I. INTRODUCTION

Bluetooth low energy (BLE) is a representative communication standard, that is widely utilized in Internet of Things (IoT) sensor devices. Ultra-low-power BLE transceivers for IoT applications have been actively studied and developed ^(1-19). Most BLE receivers (RXs) employ a low-intermediate frequency (IF) single-quadrature architecture to achieve low 1/$\textit{f}$ noise with a channel bandwidth of 1 MHz. However, low-IF RXs suffer from a so-called image problem. When the sensitivity degradation, caused by the image signal, is less than 0.3 dB, an image rejection ratio (IRR) greater than 21 dB is required from a carrier-to-image interference ratio ($\textit{C}$/$\textit{I}$$_{\mathrm{Image}}$) of ${-}$9 dB ^(17,20). In general, quadrature signals for single-quadrature mixing are generated by a quadrature voltage-controlled oscillator or divide-by-two circuit in the local oscillator (LO) path ^(1-12). Recently, owing to less restrictive IRR requirements, numerous studies on BLE RXs, whose quadrature signals are provided by a quadrature low-noise amplifier (LNA) or a quadrature mixer in the RF path or quadrature LO buffer in the LO path, have been published ^(15-18,21). In these studies, the quadrature LNA and mixer have a lower voltage gain and degraded noise figure (NF) instead of providing quadrature signals without additional power consumption. Moreover, in terms of the RX sensitivity, it is more advantageous that the quadrature mixer provides the quadrature signals rather than the quadrature LNA. Kwon $\textit{et al.}$ introduced the in-phase/quadrature (I/Q) down-conversion active mixer to generate quadrature signals ⁽²¹⁾. The quadrature mixer employs a common-source (CS) amplifier with capacitive degeneration and a compensating resistor. In addition, a current-bleeding circuit is used to reduce the 1/$\textit{f}$ noise of the switching stage. If the current consumed by the current-bleeding circuit can be reused to boost the overall transconductance ($\textit{g}$$_{\mathrm{m}}$) and generate quadrature signals, the performances of the I/Q mixer can be improved.

In this paper, a 2.4 GHz low-power BLE RX front-end using a new I/Q down-conversion active mixer with a current-reused quadrature transconductor is proposed. Section II introduces the new I/Q down-conversion active mixer architecture with a current-bleeding circuit that performs $\textit{g}$$_{\mathrm{m}}$-boosting and generates quadrature signals. A detailed circuit implementation of the BLE RX front-end is presented in Section III. Section IV discusses the simulation results. Finally, Section V concludes this paper.

II. A New I/Q Down-conversion Mixer Architecture with Current-reused Quadrature Transconductor

This section introduces a new I/Q down-conversion active mixer architecture with a current-reused quadrature transconductor. Fig. 1 illustrates the conventional I/Q mixer architecture with a quadrature transconductor, which was introduced in ⁽²¹⁾, and our new I/Q mixer architecture with a current-reused quadrature transconductor.

In the conventional I/Q mixer architecture, the current-bleeding circuit is only used to reduce the 1/$\textit{f}$ noise of the switching stage, without reusing the current flowing into the current-bleeding circuit to perform additional functions. In the proposed I/Q mixer architecture, the current-bleeding circuit reduces the current flowing into the switching stage and reduces its 1/$\textit{f}$ noise. It also performs voltage-to-current conversion and provides quadrature signals. Therefore, the proposed I/Q mixer architecture can improve the conversion gain and noise performance. The output currents $\textit{I}$$_{\mathrm{RF,I}}$ and $\textit{I}$$_{\mathrm{RF,Q}}$ in the proposed current-reused quadrature transconductor of the I/Q mixer can be expressed as

where $\textit{g}$$_{\mathrm{mN1}}$, $\textit{g}$$_{\mathrm{mN2}}$, $\textit{g}$$_{\mathrm{mP1}}$, and $\textit{g}$$_{\mathrm{mP2}}$ denote the transconductances of $\textit{M}$$_{\mathrm{N1}}$, $\textit{M}$$_{\mathrm{N2}}$, $\textit{M}$$_{\mathrm{P1}}$, and $\textit{M}$$_{\mathrm{P2}}$, respectively, and $\textit{C}$$_{\mathrm{N}}$ and $\textit{C}$$_{\mathrm{P}}$ denote the source degeneration capacitors. When $\textit{g}$$_{\mathrm{mN1}}$ $\textit{= g}$$_{\mathrm{mN2}}$ $\textit{=}$ $\textit{g}$$_{\mathrm{mN}}$, $\textit{g}$$_{\mathrm{mP1}}$ $\textit{=}$ $\textit{g}$$_{\mathrm{mP2}}$ $\textit{=}$ $\textit{g}$$_{\mathrm{mP}}$, $\textit{${\omega}$C}$$_{\mathrm{N}}$ $\textit{=}$ $\textit{g}$$_{\mathrm{mN2}}$, and $\textit{${\omega}$C}$$_{\mathrm{P}}$ $\textit{= g}$$_{\mathrm{mP2}}$, the output currents $\textit{I}$$_{RF,I}$ and $\textit{I}$$_{RF,Q}$ exhibit the following quadrature relationships: |$\textit{I}$$_{\mathrm{RF,I}}$/$\textit{I}$$_{\mathrm{RF,Q}}$| = 1, and ${\angle}$($\textit{I}$$_{\mathrm{RF,I}}$/$\textit{I}$$_{\mathrm{RF,Q}}$) = 90$^{\circ}$. The overall transconductance of the current-reused quadrature transconductor can be expressed as

Based on (3), the proposed current-reused quadrature transconductor boosts the overall transconductance by reusing the current-bleeding circuit. Fig. 2 depicts the simulated conversion gain and NF of the proposed and conventional I/Q mixers. The proposed architecture achieves a 3.7 dB higher conversion gain and a 3.1 dB lower NF.

