A 60GHz 16Gb/s 16QAM Low-Power Direct-Conversion Transceiver Using Capacitive Cross-Coupling Neutralization in 65nm CMOS Hiroki Asada, Keigo Bunsen, Kota Matsushita, Rui Murakami, Qinghong Bu, Ahmed Musa, Takahiro Sato, Tatsuya Yamaguchi, Ryo Minami, Toshihiko Ito, Kenichi Okada, Akira Matsuzawa Tokyo Institute of Technology, Japan
Outline Motivation Previous Work Challenges for 60GHz Transceiver Capacitive Cross-Coupling Neutralization Transceiver Design Measurement Results Conclusion 2
Motivation 60GHz CMOS direct-conversion transceiver for multi-gbps wireless communication IEEE 802.15.3c specification 57.24GHz - 65.88GHz 2.16GHz/ch x 4channels QPSK 3.5Gbps/ch 16QAM 7.0Gbps/ch 3
Previous work 1 Direct-conversion transceiver by UCB[1] 90 o hybrid is used to generate I/Q signal 4Gb/s for QPSK(Ch2) 16QAM is unsupported [1] C. Marcu, et al., ISSCC 2009, pp. 314-315 4
Previous work 2 Direct-conversion transceiver by Tokyo Tech[2] Quadrature LO is used to generate I/Q signal 11Gb/s for 16QAM(Ch2) Not fully-balanced design large I/Q mismatch LNA I Mixer I+ I- 60GHz QILO Rx input Q Mixer Q+ Q- 20GHz PLL Tx output PA I Mixer Q Mixer I+ I- 60GHz QILO PLL 36MHz REFCLK [2] K. Okada., et al., ISSCC 2011, pp. 160-161 Q+ Q- 5
Gain Flatness at RF band f LO (60GHz) CG 2dB Down-conv. 2dB 1.76GHz-BW Gain Flatness 0dB 1dB 2dB EVM - -22dB -18dB 0 0.88GHz Constellation 6
Parasitic Capacitance Parasitic Capacitance 35 30 C GD = CGD Im[ Y 12 ] jω Max Gain[dB] 25 20 15 10 5 0 W=40um 0 20 40 60 80 100 120 Frequency[GHz] lower MAG Parasitic Capacitances causes low reverse isolation and low gain. 7
Capacitive Cross-Coupling IN C x C x OUT A cross-coupled capacitor between gate and drain of the opposite-side transistor works as negative capacitor. MAG is improved about 5dB at 60GHz MAG, MSG [db] 40 35 30 25 20 15 10 5 0 w/ CCC w/ CCC(calc.) w/o CCC 0 20 40 60 80 100 120 Frequency [GHz] [3] W.L. Chan, et al., ISSCC 2009 8
Direct-Conversion Architecture Fully-balanced direct-conversion transceiver Capacitive cross-coupling neutralization Baseband LNA IMixer BB LNA RF Amp. LO Buf. I+ I- VGA, ADC 36MHz Ref. Clk Rx input 60GHz 20GHz PLL LNA LO Buf. RF Amp. QMixerBB LNA Q+ Q- VGA, ADC PLL IMixerBB Amp. RF Amp. LO Buf. I+ I- DAC 19.44GHz, 20.16GHz, 20.88GHz, 21.60GHz Tx Output PA RF Frontend 60GHz LO Buf. RF Amp. QMixerBB Amp. Q+ Q- DAC Capacitive Cross Coupling Amplifier 9
Up-Conversion Mixer Double-balanced Gilbert mixer Capacitive cross-coupling neutralization MIM TL TL from LO LO p LO n RF p RF n to PA IF p IF n from BB I/Q 10
Mixer Core Layout Mixer core excluding intersection LO line and RF line cross in matching network Mixer core including intersection bad symmetrical property RF+ RF+ LO+ LO+ LO- RF- Symmetrical core (Not Good) RF- LO- Asymmetrical core (Good) 11
Mixer Core Layout Symmetrical core needs crossed and complicated matching network. Asymmetrical core can realize simple matching network. RF+ IF+ IF- RF+ IF+ IF- LO+ LO- RF- LO+ LO- Symmetrical core RF- Asymmetrical core 12
SRR measurement Asymmetrical core shows higher Sideband Rejection Ratio(SRR) and low I/Q mismatch Symmetrical core Asymmetrical core SRR Amplitude Error Phase Error -24.5 [db] 0.04[dB] 6.8[deg] -42.3[dB] 0.02[dB] 0.9[deg] Simple layout of mixer can make I/Q mismatch negligible. 13
