hacking the backplane: optimizing backplane …©eric bogatin 2006 slide -1 hacking the backplane:...
TRANSCRIPT
© Eric Bogatin 2006
Slide -1
www.BeTheSignal.com
Hacking the Backplane:Optimizing Backplane Performance with
Measurement Based Models Using Agilent ADS
Dr. Eric BogatinSignal Integrity Evangelist
Bogatin Enterprises
Fall 2006
© Eric Bogatin 2006
Slide -2
www.BeTheSignal.com
Hacker (from: www.iwriteiam.nl/HackerDef.html)
hacker:hacker: [originally, someone who makes furniture with an axe] n. …6. An expert or enthusiast of any kind. One might be an astronomy hacker, for example. 7. One who enjoys the intellectual challenge of creatively overcoming or circumventing limitations.
© Eric Bogatin 2006
Slide -3
www.BeTheSignal.com
Hacking the Backplane: The Process
• Goal “improving backplane performance”(applies to any backplane)
• Perform 4 port VNA/TDR measurements• Synthesize the circuit model• Optimize parameter values• Validate the model• Identify the weak links• Optimize the design for enhanced performance• Evaluate enhanced performance
© Eric Bogatin 2006
Slide -4
www.BeTheSignal.com
For More Information
www.BeTheSignal.comFeature articles and columnsSignal integrity public classesOnline lecturesResources Special offer: use coupon code ADS651 to view for free OLL-651 from the www.BeTheSignal.com web site
Published by Prentice Hall, 2004
© Eric Bogatin 2006
Slide -5
www.BeTheSignal.com
Example Backplane:Tyco HM-Zd Legacy Backplane with
XAUI Test Cards
~ 16 inch backplane traces
~ 2
inch
dau
ghte
r ca
rd tr
aces
Total length ~ 20 inches
~ 2
inch
dau
ghte
r ca
rd tr
aces
© Eric Bogatin 2006
Slide -6
www.BeTheSignal.com
Differential mode
Common mode
A Secret to Minimize Confusion About Differential S-Parameters
Think:Think:Differential signalsDifferential signalsCommon signalsCommon signals
© Eric Bogatin 2006
Slide -7
www.BeTheSignal.com
Transparent Translation Between the 4-Port Matrices with PLTS and ADS
Single ended S-parameters Differential S-parameters
Differential T-parametersSingle ended T-parameters
TDR
TermTerm3
Z=50 OhmNum=3
TermTerm1
Z=50 OhmNum=1
TeTe
Z=Nu
TeTe
Z=NuML2CTL_V
CLin1
W [1]=W _1 milLength=Len inSubst="Subst1"
ADS
VNA
Frequency Domain
Time Domain
© Eric Bogatin 2006
Slide -8
www.BeTheSignal.com
TDD11Differential Reflected Signal
Reflected signalConverted to 1st order Impedance Profile
Displayed with Agilent’s PLTS
© Eric Bogatin 2006
Slide -9
www.BeTheSignal.com
Simulating Eye Diagrams from TDD21
PRBS, 5 Gbps, 211 – 1 bits
Convolution integral
=Overlay each bit, synchronous with the clock
+Displayed with Agilent’s PLTS
Impulse Response
© Eric Bogatin 2006
Slide -10
www.BeTheSignal.com
De-constructing Performance Path
• ProsRelate physical features with performanceCan extract design information and material propertiesSimulates very fastDon’t need a lot of geometry information or vendor information
