A tunable passive mixer for SAW-less front-end with reconfigurable voltage conversion gain and intermediate frequency bandwidth①

0　Introduction

Recently, various kinds of wireless communication standards for handset mobile terminals are emerging and the whole industry is making great effort to enable mobile terminals to support more and more communication applications[1]. As an integral part of mobile platform, the radio frequency (RF) front-end needs to be reconfigurable for cellular communications and short range communications such as wireless local area networks (WLAN). Conventional multi-mode multi-standard (MMMS) systems with a large number of surface acoustic wave (SAW) filters are cumbersome for commercial applications. Thus, studying SAW-less MMMS RF front-end becomes an important trend[2,3].

2)在船舶交通队列特性的影响下，船舶减速行为具有传递性，航道内船舶减速后，可能会造成后续船舶连续减速，由此导致航道整体通航能力降低。

One way to implement SAW-less RF front-end is depicted in Fig.1. Use a high Q impedance translation band-pass filter (BPF) by adopting a passive mixer following with a base-band (BB) low-pass filter (LPF). And the gain and intermediate frequency (IF) bandwidth (BW) should be adjusted for MMMS applications, which is a big challenge. Recently, some SAW-less RF front-ends have been reported, while the gain and IF BW of these designs are fixed[4-6]. The low noise amplifiers (LNAs) could be blocked by interference signals. Another front-end structure can adjust gain and IF BW by changing transimpedance amplifier (TIA)[7-9], while the operational amplifier of TIA needs quite large gain-bandwidth.

Fig.1　Block diagram of a SAW-less RF front-end

In this paper, a tunable voltage-mode passive mixer is proposed with a cross-coupled common gate (CC-CG) LNA to handle interference signals. It can be flexibly reconfigurable in terms of VCG and IF BW performance, which is a great advantage for SAW-less MMMS RF front-end.

The paper is organized as follows. Section 1 discusses the high Q RF BPF based on passive mixer impedance translation model. Section 2 analyzes the design considerations for CC-CG LNA. In Section 3, a transistor-level of the SAW-less RF front-end is presented. The measurement results are shown in Section 4. Finally, conclusions are drawn in Section 5.

1　Tunable high-Q RF BPF based on passive mixer

A high-Q RF filter forms the core of the SAW-less RF front-end, and the main body is made of a pair of double balanced mixing switches as depicted in Fig.2(a). This mixer can be used as an N-path filter if its output connects with a BB capacitor[10]. A 4-phase or 8- phase local oscillator (LO) would provide additional benefits in harmonic rejection and noise. However, 4-phase or 8-phase LO will increase power consumption and make the circuit design more complicated. For simplicity, the mixer of this study is driven by a 50% duty cycle LO.

1.1　Reconfiguration analysis

The LNA acts like a transconductance stage. In order to analyze the performance of high-Q RF BPF with LNA, the circuit is simplified in Fig.2(b)[11]. Current IRF(ω) flows out from LNA output which can be expressed as

双冷源新风机组承担室内的湿负荷的处理，送风含湿量8 g/kg(A)时，对应的露点温度仅为10.5 ℃，新风送风温度仅为11 ℃。过低的送风温度易造成新风口结露，同时无法对送风温度进行调节，影响舒适度和过渡季节的使用。

The theoretical predictions are verified by implementing the passive mixer with CC-CG LNA in TSMC 0.18 μm RF CMOS technology. As explained in Section 2, the circuit is shown in Fig.6. CC-CG LNA uses a pair of cross-coupling capacitor CC1 and CC2 to implement gm boosting. Metal-insulator-metal capacitor is recommended that it makes less thermal noise than MOS capacitor. Differential transistors M3 & M4 are biased making gmi=16.7ms, which causes the input impedance to become 30Ω. It is a good compromise for matching and NF. Cascade devices M1 and M2 are added to improve the reverse isolation. L1 and L2 are off-chip choke inductances to offer DC path. Supply voltage is lifted to 2.5V to compensate the voltage across resistance load RL1 and RL2.

