Beruflich Dokumente
Kultur Dokumente
Acknowledgements
Renato Rimolo-Donadio, Xiaomin Duan,
Sebastian Mller, Miroslav Kotzev,
Heinz-Dietrich Brns
Christian Schuster 2
Abstract
This presentation will give an introduction to the fundamentals of
signal and power integrity engineering for high-speed digital
systems with a focus on packaging aspects. The presentation is
intended for an audience that has little or no formal training in
electromagnetic theory and microwave engineering.
Topics that will be addressed include lumped discontinuities,
transmission line effects, crosstalk, bypassing and decoupling, via
and power plane effects, return current issues, and measurement
techniques for Gbps links.
More information on current research projects at the Institute of
Electromagnetic Theory can be found at:
http://www.tet.tuhh.de/
Christian Schuster 3
Via
PCB
Power Plane
Ground Plane
DC Power Supply
Receiver
Christian Schuster 4
Christian Schuster 11
Outline
(1)
(2)
Signal Integrity
(3)
Power Integrity
(4)
(5)
Measurement Techniques
(6)
Wrapping Up
Christian Schuster 12
(1)
Christian Schuster 13
Christian Schuster 14
Downtown
Founded 1978
Christian Schuster 15
Christian Schuster 16
What We Do at TUHH
B 0
B
E
t
D
H J
t
Maxwells Equations
Printed circuit board layout
Christian Schuster 17
(2)
Signal Integrity
Christian Schuster 18
Christian Schuster 19
Cable
Daughtercard
Housing / Chassis
IC (Transmitter)
IC (Receiver)
Package / Module
Socket
Connector
Connector
Backplane / Motherboard
Christian Schuster 20
Interconnect
(Link)
Connector
Christian Schuster 21
Effect of Interconnects
The ideal interconnect will simply delay the signal:
Tx
Rx
t
Rx
Tx
Christian Schuster 22
Effect of Interconnects
The deviations in timing and amplitude are in general called:
Christian Schuster 23
JITTER
Signal Bandwidth
f max
s(t )
1
0.3 .. 0.5
TR rise time
Maximum Frequency
s / 2
TB
TR
1
0.5
f0
Fundamental Frequency
Christian Schuster 24
2. Manage discontinuities
3. Reduce Coupling
4. Limit attenuation
5. Equalize signals
Christian Schuster 25
Effect of Terminations
Lets use the following interconnect (link) model:
Z 0, , l
ZS
u0
Transmitter
ZL
u1
u2
Interconnect
Christian Schuster 26
Receiver
??
Z0
87
5.98 h
ln
r 1.41 0.8 w t
(h = height of dielectric,
w = conductor width,
t = conductor thickness)
Stripline
(symmetric)
Z0
60
1.9 h
ln
r
0.8 w t
(h = height of dielectric,
w = conductor width,
t = conductor thickness)
Metal
Christian Schuster 27
Dielectric
Operating frequency
5 GHz
Corrsponding wavelength
3 cm
Delay 5 ns
up to 25 wavelengths on a trace!
Printed circuit board layout
Christian Schuster 28
Effect of Terminations
Lets use the following interconnect (link) model:
Z 0, , l
ZS
u0
ZL
u1
u2
u2
const. and max. !
u0
Christian Schuster 29
??
Effect of Terminations
Z0
a
ZS Z0
Z 0, , l
ZS
ZL
input acceptance
Christian Schuster 30
Effect of Terminations
H exp( l )
Z 0, , l
ZS
ZL
input acceptance
TL transfer function
Christian Schuster 31
Effect of Terminations
ZL Z0
rL
ZL Z0
Z 0, , l
ZS
t L 1 r L
ZL
input acceptance
TL transfer function
load transmission
load reflection
Christian Schuster 32
Effect of Terminations
ZS Z0
rS
ZS Z0
tS 1 rS
Z 0, , l
ZS
ZL
input acceptance
TL transfer function
source transmission
load transmission
source reflection
load reflection
Christian Schuster 33
Effect of Terminations
Z 0, , l
ZS
ZL
u2
a H tL
??
