Integrated Passive Device (IPD) Technology for Wireless Applications

Ultra-Wide-Band (UWB) Band-Pass-Filter Using Integrated Passive Device (IPD) Technology for Wireless Applications June 17, 2009 STATS ChipPAC D&C YongTaek Lee Rev01

Agenda Introduction Design and characterization for flip-chip IPD Design and characterization for Wire bonding IPD The simple triple wire-bond and philips/tu hl Delgt tripe wire-bond inductance model Conclusions 2

Introduction Most common applications of Integrated passives devices (IPDs) are in the front-end of wireless systems, between the antenna and transceiver. Integrated passives devices (IPDs) based on semiconductor 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 inductors and large value resistors. The Ultra Wide Band (UWB) band-pass-filter developed in this paper has the smallest size while achieving equivalent electrical performance. In this paper, an UWB Band-pass-filter is made using lumped integrated t passive devise technology on a silicon substrate t for wireless applications. 3

The technology gy of UWB UWB offers great potentials for home networking, wireless sensors, and location aware system. system UWB Characteristics and signal. ¾ Low power ¾ High data rate WLAN 2G 2G ZigBee Bludtooth VDSL ADSL Wireline PAN USB1.1 0.1 BW < 1% 1% < BW < 20% BW > 20% 4G 802 15 3 802.15.3 0.01 Narrowband Wideband Ultra-Wide-Band Transmis ssion Pow wer Cellular (Mobile) WLAN ¾ BW = (fh - fl)/fc = 2(fH - fl)/(fh + fl) 1 10 USB2.0 50 100 400 Carrier based system (30KHz) Wireless LAN (5MHz) UWB(a few GHz) 1000 Frequency q y Data Rate (Mbps) 4

Integrated Passive Device Process Description In the STATS ChipPAC s silicon process, a specially treated silicon substrate t is used to grow dielectric i layer and metal layer. Capacitors: < 100pF, inductors: 20nH (Q 25-45), Resistors: < 100k Ohm. Component surface Component surface WIRE-BOND FLIP-CHIP Figure1: Thin film Integrated Passive Device (IPD) structure ( not in scale). Resistor Inductor Capacitor 5

RF Product Design Examples GSM LPF, Balun. DCS LPF, Balun. 11b/g BPF, Balun. 11a BPF, Balun. WiMax BPF, Balun. Diplexers: GSM/DCS, 11b/a, WiMax. Compact designs: Balanced filter, Balanced diplexers. RF Modules (CSMP) and RF SiP. IPD on wafer being probed IPD on substrate 6

IPD RF Applications (side by side with a die) IPD product databook available from the website. IMS2008, Atlanta 7

UWB Band-Pass-Filter for Flip-Chip p 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. 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). PORT03 C05 C08 PORT06 C04 C01 L01 C02 C07 PORT02 PORT05 L02 L03 C03 C06 PORT01 PORT04 Figure 2: Circuit topology for UWB band- pass-filter (BPF). Figure 3: Flip-chip IPD layout of the UWB band-pass-filter for EM simulation. 8

UWB Band-Pass-Filter for Flip-Chip p IPD To meet electrical performance and size target a general design methodology (Figure 4) was followed. Step: 1) Create circuit model for IPD Step: 2) Generate physical layout to fit available space, and perform EM simulation Step: 3) Optimize as required to meet specifications. Figure 4: Design methodology of integrated passives. 9

UWB Band-Pass-Filter for Flip-Chip p IPD The simulated characteristics of the UWB band-pass-filter are shown in Figure 5. The insertion loss from 7GHz to 9GHz is 1.8dB and the return loss is greater than 15dB in EM-simulation. s (S 11 db) loss(s 21 db) and return loss and insertion l Passba Passba 0.0-20.0-40.0-60.0-80.0 Passband retrun loss S 11 Passband insertion loss S 21-100.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 Frequency (GHz) Figure 5: S 11 and S 21 parameters for the UWB band-pass-filter in simulation. 10

Fabrication Passive integration on silicon substrate. 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 +85 C. Storage temperature: -40 to +85 C. Figure 6: UWB band-pass-filter flip-chip die. 11

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 12

Characterization on Laminate Board S-Parameters were measured with wire bonding IPD die through probing on the G-S-G G patterns on the test board. Verify the response in package. Temperature controller used for IPD temp characterization. IPD Flipped Probe Station G-S-G patterns on the test board Temperature Controller 13

