View
4
Download
0
Category
Preview:
Citation preview
Nanya CCL 1
The Signal Integrity Study with Fiber Weave Effect
Nan Ya Plastics Corp.
Speaker: Peter Liang
Electro Material Div. Copper Clad Laminate Unit
Nanya CCL 2
Outline:
-Demand of High Data Rate For Transmission Line Design
-What Influences Skew
- How to Reduce Skew
- Conclusion
Nanya CCL 3
Applications of High Speed TransitionApplications Challenge
High-End Servers
High-End Routers
Switches
Optical Transport
High-End Datastorage
Wireless Base Station
Higher Data Rate
-Rotate the image
-Spread glass cloth
-Low Dk cloth
- others1- 3 Gbps =>4 Gbps => 10 Gbps => Higher
DemandsReal time & Mass Data Transmission and Storage
Solution
Nanya CCL 4
Standard Year
2006
Year
2009
In development / Next generation
Ethernet 1Gbps 10Gbps 40Gbps, 100Gbps
Fibre Channel 1,2,4 Gbps 8Gbps, 16Gbps 40Gbps
Fibre channel over Ethernet N/A 10GFCoE 40GFCoE,
SATA/SAS 1.5Gbps / 3Gbps 3Gbps / 6Gbps 6Gbps / 12Gbps
PCI-Express PCIe1.0 1Gbps PCIe2.0 5Gbps PCIe3.0 8Gbps
Infiniband 2.5Gbps 10Gbps 40Gbps
What is the Demand of High Speed Transmission
5
Skew Effect in Differential Line
Is the impedance variation always small enough in differential line for wide frequency range?
Differential signals is out-of-phase ?
Substrate material uniformity ?
Signal coupling effect ?
From DesignCon 2005:The Impact of PCB Laminate Weave on the Electrical Performance of Differential Signaling at Multi-Gigabit Data Rates
Nanya CCL 6
An uniform group delay and a small phase difference result in a superior skew characteristic in the differential line, indicating a low time delay between the differential signals.
Skew Effect in Differential Line
Line 1
Line 2
WHY no uniform Group Delay?
WHY no Out-of-Phase?
180°°°°
From DesignCon 2005:The Impact of PCB Laminate Weave on the Electrical Performance of Differential Signaling at Multi-Gigabit Data Rates
Nanya CCL 7
The Effect of Fibric
The traces at resin rich or resin poor have different Dk value. This may cause different impedance and reflection result.
Epoxy only
Weave bundle
Resin rich
Resin poor
From Intel Fiberweave Effect : Practical Impact Analysis and Mitigation Strategies
Nanya CCL 8
From ALTERA: PCB Dielectric Material Selection and Fiber Weave Effect on High-Speed Channel Routing
Skew Effect
The trace with skew can make the signal differ from design.
Nanya CCL 9
From Intel :Fiberweave Effect: Practical Impact Analysis and Mitigation Strategies
Rotate the image to Reduce the Skew Effect
The most popular way to reduce the skew effect is to rotate the image.
1. Higher cost2. Material utility ratio
become low3. Hard to design
Nanya CCL 10
Using Spreading Cloth to Reduce the Skew Effect
The other way to reduce the skew effect is to use spreading cloth.
Resin rich Resin poor Resin rich Resin poor
Non-spreading glass cloth Spreading glass cloth
How to gauge the fabric‘s spread that is enough to solve skew effect?
Nanya CCL
EqnTest_2=0.5*(S(2,1)-S(2,3)-S(4,1)+S(4,3))
EqnTest_1=0.5*(S(6,5)-S(6,7)-S(8,5)+S(8,7))
m1freq=dB(Test_1)=-24.132
20.00GHz
m2freq=dB(Test_2)=-19.888
20.00GHz
2 4 6 8 10 12 14 16 180 20
-20
-15
-10
-5
-25
0
freq, GHz
dB(T
est_
1)
m1
dB(T
est_
2)
m2m1freq=dB(Test_1)=-24.132
20.00GHz
m2freq=dB(Test_2)=-19.888
20.00GHz
Eqn Test_2=0.5*(S(2,1)-S(2,3)-S(4,1)+S(4,3))
Eqn Test_1=0.5*(S(6,5)-S(6,7)-S(8,5)+S(8,7))
m2freq=dB(Test_2)=-22.146
20.00GHz
2 4 6 8 10 12 14 16 180 20
-20
-15
-10
-5
-25
0
freq, GHz
dB(T
est_
1)
m1dB
(Tes
t_2)
m2m1freq=dB(Test_1)=-24.810
20.00GHz
m2freq=dB(Test_2)=-22.146
20.00GHz
m1freq=dB(Test_1)=-24.810
20.00GHz
11
∆∆∆∆Loss=4.244 dB@20 GHz
Non-spreading glass cloth
∆∆∆∆Loss=2.665 dB@20 GHz
Some Test Results by NanYa
spreading glass cloth
Using spreading glass cloth can minimize the gap between pattern 1 and pattern 2. But it also can’t ignore the FWE effect.
