Post on 17-Apr-2018
The “Ground” Myth
Dr. Bruce ArchambeaultIBM Distinguished Engineer
IEEE Fellow
IBMResearch Triangle Park, NC
Barch@us.ibm.com
October 2007
IEEE
October 2007 Dr. Bruce Archambeault 2
Outline
• Electromagnetics• Skin Effect• Inductance• Ground• Return Current Path
October 2007 Dr. Bruce Archambeault 3
ElectromagneticsIn the Beginning
• Electric and Magnetic effects not connected
• Electric and magnetic effects were due to ‘action from a distance’
• Faraday was the 1st to propose a relationship between electric lines of force and time-changing magnetic fields– Faraday was very good at experiments and
‘figuring out’ how things work
October 2007 Dr. Bruce Archambeault 4
Maxwell• Maxwell was
impressed with Faraday’s ideas
• Discovered the mathematical link between the “electro”and the “magnetic”
• Scotland’s greatest contribution to the world (next to Scotch)
October 2007 Dr. Bruce Archambeault 5
Maxwell’s Equations –Differential Form
tBE
tDJH
∂∂
∂∂
−=×∇
+=×∇
A difference in Magnetic Fieldacross a small piece of space
A difference in Electric Fieldacross a small piece of space
A change in Electric FluxDensity with respect to time
A change in Magnetic FluxDensity with respect to time
October 2007 Dr. Bruce Archambeault 6
Maxwell’s Equations are not Hard!
• Change in H-field across space ¡ Change in E-field (at that point) with time
• Change in E-field across space ¡ Change in H-field (at that point) with time
• (Roughly speaking, and ignoring constants)
October 2007 Dr. Bruce Archambeault 7
Current Flow• Most important concept of EMC• Current flow through metal changes as
frequency increases• DC current
– Uses entire conductor– Only resistance inhibits current
• High Frequency– Only small part of conductor (near surface) is used– Resistance is small part of current inhibitor– Inductance is major part of current inhibitor
October 2007 Dr. Bruce Archambeault 8
Skin Depth• High frequency current flows only near the
metal surface at high frequencies
Frequency Skin Depth Skin Depth 60 Hz 260 mils 8.5 mm 1 KHz 82 mils 2.09 mm 10 KHz 26 mils 0.66 mm
100 KHz 8.2 mils 0.21 mm 1 MHz 2.6 mils 0.066 mm 10 MHz 0.82 mils 0.021 mm
100 MHz 0.26 mils 0.0066 mm 1 GHz 0.0823 mils 0.0021 mm
µσπδ
f1=
October 2007 Dr. Bruce Archambeault 9
Current Migrates to Outer Portions of the Conductor at High Frequencies
Resistance is determined by the area of the copper conductor actually used at each frequency!
October 2007 Dr. Bruce Archambeault 10
S21 Skin Effect Loss ONLY-- Comparison for Microstrip (1 cm Long)Various Widths -- 50 ohms (FR4, 4.2, 0.021) Trace Thickness = 0.6mil
-1
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1 1 10 100
Frequency (GHz)
Loss
(dB
)
1 Mil Wide uStrip
2 Mil Wide uStrip
4 Mil Wide uStrip
6 Mil Wide uStrip
8 Mil Wide uStrip
October 2007 Dr. Bruce Archambeault 11
S21 Dielectric Loss ONLY-- Comparison for Microstrip (1 cm Long)Various Widths -- 50 ohms (FR4, 4.2, 0.021) Trace Thickness = 0.6mil
-1
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1 1 10 100
Frequency (GHz)
Loss
(dB
)
1 Mil Wide uStrip
2 Mil Wide uStrip
4 Mil Wide uStrip
6 Mil Wide uStrip
8 Mil Wide uStrip
October 2007 Dr. Bruce Archambeault 12
S21 Loss Comparison Between Stripline and Microstrip (1 cm Long)6 mil wide -- 50 ohms (FR4, 4.2, 0.021) Trace Thickness = 0.6mil
-2
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.1 1 10 100
Frequency (GHz)
S21
Mag
(dB
)
6 mil Wide uStrip6 mil Wide StripLine
October 2007 Dr. Bruce Archambeault 13
S21 Loss Comparison for 6 Mil Wide Microstrip Loss (1 cm Long)Various Dielectric Loss -- 50 ohms (FR4, 4.2) Trace Thickness = 0.6mil
-1
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1 1 10 100
Frequency (GHz)
S21
Mag
(dB
)
Skin OnlyDiel (.010) OnlyDiel (.021) OnlyDiel (.029) OnlyDiel (.021) + Skin
October 2007 Dr. Bruce Archambeault 14
At High Frequencies
• Resistive loss and dielectric loss are present
• Inductance will usually dominate
October 2007 Dr. Bruce Archambeault 15
Inductance
• Current flow through metal => inductance!