(1)

where ω is RF angular frequency, Gm,eff is the effective transconductance of LNA. Denote ZRF(ω) as the output impedance of LNA, RSW is the mixer switch on-resistance, ZBB(Δω) is the impedance of CBB, η is the scaling factor for ZBB(Δω), and Zsh(ω) represents the power dissipation due to BB signal re-up conversion to the RF side, which is expressed as

(2)

And the input impedance of the mixer is

ZMIX(Δω)=RSW+ηZBB(Δω)//Zsh(ω)

(3)

The LNA gain can be approximated as

VOUT(Δω)=IRF·(ZRF(ω)//ZMIX(ω))

(4)

Fig.2　High-Q RF BPF model with LNA

The following assumptions can be made for out-of-band interferers:

Δω>>BW/2→ZBB(Δω)=

Therefore, the VCG for out-of-band interferers is simplified as

从孵化器内企业来源看，本土孵化器孵化的企业占51.77%，东莞其它镇区迁来的企业占26.81%，深圳迁来的企业占14.57%。

VCG(ω)

(6)

where ZRF(ω) is hardly changed around input frequency. So VCG is proportional to ZBB(Δω), that means the impedance of BB capacitor with LPF profile transfers a high-Q BPF at LNA output and the center of BPF is just at LO frequency (Fig.3(a)). Furthermore, if the value of CBB can be adjusted, the BW of BPF will change accordingly (Fig.3(b)).

Worthy to mention that RSW is only in the denominator of Eq.(4) and it doesn’t affect other parameters. If the value of RSW can be changed, the VCG could be reconfigured easily. The RSW can be presented as

(7)

where w and L are the width and length of switching MOS transistor. VTH is the threshold voltage; COX, the gate oxide capacitance per unit area; μn, the mobility of charge carriers and VGS is gate-source voltage of the switching MOS. Therefor RSW is controlled by VGS. If RSW is reduced linearly by changing the gate voltage (VBLO) of switching MOS, VCG will increase gradually (Fig.3(c)).

刚开始他很为何冰不值，一度考虑过要不要把真相告诉她，但这就违反了他与那女人签下的保密协议。后来他向那笔钱妥协了，开始认真考虑要用怎样的手段拆散他们。

Fig.3　High-Q RF filter characteristic under different conditions

FCC-CG

墩台翻模技术是将模板和工作平台作为两个相互不影响，独立的系统，一方面使墩台滑模施工期间的连续性要求得到了解决，另一方面也解决了墩台爬模施工对平台造成的不良影响。

ALNA(ω)=IRF·ZRF(ω)//ZMIX(ω)=IRF·ZRF(ω)//

(8)

The following assumptions could be also made for in-band signal:

(9)

where Δω=ω-ωLO. Based on this, the output can be deduced:

(10)

It means that a small RSW will decrease LNA gain and improve linearity due to smaller voltage swing at the output of LNA. Therefore, a small value of RSW will results in a larger VCG and higher linearity, that is very critical.

“你别管我疯没疯。你不觉得我开的条件太诱人了吗？这还用想吗？你如果赢了我，等于白白地得到了这五十万。五十万哪！”

1.2　Noise analysis

The noise sources are introduced in the architecture of the proposed high-Q RF BPF in Fig.4. The thermal noise of source resistor RS and mixer on-resistance RSW are 4kTRS and 4kTRSW separately, where k denotes the Boltzmann constant and T is the absolute temperature. is the noise current flows from LNA. The noise of mixer equivalent resistance is also white noise, so it will be down-converted by all the harmonics of the fLO to baseband output. The noise of mixer at output is given by Ref.[10].

