16

9 Emission standards Reference Description (Emission Standard) CISPR13 /EN55013 CISPR14 /EN55014 CISPR15 /EN55015 Limits and methods of measurement of radio interference characteristics of sound and TV broadcast receivers and associated equipment. Limits and methods of measurement of radio interference characteristics of electric motor operated and thermal appliances for household and similar purposes, electric tools and similar electrical apparatus. (Latest revision cover all electrical household appliances) Limits and methods of measurement of radio interference characteristics of electrical lighting and similar equipment. CISPR22 /EN55022 IEC /EN IEC /EN EN EN FCC 15B Limits and methods of measurement of radio interference characteristics of information technology equipment. Limits for harmonic current emission (<= 16A per phase) Limitation of voltage fluctuations and flicker in low voltage supply system (<=16A per phase) Generic residential emission standards Generic industrial emission standards USA National EMI Standard 17 Immunity standards Reference Description(Immunity Standards) CISPR20 /EN55020 IEC /EN IEC /EN IEC /EN IEC /EN IEC /EN IEC /EN EN EN Limits and methods of measurement of immunity characteristics of sound and TV broadcast receivers and associated equipment. Electrostatic discharge immunity test Radiated radio frequency electromagnetic field immunity test Electrical fast transient immunity test Surge immunity test Immunity to conducted disturbances induced by radio frequency fields above 9kHz. Voltage dips, short interruptions and voltage variations immunity test. Generic residential immunity standard Generic industrial immunity standard 18

10 Measurement use what type of equipment? Spectrum Analyzer 0 Cheap 1 One is enough 10kHz 1GHz 2 Less accurate EMI Receiver 0 Expensive 1 Two are needed 10kHz to 30MHz, 30MHz to 1GHz 2 More accurate 19 Measurement Set Up Conducted (< 30MHz) Vertical ground plane 0.4m(CISPR) 1m(FCC) EUT 0.8m 0.8m(CISPR) Excess lead bundle Ground plane 2m x 2m(CISPR), 2.5m x 3m(FCC) To receiver/ Spectrum Analyzer 20

11 Measurement Set Up Radiated EMI(> 30MHz) Measurement distance L = 3M or 10M Rotate to obtain Max. reading EUT 0.8m 1m 4m to obtain Max. reading To Receiver or Spectrum Analyzer with correction factor Ground plane 21 Radiated Set Up Open site Grounding 1m L 3 1/2 L Max. EUT dimension Minimum ground plane 0.5m Boundary of area to be free of reflective objects 2L Max. antenna dimension 22

12 Radiated Set Up - diagnosis Anechoic chamber filled with ferrite absorbers to absorb EM reflections Small TEM cell provides convenient way to diagnose EMI problem 23 Measurement Limit level Average limit (B) Quasi peak limit (B) 0.15MHz 0.5MHz CISPR22 conducted emission limit 56 dbuv 46 dbuv 630 uv uv 66 dbuv 56 dbuv 2000 uv 630 uv 0.5MHz 5MHz 46 dbuv 200 uv 56 dbuv 630 uv 5MHz 30MHz 50 dbuv 316 uv 60 dbuv 1000 uv Can you make an amplifier which receives signal from 150kHz to 1GHz at 200uV signal level?? CISPR22 radiated emission limit Quasi peak limit(10m class A) 30MHz 230MHz 40dBuV/m 100 uv/m 230MHz 1G 47dBuV/m 223 uv/m Quasi peak limit(10m class B) 30dBuV/m 31uV/m dBuV/m 70 uv/m

13 Measurement about error According to "EMC for product Designers - Tim Williams RF measuring receiver - +/- 2.5dB (worse for spectrum analyzer) Impedance mismatch Antenna +/- 1dB +/- 4dB Antenna cable +/- 2.5dB Anechoic chamber Test engineer +/-3dB +/- 4dB Total = 17 db! 25 Measurement short summary Spectrum Analyzer is cheaper for in house pre-compliance test. Conducted measurement is rather simple. Radiated measurement is difficult. Different test Laboratory will give you different results. 26

14 Most Important Concept - LOG Equipment reading are in LOG scale, it lies in our mind that Measured Reading = 20Log( ) dbuv 6 10 Can you imagine what a number will become when it is divided by 10-6, then take LOG, and finally multiplied by 20?? 27 Most Important Concept - LOG & Sum 500uV = 54 dbuv 1000uV = 60 dbuv 2000uV = 66 dbuv 500 uv uv uv = 3500 uv = 71 dbuv Generic human intuitive comparison Compare linear result 3500 / 500 = 7 times Compare LOG result 71/54 = 1.3 times 28

15 Most Important Concept - LOG & Subtract 3500uV = 71 dbuv 2000uV = 66 dbuv 1000uV = 60 dbuv 3500 uv uv uv = 500 uv = 54 dbuv Generic human intuitive comparison tells us Compare linear result 500 / 3500 = times Compare LOG result 54/71 = 0.76 times 29 LOG Reading an example Assume that noise is produced by four contributors transformer MOSFET PCB Voltage (dbuv) Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level 30

16 Example eliminate effect of the transformer Conclusion Little effect MOSFET Voltage (dbuv) PCB Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level Xformer noise eliminated 31 Example eliminate effect of the MOSFET Conclusion Very Little effect 90 Xformer PCB Voltage (dbuv) Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level MOSFET noise eliminated 32

17 Example eliminate effect of the PCB Conclusion No effect Xformer MOSFET Voltage (dbuv) Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level PCB noise eliminated 33 Example eliminate the little and the very little Conclusion Hopeless, still exceeds the limit PCB Voltage (dbuv) Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level Xformer & MOSFET noise eliminated 34

