Ultra-Wide-Band (UWB) Band-Pass-Filter Using Integrated Passive Device (IPD) Technology for Wireless Applications J 17 2009 June 17, 2009 STATS ChipPAC D&C YongTaek Lee STATS ChipPAC D&C YongTaek Lee Rev01 Rev01
Ultra-Wide-Band (UWB) Band-Pass-Filter Using Integrated Passive Device (IPD) Technology for Wireless Applications
J 17 2009June 17, 2009
STATS ChipPAC D&CYongTaek Lee
STATS ChipPAC D&CYongTaek Lee
Rev01 Rev01
Agenda
• Introduction
• Design and characterization for flip-chip IPD
D ig d h t i ti f Wi b di g IPD• Design and characterization for Wire bonding IPD
• The simple triple wire-bond and philips/TU Delgt tripe i b d i d t d lwire-bond inductance model
• Conclusions
22
Introduction
• Most common applications of Integrated passives devices (IPDs) are in the front-end of wireless systems, between the antenna and t i transceiver.
• Integrated passives devices (IPDs) based on semiconductor processes offer the advantage of excellent parameter control and processes offer the advantage of excellent parameter control, and allow simplified and compact module design.
• IPD processes can be used to make high density capacitors, high Q p g y p , g Qinductors and large value resistors.
• The Ultra Wide Band (UWB) band-pass-filter developed in this paper h th ll t i hil hi i i l t l t i l has the smallest size while achieving equivalent electrical performance.
• In this paper an UWB Band-pass-filter is made using lumped • In this paper, an UWB Band pass filter is made using lumped integrated passive devise technology on a silicon substrate for wireless applications.
33
The technology of UWB
• UWB offers great potentials for home networking, wireless sensors, and location aware system.
• UWB Characteristics and signal.
Low power BW = (fH - fL)/fc = 2(fH - fL)/(fH + fL) High data rate Narrowband BW < 1%
Wideband 1% < BW < 20%Ultra-Wide-Band BW > 20%
C i b d t (30KH )Trans
2G 2G 4GCellular(Mobile)WLAN
Carrier based system (30KHz)
Wireless LAN (5MHz)
smission P
ZigBee Bludtooth
802.15.3
ADSL VDSL
WLAN
Wireline
UWB(a few GHz)
Frequency
Power
0.01 0.1 1 10 50 100 400 1000
USB1.1 USB2.0WirelinePAN
44
FrequencyData Rate (Mbps)
Integrated Passive Device Process Description
• In the STATS ChipPAC’s silicon process, a specially treated silicon substrate is used to grow dielectric layer and metal layer. g y y
• Capacitors: < 100pF, inductors: 20nH (Q 25-45), Resistors: < 100K Ohm.
Component Componentsurface
Component surface
WIRE-BOND FLIP-CHIP
Figure1: Thin film Integrated Passive Device (IPD) structure ( not in scale)
Inductor CapacitorResistor
55
structure ( not in scale).
RF Product Design Examples
• GSM LPF, Balun.
• DCS LPF Balun• DCS LPF, Balun.
• 11b/g BPF, Balun.
• 11a BPF Balun• 11a BPF, Balun.
• WiMax BPF, Balun.
• Diplexers: GSM/DCS 11b/a WiMax• Diplexers: GSM/DCS, 11b/a, WiMax.
• Compact designs: Balanced filter, Balanced diplexers.p
• RF Modules (CSMP) and RF SiP.
66
IPD on wafer being probedIPD on substrate
IPD RF Applications (side by side with a die)
IPD product IPD product databook available from the website.
77
IMS2008, Atlanta
UWB Band-Pass-Filter for Flip-Chip IPD
• Figure 2 shows a circuit topology for the band-pass filter.
• Flip-chip IPD layout of the UWB band-pass-filter for EM simulation .Flip chip IPD layout of the UWB band pass filter for EM simulation .Two bumps are for UWB band-pass-filter input and output.
Four bumps are just for electrical ground.
The UWB band-pass-filter of flip-chip die has a size of 1.4mm x 1.2mm x 0.40mm (including bump height).
PORT06
C05C08
PORT03
C04
PORT06
PORT05
C02L01 C07C01
PORT02
PORT03
C03
L02
C06
L03
C03 C06
PORT01 PORT04
Figure 2: Circuit topology for UWB band- Figure 3: Flip-chip IPD layout of the
88
pass-filter (BPF). UWB band-pass-filter for EM simulation.
UWB Band-Pass-Filter for Flip-Chip IPD
• To meet electrical performance and size target a general design methodology (Figure 4) was followed.gy ( g )
Step: 1) Create circuit model for IPDStep: 2) Generate physical layout to fit available space, and perform EM simulationStep: 3) Optimize as required to meet Step: 3) Optimize as required to meet specifications.
