Reducing Emissions in DC-DC Switched Mode Power Supplies

Scott Mee – Johnson ControlsJim Teune – Gentex

1

Outline

Overview of SMPS designs and basic emissions issues

Root Causes of Emissions

Design Strategies for Reducing Emissions

Schematic Design

Component Selection

Layout Considerations

Trade-offs between EMI and other requirements

Hardware Demonstration

2

Power SuppliesLinear vs. switching

Linear supplies

Typically used when the input and output voltage levels are similar

Large voltage drops and high current output cause low efficiency

Low efficiency = higher heat

Quiet from RF emissions point of view

Switching supplies

Preferred for applications where efficiency is important

Buck step down i.e. 12Volts to 5Volts logic level

Boost step up i.e. 12Volts to 40Volts LED lighting level

Sudden changes in voltage & current cause EMC problems

Circuit uses a switch, inductor and diode to transfer energy from input to output

3

Buck SMPS

Vout = VIN x D, where D = tON/(tON + tOFF)

4

Buck SMPS

Charge phase

5

Buck SMPS

Discharge phase

6

Buck Circuit Voltages and Currents

Iind

Imax

Imin

Switch StateVo

ltage

Vin

Vout

Vind

0 volts

7

Boost SMPS

Vout = VIN / (1 – D), where D = tON/(tON + tOFF)

8

Boost SMPS

Charge phase

9

Boost SMPS

Discharge phase

10

Switch State

Boost Circuit Switching Voltages and Currents

Iind

Idiode

Imax

Imin

Volta

ge L

evel

Timing

Vout

Vin

Vind

0 volts

11

Trapezoidal Periodic Signals

( ) ( )∑∞

=

∠++=1

00 cos2n

nn ctncctx ω( )

rfjn

r

r

nre

n

n

n

n

TAc ττ

τω

τω

τω

τωτ ττω =

⎟⎠⎞

⎜⎝⎛

⎟⎠⎞

⎜⎝⎛

= +− ,

21

21sin

21

21sin

2

0

0

0

00

frTAc τττ

== ,0

Fourier CoefficientsFourier Series

12

To be conservative we might choose a point, 3 times this second breakpoint this is approximately

Bandwidth of Periodic Waveforms

Bounds on frequency spectrum Above the 2nd break point, the harmonics drop off at a rate of -40dB/decade.

rπτ3

rπτ1

HzBWrτ

1=Bandwidth of a periodic signal

rτ1

13

Noise Sources in SMPS

Switching characteristics

dv/dt & di/dt

Fundamental frequency

Harmonic series

Resonances

Step response to the RLC network Ringing

Secondary effects

Power surges at input

Ripple on power bus

Ripple on system wiring

Output ripple

Magnetic fields

14

15

BUCK SupplyEmissions

Investigation

Emissions InvestigationBUCK SMPS Circuit

16

Emissions InvestigationBUCK Voltage Measurements

Switch Output Voltage

Voltage at Input to SMPS

17

1

3

Emissions InvestigationBUCK Voltage Measurements – Zoom

Switch Output Voltage

Voltage at Input to SMPS

18

Time domain

Frequency domain

19

Emissions InvestigationNarrow Band vs. Broadband

Conducted Emissions (150kHz – 2MHz)

Before Techniques Applied After Techniques Applied

Emissions InvestigationSuccess Stories – 70kHz SMPS

20

Improvement came from- Front-end filtering (L/C filter)- Slew rate controls- Layout improvements

Conducted Emissions (150kHz – 2MHz)

Before Techniques Applied After Techniques Applied

Emissions InvestigationSuccess Stories – 150kHz SMPS (Low band)

21

Improvement came from- Front-end filtering (L/C filter)- Layout improvements

CISPR 25 – Radiated Emissions (25MHz – 200MHz)

Before Techniques Applied After Techniques Applied

Emissions InvestigationSuccess Stories – 150kHz SMPS (High band)

22

Improvement came from- Diode snubber- Diode switching changes- Layout improvements