18 Example already in a professional way? Start PCB effect has no contribution Isolate various causes Ignore Y Only Xformer and MOSFET contribute Investigate each cause Stupid & unnecessary N Eliminate 35 Wait a minute Vn 0.7Vn Xformer MOSFET Voltage (dbuv) Vn PCB Vn Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level Xformer only MOSFET only PCB only Aux. only 36

19 We re wrong twice eliminate the No Effect Voltage (dbuv) Aux Frequency (Log scale) CISPR22 conductive class B limit line Original Noise Level Aux. only Wrong three times? Xformer MOSFET PCB Bingo! PCB is important PCB MOSFET Xformer Bingo! Xformer is important 38

20 Wrong Concepts There is a dominant noise mechanism or path Wrong! Some other noise mechanism can be ignored as it shows no effect Wrong! Yesterday EMI solution did not work in today s problem Wrong! 39 Right Concept You will not see a result until you have eliminated the LAST problem" Peter Bardos 40

21 Remember Two equal noise level add together only make a 6 db difference Measured + Measured Reading = 20Log( ) dbuv 6 10 Measured Reading = (20Log( ) + 20Log(2)) dbuv 6 10 Measured Reading = (20Log( ) + 6) dbuv Take a break 42

22 Picking up the noise - Line Impedance Stabilization Network ac Only Simple Mission 50 Ω to Earth 50Ω Line 1 Line 2 50Ω Earth EUT 43 a little bit more detail EUT Live 50uH Line 1 Line 2 50uH Neutral 0.25uF 0.25uF 10uF 1k 50 Ohm receiver 50 Ohm receiver 1k 10uF Earth 44

23 detects line to line noise Input C1 Earth Input L1 To 50Ohm receiver Terminated by 50Ohm Idiff Line1 Differential mode noise Line detects line to earth noise Input L1 Line1 Earth C1 To 50Ohm receiver Icom1 Common mode noise Input Terminated by 50Ohm Icom2 Line2 46

24 Picking up the noise in space - Antenna Antenna factor 20dB Log periodic 10dB biconical MHz (log scale) 47 Predicting noise to space- Absorption clamp Move alone the input cable EUT Current transformer Ferrite rings absorber Measured cable Directional device Ferrite rings Absorbing clamp To receiver 48

25 Nature of EM Noise Time varying Amplitude varying Frequency varying It changed After we have measured it! 49 Nature of EM noise Worse, normal and good Peak measurement the worse, measure the peak noise within a period of time. Average measurement the good, measure the averaged noise within a period of time Quasi-peak measurement the normal, between peak and average measurements. 50

26 Peak, Quasi-peak, average Noise after detector Peak result Quasi-peak result Average result Time 51 Peak detector simplified Mixer Resolution Bandwidth Amp. Peak or Envelope Detector Input To meter Y-axis VCO Osc. Select measuring frequency X-axis Simplified Peak EMI receiver 52

27 Quasi-peak detector simplified Mixer Resolution Bandwidth Amp. Quasi-Peak Detector 0.5Hz LPF Input To meter Y-axis VCO Osc. Select measuring frequency X-axis Quasi-Peak EMI receiver 53 Quasi-peak constant QP-detector kHz MHz MHz 0.3 1GHz 6dB bandwidth 0.2kHz 9kHz 120kHz 120kHz Charge time 45mS 1mS 1mS 1mS Discharge time 500mS 160mS 550mS 550mS What happens if the noise occurrence time is less than the charge time? 54

28 Average detector simplified Mixer Resolution Bandwidth Amp. envelop Detector 0.5Hz LPF Input To meter Y-axis VCO Osc. Select measuring frequency X-axis 55 Peak constant? QP-detector kHz MHz MHz 0.3 1GHz 6dB bandwidth 0.2kHz 9kHz 120kHz 120kHz Charge time???? Discharge time???? Don t forget every detector has a charging time constant. 56

29 Detector how long does a signal exist 9kHz Period for detection at a certain frequency VCO Osc. T charge = f stop BW f initial T For conducted EMI range and sweep is 1s sweep 1s T charg e = 0. 3 ms 57 Detector fast and slow sweep Sweep time = 10 s Sweep time = 1 s 58

30 Short summary, spectrum, receiver, antenna, absorption clamp are introduced. Peak detector is used for screen display as it gives fastest response. Quasi-peak and average are used in most standards. You can cheat the detector or display screen if the noise pop out faster than the charge time. Don t be cheated by the screen if sweep time is too fast Any Short Question? 60

31 Invisible coupling capacitive 16cm 2 450Ω 2cm - 1.9pf V mc 450 Ω 1.9pf V mc 15pf 15pf Z s T I spectrum analyzer Z s spectrum analyzer Does a capacitive object work like a capacitor in a circuit? 61 Capacitor is capacitor V mc = 1V 450 Ω 15pf 15pf Z s R Zs C Zs 1.9pf spectrum analyzer T1 O/P Volt.(dBuV) Frequency(Hz) Simulated result Measured result L Zs Mutual Capacitive Coupling Z s 62

32 Capacitor of course is capacitor The above experiment actually does not intend to prove a two plate structure is a capacitor. It proves two separate physical objects can have effect very much like a simple capacitor coupled together and can give accurate result even the two separate objects connected to a more complicated circuit. Don t forget a power supply is a complicated circuit, with a lot of separate parts and traces Invisible coupling inductive 500cm 2 220p f 1.5Ω r=6cm 350nH 50Ω 0.5cm 40pf V mi I mi 10cm Z s T I spectrum Do I pick up analyzer something here? R Zs C Zs L Zs Z s 64

33 Mutual inductive coupling transformer R cm r cm Lm= Z M R R 2 r _ line_ loop= u0 ( ) ( ) Inductive coupling equivalent model V mi I mi 220pf I mi 1.5 Ω 350nH 1:1 50 Ω 40pf V mi Lm = 25 nh spectrum analyzer Z s 66