(11)

where an is the Fourier coefficient which depends on phase number of LO. Thus NF of the circuit from Fig.4 can be obtained as

NF=

(12)

Mixer stage consists of a pair of double balanced mixing switches. One key issue is the size of switch transistors (M5-M8). A high W/L ratio results in a small RSW, and it would be beneficial to improve linearity and gain. Considering the input frequency of the mixer, the RF transistors implemented in the deep N-well are used as switches due to their attractive high frequency performance. Capacitor C3 and C4 are used for AC current coupling from LNA stage. Zero DC current of switch transistors improve the flicker noise performance. The IF BW must be reconfigurable for MMMS applications. A capacitor array is taken as CBB which consists of four capacitors in parallel and each capacitor is controlled by digital signal. CBB is various from 2pF to 46pF, and the value of BW is changed from 4MHz to 66MHz. As mentioned in part 2, the VCG is inverse to RSW. The value of RSW is determined by the gate bias voltage of switch transistors. If adjusting the VBLO in constant then the VCG could be fine tuned. External LO is 0dBm sinusoidal signal which is transferred into a differential signal by off-chip balun.

(13)

As Eq. (13) shown, the best way to reduce the noise is to increase Gm,eff . In addition, increasing the load of LNA and decreasing the parastic capacitor at LNA output also could improve the noise performance of the whole circuit.

Fig.4　The simplified circuit model with main noise sources

2　CC-CG LNA design

High-Q BPF works at LNA output. When out-of-band interference enters into LNA, it should not be amplified, otherwise the LNA will be blocked. CG LNA converts a voltage input signal to a current signal at output and interference can be filtered at LNA output.

CC-CG LNA is used in Fig.5, because it has gm boosted effect[12]. When CC≫Cgs, the effective transconductance Gm,eff≈2gmi, where CC，Cgs and gmi are the cross-coupled capacitor, gate-to-source capacitor, transconductance of M1 and M2 respectively. The approximate NF of CC-CG LNA is

Back to Fig.2(b), the LNA gain can be obtained as follows:

(14)

And for traditional CG LNA

(15)

where γ and α are the bias-dependent noise parameters, RS, source resistance, and RL, load resistance. Hence, compared with traditional CG LNA, it has lower NF for the same power dissipation. Size and bias of M1 and M2 can be adjusted to achieve input matching. The perfect input matching conditions occurs at Rs=1/2gmi. Decreasing the input impedance, for example, to 30 Ω will result in S11≈-12dB, but gmi increases to 16.7ms. That means higher gain and lower NF.

The proposed circuit is fabricated in TSMC 0.18μm RF CMOS technology. Fig.7 shows the chip photograph of the proposed circuit. The die area is 0.7mm×0.63mm including pads. The measured DC current is 4.9mA from a 2.5V supply voltage. All measurements are performed using an assembled printed circuit board (PCB). RF signal are applied using a commercial wideband balun. The insertion loss from external components and transmission lines is measured at the desired frequencies and used for compensating the measurement results.

Another interesting characteristic of CC-CG LNA is that it can reduce second-order distortion. And its third order inter-modulation point is given[13]:

(16)

where ZL is the impedance of external inducter. Thus lager gmi can also help to increase AIP3. In order to further boost compression point of LNA, a 2.5V supply voltage[14] is used. Make sure no device terminal suffers more than the reliability specification of the technology. Lext is external choke inductance to offer DC path for LNA, and it can also eliminate the parasitic capacitor (Cpad) at input pad. Cascade devices M3 and M4 improve the reverse isolation. RL can offer large output impedance over wide BW.

Fig.5　Schematic of CC-CG LNA

3　Circuit implementation

IRF(ω)=Gm,eff·VRF(ω)

颜晓晨和沈侯虽然在一个学院，可是专业不同，颜晓晨是游离在班级之外的人，沈侯也是游离在班级之外的人，两人完全无交集，就算有学院必修课，可全院两百多人，混到大学毕业，仍会有很多人叫不出名字。本来，他们的生活应该是两条平行线，可就是因为代写作业和论文，颜晓晨进入了沈侯的视线。从那之后，沈侯不想做的作业，要完成的论文，期末考试前复印笔记、勾重点……沈侯都会找颜晓晨，颜晓晨从来不拒绝，但只第一次收了他四千块钱，之后，无论如何，她都不要钱。因为颜晓晨不肯要钱，沈侯也不好意思总找她代写，只能变得勤快点，借了作业来抄，一来二去，有意无意地，变成了颜晓晨帮他辅导功课，沈侯也渐渐地不再玩游戏。

Same as Eq.(9), for in-band signal NF can be approximated to

The simple buffer is implemented for test. The common source stage with 60Ω resistance load can offer good matching at output. The gate of M9 and M10 are high impedance that wouldn’t affect the gain and BW.