2
u0
1 H r L rS
Christian Schuster 34
Effect of Terminations
Z 0, , l
ZL Z0
ZS
u2
aH
u0
ZL
ZS Z L Z0
Christian Schuster 35
u2 1
H
u0 2
Effect of Terminations
Matched interconnect:
Voltage
Time
TD
Mismatched Interconnect:
Voltage
Time
2 TD
Christian Schuster 36
Effect of Terminations
1 Z S 10 , Z 0 50 , Z L 1k
zero losses
2 Z S 50 , Z 0 50 , Z L 100
zero losses
3 Z S 50 , Z 0 50 , Z L 50
zero losses
4 Z S 100 , Z 0 50 , Z L 100
3
zero losses
5 Z S 10 , Z 0 50 , Z L 1k
non-zero losses
6 Z S 50 , Z 0 50 , Z L 50
non-zero losses
(all lines have a delay of 0.1 ns)
Christian Schuster 37
Matching Terminations
(2 TD TR )
!
Christian Schuster 38
Data
Equalizer
+ Slicer
.
.
Data
Equalizer
Serializer
Tx
CDR
Deserializer
Clock
Rx
Christian Schuster 39
Interconnect
(Link)
Connector
Christian Schuster 40
Signal
Out
Signal
In
Source
Voltage
u1
50
Tx-Output
2.5
nH
Bond Wire
Christian Schuster 41
50
Rx-Input
u2
Received
Voltage
Source
Voltage
u1
50
Tx-Output
50
1 pF
Via
Christian Schuster 42
Rx-Input
u2
Received
Voltage
u2(t) / u1(t)
Magnitude of u2 / u1
Frequency Response
t 1/w0 = 25 ps
f0 6.37 GHz
Frequency [GHz]
Time [ps]
Christian Schuster 43
Z, ,l
Z0
c
f
2.952GHz
4l
Frequency Response
(Scattering Parameters)
Christian Schuster 44
Port1
Port2
2nH
2nH
2nH
P=1cm
P=15cm
P=5cm
P=1cm
Christian Schuster 45
300fF
Z=48
300fF
Z=52
300fF
Z=48
300fF
Z=49
Managing Discontinuities
Avoid them!
!
Christian Schuster 46
Interconnect
(Link)
Connector
Christian Schuster 47
Effect of Coupling
Consider two transmission lines in close proximity:
(1) Input
Christian Schuster 48
(2) Output
Effect of Coupling
Consider two transmission lines in close proximity:
(1) Input
(2) Output
IC
UL
NEXT =
Near End Crosstalk
FEXT =
Far End Crosstalk
Christian Schuster 49
Effect of Coupling
For weak coupling (kL,C 0.25) it is found approximatively:
(1) Input
(2) Output
Effect of crosstalk
is usually small.
INPUT
U max
TR
TD
TR
2 TD TR
Christian Schuster 50
TD TD TR
Effect of Coupling
For weak coupling (kL,C 0.25) it is found approximatively:
NEXT
U max
FEXT
max
kC k L TD
INPUT
U
max
2
T
R
INPUT
kC k L U max
(TD 0.5 TR )
(TD 0.5 TR )
kC kL TD
INPUT
U max
2
TR
It should be noted that these formulas do not take into account losses
on the lines or reflections from load mismatches.
Christian Schuster 51
diameter = d
C12
ln(1 (2h / a) 2 )
kC
C12
C11
2 ln( 4h / d )
Christian Schuster 52
Reducing Coupling
Christian Schuster 53
Differential Signaling
In a SINGLE-ENDED link
there is a common
(global) reference against
which the signal is
measured ("ground").
In a DIFFERENTIAL link
the reference is the
negative of the signal
itself (which has to be
transmitted as well).