• Fundamental element in EVERYTHING• Loop area first order concern• Inductive impedance increases with
frequency and is MAJOR concern at high frequencies
fLX L π2=
October 2007 Dr. Bruce Archambeault 16
Current Loop => Inductance
Courtesy of Elya Joffe
October 2007 Dr. Bruce Archambeault 17
Inductance Definition
• Faraday’s Law∫ ∫∫ ⋅
∂∂−=⋅ SdtBdlE
tBAV
∂∂−=
VB
Area = A
• For a simple rectangular loop
The minus sign means that the induced voltage will work against the current that originally created the magnetic field!
October 2007 Dr. Bruce Archambeault 18
Self Inductance
• Isolated circular loop
−≈ 28ln
00 r
aaL µ
+−+−+
+++
= 22
0 1121121
1ln
2 ppp
ppaLπµ
• Isolated rectangular loop
Note that inductance is directly influenced by loop AREA and only less influenced by conductor size!
radiuswiresideoflengthp =
October 2007 Dr. Bruce Archambeault 19
Partial Inductance
• Simply a way to break the overall loopinto pieces in order to find total inductance
L3
L4
L2
L1
L total=Lp11+ Lp22 + Lp33 + Lp44 - 2Lp13 - 2Lp24
October 2007 Dr. Bruce Archambeault 20
Decoupling Capacitor Mounting
• Keep as to planes as close to capacitor pads as possible
Height above Planes
Via Separation
Inductance Depends on Loop AREA
October 2007 Dr. Bruce Archambeault 21
Important Points About Inductance
• Inductance is everywhere• Loop area most important• Inductance is everywhere
October 2007 Dr. Bruce Archambeault 22
‘Ground’
• Ground is a place where potatoes and carrots thrive!
• ‘Earth’ or ‘reference’ is more descriptive • Original use of “GROUND”• Inductance is everywhere
fLX L π2=
October 2007 Dr. Bruce Archambeault 23
What we Really Mean when we say ‘Ground’
• Signal Reference• Power Reference• Safety Earth• Chassis Shield Reference
October 2007 Dr. Bruce Archambeault 24
‘Ground’ is NOT a Current Sink!
• Current leaves a driver on a trace and must return (somehow) to its source
• This seems basic, but it is often forgotten, and is most often the cause of EMC problems
October 2007 Dr. Bruce Archambeault 25
‘Grounding’ Needs Low Impedance at Highest Frequency
• Steel Reference Plate– 4 milliohms/sq @ 100KHz– 40 milliohms/sq @ 10 MHz– 400 milliohms/sq @ 1 GHz
• A typical via is about 2 nH– @ 100 MHz Z = 1.3 ohms– @ 500 MHz Z = 6.5 ohms– @ 1000 MHz Z = 13 ohms– @ 2000 MHz Z = 26 ohms
October 2007 Dr. Bruce Archambeault 26
Single-point ‘Ground’?Vs
Multi-point ‘Ground’?
• Which do I want? • Which do I get?
October 2007 Dr. Bruce Archambeault 27
Single Point ‘Ground’ Myth
• At high frequencies, inductance makes this impossible!
• At high frequencies, parasitic capacitance makes this impossible!
• Depends on the amount of ‘Ground’ error your system can stand…...
October 2007 Dr. Bruce Archambeault 28
Single-Point Ground Concept
GND via
Board A Board B
Metal Enclosure
GND via
GND viaGND via
GND via
GND planes
October 2007 Dr. Bruce Archambeault 29
Reality Overcomes Single-Point Ground Intentions
GND via
Board A Board B
Metal Enclosure
GND via
GND viaGND via
GND via
GND planes
October 2007 Dr. Bruce Archambeault 30
Where did the Term “GROUND”Originate?
