Large-Signal Nonlinear Model of a Highly Integrated
Quad-Band GSM Transmit Front-End Module
Y. Tkachenko, S. Boerman, H-C. Chung, P. DiCarlo, M. Gerard, J. Gering, J. Hu, A. Klimashov, K. Kwok, P. Reginella, S. Sprinkle, C. Wei and Y. Yang
Skyworks Solutions, 20 Sylvan Road, Woburn, MA 01801, Gene.Tkachenko@skyworksinc.com
Abstract
A large-signal nonlinear model was developed for a highly integrated quad-band (UGSM/EGSM/DCS/PCS) transmit front-end module (TX-FEM). The model includes separately validated large-signal nonlinear models of HBT amplifiers; large-signal nonlinear PHEMT switch models; signal nonlinear Si BJT PA controller model; large-signal EM-co-simulated model of a dual-band directional coupler-detector, including large-signal GaAs MESFET diodes; and Momentum or lumped element based model of the multi-chip module (MCM) level passive networks. To authors’ best knowledge this is the highest complexity large-signal transistor-based model of a radio sub-system ever reported.
INTRODUCTION
Higher levels of PA and transmit module complexity [1-3] and aggressive design cycles for new products are driving the need for more accurate models of these RF sys-tems, including models of all active and passive compo-nents placed into a module and their interactions. Robust nonlinear models are required that are capable of predicting large-signal module parameters critical for meeting the product specifications, such as output power, efficiency, nonlinearity, etc. Such models provide valuable insight into MCM operation and serve as powerful tools in their design and optimization.
In this paper we present a model of a quad-band (UGSM/EGSM/DCS/PCS) TX-FEM, which integrates two power amplifiers, a PA controller (PAC), T/R switches with controller, a dual-band directional detector/coupler, a diplexer, matching networks and harmonic filters in a sin-gle, 50 : input and output, 9x10x1.5mm package [3]. The module schematic is shown in Figure 1. It employs InGaP/GaAs HBT, AlGaAs/InGaAs/AlGaAs PHEMT, GaAs epi MESFET Schottky/passive, and Si Schot-tky/bipolar/CMOS semiconductor technologies. The mod-ule delivers 34dBm Pout with 45% PAE at UGSM/EGSM and 31dBm Pout and 36% PAE at DCS/PCS, while meet-ing a VSWR>20:1 open loop ruggedness spec.
HBT PA models
The three-stage InGaP/GaAs HBT PA IC models are developed using scalable large area HBT array models of
up to 11,000 um2 in area developed in-house. These array
models are based on the VBIC core unit cell model with extrinsic passive elements describing parasitic elements
associated with combiner networks for multiple cells [4], as shown in Figure 2a. Extrinsic element values are extracted for individual array sizes and then fitted for different array sizes using table-based approach and resulting in a single model for a variety of cells.
The model is validated for different unit cell types and different array sizes using IV, over temperature, s-parameter, power, PAE, load-pull and non-linearity (inter-modulation distortion and ACPR) characteristics. Figures 2b-e demonstrate some of the simulated characteristics using the HBT array model, which are in good agreement with the measured data: Figure 2b compares the measured
vs. modeled output IV characteristics of the 960um2 HBT
array under forced Ib operation; Figure 2c shows measured
and modeled bias-dependent S-parameters of the 960um2
HBT array; Figure 2d demonstrates good agreement be-tween the measured and modeled power, gain, PAE and Ic
characteristics of a 3600um2 HBT array under a particular
bias, drive, input and output loading conditions during on-wafer power sweep; Figure 2e shows the model prediction of the 3rd and 5th order inter-modulation distortion products under a load-pull sweep across the whole Smith chart being in good agreement with measurements.
The models of PA IC’s with the Si PA controller were validated separately in a generic quad-band module envi-ronment without any post-PA components, except the out-put matching networks. Si BJT models of the PAC were based on the models provided by the foundry. Figure 3 shows reasonable agreement between the measured and modeled PA module Gain, Pout and PAE characteristics vs. the swept Si controller input voltage.