34 Mutual inductive coupling = transformer? pf V mi 1.5Ω r=6cm 350nH I mi 10cm 50Ω Z s T I 500cm 2 0.5cm 40pf spectrum analyzer Normalized gain(db) Frequency(Hz) Simulated result Measured result Mutual Inductive Coupling (f) 67 As an EMI engineer You don t see this You see this 68

35 Reminder Any two nodes, two objects, two components, two traces etc, has an invisible capacitance and capacitive coupling effect Any two segments, two loops, two traces, two paths etc, produce an invisible transformer and have inductive coupling effect. Is there any combination of capacitiveinductive effect? 69 Capacitive-Inductive effect Simple idea of Electro-Magnetic wave propagation. V & I propagate 70

36 Capacitive coupling trace to trace Say A2 A1 C 1-2 = 0.1pf Line 1 C 1-2 V S1 = 300V Line 2 Input circuit & Bridge V bulk V S1 C bulk is blocked by Input circuit & Bridge 71 Capacitive coupling trace to trace V Line 1 50 Ω 50 Ω Line 2 C V = πfjC V S1 As C 1-2 = 0.1pf V S1 = 300V At 150kHz V = 1400uV EN55022 B limit at 150kHz is 630uV 72

37 Capacitive coupling C 1 C 2 C 1 = C 2? C 1 > C 2? C 1 < C 2? 73 Capacitive coupling which is bigger C 1 C 2 C 1 << C 2 Imagine C 2 approximate as the sum of all the capacitors 74

38 Capacitive coupling trace to ground Input circuit & Bridge V S1 A1 Say C s = 0.1pf V S1 = 300V C s C s has a considerable length between the plates but the surface on earth is very large Earth 75 Capacitive coupling trace to ground V V = 50 Ω 50 Ω C s 50 / / 2 + 2πfjC s 300 V=300V Assume C s = 0.1pf V S1 = 300V At 150kHz V = 700uV EN55022 B limit at 150kHz is 630uV 76

39 HVHF circuits High Voltage High Frequency 77 HVHF circuits reduce to a size As Small As Possible ASAP RIGHT Wrong RIGHT 78

40 Capacitive coupling PCB trace rule Good small trace area, symmetrical, good separation Bad large trace area, asymmetrical, not enough separation and long connection wire Inductive coupling trace to trace C X C X V S1 V ind V S1 I' I' Although the loop has low voltage, it generates other kind of noise. 80

41 Inductive coupling trace to trace induction C X V ind V S1 Say L m = 1nH I = 150kHz 1nH I' V ind = I 2 π f L m V ind = 940 uv C x gives a low impedance path for the induced current to flow 81 Inductive coupling trace to trace result V 50 Ω 50 Ω C X V ind Say C X = 0.1uF At 150kHz V = 460uV V 50 = πfjC 940 X uv EN55022 B limit at 150kHz is 630uV 82

42 Inductive coupling the trace itself Self inductance Line 1 Line 1 I I C X V S1 C X V S1 Line 2 Line Inductive coupling trace induced voltage Line 1 Ls I Let say L m = 10nH I = 150kHz C X V S1 V ind = I 2 π f Ls V ind = 9400 uv Ls V ind EN55022 B limit at 150kHz is Line 2 630uV 84

43 Inductive coupling PCB trace rule Good small loop area, good separation between loops, node terminated at capacitor terminals Bad large loop area, short separation between loops, nodes are not terminated at capacitor terminals 85 Typical capacitor and inductor parameter values C = 0.085A/d pf L = 0.002l(ln(2l/r)-0.75) uh Lm = 0.002l(ln(2l/D)-1+D/l) uh 1cm 1cm d=1cm l=1cm D=1cm l=5cm C = pf L awg18 =5.8nH Lm awg18 =15nH 86

44 Short conclusion Very small capacitor may cause serious capacitive coupling effect Very small inductor may cause serious inductive coupling effect, either mutual or self inductive coupling. Very small pieces of object may cause devastating magnitude of capacitance or inductance Any Short Question? 88

45 An obvious noise source input ripple current C 2 L 3 C 1 Well known equivalence L 3 R L3 C 2 C L3 C 1 L C2 L C1 I S1 V _1 R C2 R C Input ripple current simulation Don t be scared by complicated circuit. You don t need to calculate for the answer. You ONLY need to simulate it s result. SPICE, Math software... etc may help Voltage (dbuv) Frequency (Log scale) CISPR22 conductive class B limit line V noise level (g) 90

46 Input ripple current typical result V _1 200 uh pf 0.47 uf 5 nh 20 nh uf 0.5 Voltage (dbuv) I S1_ini = 0.5A I S1_peak = 0.8A t r =50 ns t f =50 ns t on = 2 us T = 6.6 us Frequency (Log scale) CISPR22 conductive class B limit line V noise level 91 Input ripple current ideal C uh pf 0.47 uf 5 nh C uf Voltage (dbuv) I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns Frequency (Log scale) CISPR22 conductive class B limit line V noise level 92

47 Input ripple current ideal C 2 C 2 C uh pf 100 uf 0.47 uf 20 nh Voltage (dbuv) I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns Frequency (Log scale) CISPR22 conductive class B limit line V noise level 93 Input ripple current ideal L 3 C 2 L 3 C uh 90 V _ uf 5 nh nh 100 uf 0.5 Voltage (dbuv) I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns Frequency (Log scale) CISPR22 conductive class B limit line V noise level 94

48 Input ripple current swap C 1, C 2 L 3 less rms current C uh 0.5 C 1 90 Not much difference V _1 50 pf 100 uf 20 nh nh 0.47 uf 0.05 Voltage (dbuv) I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns Frequency (Log scale) CISPR22 conductive class B limit line V noise level 95 Input ripple current ideal C 1 again C uh 0.5 C pf 100 uf 20 nh 0.47 uf Voltage (dbuv) V _1 0.5 I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns Frequency (Log scale) CISPR22 conductive class B limit line V noise level 96

