复旦大学：微电子工程教学资源（参考论文）A Low Noise High Linearity CMOS Upconversion Mixer

团购合买资源类别：文库，文档格式：PDF，文档页数：4，文件大小：1.08MB

A Low Noise High Linearity CMOS Upconversion Mixer Liming Jin,Zhangwen Tang*and Hao Min State Key Laboratory of ASIC and System,Fudan University,Shanghai 201203,P.R.China Abstract-In this paper,a low noise high linearity mixer conclusions are given in Section V. is presented,exploiting a switched transconductor topology.Its noise figure (NF)and linearity are analyzed II.MIXER TOPOLOGY particularly.The mixer chip is implemented in 0.18-um A.Topologies Comparison RF CMOS process.The measurement result shows that Figure.1 (a)shows the conventional Gilbert mixer.It the conversion gain of the mixer is about 8dB,the SSB consists of a transconductance stage (M5,M6),switch NF is about 1ldB,and the input-referred third-order pairs(MI~M4)and load impedances.The low frequency intercept point (IIP3)is about 10.5dBm.The chip voltage signal is converted to the current signal by the consumes 10mA at 1.8V power supply and the size of the transconductance stage,and then switched between the whole chip is 0.63mmx0.78mm. two signal paths by the switch pairs.Finally the load impedances convert the current signal back to the voltage I.INTRODUCTION signal. Upconversion mixers are important modules in RF Figure.1 (b)shows a switched transconductor transceivers,converting the low frequency baseband(BB) upconversion mixer topology[1].It also consists of signal to RF band.The NF,which determines the transconductance stages (M1-M4),switch pairs sensitivity of the circuit,is a key performance of the (M5~M8)and load impedances.But the relative position upconversion mixer.The IIP3 determines the maximum of the former two is changed,and the switch pairs are signal level that the mixer can handle.A mixer with low replaced with inverters.In the positive half cycle of LO, NF and high IIP3 has larger dynamic range.This paper is the output of left inverter(consists of M5 and M6)is at organized as follows.In Section II,advantages of the low voltage level,while the output of right inverter is switched transconductor mixer topology are analyzed high.So,the left transconductance pair of MI and M2 is compared with the conventional one.In Section III,the switched on,while the right pair is off.In the negative schematic of an upconversion mixer is introduced. half cycle of LO,the situation is opposite.The frequency Simulation and measurement results of the designed translating principle is just the same as the conventional mixer are presented in Section IV.Finally,the mixers VDD VDD o RF+RF- oRF+RF-o 8B+0◆ B VDD BB+M5 oBB VSS白 (a)Conventional topology (b)Switched transconductor topology Figure.1.Mixer topologies *Corresponding author.Email:zwtang@fudan.edu.cn 1-4244-1132-7/07/$25.00©20071EEE

A Low Noise High Linearity CMOS Upconversion Mixer Liming Jin, Zhangwen Tang* and Hao Min State Key Laboratory of ASIC and System, Fudan University, Shanghai 201203, P.R. China Abstract – In this paper, a low noise high linearity mixer is presented, exploiting a switched transconductor topology. Its noise figure (NF) and linearity are analyzed particularly. The mixer chip is implemented in 0.18-μm RF CMOS process. The measurement result shows that the conversion gain of the mixer is about 8dB, the SSB NF is about 11dB, and the input-referred third-order intercept point (IIP3) is about 10.5dBm. The chip consumes 10mA at 1.8V power supply and the size of the whole chip is 0.63mm×0.78mm. I. INTRODUCTION Upconversion mixers are important modules in RF transceivers, converting the low frequency baseband (BB) signal to RF band. The NF, which determines the sensitivity of the circuit, is a key performance of the upconversion mixer. The IIP3 determines the maximum signal level that the mixer can handle. A mixer with low NF and high IIP3 has larger dynamic range. This paper is organized as follows. In Section II, advantages of the switched transconductor mixer topology are analyzed compared with the conventional one. In Section III, the schematic of an upconversion mixer is introduced. Simulation and measurement results of the designed mixer are presented in Section IV. Finally, the conclusions are given in Section V. II. MIXER TOPOLOGY A. Topologies Comparison Figure.1 (a) shows the conventional Gilbert mixer. It consists of a transconductance stage (M5, M6), switch pairs (M1~M4) and load impedances. The low frequency voltage signal is converted to the current signal by the transconductance stage, and then switched between the two signal paths by the switch pairs. Finally the load impedances convert the current signal back to the voltage signal. Figure.1 (b) shows a switched transconductor upconversion mixer topology[1]. It also consists of transconductance stages (M1~M4), switch pairs (M5~M8) and load impedances. But the relative position of the former two is changed, and the switch pairs are replaced with inverters. In the positive half cycle of LO, the output of left inverter (consists of M5 and M6) is at low voltage level, while the output of right inverter is high. So, the left transconductance pair of M1 and M2 is switched on, while the right pair is off. In the negative half cycle of LO, the situation is opposite. The frequency translating principle is just the same as the conventional mixers. VDD VDD VSS LO+ LOBB+ BBM5 M6 M7 M8 RF+ RFZL ZL M1 M4 M2 M3 (a) Conventional topology (b) Switched transconductor topology Figure.1. Mixer topologies *Corresponding author. Email:zwtang@fudan.edu.cn 1-4244-1132-7/07/$25.00 © 2007 IEEE

