Practical Design Considerations for the Reduction of Conducted EMI; Part 2

Practical Design Considerations For The Reduction Of Conducted EMI; Part 2

Chris Swartz

1 Agenda

• Welcome and thank you for attending. Today I hope I can provide a overall better understanding of the practical design aspects of managing conducted EMI in power systems. The topics we will cover in Part 2 of our two part series will be: • The determination of input capacitance for our Buck Converter model continued from Part I of the series. • Measurement of insertion loss will be explained using circuit simulation. • How to determine if a given input filter network is stable with a particular DC-DC Converter. • Passive Differential and Common Mode Filter Schemes • The Picor Active EMI Filter Topology • Case histories and troubleshooting topics

2 Differential Mode Noise Ideal Buck Converter Review

• Consider the ideal buck converter shown below: This ripple current is responsible for the differential mode EMI that will be measured at the LISN.

We are concerned about these ripple currents, as they are responsible for the ripple voltage measured at the node Vin. This ripple voltage is the source for the differential mode EMI that occurs at the fundamental as well as the much of the harmonic content.

3 Input Capacitor Selection For Our Ideal Buck Converter

• The typical Buck regulator will usually require two different types of input capacitors. One type must carry all of the high frequency switching ripple current and the second type must supply input voltage hold up time in the event an input inductor is used to isolate the high frequency differential current from the source. • X5R or X7R ceramic capacitors will be used to carry the high frequency ripple current because they have the lowest ESR and ESL. In order to reduce the switching ripple voltage to an acceptable level, multiple ceramic capacitors will be used in parallel to reduce the overall ESR and ESL even further. The equations that follow are general in nature and do not include ESR and ESL when calculating the actual capacitance value required. For designs that have very high ripple current with very fast transients, a more rigorous study may be required. • The input bulk capacitor is required to prevent the input voltage from sagging below the converter under voltage lockout during a load transient. A low ESR Oscon or Aluminum Electrolytic capacitor will be required if an input inductor of a significant value is used or if the DC-DC requires EMI scans using a LISN. 4 Input Capacitor Selection - Ideal Buck Converter

• 10V to 15V input, 1.2V output at 15A max

• Normal output is 7.5A with pulse load to 15A 20ms after start up RSH=100k DCR=5m Vin LIN • Lout is 1uH IC=0

• Switching frequency is 300kHz C1 Q1 3m DCR=1m 1n VBAT 1x LOUT1 ESR=1m IC=0 12

Q2 3m 1n 1x

VOUT

RSRC=160m LOAD 1x 1x DELAY=20m COUTX COUTY RISE=10u 5.6m 400u ISTART=0 IC=0 IC=0 IFINAL=7.5 ESR=1.5m ESR=150u

5 Input Capacitor Selection - Ideal Buck Converter

Vinmin 10 Vout 1.2 .9

RSH=100k DCR=5m Fswitch 300000 Ioutmax 15 Vripplepp .075 Vin LIN IC=0

C1 V out Q1 3m DCR=1m Don 1n VBAT Vin 1x LOUT1 ESR=1m min IC=0 12

Q2 D 0.133 3m on 1n 1x

Ioutmax Don 1 Don Cin VOUT ceramic F Vripple switch pp RSRC=160m LOAD 1x 1x DELAY=20m COUTX COUTY The calculation determines the RISE=10u 5.6m 400u ISTART=0 minimum input ceramic IC=0 IC=0 5 IFINAL=7.5 Cinceramic 7.704 10 capacitance value for a ESR=1.5m ESR=150u continuous mode Buck regulator. We will use 80uF total. 6 Input Capacitor Selection - Ideal Buck Converter

I outmax RSH=100k CinRMS Vout Vinmin Vout DCR=5m Vinmin Vin LIN IC=0

Cin 4.874 Q1 RMS 3m DCR=1m 1n VBAT 1 1x LOUT1 ESR=1m IC=0 12 Vripplepp 2 3 1u Cbulk RMS Q2 Cbulk This equation is fairly accurate so 3m ESR 1n long as the inductor ripple 1x current is not a significant percentage of the total output current. Stated another way, the VOUT higher the ripple current, the less RSRC=160m LOAD 1x 1x accurate the equation is. DELAY=20m COUTX COUTY RISE=10u 5.6m 400u ISTART=0 IC=0 IC=0 IFINAL=7.5 ESR=1.5m ESR=150u