• ConsMay be difficult to find optimum topologyMay be tricky to optimize the parametersNo unique topology – no assurance you have the right one
Fit parameterized models to measured response
Circuit topology model
2 4 6 8 10 120 14
-80
-60
-40
-20
-100
0
freq, GHz
dB(S
21_m
eas)
dB(S
21_s
im)
© Eric Bogatin 2006
Slide -11
www.BeTheSignal.com
Measured Performance of the Backplane in ADS
S_ParamSP1
CalcGroupDelay=yesCalcS=yesStep=10 MHzStop=14 GHzStart=10 MHzSweepVar="freq"
S-PARAMETERS
TranTran1
MaxTimeStep=10 psecStopTime=10 nsecStartTime=0 nsec
TRANSIENT
S4PSNP1
4
1 2
3 RefTermTerm2
Z=50 OhmNum=2
TermTerm1
Z=50 OhmNum=1
TermTerm3
Z=50 OhmNum=3
TermTerm4
Z=50 OhmNum=4
V3_meas
S4PSNP2
4
1 2
3 RefVtPulseSRC2
Period=100 nsecWidth=100 nsecFall=(2.25*RT) nsecRise=(2.25*RT) nsecEdge=erfDelay=0 nsecVhigh=1 VVlow=0 V
t
RR5R=50 Ohm
RR2R=50 Ohm
RR1R=50 Ohm
RR4R=50 Ohm
Converting frequency domain data into time domain simulation
© Eric Bogatin 2006
Slide -12
www.BeTheSignal.com
Starting Place: Measured S-Parameters
2 4 6 8 10 120 14
-80
-60
-40
-20
-100
0
freq, GHz
dB(S
dd21
_mea
s)
2 4 6 8 10 120 14
-100
0
100
-200
200
freq, GHz
phas
e(Sd
d21_
mea
s)
2 4 6 8 10 120 14
-50
-40
-30
-20
-10
-60
0
freq, GHz
dB(S
dd11
_mea
s)
2 4 6 8 10 120 14
-100
0
100
-200
200
freq, GHz
phas
e(Sd
d11_
mea
s)
MeasEqnMeas5
Sdd11_meas=0.5*(S11+S33-S13-S31)Sdd21_meas=0.5*(S21+S43-S32-S41)
EqnMeas
© Eric Bogatin 2006
Slide -13
www.BeTheSignal.com
Transformed by ADS into Time Domain Single Ended
2 4 6 80 10
450
500
550
400
600
time, nsec
V1_
mea
s, m
V
2 4 6 80 10
-0
100
200
300
400
-100
500
time, nsecV
2_m
eas,
mV
2 4 6 80 10
0
20
40
-20
60
time, nsec
V3_
mea
s, m
V
2 4 6 80 10
-15
-10
-5
0
5
-20
10
time, nsec
V4_
mea
s, m
V
1 2
3 4
TDR into chan 1
NEXT FEXT
TDR TDT
© Eric Bogatin 2006
Slide -14
www.BeTheSignal.com
Eye Diagram Simulated from Frequency Domain Measured Data
(PLTS measured Eye Diagram)ADS simulation from the measured S-Parameter data
© Eric Bogatin 2006
Slide -15
www.BeTheSignal.com
Physical Features of Backplane
SMA Launch
Daughter card traces
Daughter card via
Backplane via
connector Backplane traces
SMA Launch
Daughter card traces
Daughter card via
Backplane viaconnector
Impe
danc
e, in
Ohm
s
© Eric Bogatin 2006
Slide -16
www.BeTheSignal.com
Simplifications to the Model
1. Symmetry between one line in the pair and the other –no mode conversion
2. Symmetry from one end to the other – back half same as front half
3. Negligible via stubs4. Same via model in
daughter card and backplane – different lengths
5. Same dielectric constant and dissipation factor in daughter card and backplane
TDD11 TDD22 Losses smear out far end
1 nsec/div
2 GHz/div
© Eric Bogatin 2006
Slide -17
www.BeTheSignal.com
Circuit Model Topology
V4_sim
V2_sim
Daughter Card via
Backplane viaDaughter Card via
Daughter CardSMA launch
Backplane
ConnectorConnector