23

BOOST SupplyEmissions

Investigation

Emissions InvestigationBoost Supply Case Study

12volt input & 34volt output

+12 V +34 V

24

Emissions InvestigationBoost Supply – current loop when switch is closed

Red = current flow to load, Blue = return current

+12 V +34 V

25

Emissions InvestigationBoost Supply – current loop when switch is open

Red = current flow to load, Blue = return current

+12 V +34 V

26

Emissions InvestigationBoost Supply Radiated Emissions 50MHz – 180MHz BL ON

123MHz 161MHz

27

Emissions InvestigationBoost Supply Radiated Emissions 50MHz – 180MHz BL OFF

123MHz 161MHz

28

Emissions InvestigationBoost Supply Measurement points

V1

i1+12 V +34 V

29

Emissions InvestigationBoost Supply Measurement Setup

Bench measurement setup overview LeCroy 6GHz 40GS/s

Voltage probe 500MHz 1.8pf

Current probe Langer HF magnetic field probe

30

Emissions InvestigationBoost Supply Measurement Setup

Bench measurement setup overview Voltage probe used to show when switch is open/closed

Current probe used to see shape of current flowing through the diode

31

Emissions InvestigationBoost Supply Voltage Measurement Results

Overview of switching waveforms

V1

i1

32

Emissions InvestigationBoost Supply Voltage Measurement Results

Switch turns from off to on

V1

i1

33

Emissions InvestigationBoost Supply Voltage Measurement Results

Switch turns from on to off

V1

i1

34

Emissions InvestigationBoost Supply Radiated Emissions 50MHz – 180MHz BL ON

V1

i1

123MHz 161MHz

35

Johnson Controls36

Emissions InvestigationBoost Supply Diode Current No changes / Baseline

123MHz ringing corresponds to 123MHz emissions

123MHz

Johnson Controls37

Emissions InvestigationBoost Supply Diode Current 1nf cap across diode CR6401

77 MHz ringing corresponds to 77MHz emissions

77MHz

38

Schematic Design

Schematic DesignBuck topology

12V input

5V output

+12 V+5 V

39

Input filtering Snubber

Soft-start capacitor

SnubberSlew rate control

Spread spectrum

Output filter

Schematic DesignBoost topology

Johnson Controls40

SnubberSlew rate control Output cap

Front end Pi filter

Schematic DesignSnubber Calculations

FRinging 126MHz:=

CSnubber 1500pF:=

FTuned 40MHz:=

CParasitic 168.114pF=

LParasitic 9.491 10 9−× H=

The optimum resistor to damp the overshoot is twice the inductive impedance at the new resonant frequency. This is calculated by the following equation:

RSnubber 2 2π FTuned⋅ LParasitic⋅( )⋅:=

RSnubber 4.771Ω=

CSnubber 1.5 10 9−× F=

Schematic DesignCombination Selection

43

Component Selection

Component BehaviorCapacitors, Resistors, Inductances, Ferrites

All passive components have resistance, capacitance and inductance

Component behavior is different at low and high frequencies

44

Component BehaviorCeramic Capacitors

^Z

Capacitive Inductive (ESL)

f

-20 dB/decade

20 dB/decadeESL - Equivalent Series Inductance (L)

Capacitor has low impedance for a narrow range of frequencies

CLf

lead

res

π21

=

45

Component BehaviorElectrolytic Capacitors – Example 150uf 10V

Power supply output filter

BUCK 5V150uf

What does this mean???

46

Component BehaviorElectrolytic Capacitors – Example 150uf 10V

Capacitance measurement over frequency

HP4284A Precision LCR Meter & 16047D adapter used

47

‐20

0

20

40

60

80

100

120

140

160

180

10 100 1000 10000 100000 1000000

EPN_1214353_SUNCON 150uF 20% 10V

Capacitance (uF)Capacitance (u

F)

Frequency (Hz)

Medium Wave BandLong Wave Band

Component BehaviorElectrolytic Capacitors – Example 150uf 10V

No real capacitance after a few kHz

48

Component BehaviorInductors

f

^Z

Resistive CapacitiveInductive

Rpar

0 dB/decade

-20 dB/decade

20 dB/decade

LRpar

π2 parLCπ21

Inductance resonates with parasitic capacitance between windings of the inductor

Saturation can happen

parLCfres

π21

=

49

Component BehaviorInductors

Resonant Frequency Saturation Curves

10uH goes resonant at 30MHz

10uH has only ~180ohms impedance at 300kHz Our typical

use cases are borderline saturation

50

Low profile is very important!