DCS/PCS PAMMIC UGSM/EGSM PA MMIC V_control Band Select RF In (Low Band) RF In (High Band) Bias Controller Vcc1(HB) Output Matching Network Output Matching Network Vcc2(HB)Vcc3(HB) Vcc1(LB) Vcc2(LB) Vcc3(LB) Detector/ Coupler TX Enable Antenna RX DCS RX PCS RX UGSM RX EGSM Switch Controller T/R RX Upper/Lower Vdet Vref Detector Enable DCS/PCS PAMMIC UGSM/EGSM PA MMIC V_control Band Select RF In (Low Band) RF In (High Band) Bias Controller Vcc1(HB) Output Matching Network Output Matching Network Vcc2(HB)Vcc3(HB) Vcc1(LB) Vcc2(LB) Vcc3(LB) Detector/ Coupler TX Enable Antenna RX DCS RX PCS RX UGSM RX EGSM Switch Controller T/R RX Upper/Lower Vdet Vref Detector Enable
Figure 1. TX-FEM block diagram.
Figure 2a. HBT array model diagram.
0 5
Figure 2b. Modeled vs measured IV characteristics of a
960um2 HBT array under forced Ib conditions.
Ga in q56 x 60LP_ P m ax ..p ae ff Pin f1 f3 q56 x 60LP_ P m ax .. p 2 DC1. i, A DC1. D C. V b, V B _m eas
Figure 2d. Measured (symbols) vs. modeled (lines) Gain, Pout,
PAE and Ic characteristics of a 3600um2 HBT array under
on-wafer power sweep test. 900 MHz, Vc=3.2V, Ic0=420mA. Gamma_Load=0.81<177.7, Gamma_Source=0.68<164.7 freq (400.0MHz to 20.20GHz) S m eas (2, 2) S m odel (2,2) freq (400.0MHz to 20.20GHz) S m eas (1, 1) S m odel (1,1) Core HBT Model
with _M=No. transistors in array
1 2 3 4 0.00 0.05 0.10 0.15 0.20 -0.05 0.25 VC DC. IC1. Ic _m eas 1 2 3 4 0 5 1.1 1.2 1.3 1.0 1.4 VC V -30 -20 -10 0 10 20 30 freq (400.0MHz to 20.20GHz) Sm ea s( 2, 1) S m od el (2, 1) -0 .1 5 -0 .1 0 -0 .0 5 0. 00 0.05 0.10 0.15 -0 .2 0 0. 20 freq (400.0MHz to 20.20GHz) S m ea s( 1, 2) S m od el (1, 2)
Figure 2c. Measured S-parameters of a 960um2 array vs.
large-signal HBT array model. Vcb=0.5V, Ib=8e-6 to 1.6mA.
-3 m1 Pin=-3.000 Gain=21.478 -25 -20 -15 -10 -5 0 5 10 -30 15 0 10 20 30 -10 40 10 20 30 40 50 60 0 70 Pin m1 PAE Po ut pin q56x 60 LP_Pm a x. .pou t q56x 60 LP_Pm a x. .gain -25 -20 -15 -10 -5 0 5 10 -30 15 0.35 0.40 0.45 0.50 0.55 0.30 0.60 re al (Ic c .i[: :,0 ]) pin q56 x 60LP_ P m ax .. i2 -25 -20 -15 -10 -5 0 5 10 0 15 -120 -100 -80 -60 -40 -20 0 20 -140 40 Pin f2 pin q56 x 60LP_ P m ax .. p 3 q56 x 60LP_ P m ax .. p 1 10 20 30 40 50 60 70 80 90 100 0 1 0 -35 -30 -25 -20 -15 -10 -5 0 -40 5 k IM 5 H IM 5 L IM 5 H I_ m e a s IM 5 L O _m ea s 10 20 30 40 50 60 70 80 90 100 0 110 -40 -35 -30 -25 -20 -15 -10 -5 0 5 -45 10 k IM 3 H IM 3 L IM 3 H I_ m e a s IM 3 L O _m ea s 1
Figure 2e. Measured (symbols) vs. modeled (lines) two-tone inter-modulation distortion characteristics (IM3 and IM5 in
dBm) of the 960um2 HBT array during loadpull test at 900
MHz, 3.2V, Pin=-2dBm. Index k specifies a particular load state, with the entire state space covering roughly a
VSWR=7:1 circle on the Smith chart.
Figure 3. Measured (symbol) vs. modeled (line) GSM and DCS characteristics of a generic quad-band PA module. 900
and 1750 MHz, Vcc=3.2V, Pin=6dBm.