Fig.6　Entire schematic of proposed front-end including buffer

4　Measurement results

政府的参与和支持是大型体育赛事运作成功的外部保证。2011年第九届全国民族运动会的举办得到了贵州省政府与贵阳市政府的高度重视，许多与运动会有直接关系的事情基本都是“特殊事情，特殊对待，优先办理”，政府的积极支持是大型体育赛事运作成功的主要原因。2010年四川成都“IDSF世界杯标准舞暨中国成都国际体育舞蹈公开赛”的赛事组织中，四川省政府与成都市政府不仅在筹备时期就给予了一定的财政拨款作为赛事筹备的运作经费，并且主要的市政府领导也在赛组委中担任了重要职务。在这个方面，就生动地诠释了“政府搭台，部门主导，商企唱戏，共谋发展”。

The VCG measurement is performed with the LO frequencies from 500MHz to 2.5GHz in double sideband in Fig.8. With various fLO, for example 1.5GHz, the IF BW is fixed at 4MHz and the frequency of input signal is changed from 1.5001GHz to 2GHz, so that the output IF frequency can be swept from near dc to 500MHz as high sideband VCG. Then the input signal is varied from 1G to 1.4999G to plot the low sideband VCG. The rejection is larger than 25dB at 100MHz offset frequency in the wideband range and the maximum rejection larger than 40dB.

Fig.7　Die micrograph of the test chip

The IF BW should be flexibly reconfigurable for multimode application. Fig.9 shows the measured VCG with 1.5GHz LO frequency. The IF BW depends on the value of CBB and it can be adjusted from 4MHz to 66MHz for different CBB. This SAW-less RF front-end can offer 15 kinds of IF BW to adapt different communication standards.

Fig.8　　　　Measured VCG curves of proposed circuit with different LO frequency

Fig.9　Measured IF bandwidth with different CBB

Fig.10(a) shows measured 1-dB Compression curves at 4MHz IF BW (CBB is 46pF) and 1.5GHz LO frequency. Then 1-dB Compression points are found under various offset frequencies (Δf=fin-fLO) to compare the variation of linearity. Fig.10(b) depicts measured 1-dB compression points of 4 different IF BW. It clearly shows that the linearity of out-of-band is much better than in-band. So the circuit can resist higher interference signals at out-of-band frequencies.

Fig.10　Measured 1-dB compression point

Since we are only interested in intermodulation products that fall in the channel band, measurements are carried out with the two tones located at frequency offsets Δf+600kHz and 2Δf, such that the third-order intermodulation (IM3) always falls at 1.2MHz. Fig.11 shows the measured IIP3 with a 1.5GHz LO and four different CBB. IIP3 is measured for various Δf showing that with a larger CBB engaged, the circuit could have better linearity under same interferers. With increasing Δf, the out-band IIP3 is improved to +0dBm.

洞中开始变得阴冷。二人收拾出一块空地，又弄了些干草干枝，钻木燃起一团篝火。篝火的烟气飘出洞外，转眼被山风吹散。

1.2 基本概念 试题的第(1)小题： 水稻对Mp表现出的抗病与感病为一对相对____________。为判断某抗病水稻是否为纯合子，可通过观察自交子代____________来确定。

In Section 3, the fact that different RSW can affect the performance significantly is discussed. Fig.12 depicts the measured results to verify this concept. By increasing the gate voltage of switch transistors (VBLO), the RSW can be decreased. Adjusting VBLO from 0.95V to 1.3V with 1.5GHz LO frequency, the VCG is increased from 5dB to 17dB which is shown in Fig.12(a). The noise performance is illustrated in Fig.12(b) which has not much changed. The in-band IIP3 and 1-dB compression point are shown in Fig.12(c). They are improved almost 10dB from large RSW to small RSW.