49 Input ripple current ideal C 2 again C uh 0.5 C 1 90 V _1 50 pf 100 uf 5 nh 0.47 uf 0.05 Voltage (dbuv) I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns Frequency (Log scale) CISPR22 conductive class B limit line V noise level 97 Input ripple current short summary Realistic equivalent circuit has difference with ideal elements. Equivalent series resistance of a capacitor contribute to low frequency EMI. Equivalent series inductance of a capacitor contribute to high frequency EMI. Perfect the components with largest value is very effective. E.g. 100 uf 98

50 Input ripple current two stage design C 2 L 3 C 1 Add one more stage? L 4 L 3 Wait a minute! C 5 C 2 C Input ripple current make use of leakage There is a Common Mode Choke (CMC) anyway. Make use of it s leakage inductance! C 5 L d4 C 2 L 3 C

51 Input ripple current two stage equivalence 50 uh uh pf 50 pf 0.47 uf 0.1 uf 100 uf 5 nh 20 nh 5 nh V _ mh 50 uh I S1_ini = 0.5A I S1_peak = 0.8A tr=50 ns tf=50 ns 10 mh Input ripple current two stage result Voltage (dbuv) Frequency (Log scale) CISPR22 conductive class B limit line V noise level Single Stage Filter Voltage (dbuv) Frequency (Log scale) CISPR22 conductive class B limit line V noise with P & CM filter Two Stage Filter 102

52 Input ripple current common trick Voltage (dbuv) Frequency (Log scale) Voltage (dbuv) CISPR22 conductive class B limit line V noise with P & CM filter fs = 150kHz, 8 db margin Frequency (Log scale) CISPR22 conductive class B limit line V noise with P & CM filter fs = 130kHz, 30 db margin Input ripple current Disappears Tcon Beyond conduction time t Tcon During conduction time t=tcon 104

53 Input ripple current average result V _1 V V 200 uh pf 0.47 uf 100 uf 20 nh 5 nh _ peak = V _ quasi peak = _ dc V _ dc V _ average = V 2Tcon _ dc T line Voltage (dbuv) Assuming conduction time Tcon = 3 ms Frequency (Log scale) CISPR22 conductive class B limit line V noise of DC supply Averaged noise with ac supply Voltage picked up at with P filter At different time different equivalent circuit Beyond conduction time 0<t<π and t Tcon During conduction time 0<t<π and t=tcon Beyond conduction time π<t<2π and t Tcon During conduction time π<t<2π Apply to all and t=tcon analysis!! 106

54 L 3 Vin Input ripple current stability R L3 C 1 Pin As input of converter is a power source Pin. Approximate the differential Eqn. d v dt 2 C1 2 R + L L3 3 Pin dv 2 v dt C1 C1 C1 vc1 Vin + = L C L C big C 1 is good big R L3 is good small Vin is bad large L 3 is bad Keep first order coefficient > 0 R L L 3 > 3 v Pin 2 C1 C Input filter stability issue V in = v C1 approximately 100 uh R L3 =? R L L 3 > 3 v Pin 2 C1 C 1 100V 100 uf 0.1 uf 100W RL3 > 10Ω!! Usually winding resistance RL3 = 0.2 Ω 108

55 Input filter why usually stable? Because skin effect increases the winding resistance WRONG Because proximity effect increases the winding resistance WRONG Because parasitic capacitor inhibits the oscillation WRONG Because the inductor is no longer an inductor Meaningless Input filter what ac inductor is Inductance Resistor reflecting core losses Resistor reflecting skin / proximity losses Resistor reflecting dc winding losses Frequency dependent Frequency dependent Depend on Frequency & winding structure Wire size dependent DOMINANT! 110

56 Input filter core losses issue Iron powder core Ferrite core Every engineer know core losses is significant, don t ignore the ac resistor reflecting core losses Input filter why usually stable? again Because core losses of the filtering inductor provide extra resistance RIGHT. Because the higher the inductance, the higher the core losses resistance RIGHT. Because the capacitor C 1 is usually bigger than the example RIGHT. Because the oscillating frequency may fall out of regulating region of the converter, and becomes a circuit with positive resistance MAY BE

57 Input filter = Input magnetic receiver Recall the idea of inductive coupling and mix it with the filter. Magnetic field Magnetic Noise voltage V M pick up Many loops to pick up noise field Input filter = Input electric receiver Recall the idea of capacitive coupling and mix it with the filter. Electric field Noise voltage V E pick up 114

58 Input filter summary Leakage inductance can be made used for ripple current filtering. Choose low ESR for big capacitor. Conduction angle or boost PFC yield lower average value than seen on screen. Input filtering circuit is also a receiver circuit. Core losses resistor is introduced Invisible path drain to earth capacitor Simplified circuit showing drain to earth capacitor causing invisible current flow Equivalent circuit using equivalent input resistance and equivalent voltage source 116

59 Drain to earth capacitor equivalent model Basic equivalent circuit Input capacitor is large Conduction period only Input lines are paralleled Drain to earth capacitor how large? 0.31pf Let the total surface area attach to the drain is equal to a sphere with 1cm 2 surface area. Equivalent radius r r = 0.28cm As the self capacitance of a sphere of r cm C sphere = 4 π0.0885r pf The equivalent drain self capacitance is C drain = 0.31 pf 118

60 Drain to earth capacitor simulation 25 uh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 25 Ω Fails just because of 0.31 pf! 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 0.31pf 10uV 3.0uV 200KHz 400KHz800KHz 2.0MHz4.0MHz8.0MHz 12MHz 20MHz V(R5:1) Frequency Filtering doesn t t work? 50 pf 20 mh 0.5 Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us What s wrong? It must be the CMC stray capacitor. 2x4700 pf 0.31pf 25 uh 25 Ω 1 10 nh Frequency 120