PHEMT SP3T models
SP3T switch models are based on the previously re-ported in-house developed single-gate and multi-gate large-signal nonlinear PHEMT models [5-6]. These models take into account charge conservation, dispersion and self-heating and incorporate accurate models of leakage cur-rents. As shown in Figure 4, the critical capacitance nonlin-earities in the device are well predicted by the model. Figure 5 demonstrates good prediction of the insertion loss
and isolation by the model. The 2nd and 3rd harmonic
be-havior vs. control voltage and power is also predicted well.
Figure 4. Measured vs. modeled PHEMT Cgs characteristics.
Coupler-Detector Model
Large-signal model of the dual-band 900/1800 MHz directional coupler-detector was constructed based on the
large-signal GaAs epi-based Schottky diode models for the reference and detector diodes and Momentum-based EM model of the directional couplers, which were also built on the GaAs die (Figure 6). Measurement-based s-parameter model library was constructed for the MCM level resistor, inductor and capacitor components. Models for the external biasing and matching components of the coupler-detector were taken from this library. Figure 7 shows detected volt-age vs. input power simulations based on the models of the two different coupler-detector designs, which were in agreement with the measurements. While the detected volt-age characteristics for these two designs are relatively
simi-lar, as shown in Figure 8, the difference in the 5th GSM
harmonic distortion between them can be as much as 20dB.
This result was in agreement with the overall 5th harmonic
level of the entire TX-FEM and helped pinpoint the source of distortion being the coupler-detector, not PA or a switch, and select the best design for the overall module perform-ance. 0.00E+00 2.00E-13 4.00E-13 6.00E-13 8.00E-13 1.00E-12 1.20E-12 1.40E-12 1.60E-12 -3 -2 -1 0 1 Vds (V) Cg s (F ) 0.5 1.0 1.5 0.0 2.0 -60 -40 -20 0 20 -80 40 10 20 30 40 50 60 70 80 90 0 100 basev Po u t Gai n PA E G S M_meas .. gai n G S M_meas .. pou t
Figure 5. Measured vs. modeled insertion loss and isolation characteristics of an SP3T PHEMT switch.
Figure 6. Large-signal GaAs coupler-detector model using Momentum EM co-simulated coupler model and large-signal diode models. Library-based bias and match components are
added externally. G S M_ me a s.. p a e D C S _ m e a s.. p a e 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0.0 2.0 -60 -40 -20 0 20 -80 40 10 20 30 40 50 60 70 80 90 0 100 basev Po ut Ga in P A E D C S_ me a s..g ain D C S _ me a s..p o ut RE0783_1_coup RE0783_1_COUP_1 ModelType=RF R e f MCM_RF_WIRE X4 PLEN=500 u LoopHt=175 u Type=WB Port VDET_EN_P4 Num=4 MCM_RF_WIRE X5 PLEN=500 u LoopHt=175 u Type=WB Port VREF_EN_P5 Num=5 MCM_RF_WIRE X6 PLEN=500 u LoopHt=175 u Type=WB Port VREF_P6 Num=6 MCM_RF_WIRE X7 PLEN=500 u LoopHt=175 u Type=WB Port BYPASS_P7 Num=7 MCM_RF_WIRE X3 PLEN=500 u LoopHt=175 u Type=WB Port VDET_P3 Num=3 GaAs_Schottky_diode_PP X17 PP GaAs_Schottky_diode_PP X16PP GaAs_Schottky_diode_PP X15 P P GaAs_Schottky_diode_PP X14 P P MCM_RF_WIRE X9 PLEN=500 u LoopHt=175 u Type=WB Port HB_out_P9 Num=9 Port LB_out_P8 Num=8 MCM_RF_WIRE X8 PLEN=500 u LoopHt=175 u Type=WB MCM_RF_WIRE X2 PLEN=500 u LoopHt=175 u Type=WB Port LB_in_P2 Num=2 MCM_RF_WIRE X1 PLEN=500 u LoopHt=175 u Type=WB Port HB_in_P1 Num=1 R R1 R=93.47 Ohm C C9 C=0.1 pF
-80 -60 -40 -20 0 20 -100 40 0.0 0.5 1.0 1.5 2.0 -0.5 2.5 Pin re al (V de lta _V 14 [n f,: :,0 ]) re al (V de lta _V 10 P [n f,::,0 ])
Figure 7. Detected voltage vs. input power simulated for two different coupler-detector designs using large-signal models.