Fig.11　　　　Measured in-band and out-band IIP3 at 2GHz LO with four CBB settings

Considering the wideband applications, the VCG, noise and in-band linearity are measured with varied LO frequency from 0.5GHz to 2.5GHz. Fig.13(a) shows in-band VCG with different LO frequency. Due to parasitic at LNA the in-band VCG is falling from 21dB to 14dB. Fig.13(b) illustrates the NF for different LO frequency. The NF is smaller than 5dB when LO frequency belows 1.5GHz and the deterioration of the noise is due to smaller gain. The in-band IIP3 and 1-dB compression point is depicted in Fig.13(c). Two tons with Δf=500kHz are applied to the input of the LNA with power -20dBm. The maximum measured IIP3 is larger than 0dB at 2.5GHz LO frequency and larger than -13dB for all frequencies. The measured 1-dB compression point shows similar trend to that of IIP3 with respect to LO frequency.

Fig.12　　　　Measured VCG, NF, in-band IIP3 and 1-dB compression point with fLO=1.5GHz, while the VBLO is changed from 0.95V to 1.3V

Fig.13　Simulated and measured VCG, NF, in-band IIP3 and 1-dB compression point with different fLO

The measured S11 and S22 are plotted in Fig.14. As shown, S11 is less than -10dB from 0.5GHz to 2.5GHz which validates the broadband input matching. The measured S22 is also below -10dB over the frequency range of 10-100MHz.

作为典型的东北城市，长春应如何依据入室盗窃案件发案特点和规律做好防范工作，将是其能否跻身全国安全城市的重要指标。具体来说公安机关应从以下几个方面做出努力。

与VPA单独处理比较，联合miR-34c-5p抑制剂处理显著逆转了VPA对ATG4B mRNA（图6）与蛋白水平（图7）的下调作用。

Table 1 summarizes the measured results of the proposed circuit and gives a comparison with other recently published wideband front-ends.

注水指示曲线为折线形，在注水井井底压力为60MPa时出现拐点。如果注水压力继续增加，吸水量大量增加，这种曲线明显具有裂缝吸水特征。产生这种情况的可能原因是，注水压力增加时油层产生微小裂缝，使油层吸水量增大。所以注水井井底压力的上限以60MPa为最佳。高于地层破裂压力注水时地层压力水平大幅度提高，油井没有出现暴性水淹主要原因是地层压力还没有达到极限。这说明高压注水［3］在一定的时间范围内可行。

Fig.14　Measured S11 and S22 of proposed circuit

5　Conclusion

This paper reports the design of a tunable mixer for multi-mode multi-standard (MMMS) applications, and the circuit is implemented in TSMC 0.18μm RF CMOS process. The adjustable mixer switches have been proposed in order to reconfigure the VCG flexibly. A LO frequency-tunable high-Q filter transfering at LNA output can reject out-of-band interferences. Furthermore the IF BW can be reconfigurable by using different CBB. Measurement results indicate that the designed mixer can provide low NF and high linearity over wideband in the applications of MMMS wireless communications.

Table 1　Performance comparison

　TheproposedworkESSCIRC2010[6]JSemicond2016[15]TMTT2010[16]TMTT2012[17]Technology0．18μmCMOS90nmCMOS0．18μmCMOS0．18μmCMOS0．13μmCMOSRFfrequency(GHz)0．5-2．50．4-30．7-2．31．5-2．30．6-3Gain(dB)5-17164-2222．5-2542-48IFbandwidth(MHz)3-661012NA0．8-12MinNF(dB)3．72．887．73MaxIIP3(dBm)0118．58-14Supplyvoltage(V)2．521．81．81．2Power(mW)12．313151030NoteLNA+MixerLNA+MixerLNTA+Mixer+TIALNTA+Mixer+TIALNTA+Mixer+TIA

References

[ 1] Lin F, Mak P, Martins R P. Wideband receivers: design challenges, tradeoffs and state-of-the-art[J]. CircuitsandSystemsMagazine, 2015, 13(1):12-24

