MICAS Department of Electrical Engineering (ESAT)
Design of EMI-Suppressing Power Supply Regulator for Automotive electronics
October 11th, 2006
Junfeng Zhou
Promotor: Prof. Wim Dehaene
KULeuven ESAT-MICAS
MICAS Department of Electrical Engineering (ESAT)
Part I: Introduction
Part II: Low Noise Power Supply–EMI-Suppressing Regulator
• Principles
• Design
• Simulation
• Calculation
• Chip Details
• Measured Results
Part III: Possible Improvements Part IV: Future work
Outline
MICAS Department of Electrical Engineering (ESAT)
Part I: Introduction
Electro-Magnetic Interference (EMI ( EMI ) and radiated emission ) radiated emission have become a major problem for automotive electronics,
Radiated emission is mostly a consequence of di/dt di/dt on the
supply lines.
Although the detailed calculation of EMI noise is rather difficult , we can use the di/dt di/dt as the index, since the current loop
contributes the EMI.
MICAS Department of Electrical Engineering (ESAT)
Part II: EMI-Suppressing Regulator
Previous research on Low Noise Logic Families shows that 2 major problems still remain:
• Static power consumption
• New logic family standard cell library must be
designed and characterised. (large NRE cost, risk)
? Any global approach ?
MICAS Department of Electrical Engineering (ESAT)
Diagram of EMI-Suppressing Regulator
1. Current source 1. Current source
ensures the ensures the major di/dt major di/dt
reduction reduction 2. Slow varying 2. Slow varying
is key to EMC is key to EMC success
success
3. Minimize the Minimize the static current static current
Principles
MICAS Department of Electrical Engineering (ESAT)
EMI-Suppressing Regulator – basic structure
Determine the switching speed, Determine the switching speed,
Hence determine the di/dt Hence determine the di/dt
Energy reservoir
Energy reservoir
MICAS Department of Electrical Engineering (ESAT)
Why new structure ?
1. Simple 1. Simple
2. Driving capability 2. Driving capability 3. Miller effect on 3. Miller effect on
compensation compensation capacitor
capacitor
4. Cascode device 4. Cascode device
MICAS Department of Electrical Engineering (ESAT)
Functionality Simulation &
Comparison with standard CMOS
di/dt and FFT comparison with standard CMOS
w/o EMI-SR and SR
I
VbatIVbat: di/dt p-p = 1.0x107 A/s
w/o EMI-SR and SR, di/dt p-p =1.51x1011 A/s
44dB
PSD comparison
di/dt comparison
MICAS Department of Electrical Engineering (ESAT)
Maple calculation
[ / ] di A s dt
[ / ] di A s dt
[ ]s
An input current step of 1 mA and 100-ps rise time was used for the calculation and simulation
10 ( / )
7di A s
dt
[ ] I A
[ ] I A
[ ]s
[ ]s
di/dt I
stimulus
MICAS Department of Electrical Engineering (ESAT)
Stability analysis
Approximation Approximation : :
1 - * 1
2 - 1
tan
*(1 )
3 - 1
* 1
1 - *( 1- )
1 1
1
2 ( 1 )
* tan p Gota
Caux Av gm Gload
p C k
Gaux Caux p Cgs
Caux Gaux gm Z Caux gm Gaux
Av gm
Gds Gload
p gm Gload Caux
GBW Gm C k
p1
p2 p3
z1
>3 for > 72° phase margin
Stability ~ C
aux/C
tankp4
i i
ininMICAS Department of Electrical Engineering (ESAT)
Stability analysis – Simulation vs.
Calculation
Spectre simulation Spectre simulation
Maple calculation Maple calculation
Raux=1.852K , Caux=20p,Ctank=100p
Stability vs. Iload
φ>72°
φ>72°
Iload =192.7u A
Stability vs. Iload (26.7u A ~ 72m A) Stability vs. Iload (26.7u A ~ 72m A)
φ ≥60°
Worst case
MICAS Department of Electrical Engineering (ESAT)
Current TF analysis
1
2 1
tan
* ( 1 )
3 * 1
p Gm
Caux
p gm
C k
Gaux Cgs Caux
p Caux Cgs
* * 1
1 * * 1 * * 1 1* * * 1* 1 * 1*
* * 1 * * 1 1* * * 1* 1 * 1*
2 * 1*( 1)
Gaux Gm gm
Z Gaux Caux gm Gm Caux gm Cgs Gm Gaux Gaux Gds Cgs Gaux Gds Caux
Gaux Caux gm Gm Caux gm Cgs Gm Gaux Gaux Gds Cgs Gaux Gds Caux
Z Caux Cgs Gaux Gm Gds
H(s)=I
VDD(s)/I
in(s) (i.e. di/dt attenuation)
dominant pole
second pole
third pole
High frequency zero
left half-plane zero
MICAS Department of Electrical Engineering (ESAT)
Current TF- simulation vs.
