Millimeter-Wave Amplifiers for E- and V-band Wireless Backhaul Erik Öjefors Sivers IMA AB

Transcription

1 Millimeter-Wave Amplifiers for E- and V-band Wireless Backhaul Erik Öjefors Sivers IMA AB THz-Workshop: Millimeter- and Sub-Millimeter-Wave circuit design and characterization 26 September 2014, Venice

2 Outline mm-wave wireless backhaul for mobile networks Dotseven technology Low-noise amplifiers Power amplifiers Conclusions 2

3 Millimeter-Wave Wireless Backhaul 3G/4G mobile networks need more backhaul capacity fiber not always an option Sivers IMA provides radio-frontend products for mm-wave radio-link backhaul since 2008 Unlicensed V-band (60 GHz) for short distances E-band (71-76 GHz/81-86 GHz) for medium range 3

4 The mm-wave Transceiver Antenna Diplexer LNA RX I/Q LO xn GHz PA Up- / Down-converter LO xn GHz TX I/Q Full-duplex FDD (separate TX, RX, external diplexer) telecom grade Switch from III-V MMICs to custom integrated Si/SiGe ICs LNA and PA critical blocks: RX NF < 8 db, TX PSAT > 20 dbm 4

5 The mm-wave Transceiver Present Solution TX IF Waveguide TX module TX compression GHz LO Psat = 20 dbm at Waveguide port SiGe Upconverter GaAs PA TX RX Dotseven goals: Replace GaAs PA Improve SiGe LNA SiGe RX Gain / NF Performance LNA Gain > 20 db SiGe RX Chip NF 6.5 db Measurements results refer to performance at the RX chip input bond pad 5

6 Dotseven in the IHP SG13 Technology Comparison of the IHP 130-nm SiGe BiCMOS technologies SG13S f t /f max = 250/300 GHz, typical performance of advanced industrially available SiGe HBT technologies SG13G2 f max = 300 / 500 GHz, Dotseven starting point Dotseven target f max > 700 GHz Faster technologies may benefit mm-wave telecom transceivers by providing: LNAs with lower NF, < 4 db needed to compete with III-Vs MMICs Efficient power amplifiers with Psat > 20 dbm 6

7 mm-wave LNA design E-band Low-Noise Amplifiers 7

8 mm-wave LNA design Common-Emitter HBT Simplified Noise Model f max and NF B Vbn Rb C ZS ZL ZS Ibn Icn ZL NF minimized by improving ft and minimizing Rb (and Re) of the device. Better signal-to-shot-noise ratio at the collector node with higher ft Lower Rb minimizes Vbn, and lets us reduce Zs Re(Zs) = Ropt, best compromise between noise from Vbn and Ibn. E Problem: Re(Zopt) Re(Zin), noise match yields poor return loss! 8

9 mm-wave LNA design Simultaneous Noise and Impedance Matching 1. Select current density for best NF Lb 2. Scale device for Re(Zopt) = 50 Ohm Zs Zopt Zin Ldeg 3. Increase Re(Zin) to 50 Ohm by degeneration using Ldeg 4. Cancel out Imag (Zin) by Lb Method described in detail by Voinigescu et al (IEEE JSSC, 1997) Parasitic inductance in base and emitter leads absorbed by Lb/Ldeg. Negative feedback difficult to apply when operating close to ft/fmax 9

10 mm-wave LNA design Common-Emitter Stage Cascode VCC VCC Vbias Ibias Ibias ZS ZL ZS ZL Lowest NF Miller capacitance complicates matching Low gain and isolation Higher gain, slightly higher NF Higher gain improves cascaded system NF Decoupled input and output 10

11 300 um mm-wave LNA design Dotseven LNA Design V B 2:1 xfmr VCC RF Out Input Input stub xfmr Output Ibias Q2 320 um RF In ESD TL Stub Rbias Q1 RF In ESD protection shunt stub at input Input series inductance replaced by stub Real part of impedance levels retuned with stub Wide-band transformer output match Q1-Q2: (8 x 0.96 x 0.12) mm 2, biased at I C = 7 ma 11

12 S-param (db) mm-wave LNA design Single-Stage LNA S-Parameters S Meas. Sim. S22 S11 P1 P2 f (GHz) S12 >10 db gain over GHz frequency range Input return loss >10 db 12

13 mm-wave LNA design E-Band NF Measurement Setup E-band noise source ENR = 14 db, switched hot/cold Downconverter LO IF Spectrum Analyzer 13

14 NF [db] mm-wave LNA design NF Results Two-Stage Cascaded E-Band LNA Two-stage LNA In Out Meas [db] Sim [db] Simulation and measurements in good agreement Target NF < 4 db (Commercial E- band GaAs LNA: NF = 3-5 db) Freq. [GHz] 14

15 E-Band Power Amplifiers Power-Combined E-band Amplifiers 15

16 E-Band Power Amplifiers Large-Device Design for >20 dbm Output Power Zload << 50W Limited voltage swing requires high current for high output power Large or multiple parallel power devices Low load impedance, transformation losses, distributed parasitics Potential electro-thermal stability issues Power-Combining Amplifier Zload Modular approach Good dc stability, individually biased amplifiers Power-combining losses Space-consuming design, less heat concentration 16

17 Power-Combined E-Band PAs Power-Amplifier Unit Cell VCC VCC T2 RF Out In T1 T2 Out VCC Q3 V B Q4 Differential cascode Q1-Q4: (16 x 0.96 x 0.12) mm 2, biased at I CQ = 20 ma VCC = 3.3V RF In T1 Q1 Q2 Current-summing combining req. high unit cell impedance T1 2:1 ratio, Zin = 100 W T2 1:1 ratio, Zload = 100 W 17

18 Power-Combined E-Band PAs Single cell Sim Psat = 16 dbm Microstrip in-phase combiner 4-way comb. sim Psat = 21 dbm 8-way comb. sim Psat = 24 dbm Electrically short, currentcombining network Lumped matching circuit Combining network EM-simulated IL = 1 db 18

19 E-Band Power Amplifiers Single-Stage, 8-way Combined PA, S-parameters >10 db gain GHz 8-GHz shift of measured input tuning vs simulation, bias dependent Biased at 3.3V, 330 ma 19

20 E-Band Power Amplifiers Power Sweep of an 8-way Combined Amplifier incl. Driver at 76 GHz Meas Psat: 22 dbm Sim: 24 dbm --- Measured Simulated Two-stage cascade of a 4- way-combined driver with an 8-way-combined PA Mid-band (76 GHz) Psat = 22 dbm, 1-2 db roll-off at band edges DC bias 3.3V, 500 ma incl. Driver Further details to be presented at EuMIC 2014, Rome 20

21 Conclusions mm-wave amplifiers are critical components in wireless backhaul transceivers E-band LNA NF < 4 db possible with Dotseven technology, comparable to commercial III-V LNAs PA Psat > 20 dbm possible, faster technology could improve the bandwidth Low power consumption important (Power-overEthernet, cooling), amplifier efficiency will benefit from higher f max 21

22 Acknowledgement This work is part of the: DOTSEVEN project supported by the European Commission through the Seventh Framework Program for Research and Technological Development The author would like to acknowledge the support of Bernd Heinemann and Holger Rücker, IHP 22