B.Noise Figure VDD The flick noise and thermal noise current generated by a MOSFET are as equation(1)and(2)respectively [2]. RF-o They are both functions of gm.When the MOSFET is cut off,the noise can be negligible.Since the LO switches the transconductance stages periodically, only half of the transconductor MOSFETs are VDDT switched on at one time.Though the number of the othsho transconductor MOSFETs doubles in the switched transconductor topology,the noise introduced is just w网MPL&-48412 the same as the conventional topology. VSs专 K 1 Fu -C.WL79 (1) Figure.2.Schematic of the upconversion mixer m=4kTygm (2) NMOS of corresponding inverter is at linear region, acting as a resistor Ro.The output differential current can In the switched transconductor topology,the noise be derived as equation(5)[1],where the gi,gz and gs current generated by the switches is in common mode, denote the Taylor coefficients derived from the I-V of the because the two MOS transistors from one MOSFET. transconductantor pair are working at the same state. Then the noise will be cancelled in the differential i。=gVae+ 93- 9R。） 41+2g,R。) (5) outputs.But in the conventional topology,the situation is totally different.Each switch transistor will contribute Though g3 is negative when the MOSFET is in strong notable output noise [3].That's why the switched inversion,the deterioration of the linearity resulted from transconductor topology has lower NF. Ro can be negligible if Ro is small enough. The approximate SSB NF of Gilbert mixers in Figure.1(a)has been derived as equation(3)[4],where III.UPCONVERSION MIXER DESIGN the a and c are LO related parameters,and Y,rg and gm Figure.2 shows the schematic of the low noise high denote the noise factor,gate resistance and linearity upconversion mixer.To achieve better switching, transcondcutrance respectively. the size of the inverter should be increased.Out of question,that will lead to larger power dissipation.To NFssa solve this problem,two inverters with small size are =Q c2 2s+s9s9a+4G+4nG+1VZ（③) cascaded to replace the single inverter switches.And the inverters with small size make the VCO design easier, c2gisRs reducing the driving capability requirement. LC tank is employed as output loads,as the output Since the noise generated by the switch is cancelled, signal is narrow RF band.There's no expense of any DC the SSB NF of the switched transconductor topology can voltage headroom,So that the mixer can work in the low be derived as equation(4)[1]. voltage power supply.In this design,a center-tapped differential inductor [5]is implemented,instead of two 2(71+rogm)gma+1Z NFssa (4) independent inductors to saves the chip area.The C2 c2giRs differential equivalent parallel resistance of the LC tank can be as high as Ik ohms,which reduce the requirement C.Linearity of gm to achieve given conversion gain(CG).According When the transconductance stage is switched on,the to equation (6),the over drive voltage of the common-source node of it is at low voltage.So,the transconductance NMOS can be large,leading to good