61 25 uh 25 Ω Filtering parasitics not crucial 5 pf 20 mh 0.5 2x4700 pf 1 10 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 0.31pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV What s wrong? Beware of the impedance 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(C6:2) Frequency Filtering right filter design 50 pf 20 mh 0.5 Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 30mV 10mV 3.0mV 1.0mV DONE! 2x4700 pf 0.31pf 300uV 100uV 25 uh 25 Ω 1 10 nh 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R5:1) Frequency 122

62 Filtering wrong filter design again 25 uh 25 Ω 500 pf 20 mh 0.5 2x4700 pf 1 10 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 0.31pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV Don t always blame the parasitic elements 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R5:1) Frequency uh 25 Ω Filtering how small it can be? 5 pf 2 mh 0.5 2x2200 pf 1 10 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 0.31pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV A very small filter can filter out the noise!! 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R5:1) Frequency 124

63 Filtering simplified view Filtering get back to the real circuit During bridge conduction period Tcon 126

64 Filtering get back to the real circuit L1 L2 G Cy Cy During bridge conduction period Tcon Practical Look L1 L2 G 2Cy Filtering are the two practical circuit equivalent? L1 L1 L2 L2 G Cy Cy G Cy Cy Practical Look L1 During bridge non-conduction period t Tcon L1 L2 L2 G 2Cy G 2Cy 128

65 Two different filters go to their equivalence L1 L2 G Cy Cy During bridge non-conduction period t Tcon L1 L2 G 2Cy Two different filter Cy on the left hand side 2 pf Not bad during nonconduction period 30mV 10mV 3.0mV L1 1.0mV 300uV L2 100uV 30uV 10uV G Cy Cy 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(C5:1) Frequency 130

66 Two different filter Cy on the right hand side L1 L2 G 2 pf Better during nonconduction period 2Cy 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency Two different filters left and right High Z loop 2 pf High Z loop High Z loop 2 pf Low Z loop 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(C5:1) Frequency 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency 132

67 Invisible path Secondary to earth capacitor Sec. to earth capacitor circuit model 0.31pf 134

68 Sec. to earth capacitor how big? C ww 3cm 3cm 0.15cm C = 0.085A/d pf C ww = 53 pf r = 1.5cm C sphere = 4 π0.0885r pf C se = 1.67 pf Sec. to earth capacitor rare noise Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 53 pf 0.31pf 1.67 pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV Greatly exceeds limit usually come across in bad designs. 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R2:2) Frequency 136

69 Sec. to earth capacitor with filter 5 pf 2 mh 0.5 2x2200 pf 25 uh 1 25 Ω 10 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 53 pf 1.67 pf 0.31pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV Still exceeds limit even after filter is added. 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R2:2) Frequency Sec. to earth capacitor Pri.. To Sec. Capacitor The primary to secondary capacitor can by pass the noise current generated from the primary side

70 Pri.. To Sec. Capacitor circuit model 0.31pf C ps 56 pf 1.6 pf Provide a low impedance path to shunt or by pass the current flowing to earth C ps 1.6 pf 56 pf 0.31pf Or you can interpreter Cps as dividing down the noise voltage Pri.. To Sec. Capacitor save power 5 pf Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 1100 pf Much better solution as no extra power losses is produced. 2 mh uh 2x2200 pf pf 25 Ω 10 nh 53 pf 1.67 pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R2:2) Frequency 140

71 Pri.. To Sec. Capacitor drawback Increase leakage current!! Capacitance is limited Sec. to earth capacitor by pass elsewhere Any point on this side provides by pass effect 142

72 Sec. to earth capacitor by pass inside? How to split the stray capacitor? Sec. to earth capacitor F.5 physics a a a a 144

73 Sec. to earth capacitor add one more plate Sec. to earth capacitor better way to by pass 146

74 Sec. to earth capacitor use both methods Leakage capacitor Sec. to earth capacitor another noise source Secondary winding also act as a noise source 148

75 Sec. to earth capacitor noise source on Sec. side 0.31pf Sec. to earth capacitor final solution By pass shielding capacitor connected to primary quiet node or 0V By pass shielding capacitor connected to secondary quiet node or 0V 150

76 By pass or simply short circuit By pass any noise going right By pass any noise going left 0.31pf Don t t forget the leakage capacitors 0.31pf Leakage capacitors 152

77 Shielding, by pass or short circuit? Does it shield all the noise going out? Revision again - 1 Large Metal 154

78 Revision again - 2 Large Metal Don t t float the metal plate Large Metal 0V 0V 156

79 By pass and short circuit 0V By pass all the noise current back to 0V The victim would see any noise current flow in 0V Short to 0V 0V The victim would only see the 0V of the noise source Wait a minute Zo Zs Zo Shielding - Usually come across characteristic impedance, attenuation, reflection losses The Truth is - Plane wave propagation, or noise source is located at a far distance from the shield

80 Still use by pass and short circuit concept C couple C self By passing noise current C self If far enough, C self >> C couple Like a short circuit 0V C couple 0V Short summary Lump element circuit is possible to simulate EMI behavior. Parasitic elements is not enough to explain all cases. (More concept will be introduced.) So call electric field shielding must be terminated one end to a quiet node or 0V of the noise source. NO FLOATING! So call electric field shielding is a kind of by pass, shunt or short circuit equivalent in a lump element model

81 161 Input terminal conduction period During bridge conduction period t = Tcon 162

82 Input terminal non-conduction period During bridge non-conduction period t Tcon Input terminal coupling capacitor 0.01 cm 0.1 cm 5 cm Input trace 0.4 cm 164