TLINTL36
F=(1/TD_SMA) GHzE=360Z=Z_SMA Ohm
ML2CTL_VCLin6
ReuseRLGC=noRLGC_File=Layer[2]=2Layer[1]=2W[2]=W2_DC_1 milS [1]=S_DC_1 milW[1]=W1_DC_1 milLength=Len_DC_1 inSubst="Subst1"
TLINTL37
F=(1/TD_launch) GHzE=360Z=Z_launch Ohm
TLINTL38
F=(1/TD_launch) GHzE=360Z=Z_launch Ohm
CLINPTL40
Ao=0Ae=0Ko=Dk1_DC_1Ke=Dk1_DC_1L=Len_DCvia milZo=(Z11_DCvia-Z12_DCvia) OhmZe=(Z11_DCvia+Z12_DCvia) Ohm
ML2CTL_VCLin7
ReuseRLGC=noRLGC_File=Layer[2]=2Layer[1]=2W[2]=W2_CON_1 milS [1]=S_CON_1 milW[1]=W1_CON_1 milLength=Len_CON_1 inSubst="Subst2"
ML2CTL_VCLin8
ReuseRLGC=noRLGC_File=Layer[2]=2Layer[1]=2W[2]=W2_BP_1 milS [1]=S_BP_1 milW[1]=W1_BP_1 milLength=Len_BP_1 inSubst="Subst3"
CLINPTL44
Ao=0Ae=0Ko=Dk1_BP_1Ke=Dk1_BP_1L=Len_BPvia milZo=(Z11_BPvia-Z12_BPvia) OhmZe=(Z11_BPvia+Z12_BPvia) Ohm
ML2CTL_VCLin9
ReuseRLGC=noRLGC_File=Layer[2]=2Layer[1]=2W[2]=W2_CON_1 milS [1]=S_CON_1 milW[1]=W1_CON_1 milLength=Len_CON_1 inSubst="Subst2"
CLINPTL45
Ao=0Ae=0Ko=Dk1_DC_1Ke=Dk1_DC_1L=Len_DCvia milZo=(Z11_DCvia-Z12_DCvia) OhmZe=(Z11_DCvia+Z12_DCvia) Ohm
ML2CTL_VCLin10
ReuseRLGC=noRLGC_File=Layer[2]=2Layer[1]=2W[2]=W2_DC_1 milS [1]=S_DC_1 milW[1]=W1_DC_1 milLength=Len_DC_1 inSubst="Subst1"
TLINTL47
F=(1/TD_launch) GHzE=360Z=Z_launch Ohm
TLINTL48
F=(1/TD_launch) GHzE=360Z=Z_launch Ohm
TLINTL49
F=(1/TD_SMA) GHzE=360Z=Z_SMA Ohm
TLINTL50
F=(1/TD_SMA) GHzE=360Z=Z_SMA Ohm
RR41R=50 Ohm
RR42R=50 Ohm
TLINTL35
F=(1/TD_SMA) GHzE=360Z=Z_SMA Ohm
RR40R=50 Ohm
RR16R=50 Ohm
CC4C=C_DCvia pF
CC26C=C_DCvia pF
CC5C=C_DCvia pF
CC27C=C_DCvia pF
CC7C=C_BPvia pF
CC28C=C_BPvia pF
CC6C=C_BPvia pF
CC29C=C_BPvia pF
CLINPTL42
Ao=0Ae=0Ko=Dk1_BP_1Ke=Dk1_BP_1L=Len_BPvia milZo=(Z11_BPvia-Z12_BPvia) OhmZe=(Z11_BPvia+Z12_BPvia) Ohm
VtPulseSRC3
Period=100 nsecWidth=100 nsecFall=(2.25* RT) nsecRise=(2.25*RT) nsecEdge=erfDelay=0 nsecVhigh=1 VV low=0 V
t
SMA Launch
Daughter card traces
Daughter card via
Backplane via
connector Backplane traces
SMA Launch
Daughter card traces
Daughter card via
Backplane viaconnector
SMA LaunchDaughter card traces
Daughter card via
Backplane via
connector
Backplane traces
SMA LaunchDaughter card traces
Daughter card viaBackplane
via
connector
© Eric Bogatin 2006
Slide -18
www.BeTheSignal.com
Very Simple Differential Via Model
Ze = Z11 + Z12Zo = Z11 – Z12Dke = Dko = DkLen = 90 mils (daughter card)
Len = 190 mils (backplane)
Len
uniform ideal differential pair model
Negligible stub
Cvia
Ideal, lossless differential pair
© Eric Bogatin 2006
Slide -19
www.BeTheSignal.com
Optimizing Circuit Model Parameters
• Outputs:T11 (TDR) (sensitive to single ended impedance profile)T21 (TDT) (sensitive to len, dielectric constant)T31 (NEXT) (sensitive to line to line coupling)T41 (FEXT) (should be 0, sensitive to via structure)TDD11 (DTDR) (sensitive to diff impedance profile)SDD21 (sensitive to impedance profile and losses)