• Original Teletype connections• Lightning Protection
October 2007 Dr. Bruce Archambeault 31
Ground/Earth
TeletypeReceiver
TeletypeTransmitter
October 2007 Dr. Bruce Archambeault 32
Ground/Earth
TeletypeReceiver
TeletypeTransmitter
October 2007 Dr. Bruce Archambeault 33
Lightning striking houseFIG 7
Lightning
October 2007 Dr. Bruce Archambeault 34
Lightning effect without rod
October 2007 Dr. Bruce Archambeault 35
Lightning effect with rod
Lightning rodLightning
October 2007 Dr. Bruce Archambeault 36
What we Really Mean when we say ‘Ground’
• Signal Reference• Power Reference• Safety Earth• Chassis Shield Reference
Circuit “Ground”
Chassis “Ground”
Digital “Ground”
D
Analog “Ground”
A
October 2007 Dr. Bruce Archambeault 37
October 2007 Dr. Bruce Archambeault 38
Schematic with Return Current Shown
IC1 IC2 IC3
Return currents on ground
Signal trace currents
October 2007 Dr. Bruce Archambeault 39
Actual Current Return is 3-Dimensional
Ground Layer
Signal Trace
IC
Ground Vias
Ground Layer
Signal TraceICGround Via
BOARD STACK UP:
Ground Layer
Signal TraceCURRENT LOCATION:
October 2007 Dr. Bruce Archambeault 40
Low Frequency Return Currents Take Path of Least Resistance
Ground Plane
DriverReceiver
October 2007 Dr. Bruce Archambeault 41
High Frequency Return Currents Take Path of Least Inductance
Ground Plane
DriverReceiver
October 2007 Dr. Bruce Archambeault 42
Microstrip Transmission Line
Stripline Transmission Line
DielectricReference Planes
Signal Trace
Traces/nets over a Reference Plane
October 2007 Dr. Bruce Archambeault 43
Signal Traces
Reference Planes
(Power, “Ground”, etc.)
Traces/nets and Reference Planes in Many Layer Board Stackup
October 2007 Dr. Bruce Archambeault 44
Microstrip Electric/Magnetic Field Lines(8mil wide trace, 8 mils above plane, 65 ohm)
Electric Field Lines
Vcc
October 2007 Dr. Bruce Archambeault 45
Microstrip Electric/Magnetic Field LinesCommon Mode
8 mil wide trace, 8 mils above plane, 65/115 ohm)
Electric Field Lines
Vcc
October 2007 Dr. Bruce Archambeault 46
Microstrip Electric/Magnetic Field LinesDifferential Mode
8 mil wide trace, 8 mils above plane, 65/115 ohm)
Electric Field Lines
Vcc
October 2007 Dr. Bruce Archambeault 47
Electric/Magnetic Field LinesSymmetrical Stripline
Vcc
GND
October 2007 Dr. Bruce Archambeault 48
Electric/Magnetic Field LinesSymmetrical Stripline (Differential)
Vcc
GND
October 2007 Dr. Bruce Archambeault 49
Electric/Magnetic Field LinesAsymmetrical Stripline
Vcc
GND
October 2007 Dr. Bruce Archambeault 50
Electric/Magnetic Field LinesAsymmetrical Stripline (Differential)
October 2007 Dr. Bruce Archambeault 51
What About Pseudo-Differential Nets?
• So-called differential traces are NOT truly differential– Two complementary single-ended drivers
• Relative to ‘ground’– Receiver is differential
• Senses difference between two nets (independent of ‘ground’)
• Provides good immunity to common mode noise• Good for signal quality/integrity
October 2007 Dr. Bruce Archambeault 52
Pseudo-Differential Nets Current in Nearby Plane
• Balanced/Differential currents have matching current in nearby plane– No issue for discontinuities
• Any unbalanced (common mode) currents have return currents in nearby plane that must return to source!– All normal concerns for single-ended nets
apply!
October 2007 Dr. Bruce Archambeault 53
Pseudo-Differential Nets
• Not really ‘differential’, since more closely coupled to nearby plane than each other
• Slew and rise/fall variation cause common mode currents!