Figure 4: Design methodology of integrated passives.
99
UWB Band-Pass-Filter for Flip-Chip IPD
• The simulated characteristics of the UWB band-pass-filter are shown in Figure 5.g
• The insertion loss from 7GHz to 9GHz is 1.8dB and the return loss is greater than 15dB in EM-simulation.
B)
1dB
) 0.0et
urn
loss
(S11
dBns
ertio
n lo
ss(S
2
-40.0
-20.0
Pass
band
rePa
ssba
nd in
-80.0
-60.0Passband retrun loss S11
Passband insertion loss S21
Frequency (GHz)
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0-100.0
Figure 5: S11 and S21 parameters for the UWB band-pass-filter in simulation.
1010
g 11 21 p p
Fabrication
• Passive integration on silicon substrate.
• Low insertion loss in pass band• Low insertion loss in pass band.• Eutectic Sn/Pb or lead-free solder
bump.• Low profile, 0.40mm height.• Directly flipped on PCB.
Operating temperature: 40 to • Operating temperature: -40 to +85°C.
• Storage temperature: -40 to +85°C.
Figure 6: UWB band-pass-filter flip-chip die.
1111
Wafer Level Test
• Membrane probes for non-destructive wafer-level test.
• PCM for line width, leakage current, capacitance, inductor Q, etc.Ensure yield before wafers/dies shipped out
RF Wafer Membrane Probe
• Ensure yield before wafers/dies shipped out.
1212
Characterization on Laminate Board• S-Parameters were measured with wire bonding IPD die through
probing on the G-S-G patterns on the test board.Verify the response in package.Temperature controller used for IPD temp characterization.
IPD FlippedProbe Station
G-S-G patterns on the test boardTemperature Controller
1313
UWB Band-Pass-Filter Results for Flip-Chip IPD
• Typical characteristics of the manufactured UWB band-pass-filter are shown in Figure 7.g
• The insertion loss is 1.7dB (minimum) and return loss is 15dB.• The manufactured UWB band-pass-filter has low pass-band insertion
loss and small si eloss and small size.
B) 0.0
B)
0.0
on L
oss
(dB
-40.0
-20.0
Insertion Loss n Lo
ss (d
B
-15.0
-10.0
-5.0Return Loss
Inse
rtio
-80.0
-60.0
Ret
urn
-30.0
-25.0
-20.0
Frequency (GHZ)0.0 5.0 10.0 15.0 20.0
-100.0
Frequency (GHz)0.0 5.0 10.0 15.0 20.0
-35.0
Figure 7: Typical characteristics of the manufactured UWB band-pass-filter.
1414
g y
UWB Band-Pass-Filter Results for Flip-Chip IPD
• Typical characteristics for the filp-chip UWB band-pass filter.
• In high frequency applications our simulation scheme is very • In high frequency applications, our simulation scheme is very suitable for designing IPD products.
Specification Units Min Typ Max Design Measurement
Table 1: Typical characteristics for the flip-chip UWB band-pass filter.
Passband frequency MHz 7000 9000
Passband insertion loss dB 2.0 1.8 2.0
P b d t l dB 10 15 15Passband return loss dB 10 15 15
Attenuation, 824-915 MHz dB 55 55 50
Attenuation, 1710-1910 MHz dB 35 60 53
Attenuation, 2400-2500 MHz dB 30 55 55
Attenuation, 4900-5900 MHz dB 15 16 18
Attenuation , 12000-15000 MHz dB 15 30 20
1515
UWB Band-Pass-Filter for Wire Bonding IPD• The circuit design and fabrication of the wire-bond UWB band-pass-
filter IPD are similar to those described above for the flip-chip device. device.
• The device can be mounted directly on a PCB or laminate substrate using conventional wire-bonding techniques.
• The UWB band-pass-filter of wire-bonding die has a size of 1.2mm x 1.0mm x 0.25mm.
1616
Figure 8: Wire bonding IPD layout of the UWB band-pass -filter for EM simulation.
UWB Band-Pass-Filter for Wire Bonding IPD
• The circuit-level simulation was done using a simple inductance model (0 35nH) for each of the triple wire bonds model (0.35nH) for each of the triple wire-bonds.
S 11 d
B)
ss (S
21dB
)
-20.0
0.0 Table 2: Typical characteristics for the UWB band-pass filter.
Specification Units Min Typ Max Design
nd re
turn
loss
(nd
inse
rtio
n lo
s
-60.0
-40.0
Passband return lossPassband insertion loss
Passband frequency MHz 7000 9000
Passband insertion loss dB 2 1.6
Passband return loss dB 10 15
Attenuation 824 915 MHz dB 55 50
Pass
ban
Pass
ban
-120.0
-100.0
-80.0 Attenuation, 824-915 MHz dB 55 50
Attenuation, 1710-1910 MHz dB 35 53
Attenuation, 2400-2500 MHz dB 30 55
Attenuation, 4900-5900 MHz dB 15 18
Frequency (GHz)
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
Figure 9: S11 and S21 parameter for UWB band-pass-filter in simulation.