• Strategy:Start at the beginning of the modelOptimize parameter values- try for TDD11 best fit
© Eric Bogatin 2006
Slide -20
www.BeTheSignal.com
0.5 1.0 1.5 2.00.0 2.5
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
54 6
-15
-10
-5
0
5
-20
10
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
0
20
40
60
-20
80
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
Sculpting the Daughter Card Parameters – dielectric thickness
ParamSweepSweep1
Step=1Stop=11Start=7SimInstanceName[6]=SimInstanceName[5]=SimInstanceName[4]=SimInstanceName[3]=SimInstanceName[2]=SimInstanceName[1]="Tran1"SweepVar="H1_DC_1"
PARAMETER SWEEP
Sweeping H1_DC: 7 to 11 mils, step 1 milImpacts V11, V31, Vdd11Optimized value of H1_DC ~ 9.2 mils
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -21
www.BeTheSignal.com
0.5 1.0 1.5 2.00.0 2.5
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
0 .5 1.0 1.5 2.00.0 2.5
0
20
40
60
-20
80
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
54 6
-15
-10
-5
0
5
-20
10
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
Sculpting the Daughter Card Parameters – separation
Sweeping S1_DC: 10 to 20 mils, step 2 milsImpacts V31, Vdd11Optimized value of S1_DC ~ 15 mils
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -22
www.BeTheSignal.com
Sculpting the Backplane Parameters – line width
2 4 6 80 10
0.9
1.0
1.1
0.8
1.2
time, nsecV
dd11
__m
eas
Vdd
11__
sim
2 4 6 80 10
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
W1_BP = 8 mils to 16 mils, step 2
Optimized value ~ 11.6
TDR
DTDR
VARVAR14
S_BP_1=16W2_BP_1=W1_BP_1W1_BP_1=11.6Len_BP_1=16.7
EqnVar
VARVAR16
T2_BP_1=1Df2_BP_1=Df1_BP_1Df1_BP_1=Df1_DC_1Dk2_BP_1=Dk1_BP_1Dk1_BP_1=4H2_BP_1=H1_BP_1H1_BP_1=16
EqnVar
© Eric Bogatin 2006
Slide -23
www.BeTheSignal.com
Optimizing Daughter Card/Backplane Via Parameters
Ze = Z11 + Z12Zo = Z11 – Z12Dke = Dko = Dk
Cvia
0.5 1.0 1.5 2.00.0 2.5
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
0 .5 1.0 1.5 2.00.0 2.5
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
0
20
40
60
80
-20
100
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
54 6
-20
-10
0
10
20
-30
30
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
Sweeping Z11: 50 to 70 ohms, step 5
Z11 = 50 Ohms
Z11 ~ 65 Ohms
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -24
www.BeTheSignal.com
0.5 1.0 1.5 2.00.0 2.5
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
0 .5 1.0 1.5 2.00.0 2.5
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
0
20
40
60
80
100
120
-20
140
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
54 6
0
20
40
-20
60
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
Optimizing Daughter Card/Backplane Via Parameters
Ze = Z11 + Z12Zo = Z11 – Z12Dke = Dko = Dk
Cvia
Sweeping Z12: 10 to 50 ohms, step 10
Z12 = 50 OhmsZ11 ~ 65 ohmsZ12 ~ 30 Ohms
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -25
www.BeTheSignal.com
0.5 1.0 1.5 2.00.0 2.5