7 Input Capacitor Selection - Ideal Buck Converter

RSH=100k DCR=5m Vin LIN IC=0

Q1 3m DCR=1m 1n VBAT 1x LOUT1 ESR=1m IC=0 12

Q2 3m 1n 1x

VOUT Input ripple voltage measures RSRC=160m LOAD 70mV p/p versus our target of 1x 1x DELAY=20m COUTX COUTY 75mV p/p at the 15A load RISE=10u 5.6m 400u ISTART=0 condition at low line. IC=0 IC=0 IFINAL=7.5 Simulated RMS ripple current is ESR=1.5m ESR=150u within 2.9% of the calculated value at 5.019A 8 Input Bulk Capacitor Selection - Ideal Buck Converter RSH=100k DCR=5m Vin LIN IC=0

• Since we have an input inductor, C1 is isolated from Vin from 1u

a transient standpoint. Q1 3m DCR=1m 1n VBAT 1x LOUT1 ESR=1m • C1 would need to supply nearly all of the entire average IC=0 12 1u

Q2 current until the input inductor current equals the converter 3m 1n 1x average current.

• For our converter, this should be no problem since LIN is VOUT RSRC=160m LOAD 1x 1x small. At 10V input, a 50% load step is an increase of input DELAY=20m COUTX COUTY RISE=10u 5.6m 400u ISTART=0 IC=0 IC=0 IFINAL=7.5 current of 1A average. So, Vin will change by: ESR=1.5m ESR=150u

Lfilt Vin 1.1 Itr Cin

Vin 0.123 9 Input Bulk Capacitor Selection - Ideal Buck Converter

• If this converter were connected to a 50 Ohm LISN, the input

RSH=100k DCR=5m inductance would be 100uH (50uH in series with each lead). Vin LIN IC=0

In that case, our input Vin would drop 1.23V for the same 1u Cbulk C1 load transient. If the UV lockout had 1V of hysteresis, the Q1 3m DCR=1m 1n VBAT 1x LOUT1 ESR=1m converter could shut down at low line due to the step. IC=0 10 1u

Q2 C .000080 L .000101 3m in filt 1n 1x

V 0.5 Itr 1.0 VOUT

RSRC=160m LOAD 2 1x 1x 1.21 I L DELAY=20m tr filt 4 To operate with a LISN and our COUTX COUTY RISE=10u C C 4.888 10 5.6m 400u total total 1uH inductor with a 0.5V change ISTART=0 2 IC=0 IC=0 IFINAL=7.5 V in Vin to remain in our UV ESR=1.5m ESR=150u hysteresis window, an additional 410uF for bulk storage is needed. C C C 4 bulk total in C 4.088 10 bulk 10 Input Bulk Capacitor Selection - Ideal Buck Converter

• The RMS ripple current in Cbulk is:

RSH=100k DCR=5m Vin LIN IC=0

Cbulk .035 1u ESR Cbulk C1

Q1 3m DCR=1m 1n VBAT 1 1x LOUT1 ESR=1m Vripplepp IC=0 10 2 3 1u CbulkRMS Cbulk Q2 ESR 3m 1n 1x

VOUT

RSRC=160m Cbulk 0.619 LOAD RMS 1x 1x DELAY=20m COUTX COUTY RISE=10u 5.6m 400u ISTART=0 IC=0 IC=0 IFINAL=7.5 ESR=1.5m ESR=150u

11 Measuring Insertion Loss Of A Filter

• Consider the EMI filter below. It is both a common mode and differential mode filter. • First we will look at the differential mode measurement…………… Connect high bandwidth AC current probes to your network analyzer. We are intending to measure the relative loss

Use a signal injection isolation capacitor to keep DC The filter must be DC biased so that the internal bias off the network analyzer. Use a 50 Ohm termination capacitors assume the value with bias applied. Ceramic resistor as shown. capacitors capacitance value can change with DC bias. 12 Measuring Insertion Loss Of A Filter