83 Input terminal coupling capacitor value C trace π ( ε r + 1) l 14 _ trace = 10 πs 2ln( ) W + t pf S1 = 5 cm S2 = 5.4 cm W = 0.1 cm t = 0.01 cm pf εr = Input terminal simulation 30mV 10mV 3.0mV pf 1.0mV 300uV 100uV 30uV L1 L pf L1 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(C4:1) Frequency 30mV 50 Ω // 50uH x2 2 mh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us L2 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R5:1) Frequency 166

84 Input terminal imbalance impedance Input imbalance impedance simulation pf 30mV 10mV 3.0mV 100 uh 0.5 Ω pf 1.0mV 300uV 100uV 20 pf 50 uh 0.2 Ω L1 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R7:2) Frequency 10 pf 30mV 10mV 3.0mV 1.0mV Limit Exceeded!! L2 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R2:2) Frequency 168

85 Input imbalance impedance solution pf 30mV 10mV 3.0mV 1 nf 100 uh 0.5 Ω pf 20 pf 50 uh 0.2 Ω 10 pf 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R7:2) Frequency 30mV 10mV 3.0mV 1.0mV Pass in simulation? 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(L2:2) Frequency Input imbalance impedance solution Real? 0.1 uf 0.47 uf 170

86 Don t t forget Magnetic field Magnetic Noise voltage V M pick up Electric field Noise voltage V E pick up Input imbalance impedance solution hence 0.1 uf 0.47 uf 172

87 Input imbalance impedance bad CMC CMC leakage CMC leakage 0.1 uf 0.47 uf A filter is formed by the leakage inductance of the CMC to provide extra attenuation on the noise picked up by CMC Input again is everything fine? The large input X-capacitor provides a final low impedance loop with the. Magnetic couple will contribute to noise pick up by the Low impedance path 174

88 Input again mutual inductance 1 cm 2 cm 5 cm I L 2 cm I R V IR V IL Approximated line to loop mutual inductance l cm W = 1 cm W = 5 cm V I Loop W cm L = 2 cm L line_loop = 0.73 nh L W cm I line W + W = 0.002l(ln( ' uh W line _ loop ) W = 1 cm W = 7 cm L = 2 cm L line_loop = 0.53 nh L line_loop_eqv. = 0.2 nh 176

89 Input loop noise example 5V 6A Loading 25A 15A 2 us fs = 150 khz Input loop noise magnetic induction 0.1 uf 0.2 nh Ipeak =25A Iini =15 A Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us Exceed limit because of small input loop. 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency 178

90 Other noise loop? L 1 Z V0_L1 L 2 Z V0_L2 C G_V0 V M_G I eq Converter to earth Loop I l2 Loop 2 Loop 1 Earth 180

91 Converter to earth Loop capacitor Let the total surface area of the whole converter is equal to a sphere with equivalent radius r r = 5cm As the self capacitance of a sphere of r cm C sphere = 4 π0.0885r pf The equivalent drain self capacitance is C drain = 5.5 pf Converter to earth Loop typical value R r r = 1 cm R = 2 cm M The mutual inductance is Z line_circle =3.3 nh R 100 R 100 r ( line _ loop = u0 ( ) ) 182

92 Converter to earth Loop raw noise 3.3 nh Ipeak =25A Iini =15 A Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 5.5 pf Be careful, really close! 30mV 10mV 3.0mV 1.0mV 300uV 100uV 25A still can t fail it! 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency Converter to earth Loop resonate loop 220 pf 100 nh 3.3 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 0.1 Ω 5.5 pf Fail, much worse than the 25A case! 30mV 10mV 3.0mV 1.0mV 300uV 100uV 300 ua fail it by 30 db!! 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency 184

93 Resonate loop CMC choke 100 uh +0.1 //1 pf 220 pf 100 nh 3.3 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 0.1 Ω 5.5 pf Inductor only drift the resonate point. Kill the original resonant circuit but generate another one 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency Resonate loop RESISTOR Inductance Resistor reflecting core losses Resistor reflecting skin / proximity losses Resistor reflecting dc winding losses Frequency dependence Frequency dependence Frequency dependence & winding structure Wire size dependence Remember this? 186

94 Resonate loop increase resistance 100 uh +10 //1 pf Unfortunately, not much better. 220 pf 100 nh Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 0.1 Ω 5.5 pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency Increasing resistance at the right place Much better, but how? 220 pf 100 nh 10 Ω Vpeak =400 V Tr = 50 ns Tf = 50 ns Pw = 2 us Per = 6.6 us 5.5 pf 30mV 10mV 3.0mV 1.0mV 300uV 100uV 30uV 10uV 3.0uV 200KHz 400KHz 800KHz 2.0MHz 4.0MHz 8.0MHz 12MHz 20MHz V(R1:1) Frequency 188

95 Increasing resistance lossy ferrite 220 pf 100 nh 10 Ω 220 pf 100 nh 5.5 pf 5.5 pf L R (f) Ferrite beads 190

96 Invisible loop inside xformer W1 W2 V S1 W3 Xformer is tightly packed and has large stray capacitor to introduce many loops inside Short summary Capacitive coupling to input terminal causes EMI. Solutions for capacitive coupling lead to more magnetic coupling effect. The whole converter to earth has a large capacitor. Magnetic coupling causes EMI due to the loop form by the converter to earth capacitor. Resonance causes unexpected EMI level. Damping,not inductor, is the solution for resonance

97 193 How does EMI come out? L 1 L 2 G Contacted - can be understood through circuit diagram Non-Contacted - Mutual invisible capacitor & Mutual invisible inductor 194