Christian Schuster 54
(3)
Power Integrity
Christian Schuster 55
Christian Schuster 56
ZPDN
IC #2
U0
Christian Schuster 57
uIC
iGate1, iGate2,
Du
uIC = U0 - Du
d
Du (t ) R iGate1 (t ) iGate1 (t ) ... L iGate1 (t ) iGate1 (t ) ...
dt
"DC-drop or IR-drop"
Christian Schuster 58
"DI-drop or DI-noise"
2. Add decoupling
3. Add even more decoupling
4. Use several power supplies
5. Use on-chip VRMs
Christian Schuster 59
PDN Elements
High Power
DC Supply
Discrete
Decoupling
Capacitors
(various sizes)
IC incl.
Power/Ground Grid
& Integrated Decaps
Voltage
Regulator
Module
Christian Schuster 60
Package incl.
Power/Ground
Planes
PDN Impedance
In frequency domain the standard PDN model looks like this:
ZPDN ( f )
u0
Du( f ) Dumax
Du( f )
ZIC ( f )
Z PDN ( f ) Z Target
Christian Schuster 61
PDN Impedance
A typical maximum ripple for ditigal systems is:
Dumax
maximum ripple 5% to 10%
u0
With a 10% value the following numbers can be obtained for
applications of the early 1990'ies: of 2000 and on:
u0 1.2 V
u0 5.0 V
iavg 120 A
iavg 1 A
u0 / iavg 0.01
u0 / iavg 5.0
Pavg 144 W
Pavg 5 W
Z Target 0.001 = 1 m !
Z Target 0.5
Christian Schuster 62
PDN Impedance
Is 1 m hard to achieve? How about 10 m? Let's see
Example:
The PDN consists of a simple copper wire of 2 mm radius in
theform of a flat rectangle with side lengths of 5 cm and 1 cm,
respectively.
with
Z PDN R 2 (wL) 2
R 0.7 m
L 40 nH
Christian Schuster 63
Decouple!
!
Christian Schuster 64
Decoupling
Based on the simple example from before:
R
U0
ZIC ( f )
Z PDN R jwL
jwL
(R = 0.7
m,
L = 40 nH)
(for large w )
Christian Schuster 65
Decoupling
we ask what a so called "decoupling" or "bypass" capacitor does:
R
U0
Z PDN
R jwL
1 jwRC w 2 LC
1
(for large w )
jwC
ZIC ( f )
R = 0.7 m
L = 40 nH
C = 1 mF
Christian Schuster 66
Decoupling
Heuristic explanation:
R
U0
ZIC ( f )
Christian Schuster 67
Decoupling
While being beneficial at higher
frequencies decoupling increases
the PDN impedance in the vicinity
of the resonance frequency:
1 L
w0
R2
L C
Z PDN (w0 )
R = 0.7 m
L = 40 nH
C = 1 mF
( L / C R 2 )*
1 L
R
R C
R = 10 m
L = 40 nH
C = 1 mF
w0 1 / LC
Christian Schuster 69
More Decoupling
~
board-level
package-level
Amount of
charge, size
of decoupling
capacitance
chip-level
Speed of charge
delivery,
effective
frequency
Christian Schuster 70
Power/Ground Planes
Power/ground planes serve multiple purposes at the same time:
HOWEVER
Christian Schuster 71
Power/Ground Planes
they do show a resonant behavior:
Dielectric Filling
(r = 4)
Cpp 0 r
10 mil
Port
Power
Ground
A
11 nF
d
11 inch
Christian Schuster 72
11 inch
Power/Ground Planes
The resonance frequencies are given by:
f mn
c0
r r
m n
2a 2b
Christian Schuster 73
Adding Decoupling
Be wary of resonances!
!