Shorter package flux lines stay closer to the board lower emissions

Component BehaviorInductors

LQH44 Murata

1.1mm

Vishay IHLP-4040DZ EPCOS B82472P6Vishay IHLP-2020BZ

2 mm 4 mm 4.5mm 8.5mm

EPCOS B82477P4

Too tall!!

51

Component BehaviorResistor

^Z

Resistive Capacitive Inductive

f

R0 dB/decade

-20 dB/decade

20 dB/decade

parRCπ21

parleadCLπ21

Resistors are not purely resistive as frequency increases

parlead CLfres

π21

=

52

Component BehaviorDiodes

I-V graph of a real diode

SchottkeySoft startSlow startFast startEfficiency vs. heat vs. di/dt for emissions

53

Component BehaviorFerrite bead

Do not trust the curves you see in the datasheet!!!

Be sure to understand the circuit where the ferrite will be used

Impedance over frequency graphs change with DC bias

54

Component BehaviorFerrite bead

Take care when choosing a ferrite by it’s rating

Ratings are typically done at 100MHz

3 Devices Having  1000ohms @ 100MHz

55

56

Layout Design

Buck Power SupplyLayout – Component Placement

INDUCTOR

OUTPUT

Diode

Snubber

Controller

INPUT

57

Buck Power SupplyLayout – Copper Definitions

INDUCTOR

OUTPUT

Diode

Snubber

Controller

INPUT

58

GND

PWR

Buck Power SupplyLayout – Switch Closed

59

Start by drawing the path of the current

Ensure area of loop formed by current is kept small

Keep high di/dt components on same side of PCB

Allow common ground between input cap, regulator, diode, snubber and output cap

Keep snubber next to diode

Inductor GND: to fill or not to fill? (efficiency vs. EMC)

Buck Power SupplyLayout – Switch Open

60

Don’t forget there are two switch states!

Boost Power SupplySchematic

Boost Power SupplyComponent Placement

Same as BUCK supply with these additional items:

Keep switch node away from surrounding copper areas

Make switch node as small as possible

Inductor orientation/wiring makes a difference (node connected to winding on inside or outside)

Boost Power SupplyPCB Layout

TOP BOTTOM

Noisy switch nodeGround nodeAll other copper

1.72pF

Keep switch node small

Maintain spacing to surrounding copper areas

Boost Power SupplyPCB Layout

Noisy loopGround loop

dI/dt 2 loops

• L1 = 50nH• L2 = 270nH• K = 0.45 (represents poor coupling between loops; where 1 = perfect coupling)

• Lm = 52nH mutual inductance between loops

Avoid SMPS loop within a GND loop provide continuous ground fill

65

Design Trade-Offs

Recommendations for a Balanced Design

0 spacing ∞ spacing

Thermal Constraints

EMC Constraints

Common Solution

66

Thermal constraints prefer faster switching, larger copper areas and spacing

EMC constraints prefer slow switching, smaller copper and spacing

Recommended Reading

Power Electronics Technology trade magazine (www.powerelectronics.com)

http://www.ridleyengineering.com/

National Semiconductor Application Note 1149, “Layout Guidelines for Switching Power Supplies”.

Texas Instruments Application Report SLPA005, “Reducing Ringing Through PCB Layout Techniques”

Demystifying Switching Power Supplies by Raymond A. Mack

67

Thank you for your attention

Questions?

68

Hardware Demonstration

69

Hardware Demonstration

Hardware Demonstration – Buck SMPS

Base unit (no EMC components) Fully populated PCB