UWB Band-Pass-Filter Results for Flip-Chip p IPD Typical characteristics of the manufactured UWB band-pass-filter are shown in Figure 7. 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 size. Inse ertion Los ss (db) 0.0-20.0-40.0-60.0-80.0 Insertion Loss Ret turn Loss s (db) 0.0-5.0-10.0-15.0-20.0-25.0-30.0 Return Loss -100.0 00 0.0 50 5.0 10.00 15.0 20.00 Frequency (GHZ) -35.0 0.0 5.0 10.0 15.0 20.0 Frequency (GHz) Figure 7: Typical characteristics of the manufactured UWB band-pass-filter. 14

UWB Band-Pass-Filter Results for Flip-Chip p IPD Typical characteristics for the filp-chip UWB band-pass filter. In high frequency applications, our simulation scheme is very suitable for designing IPD products. Table 1: Typical characteristics for the flip-chip UWB band-pass filter. Specification Units Min Typ Max Design Measure ment Passband frequency MHz 7000 9000 Passband insertion loss db 2.0 1.8 2.0 Passband 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 15

UWB Band-Pass-Filter for Wire Bonding IPD The circuit design and fabrication of the wire-bond UWB band-passfilter IPD are similar to those described above for the flip-chip 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. Figure 8: Wire bonding IPD layout of the UWB band-pass -filter for EM simulation. 16

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. ) db) n loss (S 11 db) tion loss (S 21 d ssband return ssband insert Pas Pas 0.0 Table 2: Typical characteristics for the UWB band-pass filter. -20.0-40.0-60.0 Passband return loss Passband insertion loss -80.0-100.0-120.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 Frequency (GHz) Figure 9: S 11 and S 21 parameter for UWB band-pass-filter in simulation. Specification Units Min Typ Max Design 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 Attenuation, 1710-1910 1910 MHz db 35 53 Attenuation, 2400-2500 MHz db 30 55 Attenuation, 4900-5900 MHz db 15 18 Attenuation, 12000-15000 db 15 20 MHz 17

Fabrication The device can be mounted directly on a PCB or laminate substrate using conventional wire-bonding techniques. 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 wire bonds pads. G-S-G patterns on the test board Figure 10: UWB band-pass-filter wire-bond die and RF test board layout. 18

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. 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 are much worse than expected. s (db) Inse ertion Los 0-20 -40-60 Insertion Loss Ret turn Loss (db) 0.0-5.0-10.0-15.0-20.0 Rt Return Loss -80 0.0 5.0 10.0 15.0 20.0 Frequency (GHz) -25.0 0.0 5.0 10.0 15.0 20.0 Frequency (GHz) Figure 11: Measured characteristics of the UWB band-pass-filter of wire bonding die. 19

Tripe wire-bond inductance model Figure 12 shows a comparison of the simulated result versus measurement using the simple triple wire bond inductance model. The simple model does not account for mutual interactions between the wire bonds, which become more important at higher frequencies. It can be seen that the agreement 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 the simple inductance model used for the wire bonds. db) el data (S 11,S 21 ata (S 11, S 21 db) Simple WB mod Measurement da 0.0-20.0-40.0-60.0-80.0-100 100.0 0 Frequency (GHz) Simple WB model data Measurement data -120.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 Figure 12: Measurement versus simulation using the simple triple wire-bond inductance models for wire-bonding IPD. 20

Wire-bond shape model BONDW_Shape Shape1 Rw=12.5 um Gap=700 um StartH=250 um MaxH=350 um Tilt=50 um Stretch=100 um StopH=0 um FlipX=1 Top figure applies 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) MaxH is 350um (a 100um loop height typical.) Tilt and stretch are chosen to give the wire bond a reasonable shape. 21