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
0 .5 1.0 1.5 2.00.0 2.5
0
20
40
60
-20
80
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
54 6
-15
-10
-5
0
5
-20
10
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
Optimizing Daughter Card/Backplane Via Parameters
Ze = Z11 + Z12Zo = Z11 – Z12Dke = Dko = Dk
Cvia
Sweeping C_DCvia: 0.2 to 0.6 pF, step 0.1 pF
C_DCvia = 0.6 pFZ11 ~ 65 ohmsZ12 ~ 30 ohmsC_DCvia ~ 0.4 pF
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -26
www.BeTheSignal.com
Via Analysis
VARVAR13
C_DCvia=0.4 {o}Len_DCviaStub=100-Len_DCviaLen_DCvia=90 {o}Z12_DCvia=30 {o}Z11_DCvia=65 {o}
EqnVar
Ze = Z11 + Z12Zo = Z11 – Z12Dke = Dko = Dk
Cvia
Zodd = 65 – 30 = 35 ohmsZdiff = 2 x Zodd = 70 Ohms
© Eric Bogatin 2006
Slide -27
www.BeTheSignal.com
Optimized Parameters
0.5 1.0 1.5 2.00.0 2.5
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
0 .5 1.0 1.5 2.00.0 2.5
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
0 .5 1.0 1.5 2.00.0 2.5
0
20
40
60
-20
80
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
54 6
-15
-10
-5
0
5
-20
10
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
Good, but not perfect.
Is it good enough?
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -28
www.BeTheSignal.com
Overall Time Domain Performance
2 4 6 80 10
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd
11__
mea
sV
dd11
__si
m
2 4 6 80 10
450
500
550
400
600
time, nsec
V1_
mea
s, m
VV
1_si
m, m
V
2 4 6 80 10
0
20
40
60
-20
80
time, nsec
V3_
mea
s, m
VV
3_si
m, m
V
2 4 6 80 10
-15
-10
-5
0
5
-20
10
time, nsec
V4_
mea
s, m
VV
4_si
m, m
V
Very good time domain performance
NEXT FEXT
TDR DTDR
© Eric Bogatin 2006
Slide -29
www.BeTheSignal.com
Frequency DomainSweeping Df
2 4 6 8 10 120 14
-50
-40
-30
-20
-10
-60
0
freq, GHz
dB(S
dd11
_mea
s)dB
(Sdd
11_s
im)
2 4 6 8 10 120 14
-100
0
100
-200
200
freq, GHz
phas
e(Sd
d11_
mea
s)ph
ase(
Sdd1
1_si
m)
2 4 6 8 10 120 14
-80
-60
-40
-20
-100
0
freq, GHz
dB(S
dd21
_mea
s)dB
(Sdd
21_s
im)
2 4 6 8 10 120 14
-100
0
100
-200
200
freq, GHz
phas
e(Sd
d21_
mea
s)ph
ase(
Sdd2
1_si
m)
Df from 0.015 to 0.035 step 0.005
Df ~ 0.025
SDD11 SDD21
© Eric Bogatin 2006
Slide -30
www.BeTheSignal.com
Final Frequency Domain Response
2 4 6 8 10 120 14
-50
-40
-30
-20
-10
-60
0
freq, GHz
dB(S
dd11
_mea
s)dB
(Sdd
11_s
im)
2 4 6 8 10 120 14
-100
0
100
-200
200
freq, GHz
phas
e(Sd
d11_
mea
s)ph
ase(
Sdd
11_s
im)
2 4 6 8 10 120 14
-80
-60
-40
-20
-100
0
freq, GHz
dB(S
dd21
_mea
s)dB
(Sdd
21_s
im)
2 4 6 8 10 120 14
-100
0
100
-200
200
freq, GHz
phas
e(S
dd21
_mea
s)ph
ase(
Sdd
21_s
im)
BW of the model ~ 10 GHz
2 4 6 8 10 120 14
1E-9
2E-9
3E-9
4E-9
0
5E-9
freq, GHz
dela
y(2,
1)de
lay(
6,5)
delay
SDD11 SDD21
Very little dispersion
© Eric Bogatin 2006
Slide -31
www.BeTheSignal.com
21 22 23 24 25 26 27 28 2920 30
0.00
0.05
0.10
0.15
0.20
-0.05
0.25
time, nsec
Vdd
21_p
rbs_
mea
sV
dd21
_prb
s_si