98 V & I Only L 1 L 2G L 1 L 2G V 1 I m V n I 1 V 2 I Only terminals voltage are concerned G L 1 L 2 V L1 V line1 V L2 V line2 Z L1_G V 1 I Z 1 L1_L2 Z L2_G I m I 2 V n No matter how complicated a circuit, EMI are voltage measurements, V Line1 and V Line2, across terminals. V G V 2 V line1 = VL1-VG V Line2 = VL2 VG Is there a simple way to look at it? 196

99 Circuit theory voltage node G L 1 L 2 V I L1 L1 V L2 I L2 Z V L1_G G I G V 1 I Z 1 L1_L2 Z L2_G V 2 I m I 2 V n I I L1 I L2 G = = = n Vi V Z L1 i= 1 L1_ i n Vi V Z L2 i= 1 L2 _ i n Vi V Z G i= 1 G _ i Equivalent voltage source L 1 L 2 G Z L1_Veq Z L2_Veq Z G_Veq I L1 V ' = Z eq V L1 L1_ Veq V ' + Z 0 V L1 L1_ V 0 L 1 L 2 G Z L1_G Z V0_L1 Z V0_L2 Z L2_G V eq V eq V 0 Z L1_V0 Z L2_V0 Z G_V0 L 1 L 2 G I I L2 G V ' = Z V ' = Z eq V L2 L2 _ Veq eq V G G _ Veq V ' + Z V ' + Z 0 0 V L2 L2 _ V 0 V G G _ V 0 One voltage source and six mutual impedances 198

100 Simple circuit theory current node V M_L1_L2 I 2 V M _ L1_ L2 = m i= 1 I Z i M _ L1_ L2 _ i G L 1 L 2 Z V0_L1 Z V0_L2 V M_L1_G V M _ L1_ G = m i= 1 I Z i M _ L1_ G _ i V M_G Z L2_G I m Z L1_G V M _ G = m i= 1 I Z i M _ G _ i I Equivalent current source V M_L1_L2 L 1 G L 2 Z V0_L1 Z V0_L2 V M_L1_G V M _ L1_ L2 = I eq Z M _ L1_ L2 _ Ieq V M_G Z L2_G Z L1_G V M _ G = I eq Z M _ G _ Ieq V M _ L1_ G = I eq Z M _ L1_ G _ Ieq I eq One current source and three mutual impedance 200

101 Minimum Coupling Model Total 9 equivalent mutual coupling impedances 1 equivalent voltage source 1 equivalent current source L 1 L 2 G Z L1_Veq Z L2_Veq Z G_Veq V M_L1_L2 G L 1 L 2 Z V0_L1 Z V0_L2 V M_L1_G V eq I eq V M_G Z L2_G Z L1_G Z L1_V0 Z L2_V0 Z G_V0 L 1 L 2 G Physical meaning C G_Veq C G_Veq L 1 L 2 Z V0_L1 Z V0_L2 V eq C G_Veq Hot switching node to earth equivalent capacitor 202

102 Physical meaning C L1_Veq & C L2_Veq C L1_Veq C L2_Veq C L1_Veq &C L2_Veq L 1 L 2 Z V0_L1 Z V0_L2 V eq Hot switching node to input traces represented by equivalent capacitors Physical meaning C L1_V0 & C L2_V0 L 1 L 2 C L1_V0 C L2_V0 Z V0_L1 Z V0_L2 V eq C L1_V0 &C L2_V0 0V node to input traces have equivalent capacitors Absorbed in the input impedance

103 Physical meaning C G_V0 L 1 Z V0_L1 C G_V0 L 2 V M_G I eq Z V0_L2 C G_V0 Converter to earth equivalent capacitor Physical meaning Z M_L1_L2_Ieq G V M_L1_L2 L 1 Converter L 2 V M _ L1_ L2 = I eq Z L1_ L2 _ Ieq Z M_L1_L2_Ieq Input lines loop mutual inductive impedance 206

104 Physical meaning Z M_G_Ieq G V M _ G = L 1 L 2 Converter V M_G m i= 1 I Z i M _ G _ i Z M_G_Ieq Earth to converter loop mutual inductive impedance Physical meaning Z M_L1_G_Ieq V M _ L1_ G = I eq Z L 2 L 1 M _ L1_ G _ Ieq V M_L1_G Z M_L1_G_Ieq Cross section of Line 1, line 2 and earth loop mutual inductive impedance 208

105 Overall simulation Line 1 Line 2 L L1 L L3 L boost D 1 C 1 C 2 C PWM controller M 1 R 1 S1 C4 V 0 R o Vp = 400V tr = 150 ns tf = 20 ns D = 0.2 Fs = 150 khz L L2 tr = 150 ns D = 0.2 ns tf = 20 ns fs = 150 khz 0V Realistic parameter - C G_Veq 4 cm 0.1 cm Line 1 L L1 L L3 L boost D 1 S1 V 0 C 1 C 2 C PWM controller M 1 R o Line 2 R 1 C cm L L2 0V 4π C sphere = 1 pf C G_Veq = 0.2 pf r

106 Realistic parameter - C L1_Veq & C L2_Veq t = 0.01 cm Line 1 Line 2 L L1 L L2 C trace L L3 L boost D 1 C 1 C 2 C PWM controller M 1 R 1 S1 C π ( ε r + 1) l 14 _ trace = 10 πs 2ln( ) W + t V 0 0V R o 0.1 cm 5 cm 0.4 cm S1 = 5 cm S2 = 5.4 cm W = 0.1 cm t = 0.01 cm εr = 4 C L1_Veq = pf C L2_Veq = pf Realistic parameter - C L1_V0 & C L2_V0 L L1 Line 1 Line 2 L L2 L L3 L boost D 1 C 1 C 2 C PWM controller M 1 R 1 S1 C4 V 0 0V R o Absorbed in input impedance. 108 uh L L1 0.1 Ω 27 pf 51 uh 0.1 Ω L L2 14 pf 212