Simulation setup This is the same as the simple model, but the wire-bond inductors have been replaced by the ADS WIRESET model. Term Term1 Num=1 Z=50 Ohm Term Term2 Num=2 Z=50 Ohm S-PARAMETERS 1 BONDW_Shape 2 pad_1 S_Param Shape1 3 SP1 Rw=12.5 um Start=0 0.01GHz 4 Gap=700 um Stop=15 GHz 5 pad_2 StartH=250 um Step=5 MHz 6 MaxH=350 um Tilt=50 um 7 Stretch=100 um 8 pad_3 StopH=0 um 9 FlipX=1 pad_6 10 pad_1 6 pad_5 1 5 11 pad_4 pad_2 pad_4 2 4 12 3 Ref 13 pad_3 14 pad_5 S6P 15 SNP1 16 File="UWB_8GHZ_3WB_FINAL_runit_1_1.s6p" 17 pad_6 18 BONDW18 WIRESET1 Radw=12.5 um W2_Zoffset=0 um W5_Zoffset=0 um W8_Zoffset=0 um W11_Zoffset=0 um W14_Zoffset=0 um W17_Zoffset=0 um Cond=1.3e7 S W2_Angle=180 W5_Angle=180 W8_Angle=90 W11_Angle=90 W14_Angle=0 W17_Angle=0 View=s ide W3_Shape="Shape1" W6_Shape="Shape1" W9_Shape="Shape1" W12_Shape="Shape1" W15_Shape="Shape1" W18_Shape="Shape1" Layer="cond" W3_Xoffset=-465 um W6_Xoffset=-465 um W9_Xoffset=-285 um W12_Xoffset=465 um W15_Xoffset=465 um W18_Xoffset=465 um SepX=0 um W3_Yoffset=-185 um W6_Yoffset=180 um W9_Yoffset=365 um W12_Yoffset=365 um W15_Yoffset=0 um W18_Yoffset=-365 um SepY=0 um W3_Zoffset=0 um W6_Zoffset=0 um W9_Zoffset=0 um W12_Zoffset=0 um W15_Zoffset=0 um W18_Zoffset=0 um Zoffset=0 um W3_Angle=180 W6_Angle=180 W9_Angle=90 90 W12_Angle=90 W15_Angle=0 W18_Angle=0 W1_Shape="Shape1" W4_Shape="Shape1" W7_Shape="Shape1" W10_Shape="Shape1" W13_Shape="Shape1" W16_Shape="Shape1" W1_Xoffset=-465 um W4_Xoffset=-465 um W7_Xoffset=-465 um W10_Xoffset=285 um W13_Xoffset=465 um W16_Xoffset=465 um W1_Yoffset=-365 um W4_Yoffset=0 um W7_Yoffset=365 um W10_Yoffset=365 um W13_Yoffset=180 um W16_Yoffset=-185 um W1_Zoffset=0 um W4_Zoffset=0 um W7_Zoffset=0 um W10_Zoffset=90 um W13_Zoffset=0 um W16_Zoffset=0 um W1_Angle=180 W4_Angle=180 W7_Angle=90 W10_Angle=0 W13_Angle=0 W16_Angle=0 W2_Shape="Shape1" W5_Shape="Shape1" W8_Shape="Shape1" W11_Shape="Shape1" W14_Shape="Shape1" W17_Shape="Shape1" W2_Xoffset=-465 um W5_Xoffset=-465 um W8_Xoffset=-375 um W11_Xoffset=375 um W14_Xoffset=465 um W17_Xoffset=465 um W2_Yoffset=-275 um W5_Yoffset=90 um W8_Yoffset=365 um W11_Yoffset=365 um W14_Yoffset=90 um W17_Yoffset=-275 um Wires labeled 1-18 counter-clockwise 1 18 Wires labeled 1-18 counter-clockwise 22

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 triple wire bond model and the more accurate Philips/TU Delft model is shown in Figure 13. Figure 13: Simulation using the simple triple wire-bond inductance model versus simulation using the Philips/TU Delft triple wire bond inductance models. S 21 db) 1 db) Simple WB mo odel data (S 11, ADS WB mode el data (S 11, S 21 0-20 -40-60 -80-100 Frequency (GHz) Simple WB model data ADS WB model data -120 0 2 4 6 8 10 12 14 23

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. S 11, S 21 db) (S 11, S 21 db) ement data ( S B model data ( Measure ADS WB 0-20 -40-60 Measurement data -80 ADSWB model dldt data -100-120 0 2 4 6 8 10 12 14 Frequency (GHz) Figure 14: Measurement versus simulation using the Philips/TU Delft triple wire-bond inductance models. 24

Conclusions The design and implementation of a silicon based band-pass-filter for Ultra Wide Band applications have been presented. 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. 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-factors. 25