m
PRBS Results: 5 Gbps, BW = 10 GHz
21 22 23 24 25 26 27 28 2920 30
0
50
100
150
200
-50
250
time, nsec
V_pr
bs_f
ilter
ed, m
V
Bit stream going into the interconnect VDD21 signal out
measured
simulated with model
© Eric Bogatin 2006
Slide -32
www.BeTheSignal.com
Measured and Simulated Eye Diagram: 5 Gbps PRBS
measured simulated
© Eric Bogatin 2006
Slide -33
www.BeTheSignal.com
Hacking the Backplane
• Laminate dissipation factor: From Df = 0.025 0.01
• Via designFrom Zdiff = 2 x (65-30) = 70 Ohms 100 ohmsFrom C = 0.4 pF and 0.3 pF 0.1 pF
• Signal TracesFrom Zdiff = 106 ohms 100 Ohms
© Eric Bogatin 2006
Slide -34
www.BeTheSignal.com
Df from 0.025 0.01
2 4 6 8 10 120 14
-80
-60
-40
-20
-100
0
freq, GHz
dB(S
dd21
_mea
s)dB
(Sdd
21_s
im)
21 22 23 24 25 26 27 28 2920 30
0.00
0.05
0.10
0.15
0.20
-0.05
0.25
time, nsec
Vdd
21_p
rbs_
mea
sV
dd21
_prb
s_si
m
measured
simulated
measured
simulated
Measured response Simulated response
© Eric Bogatin 2006
Slide -35
www.BeTheSignal.com
Optimized Via Design
From 70 Ohm to 100 Ohm diff Impedance
Removal of NFPs, excess capacitance reduced to 0.1 pF
2 4 6 80 10
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd1
1__m
eas
Vdd1
1__s
im
Measured response Simulated response
© Eric Bogatin 2006
Slide -36
www.BeTheSignal.com
2 4 6 80 10
0.9
1.0
1.1
0.8
1.2
time, nsec
Vdd1
1__m
eas
Vdd1
1__s
im
Daughter Card and Back Plane Differential Impedance: 106 ohms 100 Ohms
Measured response Simulated response
Decreasing H to decrease diff impedance
© Eric Bogatin 2006
Slide -37
www.BeTheSignal.com
Hacking: Pushing All the Knobs
• Optimized performanceLower loss laminate (could be compensated with TX, RX equalization)Transparent viasImpedance matched traces
Original response Optimized response
5 Gbps
© Eric Bogatin 2006
Slide -38
www.BeTheSignal.com
Optimized Design Enables 10 Gbps Performance
Original response Optimized response
5 Gbps
10 Gbps
© Eric Bogatin 2006
Slide -39
www.BeTheSignal.com
Summary
• PLTS enables complete characterization of backplane differential channel
• ADS can be used to build simple, topology based circuit modelsIntegrate measured data
Multiple display modes: emulate TDR, DTDR, VNA, Differential VNA, PRBS, eye diagram
Multiple differential pair elements
Parameterized values for all circuit elements
Integrated 2D field solver
Quick, fast, interactive parameter optimization
Simple comparison of measured and simulated response
• Biggest impact on performance is from:Dielectric loss
Via design
Matched trace impedances
• This optimized backplane can operate at over 10 Gbps
© Eric Bogatin 2006
Slide -40
www.BeTheSignal.com
For More Information
www.BeTheSignal.comFeature articles and columnsSignal integrity public classesOnline lecturesResources Special offer: use coupon code ADS651 to view for free OLL-651 from the www.BeTheSignal.com web site
Published by Prentice Hall, 2004