,
Attenuation , 12000-15000 MHz
dB 15 20
1717
p
Fabrication
• The device can be mounted directly on a PCB or laminate substrate using conventional wire-bonding techniques.g g q
• RF test board layout shows an UWB band-pass-filter with 3 with bonds to input pad (#3) and output pad (#6).The four ground pads (#1 #2 #4 #5) are also connected with triple • The four ground pads (#1, #2, #4, #5) are also connected with triple wire bonds pads.
G S G patterns on the test boardG-S-G patterns on the test board
Figure 10: UWB band pass filter wire bond die and RF test board layout
1818
Figure 10: UWB band-pass-filter wire-bond die and RF test board layout.
UWB Band-Pass-Filter for Wire Bonding IPD
• The characteristics of the UWB band-pass-filter of wire bonding die are shown in Figure 11.g
The insertion loss is 2.4dB (Minimum) and the return loss is 7dB.
• Compared to the results for the flip chip UWB filter, these results
) 0.0
B)
0
are much worse than expected.
n Lo
ss (d
B)
-10.0
-5.0Return Loss
on L
oss
(dB
40
-20
0
Ret
urn
-20.0
-15.0
Inse
rtio
80
-60
-40 Insertion Loss
Frequency (GHz)0.0 5.0 10.0 15.0 20.0
-25.0
Frequency (GHz)0.0 5.0 10.0 15.0 20.0
-80
Figure 11: Measured characteristics of the UWB band-pass-filter of wire bonding die.
1919
g p g
Tripe wire-bond inductance model
• Figure 12 shows a comparison of the simulated result versus measurement using the simple triple wire bond inductance model.g p p
• The simple model does not account for mutual interactions between the wire bonds, which become more important at higher f i
(S11
,S21
dB
)S 2
1 dB
)
-20.0
0.0
frequencies.
It can be seen that the agreement between the two is poor. Because
e W
B m
odel
dat
a (
rem
ent d
ata
(S11
,
-80.0
-60.0
-40.0
Simple WB model dataMeasurement data
between the two is poor. Because of the good agreement in the flip-chip case, it was suspected the cause of the discrepancy was in th i l i d t d l d
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0Si
mpl
eM
easu
-120.0
-100.0
the simple inductance model used for the wire bonds.
Frequency (GHz)
Figure 12: Measurement versus simulation using the simple triple wire-bond inductance models for wire-bonding IPD.
2020
models for wire bonding IPD.
Wire-bond shape model
BONDW_ShapeShape1
Top figure applies
Shape1
StartH=250 umGap=700 umRw=12.5 um
Stretch=100 umTilt=50 umMaxH=350 umStartH 250 um
FlipX=1StopH=0 um
• StartH is 250um (top of IPD)• StopH is 0 (plane of the test board)• Gap is 700um (total length of the wire bond in the x y plane)• Gap is 700um (total length of the wire bond in the x-y plane)• MaxH is 350um (a 100um loop height – typical.) • Tilt and stretch are chosen to give the wire bond a reasonable h
2121
shape.
Simulation setup
• This is the same as the simple model, but the wire-bond inductors have been replaced by the ADS WIRESET modelhave been replaced by the ADS WIRESET model.