Test pattern #2
Test pattern #2
Test pattern #1
Test pattern #1
Nanya CCL 12
How to improve the effect of spreading for fabric?
-Use different yarn style-Denser Fabric
Nanya CCL 13
Outline:
-Demand of High Data Rate For Transmission Line Design
-What Influences Skew
- How to Reduce Skew
- Conclusion
Nanya CCL
Single Line Characteristics
14
Dk: 3.5~4.5 step 0.25Df: 0.012
Dk: 3.9Df: 0.002~0.02 step 0.005
• frequency independent Dk and Df
Under 5 GHz operation, Dk is the significant parameter for the impedance of transmission line.
0 2 4 6 8 10 12 14
46
48
50
52
54
Impe
danc
e (( ((O
hm)) ))
Frequency (GHz)
Df0.002 Df0.007 Df0.012 Df0.017 Df0.02
0 2 4 6 8 10 12 1440
45
50
55
60
Impe
danc
e (O
hm)
Frequency (GHz)
Dk3.5 Dk3.75 Dk4.0 Dk4.25 Dk4.5
Line width: 0.6 mmLine length: 100 mmPCB thickness: 0.3 mm
Line response by Dk
15
The Dk variation in PCB board leads to the characteristic changes of impedance and group delay, and the jitter of eye diagram was also degraded.
0 2 4 6 8 10 12 14-0.25
-0.20
-0.15
-0.10
-0.05
0.00
Los
s (d
B)
Frequency (GHz)
Dk3.5 Dk3.75 Dk4.0 Dk4.25 Dk4.5
0 2 4 6 8 10 12 14
-90
-80
-70
-60
-50
-40
-30
-20
Ret
urn
Los
s (d
B)
Frequency (GHz)
Dk3.5 Dk3.75 Dk4.0 Dk4.25 Dk4.5
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
Dk=3.5 Dk=4 Dk=4.5
Eye diagram (10 Gbps)
Nanya CCL
16
0 2 4 6 8 10 12 14-0.60-0.55-0.50-0.45-0.40-0.35-0.30-0.25-0.20-0.15-0.10-0.050.00
Los
s (d
B)
Frequency (GHz)
Df0.002 Df0.007 Df0.012 Df0.017 Df0.02
0 2 4 6 8 10 12 14-95-90
-85-80-75-70
-65
-60-55-50-45-40
-35
Ret
urn
Los
s (d
B)
Frequency(GHz)
Df0.002 Df0.007 Df0.012 Df0.0107 Df0.02
Line response by Df
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
Eye diagram (10 Gbps)
Df=0.002 Df=0.012 Df=0.0250 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
The high-frequency response of material determines the loss for a wide frequency range. It also results in a worse performance in High/Low margin of eye diagram.
Nanya CCL
Nanya CCL 17
Outline:
-Demand of High Data Rate For Transmission Line Design
-What Influences Skew
- How to Reduce Skew
- Conclusion
Nanya CCL 18
How to Mitigate FWE Effect1. Rotate the trace with an angle
2. Use Low Dk glass cloth (glass Dk: 7.2 -> 5.5 )
3. Increase the density of glass yarns
Nanya CCL
Experiment and Measurement (1)
Sampling OscilliscopeAgilent 86100C
Frequency Domain(up to 50 GHz)
(Network analyzer, Agilent E8363B)On-wafer test
RF probe station
Time Domain(up to 12.5 Gb/s)
Pulse pattern generatorAnritsu MP1763C
•S parameter(loss, reflection,phase)
•Impedance
•Waveforms (skew)•Eye diagram (jitter, eye opening)
19
Nanya CCL
Experiment and Measurement (2)
Line width: 4, 6, 8, 10 mil.Line length: 1, 2, 5 inch.Line gap: 4, 6, 8,10 mil.
250 µµµµm GSGSG RF probe 400 µµµµm GS/SG RF probe
Through Line
Loop Line
GSGSG
GS
GS
Measurement Items:• Through line: Dk/Df, phase, skew, impedance • Loop line: Dk/Df and impedance with coupling effect
20
Nanya CCL
Experimentfor
Single Line
Impedance Frequency response
Impedance variation
Signal reflection
Eye opening
Jitter
Over or under shoot
Hi/Lo margin
Insertion loss
Return loss
Group delay
Parameters Measurement Evaluation
Rotation
DK
Density S-parameter
Eye Diagram
21
Nanya CCL 22
Construction B yarns v.s. C yarns per inch.
(single line)
• The loss of the non-rotated transmission line can be reduced if yarns of glass is
increased to C.