[ 2] Madadi I, Tohidian M, Cornelissens K, et al. A high IIP2 SAW-less superheterodyne receiver with multistage harmonic rejection[J]. IEEEJournalofSolid-StateCircuits, 2016, 51(2):332-347

[ 3] Murphy D, Darabi H, Xu H. A noise-cancelling receiver resilient to large harmonic blockers[J]. IEEEJournalofSolid-StateCircuits, 2015, 50(6):1336-1350

[ 4] Hedayati H, Lau W A, Kim N, et al. A 1.8dB NF blocker-filtering noise-canceling wideband receiver with shared TIA in 40nm CMOS[J]. IEEEJournalofSolid-StateCircuits, 2015, 50(5): 1148-1164

[ 5] Ostman K B, Englund M, Viitala O, et al. Analysis and design of N-path filter offset tuning in a 0.7-2.7GHz receiver front-end[J]. IEEETransactionsonCircuitsandSystemsI: RegularPapers, 2015, 62(1): 234-243

[ 6] Borremans J, Mandal G, Debaillie B, et al. A sub-3dB NF voltage-sampling front-end with +18dBm IIP3 and +2dBm blocker compression point[C]. In: Proceedings of the European Solid-state Circuits Conference, Sevilla, Spain, 2010. 402-405

[ 7] Yang D, Andrews C, Molnar A. Optimized design of N-phase passive mixer-first receivers in wideband operation[J]. IEEETransactionsonCircuitsandSystemsI: RegularPapers, 2015 62(11): 2759-2770

[ 8] Zhang X, Chi B, Wang Z. A 0.1-1.5 GHz harmonic rejection receiver front-end with phase ambiguity correction, vector gain calibration and blocker-resilient TIA[J]. IEEETransactionsonCircuitsandSystemsI: RegularPapers, 2015, 62(4): 1005-1014

[ 9] Nejedl A, Sjoland H, Tormanen M. A noise-cancelling receiver front-end with frequency selective input matching[J]. IEEEJournalofSolid-StateCircuits, 2015, 50(5): 1137-1147

[10] Mirzaei A, Darabi H, Murphy D. Architectural evolution of integrated M-phase high-Q bandpass filters[J]. IEEETransactionsonCircuitsandSystemsI: RegularPapers, 2012, 59(1): 52-65

[11] Andrews C, Molnar A. A passive mixer-first receiver with digitally controlled and widely tunable RF interface[J]. IEEEJournalofSolid-StateCircuits, 2010, 45(12): 2696-2708

[12] Zhou W, Li X, Shekhar S, et al. A capacitor cross-coupled common-gate low-noise amplifier[J]. IEEETransactionsonCircuitsandSystemsII: ExpressBriefs, 2005, 52(12): 875-879

[13] Han H G, Jung D H, Kim T W. A 2.88mW+9.06dBm IIP3 common-gate LNA with dual cross-coupled capacitive feedback[J]. IEEETransactionsonMicrowaveTheoryandTechniques, 2015, 63(3):1019-1025

[14] Borremans J, Mandal G, Giannini V, et al. A 40 nm CMOS 0.4-6GHz receiver resilient to out-of-band blockers[J]. IEEEJournalofSolid-StateCircuits, 2011, 46(7):1148-1164

[15] Fan X, Tao J, Bao K, et al. A reconfigurable passive mixer for multimode multistandard receivers in 0.18μm CMOS[J]. JournalofSemiconductors, 2016, 37(8): 085001 1-8

[16] Kim N, Aparin V, Larson L E. A resistively degenerated wideband passive mixer with low noise figure and high IIP2[J]. IEEETransactionsonMicrowaveTheoryandTechniques, 2010, 58(4): 820-830

[17] Wang X, Sturum J, Yan N, et al. 0.6-3GHz wideband receiver RF front-end with a feedforward noise and distortion cancellation resistive-feedback LNA[J]. IEEETransactionsonMicrowaveTheoryandTechniques, 2012, 60(2): 387-392