calculation
Spectre simulation Maple calculation
Iload =80u A, Raux=1.852K , Ctank=100p
( ) ( )
( )
VDD in
I s
H s I s
( ) ( )
( )
VDD in
I s H s I s
dB
- 44 dB - 44 dB - 43 dB
- 43 dB
MICAS Department of Electrical Engineering (ESAT)
Maximum Attenuation
Maple calculation Maple calculation
Iload =800u A, Raux=1.852K , Ctank=100p
TF vs. Caux
1
1 tan
1
lim ( ) ( )
( )
VDD db
s in db k
p Gm
Caux
I s C
H s I s C C
Cut-off freq. ~ 1/Caux Large attenuation
requires Large Ctank and/or small Cdb1
Cascode structure !
dB
- 40 dB - 40 dB
MICAS Department of Electrical Engineering (ESAT)
Caux/Ctank and ∆V DDinput
∆VDDinput Caux = 4, 8, ..20 pF Caux = 4, 8, ..20 pF
∆ ∆ V V
DDinputDDinput~ C ~ C
auxaux/C /C
tanktankMICAS Department of Electrical Engineering (ESAT)
Design Strategy
EMI-Suppressing Regulator design principles EMI-Suppressing Regulator design principles
Stability ~ C Stability ~ C
auxaux/C /C
tanktank
Time domain ∆V Time domain ∆V
DDinput DDinput~ C ~ C
auxaux/C /C
tanktank
More stable also means a larger ∆V More stable also means a larger ∆V
DDinputDDinput
Current TF Current TF
Cut-off freq: Gm/C Cut-off freq: Gm/C
auxaux
Max. attenuation: C Max. attenuation: C
dbdb/(C /(C
dbdb+C +C
tanktank) )
Design for small C Design for small C
dbdb Similar story possible for Similar story possible for Gm, g Gm, g
mm Caution should be exercised to maintain the stability of the Caution should be exercised to maintain the stability of the
EMI-suppressing regulator while designing for higher di/dt reduction EMI-suppressing regulator while designing for higher di/dt reduction
MICAS Department of Electrical Engineering (ESAT)
EMC test chip with
EMI-Suppressing Regulator
SR1, MS-FF, No capa SR2, MS-FF, 1/2 PNMOS capa
SR3, MS-FF, PNMOS capa SR4, MS-FF, PNMOS capa, PWR SR5, MS-FF, PNMOS capa, MIMC capa
SR6, D-FF, No capa SR7, D-FF, 1/2 PNMOS capa
SR8, D-FF, PNMOS capa SR9, D-FF, PNMOS capa, PWR SR10, D-FF, PNMOS capa, MIMC capa
On-chip LDO
PD
On-chip Serial regulator
PD
SR1 RST Din CLK OUT
SR2 RST Din CLK OUT
SR9 RST Din CLK OUT
SR10 RST Din CLK OUT
GND LDO
PD
EMI regulator
Ctank
VDD_input
Emergency block
Power down block
V3v3
Vbat
MICAS Department of Electrical Engineering (ESAT)
Current source simulation results
[ ] I mA
[ ] V v
[ ] V v
I
VbatV3v3
VDD_input Vctrl
[ ] V v
Power down enable
MICAS Department of Electrical Engineering (ESAT)
Frequency simulation results
current of Vbat
di/dt p-p =8.5x104 [A/s]
9x106
load current of digital core
FFT FFT
di/dt p-p =1.8x109 [A/s]
7x103
di/dt of Vbat di/dt
of V3v3
40dB (EMI regulator) + 20dB (Serial regulator 40dB (EMI regulator) + 20dB (Serial regulator) )
MICAS Department of Electrical Engineering (ESAT)
Chip Details
EMI
Suppressing regulator
Area: 1mm x 1.1mm
Ctank =100p F Caux = 20 p F
Power transistors: Wp=5000 μm Lp= 1 μm (fixed) Technology
: AMIS 0.35μm I3T80
Supply voltage
: 12 V
Output voltage
: typ. 8V, min.5.5V
Quiescent current
: 30 μA
MICAS Department of Electrical Engineering (ESAT)
Measurement Setup
MICAS Department of Electrical Engineering (ESAT)
Measured Results (1)
MICAS Department of Electrical Engineering (ESAT)
Measured Results (2)
Load transient response
I
load(0 mA ~ 20 mA ) V
out10 ns
8 V
5.78 V
MICAS Department of Electrical Engineering (ESAT)
Measured Results (3)
Peak : ~ 5x
T
rise: ~ 7x
di/dt
peak: ~ 24x
MICAS Department of Electrical Engineering (ESAT)
Measured Results (4)