B. Noise Figure The flick noise and thermal noise current generated by a MOSFET are as equation (1) and (2) respectively [2]. They are both functions of gm . When the MOSFET is cut off, the noise can be negligible. Since the LO switches the transconductance stages periodically, only half of the transconductor MOSFETs are switched on at one time. Though the number of the transconductor MOSFETs doubles in the switched transconductor topology, the noise introduced is just the same as the conventional topology. 2 2 ,1/ 1 nf m ox K i g C WL f = (1) 2 i kT g n thm m , = 4 γ (2) In the switched transconductor topology, the noise current generated by the switches is in common mode, because the two MOS transistors from one transconductantor pair are working at the same state. Then the noise will be cancelled in the differential outputs. But in the conventional topology, the situation is totally different. Each switch transistor will contribute notable output noise [3]. That’s why the switched transconductor topology has lower NF. The approximate SSB NF of Gilbert mixers in Figure.1 (a) has been derived as equation (3) [4], where the α and c are LO related parameters, and γ, rg and gm denote the noise factor, gate resistance and transcondcutrance respectively. NFSSB ( ) 2 5 5 5 5 11 11 2 22 5 2 44 1 gm m g L m s rg g G rG Z c cg R α γ αγ + ++ + = + (3) Since the noise generated by the switch is cancelled, the SSB NF of the switched transconductor topology can be derived as equation (4) [1]. ( ) 1 11 1 2 22 1 2 1 gm m L SSB m s rg g Z NF c cg R α γ α + + = + (4) C. Linearity When the transconductance stage is switched on, the common-source node of it is at low voltage. So, the Figure.2. Schematic of the upconversion mixer NMOS of corresponding inverter is at linear region, acting as a resistor Ro. The output differential current can be derived as equation (5) [1], where the g1,g2 and g3 denote the Taylor coefficients derived from the I-V of the MOSFET. 2 3 2 3 1 4 12 1 o o RF RF o g gR i gv v g R ⎛ ⎞ = +− ⎜ ⎟ + ⎝ ⎠ (5) Though g3 is negative when the MOSFET is in strong inversion, the deterioration of the linearity resulted from Ro can be negligible if Ro is small enough. III. UPCONVERSION MIXER DESIGN Figure.2 shows the schematic of the low noise high linearity upconversion mixer. To achieve better switching, the size of the inverter should be increased. Out of question, that will lead to larger power dissipation. To solve this problem, two inverters with small size are cascaded to replace the single inverter switches. And the inverters with small size make the VCO design easier, reducing the driving capability requirement. LC tank is employed as output loads, as the output signal is narrow RF band. There’s no expense of any DC voltage headroom, So that the mixer can work in the low voltage power supply. In this design, a center-tapped differential inductor [5] is implemented, instead of two independent inductors to saves the chip area. The differential equivalent parallel resistance of the LC tank can be as high as 1k ohms, which reduce the requirement of gm to achieve given conversion gain (CG). According to equation (6), the over drive voltage of the transconductance NMOS can be large, leading to good

linearity The input referred third-order intercept point(IIP3)is %s-4=2g measured with the two-tone test at 9.5MHz and (6) 9m 10.5MHz.Figure.5 shows the output spectrum when the power level of every input tone is-11dBm.The central IV.SIMULATION AND MEASUREMENT RESULT two peaks are the two-tone signals,while the side two The chip is implemented in SMIC 0.18-um RF CMOS peaks are the third order intermodulation products. process.Figure.3 is the photograph of the chip after Figure.6 shows measured IIP3 with input level sweeping bonding and the test PCB board.The chip size including Though the output is attenuated due to impedance IO pads is 0.63mm0x0.78mm.And it consumes 10mA mismatch,the IIP3 is not impacted absolutely,since it current at 1.8V power supply on average.The chip is lies on the difference between the 1st and 3d order.The COB packaged to reduce the affect of parasitics.The measure IIP3 is 10.5dBm.while the simulated IIP3 is power supply VDD is connected to the corresponding 12dBm pad on the PCB via three bonding wires to reduce the Figure.7 shows the simulated and measured NF with parasitic inductance,which will deteriorate the RF output buffer,which will heighten NF by about 0.5dB. performance of the circuit. The measured SSB NF is about 11dB The 900MHz LO is given to measure the performance Table I gives the measurement results of the designed of the mixer chip.Figure.4 shows the simulated and mixer compared with several published works.It can be measured conversion gain.The resonant frequency of the seen that the NF and IIP3 performance of this work are LC tank is about 50MHz away from simulation,and the more competitive.It should be noticed that SSB NF is Q factor is a little smaller so that the curve is more 3dB larger than DSB NF commonly. flatten. Hkr1891587H2 Atten 10 dB Marker 891.507000MHz 1.● -76.11dBm ha地 (a) (b) VBH 39 KHz Sweep 3.2 ns (601 pts Figure.3.Photograph of(a)the chip (b)the PCB Figure.5.Spectrum of two-tone measurement 10 9.5 g (wgp) 29 -40 1P3=105dBm -e-measurement -100 e-1st order A-3rd order simulation 630 120 20 30 40 50 60 -20 .10 0 10 Input Frequency (MHz) Input level (dBm) Figure.4.Simulated and measured CG Figure.6.Measured IIP3