Wires labeled 1-18 t l k i
pad_2
pad_1 S_ParamSP1
Step=5 MHzStop=15 GHzStart=0.01 GHz
S-PARAMETERSBONDW_ShapeShape1
Stretch=100 umTilt=50 umMaxH=350 umStartH=250 umGap=700 umRw=12.5 um
1
2
3
4
5
6
7Term
counter-clockwise
1 18Wires labeled 1-18
pad_2pad_1
pad_3
pad_4pad_5
pad_6
pad_5
pad_4
pad_3
S6PSNP1Fil "UWB 8GHZ 3WB FINAL it 1 1 6 "
1 5
4
6
3 Ref2
FlipX=1StopH=0 umStretch=100 um
8
9
10
11
12
13
14
15T
Term1
Z=50 OhmNum=1 counter-clockwise
pad_6
File="UWB_8GHZ_3WB_FINAL_runit_1_1.s6p"
BONDW18WIRESET1
W18_Xoffset=465 umW18_Shape="Shape1"W17_Angle=0W17_Zoffset=0 um
W15_Xoffset=465 umW15_Shape="Shape1"W14_Angle=0W14_Zoffset=0 um
W12_Xoffset=465 umW12_Shape="Shape1"W11_Angle=90W11_Zoffset=0 um
W9_Xoffset=-285 umW9_Shape="Shape1"W8_Angle=90W8_Zoffset=0 um
W6_Xoffset=-465 umW6_Shape="Shape1"W5_Angle=180W5_Zoffset=0 um
W3_Xoffset=-465 umW3_Shape="Shape1"W2_Angle=180W2_Zoffset=0 um
Layer="cond"View=sideCond=1.3e7 SRadw=12.5 um
16
17
18
TermTerm2
Z=50 OhmNum=2
W18_Angle=0W18_Zoffset=0 umW18_Yoffset=-365 um
W17_Xoffset=465 umW17_Shape="Shape1"W16_Angle=0W16_Zoffset=0 umW16_Yoffset=-185 umW16_Xoffset=465 umW16_Shape="Shape1"W15_Angle=0W15_Zoffset=0 umW15_Yoffset=0 um
W14_Xoffset=465 umW14_Shape="Shape1"W13_Angle=0W13_Zoffset=0 umW13_Yoffset=180 umW13_Xoffset=465 umW13_Shape="Shape1"W12_Angle=90W12_Zoffset=0 umW12_Yoffset=365 um
W11_Xoffset=375 umW11_Shape="Shape1"W10_Angle=0W10_Zoffset=90 umW10_Yoffset=365 umW10_Xoffset=285 umW10_Shape="Shape1"W9_Angle=90W9_Zoffset=0 umW9_Yoffset=365 um
W8_Xoffset=-375 umW8_Shape="Shape1"W7_Angle=90W7_Zoffset=0 umW7_Yoffset=365 umW7_Xoffset=-465 umW7_Shape="Shape1"W6_Angle=180W6_Zoffset=0 umW6_Yoffset=180 um
W5_Xoffset=-465 umW5_Shape="Shape1"W4_Angle=180W4_Zoffset=0 umW4_Yoffset=0 umW4_Xoffset=-465 umW4_Shape="Shape1"W3_Angle=180W3_Zoffset=0 umW3_Yoffset=-185 um
W2_Xoffset=-465 umW2_Shape="Shape1"W1_Angle=180W1_Zoffset=0 umW1_Yoffset=-365 umW1_Xoffset=-465 umW1_Shape="Shape1"Zoffset=0 umSepY=0 umSepX=0 um
2222
W17_Yoffset=-275 umW14_Yoffset=90 umW11_Yoffset=365 umW8_Yoffset=365 umW5_Yoffset=90 umW2_Yoffset=-275 um
Tripe wire-bond inductance model• An improved circuit-level simulation was done using the Philips /TU
Delft wire-bond models in ADS.• These account for a more detailed shape of the wire bond and also
account for mutual inductances between all of the wires.• The comparison between the IPD characteristics using the simple The comparison between the IPD characteristics using the simple
triple wire bond model and the more accurate Philips/TU Delft model is shown in Figure 13.
data
(S11
, S21
dB
)a
(S11
, S21
dB
)
-40
-20
0
mpl
e W
B m
odel
dS
WB
mod
el d
ata
-80
-60
Simple WB model dataADS WB model data
Frequency (GHz)
0 2 4 6 8 10 12 14
Sim
AD
S-120
-100
Figure 13: Simulation using the simple triple wire-bond inductance model versus simulation using the Philips/TU Delft triple wire bond
2323
Philips/TU Delft triple wire bond inductance models.
Tripe wire-bond inductance model
• Figure 14 shows a comparison of the measured data versus simulation using the Philips TU/Delft wire-bond inductance model for wire-bond IPD. With the more accurate wire-bond models, the agreement is much better.
• At these high frequencies, the simple wire-bond models are not sufficiently accurate.
dB)
1 dB
) 0
sufficiently accurate.
nt d
ata
( S11
, S21
de
l dat
a (S
11, S
2
-60
-40
-20
Mea
sure
men
AD
S W
B m
od
-100
-80Measurement dataADS WB model data
Frequency (GHz)
0 2 4 6 8 10 12 14-120
Figure 14: Measurement versus simulation i th Phili /TU D lft t i l i b d
2424
using the Philips/TU Delft triple wire-bond inductance models.
Conclusions
• The design and implementation of a silicon based band-pass-filter for Ultra Wide Band applications have been presented.pp p
• Excellent filter properties are obtained from the UWB band-pass-filter of flip-chip die.
• For wire bondable IPDs working at high frequencies (such as this UWB filter), simple inductance model for multiple wires is not good enough for designs.enough for designs.
• More advanced coupled-wire models in ADS have shown better predictions of the wire behaviors.
• The IPD technology is especially well suited for UWB applications because of its excellent parameter control and enabling smaller from factorsfrom-factors.
2525