• Next, we will look at how to measure the common mode portion of the filter…………

13 Measuring Insertion Loss Of A Filter

• Let’s measure our LC filter from the Buck Regulator example shown earlier…………

V_I_IN

1K R4

H1 Filter Output Impedance 1

V_I_OUT Input Impedance 1u

1 L2

Attenuation 50 R2 C1 H2 80u IC=0 The output impedance of the AC 1 V1 input filter shows very high 1K R3 peaking at the resonant 1p R1 Source Impedance frequency of the input filter. This filter requires damping to prevent this problem.

14 Measuring Insertion Loss Of A Filter

• There are several ways to damp an input filter. A robust method is to add a series R-C in parallel with the ceramic capacitors. The idea is that at the resonant frequency of the input filter, the series R of the parallel combination is dominant.

V_I_IN

1K Cblock 4 Cin 4 R4 Cblock 3.2 10 H1 1 L filt V_I_OUT R R 0.112 1u damping C damping in 1 L2 111m Rdamping 50 R2 C1 H2 80u IC=0

Cblock 320u IC=0 AC 1 V1 1K R3 1p R1

15 Measuring Insertion Loss Of A Filter

• The damping network The damping network reduces the impedance peaking very significantly and should improve transient response of the filter

V_I_IN

1K R4

H1 1

V_I_OUT 1u Filter Output Impedance 1 L2 111m Rdamping 50 R2 C1 H2 80u IC=0

Cblock 320u IC=0 AC 1 V1 1K R3 1p R1

Attenuation

16 Measuring Insertion Loss Of A Filter

• The Ideal Buck Model input voltage before damping and after damping transient response

Un-damped Damped 17 Avoiding Input Filter Instability – Using A Case History

• A closed loop DC-DC converter exhibits a “negative input impedance”. This means that due to the internal feedback loop and voltage feed forward circuitry, as the input voltage goes up, the input current goes down. • A DC-DC converters input impedance is lowest at low line and highest at high line. The input impedance can

be approximated by (for our Ideal Buck Regulator) Vinmin RINmin RINmin 5 Vout Ioutmax

Vinmin

• This means that if the output impedance of the input filter is 5 Ohms, we now have created a negative resistance oscillator. • The onset of this instability is manifested by a sinusoidal input current and voltage oscillation that will perturb the output and create all sorts of havoc, including overshoots and possible damage. 18

Avoiding Input Filter Instability – Using A Case History

• A military customer of ours designed in our ZVS Buck Regulator (PI3302) and our MQPI-18 EMI Filter module • The filter module is ultra small and can handle 7A. It provides both common mode and differential attenuation. The customer did not need the common mode section but chose the filter because of it’s size and high frequency performance. The ZVS Buck regulator switches at 1 MHz

MIL-STD-461E LISN

50u U1 VIN VS1 PI33XX L7 PGD L1 ZVS BUCK REGULATOR SYNCO VOUT C9 C10 8u 250n U2 C13 C12 C6 C11 SYNCI BUS_P QPI_P GRM31CR71H475KA12 SDA REM C1 C2 C3 C5 28 5 50 47u 47u 47u 47u MQPI-18 X4 V1 R18 R20 SCL

ADR0 ADJ BUS_M QPI_M ADR1 5 50 SHIELD R19 R25 EN EAO C15 100u TRK

50m Ohm ESR PGND SGND C8 C14 8u 250n

50u

19 Avoiding Input Filter Instability – Using A Case History

• First, the customer measured the raw EMI signature without any filter and saw very high EMI as expected.