• A low loss transmission line can be achieved within 2 cases:
1. rotate substrate from 0-degree to 45-degree, no matter it weaves by B or C yarns.
2. direct using C-yarn substrate.
0 2 4 6 8 10-55
-50
-45
-40
-35
-30
-25
-20
60-45degree 60 70-45degree 70
Ret
urn
Los
s (d
B)
Frequency (GHz)
0 2 4 6 8 10-0.50
-0.45
-0.40
-0.35
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
Los
s (d
B)
Frequency (GHz)
60-45degree 60 70-45degree 70
B-45 degree
B
C-45 degree
C
B-45 degree
B
C-45 degree
C
Nanya CCL
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
Eye Diagram Test
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
B yarns 0-degree C yarns 0-degree
B yarns 45-degree C yarns 45-degree
23
Nanya CCL
Experimentfor
Single Line
DK
Rotation
Density
Parameter Measurement Evaluation
Impedance Frequency response
Impedance variation
Signal reflection
Eye opening
Jitter
Over or under shoot
Hi/Lo margin
Insertion loss
Return loss
Group delayS-parameter
Eye Diagram
24
Nanya CCL
W mil
Type1 1 mil.
1 i
nch
1 inch
B yarns per inch
S parameter for different Dk (1)
Dk W (mil)
4.5 6
4 6
3.5 6 0 2 4 6 8 10-50
-45
-40
-35
-30
-25
-20
-15
Rut
urn
Los
s
Frequency (GHz)
Dk4.5 Dk4 Dk3.5
0 2 4 6 8 10-0.50
-0.45
-0.40
-0.35
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
Los
s (d
B)
Frequency (GHz)
Dk4.5 Dk4 Dk3.5
• Trace places on fabric.
25
Nanya CCL 26
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
Dk=4.5
Dk=4
10 Gb/s
Eye Diagram Test
Dk=3.5
A small sensitivity to Dk will be presentedif MS line places on fabric of PCB.
Nanya CCL
W mil
Type21 mil.
1 i
nch
1 inchB yarns per inch
S parameter for different Dk (2)
Dk W (mil.)
4.5 6
4 6
3.5 60 2 4 6 8 10
-60
-55
-50
-45
-40
-35
-30
-25
-20
-15
Ret
ern
Los
s (d
B)
Frequency (GHz)
Dk4.5 Dk4 Dk3.5
0 2 4 6 8 10-0.45
-0.40
-0.35
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
Los
s (d
B)
Frequency (GHz)
Dk4.5 Dk4 Dk3.5
• Trace places on gaps and crosses.
27
Nanya CCL
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
28
Dk=4.5
Dk=4
10 Gb/s Eye Diagram Test
Dk=3.550 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
A large sensitivity to Dk will be presented if MS line places on gaps and crossesof PCB.
Nanya CCL
Experimentfor
Single Line
DK
Rotation
Density
Parameter Measurement Evaluation
Impedance Frequency response
Impedance variation
Signal reflection
Eye opening
Jitter
Over or under shoot
Hi/Lo margin
Insertion loss
Return loss
Group delayS-parameter
Eye Diagram
29
Nanya CCL 30
Construction B yarns v.s. C yarns per inch
0 2 4 6 8 10-0.5
-0.4
-0.3
-0.2
-0.1
0.0L
oss
(dB
)
Frequency (GHz)
60 70
• When the yarns of construction increased, the loss of transmission line will be reduced.
• 45 degree angle microstrip line shows the reliable performance compared to traditional vertical
or lateral transmission line layout owing to a better distribution of glass fabric and glue beneath
the microtrip line can be achieved.
• If the yarns less than B, the non-uniform of glass fabric and glue distribution caused a slow
wave effect which resulted in the unstable transmission line for vertical and lateral layouts.
0 2 4 6 8 10-45
-40
-35
-30
-25
-20
Ret
urn
Los
s (d
B)
Frequency (GHz)
60 70
BC
BC
Nanya CCL 31
B yarns C yarns
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
50 100 150 200 250 3000 350
0.0
0.2
0.4
0.6
0.8
1.0
-0.2
1.2
time, psec
AA
10 Gb/s
Eye Diagram Test
• C-yarns transmission line presents the performances of good 50-Ω impedance
and low transmission loss.
• Due to small insertion and return losses for C-yarns transmission line, a low jitter
and a low voltage variation in Hi/Lo level will be exhibited for 10-Gb/s eye
diagram test.
Nanya CCL 32
Outline:
-Demand of High Data Rate For Transmission Line Design
-What Influences Skew
- How to Reduce Skew
- Conclusion
Nanya CCL 33
Conclusion:
In order to improve skew issue our next step are as follow:
1. New technique for better spread out the glass cloth.
2. Use denser fabric.
Nanya CCL
THANKS
34
Recommended