Figure 8. 5th GSM harmonic level simulated using the
large-signal models for two different coupler-detector designs.
Figure 9. Schematic of a 60-port s-parameter block represen-tation of the post-PA output passive network on the MCM,
EM simulated in Momentum.
Figure 10. Measured vs. modeled (EM and lumped element) s-parameters of the post-PA IC output passive network.
MCM level passive network modeling
Passive elements including dc and rf lines located at various MCM levels, inter-level and grounding vias were simulated in Momentum and represented as a 60-port s-parameter block for large-signal module simulations. Fig-ure 9 shows the diagram of this highly complex geometri-cal shape. Passive SMT element models for capacitors, filters and diplexer were connected externally to the s-parameter block ports. Figure 10 compares simulated EM and lumped element model s-parameters to measured data, which was obtained by probing directly on the module.
CONCLUSIONS
Components of a highly integrated quad-band transmit front-end module large-signal model are de-scribed. Robust core HBT, PHEMT, Si BJT, FET diode cell and array models, combined with EM co-simulated or measurement-based passive IC level and module level pas-sive components, form a complete large-signal transistor level module simulation environment. This high complex-ity TX-FEM model constructed provides valuable insight into various module characteristics, including IV, s-parameters, power and harmonics and serves as essential part of the final product design. As a result, 34/31dBm out-put with 45/36% PAE in GSM/DCS bands was achieved for the entire TX module.
0 1 0 2 0 3 0 -1 0 4 0 -5 0 -4 0 -3 0 -2 0 -1 0 -6 0 0 P in dB m ('' V 10P _det _ci rc ui t'' .. V rf _l b[ nf ,: :, 5] ) 5 F @ L B O U T ACKNOWLEDGMENTS
The authors wish to thank many people at Skyworks who contributed to this work including: I. Khayo, I. Lalicevic, D. Evans, R. Burton, C. Jacobs, K. Frey, D. Britton, K. Crompton, D. Hirani, B. Bousquet, D. Bartle, R. Ramana-than, W.-J. Ho, P. Zampardi, and M. Glasbrener.
REFERENCES
[1] R. Jos, “Technology developments driving evolution of cellular
phone power amplifiers to integrated RF front-end modules,” IEEE Journal of Solid-State Circuits, No. 9, Vol. 36, Sep. 2001, pp. 1382-1389.
[2] D. Wu, M. Frank, N. Hendin, and S. Yajima, “Radio front-end
mod-ule technologies and design for multi-band/mode cellular handsets,” IMS-2003 short course digest on “RF FEM’s for cellular phone and WLAN applications,” 2003.
[3] P. DiCarlo, S. Boerman, R. Burton, H.- C. Chung, D. Evans, M.
Gerard, J. Gering, I. Khayo, L. Lagrandier, I. Lalicevic, P. Reginella, S. Sprinkle, and Y. Tkachenko, “A highly integrated quad-band
GSM TX-front-end-module,” Proceedings of the 25th GaAs IC
Sym-posium, Nov. 2003, pp. 280-283.
[4] C.-J. Wei, J. Gering, S. Sprinkle, A. Klimahsow, and Y. A.
Tkachenko,” Modified VBIC-model for InGaP/GaAs HBTs,” Digest of 2002 APMC, Kyoto, Japan, WEOF10, Nov. 2002
[5] C. Wei, D. Johnson, O. Manzhura, Y. Tkachenko and D. Bartle,
“Large-signal model of triple-gate MESFET/PHEMT for switch ap-plications,” in 1999 Internat. Microwave Symp. Dig. pp. 745-749
[6] C. J. Wei, Y. A. Tkachenko, and D. Bartle, "A new analytic model
for power E-mode PHEMTs and its optimum power operation,” IEEE Transactions on Microwave Theory and Techniques, Vol. MTT-50, Jan.2002, pp.57-61 dB m ('' V 14_det _ci rcui t'' .. V rf _l b[ nf ,: :, 5] ) 1 2 3 4 5 0 6 -40 -30 -20 -10 -50 0 freq, GHz dB (m eas_v14_dcs_out m at ch. .S (2 ,1 )) dB (m odel _v 14_d cs _out m at ch_m ic rost rip. .S (2 ,1 ) dB (m odel _v14_dc s_out m at ch_E M .. S (2 ,1 ))