107 Realistic parameter - C G_V0 L L1 Line 1 Line 2 L L2 L L3 L boost D 1 C 1 C 2 C PWM controller M 1 R 1 S1 C4 V 0 R o 0V Measured as C G_V0 = 6 pf Realistic parameter - Z M_G_Ieq Line 1 Line 2 L L1 L L2 L L3 L boost D 1 C 1 C 2 C PWM controller Vp = 400V tr = 150 ns 10 Ω tf = 20 ns D = 0.2 Fs = 150 khz M 1 R 1 S1 C4 200 PF 250 nh V 0 0V R o M line r R R 2 r _ loop = u0 ( ) ( ) r = 2 cm R = 1.9 cm The mutual inductance is Z M_G_Ieq =17 nh

108 Realistic parameter - Z M_L1_G_Ieq, Z M_L2_G_Ieq As the input loops are small and having quite a high input impedance, hence for simplicity, Just set Z M_L1_G_Ieq = 0 Z M_L2_G_Ieq = PSPICE Simulation CL1_Veq p CL2_Veq p CL1 27p C4 200p Rc4 10 RL L1 108u 1 V2 = 400 V1 TR = 150n TF = 20n PW = 2u PER = 6.6u 2 Lc4 250n 1 RL2 6 2 L2 51u 1 17nH CL2 14p TX1 2 R1 50 L1 1 50u V V 2 R2 50 L2 50u 1 CG_V0 6p 17nH CG_Veq 0.2p 0 (b) 216

109 Simulation & Measured Results S1 open Line 1 measured result S1 open Line 2 measured result S1 open Line 1 simulated result S1 open Line 2 simulated result S1 open Simulation & Measured Results S1 closed Line 1 measured result S1 close Line 2 measured result S1 close Line 1 simulated result S1 close Line 2 simulated result S1 close 218

110 S1 open Capacitive coupling only Resonate behavior between the input equivalent impedances Line 1 measured result S1 open Line 1 measured result S1 open S1 close Capacitive & inductive coupling Resonate behaviors between the input equivalent impedance Line 1 measured result S1 close Resonate behaviors of the source loop and mutual inductive coupling Line 1 measured result S1 close 220

111 Minimum Coupling Model A minimum of six capacitive mutual coupling impedance is concluded. A minimum of three inductive mutual coupling impedance is concluded. Complicated EMI waveform can be simulated by simple lump element circuit. It is a minimum model, not maximum! Good for diagnosis purpose

112 Shielding Purpose L1 L2 G Don t let any current flow into the Shield or being shielded I lock up the whole world. Ha, ha

113 Shielding wrong idea 1 L1 L2 G Put a metal enclosure to shield noise from going out - Wrong Shielding total recall 1 Large Metal No electric shield Actually, increase coupling 226

114 Shielding wrong idea 1 L1 L2 G Put a earthed enclosure to shield noise from going out - Wrong Shielding total recall 2 L 1 L 2 G Z V0_L1 Z V0_L2 V eq C G_Veq An earthed enclosure actually increases the equivalent hot node to earth capacitance C G_Veq and increases noise

115 Shielding is all about by passing or short circuit 0V By pass all the noise current back to 0V The victim would see some noise current flow in 0V Short to 0V 0V The victim would only see the 0V of the noise source V V is the point L1 L2 0V G The so call shielding should short to noise source 0V to return all the noise coupling to outside

116 0V V is the point - Xformer P-S shielding S-P shielding To secondary equivalent 0V W2 W1 W2 Output To primary equivalent 0V Bobbin W1 To primary equivalent 0V To secondary equivalent 0V V V is the point - MOSFET C Fet_input Heatsink shorted to 0V of primary circuit C heatsink_input C Fet_heatsink Input Main power MosFet Input Shorted to 0V V S1 Shorted to equivalent 0V 0V 232

117 Grounded Case? L1 L2 0V G The shielding to 0V wire is replaced by large capacitor to provide ac short circuit effect. Then the case can be shorted to earth for safety purpose Grounding wire L1 L2 0V G Poor grounding wire or parasitic impedance of the grounding capacitor force the coupling current flow back to G terminal of

118 Grounding choke L1 L2 0V G A ground choke can be used to further increase the impedance of the ground path to avoid noise current flowing in Magnetic shielding? Transformer Copper foil, Ae2=2cm 2 Written by Franki N.K.Poon Input Ae1=1cm 2 r3=1cm R=5cm l=2cm 236

119 Flux band - modeling I input I foil L input L foil L xformer V in_x + V f_x + V x_f + I S1_eq V in_f V f_in V x_in + - I mag = I S1_eq Xformers L input_xformer2 L input_xformer1 L input_foil1 L foil_xformer1 L input_foil2 L foil_xformer 2 Input to I S1_eq T input_foil T foil_xformer T input_xformer 238

120 I 1 I 1 Simulation I input I foil 38 nh 12 nh 17.5 uh L input 2.5 nh + L foil V in_x V f_x V x_f nh V in_f V f_in L xformer 0.1 uh 0.15 uh + + V x_in nh 2.5 nh I S1_eq Voltage (dbuv) Only 6 db reduction! Frequency (Log scale) CISPR22 conductive class B limit line Noise without copper foil shielding Noise with copper foil shielding Inductive couple noise coupled from main tranformer Flux concept Believe it or Not? B leakage1 B leakage1 B leakage2 B leakage

121 Flux concept air core is core u r = u o = 1 I 1 You can t take it away! It exist even in empty space!! Flux concept air core and ferrite core u r = 2000 u r =

122 I 1 I 1 Flux concept as a result B leakage1 B leakage1 B leakage2 B leakage2 Ferrite cannot block or shield the magnetic flux. It only increase flux level inside the ferrite core according to u r Flux concept general coupling Noise Source Short Ring Pick up 244