MIL-STD-461E LISN

50u U1 VIN VS1 PI33XX L7 PGD L1 ZVS BUCK REGULATOR SYNCO VOUT C9 C10 8u 250n C13 C12 C6 C11 SYNCI GRM31CR71H475KA12 SDA REM C1 C2 C3 C5 28 5 50 X4 47u 47u 47u 47u V1 R18 R20 SCL

ADR0 ADJ

ADR1 5 50 R19 R25 EN EAO

TRK

PGND SGND C8 C14 8u 250n

50u

20 Avoiding Input Filter Instability – Using A Case History

• Next, the customer installed the MQPI-18 filter measured the EMI signature again. He was thrilled until…………

MIL-STD-461E LISN

ADR0 ADJ BUS_M QPI_M ADR1 5 50 SHIELD R19 R25 EN EAO C15 100u TRK

50m Ohm ESR PGND SGND C8 C14 8u 250n

50u L8 21 Avoiding Input Filter Instability – Using A Case History

• He started to vary the line voltage. As he got to about 9.8V, the EMI plot went horrific. That’s when my phone started ringing, I think.

MIL-STD-461E LISN

ADR0 ADJ BUS_M QPI_M ADR1 5 50 SHIELD R19 R25 EN EAO C15 100u TRK

50m Ohm ESR PGND SGND C8 C14 8u 250n

50u L8 22 Avoiding Input Filter Instability – Using A Case History

• A current probe was connected to the LISN output cables. It revealed a very rich high current 2kHz sine wave. The input voltage to the converter was ringing below the UV lockout, causing the converter to turn on and off every 30ms. This resulted in the poor EMI plot, as the converter was oscillating on and off.

MIL-STD-461E LISN

ADR0 ADJ BUS_M QPI_M ADR1 5 50 SHIELD R19 R25 EN EAO C15 100u TRK

50m Ohm ESR PGND SGND C8 C14 8u 250n

50u

23 Avoiding Input Filter Instability – Using A Case History Impedance - Ohms

U1 VS1 VIN PI33XX L1 PGD ZVS BUCK REGULATOR VOUT SYNCO Converter Input Impedance

SYNCI

C5 C3 C2 C1 REM SDA 47u 47u 47u 47u SCL

ADJ ADR0

ADR1

EAO EN

TRK SGND PGND Filter Output Impedance

MIL-STD-461E LISN

50u

C9 C10 8u 250n U2 C13 C12 C6 C11 BUS_P QPI_P GRM31CR71H475KA12

28 5 50 MQPI-18 X4 V1 R18 R20

BUS_M QPI_M

5 50 SHIELD R19 R25 C15 100u

50m Ohm ESR C8 C14 8u 250n

50u

L8 2 kHz Oscillation Root Cause 24 Avoiding Input Filter Instability – Using A Case History

• The root cause of the instability was the LISN resonating with the ceramic capacitors inside the MQPI-18 filter and the PI3302 input capacitors. The resonant frequency of 2kHz was responsible for the input filter instability at low line. Adding a series R-C in parallel with the entry port provided the necessary damping of the LISN and eliminated the instability. Taking a page out of the late great Johnnie Cochrane’s book, “You must design with the LISN in mind!”

MIL-STD-461E LISN

50m Ohm ESR PGND SGND C8 C14 8u 250n

50u

L8 25 Passive Discrete EMI Filter Example

Leakage2 DMDM11 Leakage1 CM2 CM1

CM2 CM1

""XX"" "X" "X"

P1 P1 Output P1 "X"

Output S1 "X"

S1 S1

"Y" "Y" "Y" "Y"

26 9 Amp Passive Filter Example Schematic

27 The Picor QPI-21 Active EMI Filter Topology Simplified Block Diagram Common Mode Current Sense Transformer

LISN “X” Capacitors Pi Arrangement

High Speed, Wide Bandwidth Differential Amplifier With High Current Driver Injection Capacitor

Differential Inductor Filter

“Y” Capacitors Floating Bias Regulator 28 The Picor QPI-21 Active EMI Filter Topology Simplified Block Diagram The floating bias supply provides bias LISN power for the high speed OPAMP, so it can either AC source or sink high current as required to The differential reduce the inductor common mode provides noise. attenuation of “Y” capacitors the differential provide common mode ripple mode attenuation current. and a required return path for the current sensing transformer. 29 The Picor QPI-21 Active EMI Filter Topology Simplified Block Diagram The common mode inductor with current LISN sense winding can be very small (2uH) made with AC single turn windings and a multiple turn current sense winding. It The injection separates DM capacitor from CM and A stable DC bias blocks the DC applies point is provided by bias from the common mode a separate DC differential current sense amplifier (not amplifier from to the shown) which connecting to differential compensates for ground. amplifier. the DCR loss in the 30 circuit. The Picor QPI-21 Active EMI Filter Topology Simplified Operation