October 2007 Dr. Bruce Archambeault 54
Differential Voltage Pulse with Skew1 Gbit/sec with 95 psec rise/fall time
0
0.2
0.4
0.6
0.8
1
1.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (nsec)
Volta
ge
Complementary -- Line1Complementary -- Line 2Skew=2psSkew=6psSkew = 10psSkew = 20psSkew = 30psSkew =40psSkew =50psSkew =60ps
October 2007 Dr. Bruce Archambeault 55
Common Mode VoltageFrom Differential Voltage Pulse with Skew
1 Gbit/sec with 95 psec rise/fall time
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (nsec)
Volta
ge
BalancedSkew=2psSkew=6psSkew =10psSkew =20psSkew =30psSkew =40psSkew =50ps
October 2007 Dr. Bruce Archambeault 56
Common Mode CurrentFrom Differential Voltage Pulse with Skew
1 Gbit/sec with 95 psec Rise/fall Time
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (nsec)
Leve
l (m
a)
BalancedSkew=2psSkew=6psSkew =10psSkew =20psSkew =30psSkew =40psSkew =50psSkew =60ps
October 2007 Dr. Bruce Archambeault 57
Common Mode CurrentFrom Differential Voltage Pulse with Skew
1 Gbit/sec with 95 psec Rise/fall Time
50
60
70
80
90
100
110
120
130
140
150
1.E+08 1.E+09 1.E+10 1.E+11
Frequency (Hz)
Leve
l (dB
uA)
Skew=2psSkew=6psSkew =10psSkew =20psSkew =30psSkew =40psSkew =50psSkew =60ps
October 2007 Dr. Bruce Archambeault 58
Differential Voltage Pulse with Rise/Fall Variation/Unbalance1 Gbit/sec with 95 psec Nominal Rise/Fall Time
0
0.2
0.4
0.6
0.8
1
1.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (ns)
Leve
l (vo
lts) Original Pulse rise=95ps
Complementary Pulse Rise=90psComplementary Pulse Rise=80psComplementary Pulse Rise=105psComplementary Pulse Rise=115ps
October 2007 Dr. Bruce Archambeault 59
Common Mode VoltageFrom Differential Voltage Pulse with Various Rise/Fall Unbalance
1 Gbit/sec with 95 psec Nominal Rise/Fall Time
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (ns)
Volta
ge
Complementary Pulse Rise=90ps
Complementary Pulse Rise=80ps
Complementary Pulse Rise=105ps
Complementary Pulse Rise=115ps
October 2007 Dr. Bruce Archambeault 60
Common Mode CurrentFrom Differential Voltage Pulse with Various Rise/Fall Unbalance
1 Gbit/sec with 95 psec Nominal Rise/fall Time
-60
-40
-20
0
20
40
60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (ns)
Cur
rent
(ma)
Complementary Pulse Rise=90psComplementary Pulse Rise=80psComplementary Pulse Rise=105psComplementary Pulse Rise=115ps
October 2007 Dr. Bruce Archambeault 61
Common Mode CurrentFrom Differential Voltage Pulse with Various Rise/Fall Unbalance
1 Gbit/sec with Nominal 95 psec Rise/fall Time
50
55
60
65
70
75
80
85
90
1.E+08 1.E+09 1.E+10 1.E+11
Frequency (Hz)
Leve
l (dB
ua)
Complementary Pulse Rise=90psComplementary Pulse Rise=80psComplementary Pulse Rise=105psComplementary Pulse Rise=115ps
October 2007 Dr. Bruce Archambeault 62
Pseudo-Differential Net Summary
• Small amounts of skew can cause significant common mode current
• Small amount of rise/fall time deviation can cause significant amount of common mode current
• Discontinuities (vias, crossing split planes, etc) and convert significant amount of differential current into common mode current
October 2007 Dr. Bruce Archambeault 63
Return Current vs.. “Ground”
• For high frequency signals, “Ground” is a concept that does not exist
• The important question is “where does the return current flow?”
October 2007 Dr. Bruce Archambeault 64
Referencing Nets(Where does the Return Current Flow??)
• Microstrip/Stripline across split in reference plane
• Microstrip/Stripline through via (change reference planes)
• Mother/Daughter card
October 2007 Dr. Bruce Archambeault 65
• Don’t Cross Splits with Critical Signals!!!– Bad practice– Stitching capacitor required across split to
allow return current flow• must be close to crossing• must have low inductance• limited frequency effect --- due to inductance
– Major source of Common Mode current!
Microstrip/Stripline Across Split in Reference Plane
October 2007 Dr. Bruce Archambeault 66
Splits in Reference Plane
• Power planes often have splits• Return current path interrupted• Consider spectrum of clock signal• Consider stitching capacitor
impedance• High frequency harmonics not
returned directly
October 2007 Dr. Bruce Archambeault 67
Split Reference Plane Example
PWR
GND
October 2007 Dr. Bruce Archambeault 68
Split Reference Plane Example With Stitching Capacitors
PWR
GND
Stitching Capacitors Allow Return current to Cross Splits ???