linearity. 2 DS GS T m I V V g − = (6) IV. SIMULATION AND MEASUREMENT RESULT The chip is implemented in SMIC 0.18-μm RF CMOS process. Figure.3 is the photograph of the chip after bonding and the test PCB board. The chip size including IO pads is 0.63mm0×0.78mm. And it consumes 10mA current at 1.8V power supply on average. The chip is COB packaged to reduce the affect of parasitics. The power supply VDD is connected to the corresponding pad on the PCB via three bonding wires to reduce the parasitic inductance, which will deteriorate the RF performance of the circuit. The 900MHz LO is given to measure the performance of the mixer chip. Figure.4 shows the simulated and measured conversion gain. The resonant frequency of the LC tank is about 50MHz away from simulation, and the Q factor is a little smaller so that the curve is more flatten. The input referred third-order intercept point (IIP3) is measured with the two-tone test at 9.5MHz and 10.5MHz. Figure.5 shows the output spectrum when the power level of every input tone is -11dBm. The central two peaks are the two-tone signals, while the side two peaks are the third order intermodulation products. Figure.6 shows measured IIP3 with input level sweeping. Though the output is attenuated due to impedance mismatch, the IIP3 is not impacted absolutely, since it lies on the difference between the 1st and 3rd order. The measure IIP3 is 10.5dBm, while the simulated IIP3 is 12dBm. Figure.7 shows the simulated and measured NF with output buffer, which will heighten NF by about 0.5dB. The measured SSB NF is about 11dB. Table I gives the measurement results of the designed mixer compared with several published works. It can be seen that the NF and IIP3 performance of this work are more competitive. It should be noticed that SSB NF is 3dB larger than DSB NF commonly. (a) (b) Figure.3. Photograph of (a) the chip (b) the PCB 10 20 30 40 50 60 6.5 7 7.5 8 8.5 9 9.5 10 Input Frequency (MHz) Conversion Gain (dB) measurement simulation Figure.4. Simulated and measured CG Figure.5. Spectrum of two-tone measurement -20 -10 0 10 -120 -100 -80 -60 -40 -20 0 20 Input level (dBm) Output power (dBm) 1st order 3rd order IIP3=10.5dBm Figure.6. Measured IIP3

10 20 30 40 50 5 6 7 8 9 10 11 12 13 Input Frequency (MHz) SSB NF (dB) simulation measurement Figure.7. Simulated and measured NF V. CONCLUSION In this paper, an upconversion mixer exploiting a switched transconductor topology is designed and measured. The advantage of the topology in NF and IIP3 is analyzed compared with the conventional one. The measurement result shows that the designed mixer achieved high gain, low NF and high IIP3. ACKNOWLEDGMENT The authors would like to thank Dr. Haiyang Hu at Agilent Technologies providing necessary instruments for measurements, and thank Lu Yuan, Ronghua Ni, Lei Lu, Zhenyu Yang , Youchun Liao and Xi Tan for many helpful discussions. REFERENCES [1] E.A.M. Klumperink and S.M. Louwsma, “A CMOS switched transconductor mixer,” IEEE J. Solid-State Circuits, vol. 39, pp. 1231-1240, Aug. 2004. [2] B. Razavi, Design of analog CMOS integrated circuits, McGraw-Hill, 2001. [3] H. Darabi and A.A. Abidi, “Noise in RF-CMOS mixer: A simple physical model,” IEEE J. Solid-State Circuits, vol. 35, pp. 15-25, Jan. 2000. [4] M.T. Terrovitis and R.G. Meyer, “Noise in current-commutating CMOS mixers”, IEEE J. Solid-State Circuits, vol. 34, pp. 772-783, Jun. 1999. [5] Lei Lu and Zhangwen Tang, “Equivalent model and parameter extraction of center-tapped differential inductors,” Chinese journal of semiconductors, vol. 27, pp. 2150-2154, Dec. 2006. [6] Lu Liu and Zhihua Wang, “Analysis and design of a low-voltage RF CMOS mixer,” IEEE Trans. CAS II, vol. 3, pp.212-216, Mar. 2006. [7] H. Darabi and J. Chiu, “A noise cancellation technique in active RF-CMOS mixers,” IEEE J. Solid-State Circuits, vol. 40, pp. 2628-2632, Dec. 2005. Table I Summary of measurement results and performance comparison [1] [6] [7] This work Process 0.18-μm CMOS 0.18-μm CMOS 0.13-μm CMOS 0.18-μm CMOS Gain (dB) 10 6.6 1 8 NF (dB) 24 (DSB) 21 (SSB) 11.8 (DSB) 11 (SSB) IIP3 (dBm) 6 1.5 11 10.5 Power 0.6mW N/A 2mA×1.2V 10mA×1.8V Die area (mm2 ) 0.075×0.065 (core) N/A 0.08×0.15 0.63×0.78

点击进入文档下载页（PDF格式）

已到末页，全文结束

点击下载（PDF格式）

浏览记录