Peak : ~ 9x
T
rise: ~ 12x
di/dt
peak: ~ 18x
MICAS Department of Electrical Engineering (ESAT)
Measured Results (5)
di/dt TF di/dt TF
I I
AC ACinjected injected
-3dB: 1.6 MHz
-33 dB
di/dt TF di/dt TF
I I
AC ACinjected injected
-3dB: 1.8 MHz
-35 dB
MICAS Department of Electrical Engineering (ESAT)
Conclusions
A Low Noise Power Supply Techniques A Low Noise Power Supply Techniques is presented: is presented:
Control the way Control the way the current delivered to the internal digital core, the current delivered to the internal digital core, hence keep the EMI under control,
hence keep the EMI under control,
Comparable reduction on di/dt noise with low noise digital cells Comparable reduction on di/dt noise with low noise digital cells only,
only,
More power efficient More power efficient than the low noise digital cells, than the low noise digital cells,
Have similar power consumption Have similar power consumption to the conventional CMOS logic, to the conventional CMOS logic,
A global approach- A global approach- Can be adjusted to a wide range of chip size Can be adjusted to a wide range of chip size and power consumption level,
and power consumption level,
Measurement results match the simulation well. Measurement results match the simulation well.
MICAS Department of Electrical Engineering (ESAT)
Part III: Possible Improvements
Frequency
H(s)-dB
peakingMICAS Department of Electrical Engineering (ESAT)
•
z1– G
m/C
aux•
z2– parasitic zero, high frequency
•
p1– Pole at V
ctrl: G
m/C
aux•
p2– Pole at V
VDD_input: g
m/C
tank•
p3– pole caused by compensation path, high frequency
Current TF: small signal model
z1
z2
VDD
noise p1 p2 p3
1 + sω1 + sω I (s)
I (s) 1 + sω1 + sω1 + sω
MICAS Department of Electrical Engineering (ESAT)
• z1 cancel off the p1
• Make the p2 cut-off frequency
• This zero is intrinsic for this feedback topology
sacrifices dynamic noise performance
• Make Make p2p2 dominant dominant
• Advanced compensation techniques Advanced compensation techniques needed
needed
Current TF: pole-zero tracking
Frequency H(s)-dB
z1
p1
p2peaking
Options
Options
MICAS Department of Electrical Engineering (ESAT)
• Key Idea :
Achieving Stability Without Sacrificing Dynamic Supply Current Rejection
• R-C compensation
• Reduced Gm of OTA
• R
eqadded for moving the output pole high frequency, also for improvement of the dynamic di/dt rejection
Possible solution
Vref -
+
RLoad
Ctank Vctrl
VDD_input Caux
Mp
Raux OTA
“ Req”
+ -
Vbias
VDD_input
MICAS Department of Electrical Engineering (ESAT)
Current TF analysis
•
p1–
•
p2–
•
z1–
•
z2–
high frequency
z1 z2
VDD
noise p1 p2
1 + sω1 + sω I (s)
I (s) 1 + sω1 + sω
( )
m m
aux m m aux eq
g G
C g G R G
tan
m m aux eq
k
g G R G C
( )
m m
aux m m aux
g G
C g G R
1
m m aux
O db
g G R
r C
R
eqmakes
p1and
z1well separated
MICAS Department of Electrical Engineering (ESAT)
Current TF--Simulation results
Ctank: 100p Iload:
20uA~2m A Gm : 4uA/V RC: 1K
Caux: 20p F Raux: 100 K 632.5 uA
20 uA
63.25 uA 200 uA
TF vs. I
Load2 mA
-3dB
Peak
Freq
-3dBMICAS Department of Electrical Engineering (ESAT)
Part IV: Future Work
Extend the TF measurements into higher frequency,
Find a way to linearly inject the AC current,
Characterization & quantification of EME from digital circuits,
Prediction of EME of digital circuits,
Spread spectrum clock generation.
MICAS Department of Electrical Engineering (ESAT)