Christian Schuster 74
(4)
Christian Schuster 75
SignalVia
Load
Return Current
Ground Via
Load
Christian Schuster 76
viu
Cp
vi
vil
Cp
Plane
Via
Via Cross Section
i'iu
Zp
viu 1 Z pp vil
i
i
0
1
il
iu
Zpp
iil
Christian Schuster 77
vil 1
i 1 / Z
pl
il
Zp
v'il
0 viu
1 iiu
iiu
Zpp:
(Parallel Plate
Impedance)
v'iu 1
i ' 1 / Z
pu
iu
i'il
0 v'il
1 i'il
Port j (xj,yj)
(a,b,d)
Current
Open
Plane
Edges
(a,b,0)
Voltage
z
Filling with and
(0,0,0)
jwd
Z ij (w )
ab m0 n 0
(a,0,0)
y
x
2
2
2
k
xm
yn
m
n
k yn
a
b
Cm , Cn 1 for m, n 0 and
k xm
Christian Schuster 78
k w
2 otherwise
Including Striplines
Trace between planes:
2 Modes: Stripline + Parallel Plate
Stripline Mode
Christian Schuster 79
Including Striplines
h2
h1
h1
h1 h2
I ps1
V ps1
V
I ps2 k 2 Y stripline Y pp
(k k )Y stripline Y pp ps2
I
2
2
Vgs1
gs1
(
k
)
Y
Y
(
k
2
k
1
)
Y
Y
stripline
pp
stripline
pp
I
V
gs2
gs2
R. Rimolo-Donadio, H. D. Brns, C. Schuster, Including Stripline Connections into Network Parameter Based Via Models for
Fast Simulation of Interconnects, International Zurich Symposium on Electromagnetic Compatibility, Switzerland, Jan. 12-15, 2009
Christian Schuster 80
Zpp
Port n
Cavity
representation
Ztl
S-Parameter
Matrix
Zpp
Ztl
Cavity
representation
Cavities joined by
segmentation
techniques
R. Rimolo-Donadio et al., Physics-based via and trace models for efficient link simulation on multilayer structures
up to 40 GHz", IEEE Trans. Microw. Theory and Techn., vol. 57, no. 8, p.p. 2072-2083, August 2009.
Christian Schuster 81
Magnitude
[dB]
of SS12
Magnitude of
14 [dB]
Full-wave model
Frequency
Frequency [GHz]
[GHz]
Christian Schuster 82
Model
Model
FEM simulation
simulation
FEM
FIT simulation
simulation
FIT
14 differential
striplines (2D)
6 cavities
Terminations
Christian Schuster 83
|S12| [dB] - IL
Link 10 S3 Stripline
Link 17 S5 Stripline
Link 10 S3 Stripline
Link 17 S5 Stripline
Measurement Link 10
Christian Schuster 84
Measurement Link 17
Model Link 10
Model Link 17
1 GND via
2 GND vias
4 GND vias
6 GND vias
Christian Schuster 85
Effect of number
of ground vias:
1 GND vias
2 GND vias
4 GND vias
6 GND vias
Frequency [GHz]
Christian Schuster 86
(5)
Measurement Techniques
Christian Schuster 87
12 ports
Bandwidth 10 MHz 50 GHz
Christian Schuster 88
Surface Connectors
Access Vias
MICRO-PROBE
5 mm
5 mm
STRUCTURE UNDER TEST
Christian Schuster 89
MICROPROBE
STRUCTURE
UNDER TEST
Ground Vias
Signal
trace
No access vias
less distortion
probes closer to the structure under test
Christian Schuster 90
13
mm
45
Via array
Via array
many vias at tight pitch!
92
Christian Schuster 92
~ 1 cm
~ 1 mm
Signal pitch conversion from ~1 cm to ~1 mm
& easy multiport access
93
Christian Schuster 93
Interposer 1
Interposer 2
Interposer Prototype
SMP
adapters
SMP
connectors
Interposer
LGA
Test
board
Hardware courtesy of
IBM YKT (Y. Kwark)
Hardware courtesy of
IBM YKT (Y. Kwark)
96
Christian Schuster 96
(6)
Wrapping Up
Christian Schuster 97
Christian Schuster 98
SNR
System
Target
Frequency
PDN
Impedance
Target
System
Frequency
EMI
Target
System
Frequency
Christian Schuster 99
Contact Information
Prof. Dr. sc. techn. Christian Schuster
Institut fr Theoretische Elektrotechnik
Technische Universitt Hamburg-Harburg
Harburger Schloss Str. 20
21079 Hamburg, Germany
Tel: +49 40 42878 3116
WWW: http://www.tet.tuhh.de/
E-Mail: schuster@tuhh.de