October 2007 Dr. Bruce Archambeault 69
Capacitor ImpedanceMeasured Impedance of .01 uf Capacitor
0.1
1.0
10.0
100.0
1.E+06 1.E+07 1.E+08 1.E+09
Frequency (Hz)
Impe
dnac
e (o
hms)
October 2007 Dr. Bruce Archambeault 70
Frequency Domain Amplitude of Intentional Current Harmonic AmplitudeFrom Clock Net
40
60
80
100
120
140
160
0 200 400 600 800 1000 1200 1400 1600 1800 2000
freq (MHz)
leve
l (dB
uA)
October 2007 Dr. Bruce Archambeault 71
MoM Microstrip Model Current Distribution Example
October 2007 Dr. Bruce Archambeault 72
MoM Microstrip Model Current Distribution Example
October 2007 Dr. Bruce Archambeault 73
Emissions From Board
• Far field emissions not important unless it is an unshielded product
• Near field emissions above board AREimportant
• Example of emissions from board with critical net crossing split reference plane
October 2007 Dr. Bruce Archambeault 74
Near Field Radiation from Microstrip on Board with Split in Reference Plane
Comparison of Maximum Radiated E-Field for MicrostripWith and without Split Ground Reference Plane
20
30
40
50
60
70
80
90
100
110
120
10 100 1000
Frequency (MHz)
Max
imum
Rad
iate
d E-
Fiel
d (d
Buv
/m)
No-Split
Split
October 2007 Dr. Bruce Archambeault 75
With “Perfectly Connected” Stitching Capacitors Across Split
Comparison of Maximum Radiated E-Field for MicrostripWith and without Split Ground Reference Plane and Stiching Capacitors
20
30
40
50
60
70
80
90
100
110
120
10 100 1000
Frequency (MHz)
Max
imum
Rad
iate
d E-
Fiel
d (d
Buv
/m)
No-Split
Split
Split w/ one Cap
Split w/ Two Caps
October 2007 Dr. Bruce Archambeault 76
Stitching Caps with Via InductanceComparison of Maximum Radiated E-Field for Microstrip
With and without Split Ground Reference Plane and Stiching Capacitors
20
30
40
50
60
70
80
90
100
110
120
10 100 1000
Frequency (MHz)
Max
imum
Rad
iate
d E-
Fiel
d (d
Buv
/m)
No-SplitSplitSplit w/ one CapSplit w/ Two CapsSplit w/One Real CapSplit w/Two Real Caps
October 2007 Dr. Bruce Archambeault 77
Example of Common-Mode Noise Voltage Across Split Plane Vs. Stitching Capacitor Distance to Crossing Point
0
5
10
15
20
25
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Distance (mils)
Gap
Vol
tage
100MHz200MHz300MHz400MHz500MHz600MHz700MHz800MHz900MHz1000MHz
October 2007 Dr. Bruce Archambeault 78
Are Stitching Capacitors Effective ???
• YES, at low frequencies • No, at high frequencies• Need to limit the high frequency current
spectrum• Need to avoid split crossings with ALL
critical signals
October 2007 Dr. Bruce Archambeault 79
Pin Field Via Keepouts??
Return Current must go around entire keep out area --- just as bad as a slot
Return current path deviation minimal
sd
Recommend s/d > 1/3
October 2007 Dr. Bruce Archambeault 80
Changing Reference Planes Six-Layer PCB Stackup Example
Signal Layer
Signal Layer
Signal Layer
Signal Layer
Plane
Plane
October 2007 Dr. Bruce Archambeault 81
Microstrip/Stripline through via (change reference planes)
TraceVia
October 2007 Dr. Bruce Archambeault 82
How can the Return Current Flow When Signal Line Goes Through Via??
What happens to Return Current in this Region?
Return Current
October 2007 Dr. Bruce Archambeault 83
How can the Return Current Flow When Signal Line Goes Through
Via??• Current can NOT go from one side of the
plane to the other through the plane– skin depth
• Current must go around plane at via hole, through decoupling capacitor, around second plane at the second via hole!