3-Stage PA TL-based design for simulation accuracy Low-loss TL & MIM TL MIM TL TL from Mixer W=2µm x20 2µm x30 2µm x40 14
Tx Measurement CG: 16dB P DC : 181mW P sat : 6.5dBm(ch2) P 1dB : 5.4dBm(ch2) 15
BB LNA CCC amplifier with a source-follower buffer. To compensate Noise Figure from Mixer IN OUT To BB I/Q 16
Down-Conversion Mixer Parallel-line transformer Capacitive cross-coupling neutralization from LNA RF MIM TL TL from LO LO p LO n IF p IF n to BB LNA 17
4-Stage CS-CS LNA Wf=1µm (1 st & 2 nd stages) for noise opt. Wf=2µm (3 rd & 4 th stages) for gain opt. Variable gain by adjusting bias voltages MIM TL TL ESD protection to Mixer W=1µm x40 1µm x40 2µm x20 2µm x20 18
Rx Measurement Conversion Gain [db] 35 30 25 20 15 10 5 0 ch1(high-gain mode) ch1(low-gain mode) ch2(high-gain mode) ch2(low-gain mode) 57.24 58.32 59.40 60.48 61.56 RF Frequency [GHz] NF [db] 8 7 6 5 ch1 4 ch2 3 0.0 0.2 0.4 0.6 0.8 1.0 1.2 BB Frequency [GHz] CG: 17-27dB NF: <6.1dB (ch2) LO freq.: 60.48GHz (ch2) Lower cut-off freq.: 4MHz P DC : 138mW 19
60GHz Quadrature LO 36MHz ref. 20GHz PLL Injection-lock 60GHz QILO PFD CP LPF (27,28,29,30) 5 4 CML 19.44GHz 20.16GHz 20.88GHz 21.60GHz Q I 58.32GHz 60.48GHz 62.64GHz 64.80GHz Wide frequency tuning range Phase noise improvement by injection locking [4] A. Musa, et al., ASSCC 2010 20
Quadrature Injection-Locked Oscillator I n 60GHz Q n Dummy V DD Matching block INJ n INJ p 20GHz I p 60GHz Q p Phase noise :-94.2dBc/Hz@1MHz-offset Free-running frequency: 55-63 GHz 21
Die Photo LNA I Mixer LO Buffer. Q Mixer LO Buffer. Quadrature Oscillator 65nm CMOS Rx:2.5mm 2 Tx:2.3mm 2 PLL:1.2mm 2 4.2mm I Mixer LO Buffer. Buf. VCO LPF PA Quadrature Oscillator 20GHz PLL Q Mixer LO Buffer. 4.2mm 22
Package and PCB 60GHz Rx 2dBi antenna 60GHz Tx 2dBi antenna Face-up mount with a 270µm wire on a BGA package [5] R. Suga, et al., IEEE T-MTT 2010 DC supply DC supply I/Q output (Rx) 20GHz PLL I/Q input (Tx) 23
Measured Spectrum 1.760Gs/s QPSK with 25% roll-off, 3dB back-off Magnitude [db] 10 0-10 -20-30 -40 IEEE802.15.3c spectrum mask -50 55.08 56.16 57.24 58.32 59.40 60.48 61.56 Frequency [GHz] 24
Modulation Characteristics Constellation 9506 points 19912 points 13502 points 42024 points Modulation QPSK 16QAM QPSK 16QAM Data rate (BER <10-3 ) EVM (with DFE) 3.52Gb/s 7.04Gb/s 10.0Gb/s 16.0Gb/s -30.5dB -28.2dB -15.2dB -16.1dB 10Gb/s(QPSK) and 16Gb/s(16QAM) with wider-bw 25
Performance Comparison Data rate / Modulation EVM Direct conv. Power U. Toronto[6] 4Gb/s(BPSK) N/A Yes 374mW UCB [1] 4Gb/s(QPSK) 7Gb/s(QPSK) (loop-back) N/A Yes 170mW(Tx mode) 138mW(Rx mode) Tokyo Tech[2] 8Gb/s(QPSK) 11Gb/s(16QAM) -17dB (Tx Rx) Yes 186mW(Tx mode) 106mW(Rx mode) CEA-LETI[7] 3.8Gb/s(16QAM) -20.7dB(Tx) -19.2dB(Rx) No 1357mW(Tx mode) 454mW(Rx mode) SiBeam[8] 3.8Gb/s(16QAM) -19.2dB (Tx Rx) No 1820mW(Tx mode) 1250mW(Rx mode) This work 10Gb/s(QPSK) 16Gb/s(16QAM) -28.2dB (Tx Rx) Yes 181mW(Tx mode) 138mW(Rx mode) [6] A. Tomkins, et al., JSSC, vol.44, no.8, pp.2085-2099, Aug. 2009 [7] A. Siligaris, et al., ISSCC 2011., pp. 162-163 [8] S. Emami, et al., ISSCC 2011, pp. 164-165 26
Summary and Conclusion A 60GHz 16Gb/s 16QAM Low-Power Direct- Conversion Transceiver. Consideration of mixer layout. Capacitive Cross-Coupling Neutralization. Full-rate 16QAM/8PSK/QPSK/BPSK for IEEE802.15.3c Ch1(57.24-59.40GHz) and Ch2(59.40-61.56GHz) Standard 65nm CMOS Tx (181mW), Rx (138mW), and PLL (66mW) 27