LISN

31 The Picor QPI-21 Active EMI Filter Topology Simplified Operation

LISN

32 The Picor QPI-21 Active EMI Filter Topology Simplified Operation

LISN 20mA p/p

AC 1V p/p

1mV/div 250kHz 50mV/div

33 QPI-21 EMI Performance With And Without Active Loop

Active amplifier loop disabled

Active amplifier loop enabled

34 9 Amp Common Mode Filter Footprint

PCB area for a 9 Amp common mode inductor is longer and wider than the 14A QPI-21 filter, which contains both differential and common mode circuitry. In addition, the 9A common mode inductor is twice as high as the QPI-21.

35 Active Filter Example

QPI-21 • 2.3W dissipation round trip @ 14A! Passive Solution • Much smaller loop area, lower susceptibility • EMI performance is less dependent on layout • 3.5W dissipation round trip @ 9A and magnetic components parasitics • Large loop area • Critical layout is inside SiP and already done • Layout and magnetics quality are critical to high • No derating until 65 degree ambient @ 14A frequency performance • EMI performance equivalent to two stage passive • Multiple winding chokes have higher parasitic solution capacitance, which tends to limit attenuation • Many more “Y” capacitors are required • Solution grows significantly for a higher current 36

Side By Side On The Same PCB!

37 What To Do When Things Don’t Go Right With EMI

• Remember my golden rule: Theory and practice MUST match. This will help your thought process when you get frustrated. • Buy some good diagnostic tools like noise separation filters, a near field probe (these can be made fairly easily), an old analog scope (perfect for EMI) • Don’t be afraid to experiment. • Call Picor • The next few slides will explain why troubleshooting EMI can be fun and challenging, despite what your manager thinks!

38 What To Do When Things Don’t Go Right With EMI

• We were asked to test our QPI-21 Active Filter with a certain 200W power supply for a top customer to design into his high end system. The full load performance was outstanding.

39 What To Do When Things Don’t Go Right With EMI

• After the customer designed our filter into his system, he sent me the system stating he measured an EMI plot like that shown below at very light load. It is failing Class A by almost 20 dB!!

40 What To Do When Things Don’t Go Right With EMI

Adding current probe here Noise separator indicated Ripple voltage was very showed virtually no noise DM noise, removing LISN low here, did not match current leaving QPI-21. ground had no effect EMI noise level

Adding load did not change measured EMI at all

High DM noise measured at LISN input, synchronized to magnetic field at DC- DC using near field probe 41 What To Do When Things Don’t Go Right With EMI

Power Input

42 What To Do When Things Don’t Go Right With EMI

Moved Location Result

43 What To Do When Things Don’t Go Right With EMI Final Thoughts

• Always try to make sure that the EMI filter is the closest component to the power entry ports. It is very easy to create a sneak path due to stray magnetic fields that will bypass the filter. • Trust your measurements. EMI proficiency is not magic. Most conducted noise problems are measureable and traceable to a source. • Try to design for EMI compliance up front. It becomes very difficult to move components around on a dense layout.

44 In Summary………….

• A method was presented to design the correct amount of input capacitance for a typical Buck regulator and conf the results were confirmed using circuit simulation methods. We discussed the measurement of insertion loss and filter impedances. A method to damp the input filter of a Buck regulator was also discussed. • A case history was presented to illustrate the problem of input filter stability and how to correct it. • A passive common mode and differential mode filter was presented. • The Picor QPI-21 Active Filter Topology was presented with circuit simulation and theory of operation. • Finally, a case history illustrating how to troubleshoot an EMI problem was discussed.

45 Acknowledgements

• The following material was used as references in the presentation of this seminar: • EE Times 9/7/2007; Choosing the right input caps for your buck converter; Chris Cooper, Avnet Electronics • Fundamentals Of Power Electronics Second Edition; R. W. Erickson/Dragon Maksimovic