• Use displacement current between planes
October 2007 Dr. Bruce Archambeault 84
Return Current without Intentional Path
GND
PWR
October 2007 Dr. Bruce Archambeault 85
Return Current Across Reference Plane Change With Decoupling Capacitor (on Top)
Return Current
Decoupling Capacitor
Reference Planes
Displacement Current
Common-Mode Current
October 2007 Dr. Bruce Archambeault 86
RF Current @ 700 MHz with One Capacitor 0.5” from Via
October 2007 Dr. Bruce Archambeault 87
RF Current @ 700 MHz with One Capacitor 0.5” from Via
(expanded view)
October 2007 Dr. Bruce Archambeault 88
BadSignal Layer
Signal Layer
Signal Layer
Signal Layer
Reference Plane
Reference Plane
Bad
Signal Layer
Signal Layer
Signal Layer
Signal Layer
Reference Plane
Reference Plane
GoodSignal Layer
Signal Layer
Signal Layer
Signal Layer
Reference Plane
Reference Plane
Possible Routing OptionsSix-Layer Board
October 2007 Dr. Bruce Archambeault 89
Compromise Routing Option forMany Layer Boards
Good Compromise
Reference Plane
Gnd
Vcc1
Lot’s of Decoupling caps near ASIC
October 2007 Dr. Bruce Archambeault 90
Typical Driver/Receiver Currentsswitch IC loadIC
driverVDC
GND
CL
VCC
Z0, vp
GND
IC loadIC
driver
VCC
VCC
charge
logic 0-to-1
Z0, vp
IC loadIC
driver
GND
VCC
0 V
discharge
logic 1-to-0
Z0, vp
October 2007 Dr. Bruce Archambeault 91
Suppose The Trace is Routed Next to Power (not Gnd)
Vcc1
Vcc1
“Fuzzy” Return Path Area
“Fuzzy” Return Path Area
Return Path Options:
-- Decoupling Capacitors
-- Distributed Displacement Current
TEM Transmission Line Area
October 2007 Dr. Bruce Archambeault 92
Suppose The Trace is Routed Next to a DIFFERENT Power (not Gnd)
Return Path Options:
-- Decoupling Capacitors ??? May not be any nearby!!
-- Distributed Displacement Current – Increased current spread!!!
Vcc1
Vcc2
“Fuzzy” Return Path Area
“Fuzzy” Return Path Area
TEM Transmission Line Area
October 2007 Dr. Bruce Archambeault 93
Via SummaryRoute critical signals on either side of ONE reference planeDrop critical signal net to selected layer close to driver/receiver
Many decoupling capacitors to help return currents
Do NOT change reference planes on critical nets unless ABSOLUTELY NECESSARY!!Make sure at least 2 decoupling capacitors within 0.2” of via with critical signals
October 2007 Dr. Bruce Archambeault 94
Mother/Daughter Board Connector Crossing
• Critical Signals must be referenced to same plane on both sides of the connector
October 2007 Dr. Bruce Archambeault 95
Mother/Daughter Board Connector Crossing
Connector Signal Path
GNDPWRSignal
SignalP
WR
GN
DS
ignal
Signal
October 2007 Dr. Bruce Archambeault 96
Return Current from Improper Referencing Across Connector
Connector
De-cap
Signal Path
GNDPWRSignal
SignalP
WR
GN
DS
ignal
Signal
Return Path
De-cap
Displacement Current
October 2007 Dr. Bruce Archambeault 97
Return Current from Proper Referencing Across Connector
PW
RG
ND
Signal
Signal
Connector Signal Path
GNDPWRSignal
Signal
Return Path
October 2007 Dr. Bruce Archambeault 98
Board-to-Board Differential Pair Issues
Connector
PCB Plane 1
PCB Plane 2
Microstrip
Microstrip
VGround-to-Ground
noise
October 2007 Dr. Bruce Archambeault 99
Example Measured Differential Individual Signal-to-GND
Individual Differential Signals ADDED
Common Mode Noise 170 mV P-P
500 mV P-P (each)
October 2007 Dr. Bruce Archambeault 100
Measured GND-to-GND Voltage
205 mV P-P
October 2007 Dr. Bruce Archambeault 101
Antenna StructuresDipole antenna
PCB GND planes
Non-Dipole antenna
October 2007 Dr. Bruce Archambeault 102
Pin Assignment Controls Inductance for CM signals
Signal Pin Related Ground Pins
37.17 nH 25.21 nH
16.85 nH 20.97 nH
(a) (b)
(c) (d)
October 2007 Dr. Bruce Archambeault 103
How Many “Ground” Pins Across Connector ???
• Nothing MAGICAL about “ground”• Return current flow!• Choose the number of power and “ground”
pins based on the number of signal lines referenced to power or “ground” planes
• Insure signals are referenced against same planes on either side of connector
October 2007 Dr. Bruce Archambeault 104
Ground Myth Summary
• THERE IS NO SUCH THING AS “GROUND”
• Define which reference type is needed• Plan the return current path
– Avoid split reference planes & changing reference planes
• Differential signals have significant common mode, and must follow normal EMC rules
October 2007 Dr. Bruce Archambeault 105
IEEE