III. Circuit Implementation

This section presents a detailed circuit implementation of the proposed low-power BLE RX front-end. Fig. 3 illustrates the block diagram of the proposed 2.4 GHz low-IF single-quadrature BLE front-end, which consists of a current-reused push-pull LNA with inductive source degeneration and I/Q down-conversion single-balanced active mixers with a current-reused quadrature transconductor.

Fig. 4 depicts the proposed current-reused push-pull LNA with inductive source degeneration. The effective transconductance of the current-reused push-pull architecture approximately doubles with the same power consumption ⁽⁶⁾. Therefore, it can improve the voltage gain and noise performance. $\textit{C}$$_{\mathrm{EXN}}$ and $\textit{C}$$_{\mathrm{EXP}}$ are used to perform simultaneous noise and input matching. The input impedance of the proposed push-pull LNA can be expressed as

Here, $\textit{g}$$_{\mathrm{mN}}$ and $\textit{g}$$_{\mathrm{mP}}$ denote the transconductances of $\textit{M}$$_{\mathrm{N1}}$ and $\textit{M}$$_{\mathrm{P1}}$, $\textit{L}$$_{\mathrm{SN}}$ and $\textit{L}$$_{\mathrm{SP}}$ denote the source degenerated inductors, $\textit{C}$$_{\mathrm{TN}}$ = $\textit{C}$$_{\mathrm{EXN}}$ + $\textit{C}$$_{\mathrm{gsN}}$, and $\textit{C}$$_{\mathrm{TP}}$ = $\textit{C}$$_{\mathrm{EXP}}$ + $\textit{C}$$_{\mathrm{gsP}}$. $\textit{C}$$_{\mathrm{gsN}}$ and $\textit{C}$$_{\mathrm{gsP}}$ denote the gate-to-source capacitances of $\textit{M}$$_{\mathrm{N1}}$ and $\textit{M}$$_{\mathrm{P1}}$. For a simple and intuitive analysis, it is assumed that $\textit{g}$$_{\mathrm{mN}}$ = $\textit{g}$$_{\mathrm{mP}}$ = $\textit{g}$$_{\mathrm{m}}$, $\textit{r}$$_{\mathrm{oN}}$ = $\textit{r}$$_{\mathrm{oP}}$ = $\textit{r}$$_{\mathrm{o}}$, $\textit{L}$$_{\mathrm{SN}}$ = $\textit{L}$$_{\mathrm{SP}}$ = $\textit{L}$$_{\mathrm{S}}$, and $\textit{C}$$_{\mathrm{TN}}$ = $\textit{C}$$_{\mathrm{TP}}$ = $\textit{C}$$_{\mathrm{T}}$. The input impedance of the LNA can be expressed as

Input matching is accomplished at the resonant frequency with Im($\textit{Z}$$_{\mathrm{IN}}$) = 0. The resonant frequency is given by

and $\textit{Z}$$_{\mathrm{IN}}$ = $\textit{g}$$_{\mathrm{m}}$$\textit{L}$$_{\mathrm{S}}$/2$\textit{C}$$_{\mathrm{T}}$ = 50 Ω. The voltage gain of the proposed push-pull LNA from the voltage source $\textit{V}$$_{\mathrm{S}}$ with source resistance $\textit{r}$$_{\mathrm{S}}$ to $\textit{V}$$_{\mathrm{OUT}}$ can be expressed as

Here, $\textit{Q}$$_{\mathrm{IN}}$ is the $\textit{Q}$-factor of the input impedance network, and $\textit{Z}$$_{\mathrm{OUT}}$ denotes the output impedance of the LNA, whose values are given approximately as

where $\textit{V}$$^{2}$$_{\mathrm{MN1}}$ and $\textit{V}$$^{2}$$_{\mathrm{MP1}}$ represent the output-referred noise voltages generated by $\textit{M}$$_{\mathrm{N1}}$ and $\textit{M}$$_{\mathrm{P1}}$, and ${\gamma}$ denotes the noise parameter of the transistor. The proposed LNA halves the excess noise factor because its total output-referred noise power is the sum of the noise powers generated by $\textit{M}$$_{\mathrm{N1}}$ and $\textit{M}$$_{\mathrm{P1}}$, and its total voltage gain is the sum of the voltage gains provided by $\textit{M}$$_{\mathrm{N1}}$ and $\textit{M}$$_{\mathrm{P1}}$. Therefore, compared to a conventional CS LNA with inductive source degeneration and additional $\textit{C}$$_{\mathrm{EX}}$, the proposed current-reused push-pull LNA with inductive source degeneration can achieve a larger voltage gain and lower noise performance.

Fig. 5 shows the I/Q down-conversion active mixer using the proposed current-reused quadrature transconductor. The current-bleeding circuit reduces the 1/$\textit{f}$ noise of the switching stage, performs voltage-to-current conversion, and generates quadrature signals. Therefore, it can enhance the conversion gain and NF. In high frequencies, the parasitic capacitances of main transistors in the current-reused quadrature transconductor ($\textit{M}$$_{\mathrm{N1}}$, $\textit{M}$$_{\mathrm{N2}}$, $\textit{M}$$_{\mathrm{P1}}$, and $\textit{M}$$_{P2}$) can cause gain and phase mismatches. To compensate for both mismatches and increase the design degree of freedom, compensating resistors of $\textit{r}$$_{\mathrm{N}}$ and $\textit{r}$$_{\mathrm{P}}$ are added ⁽²¹⁾.

Fig. 6 shows the simulated gain and phase mismatches of the proposed I/Q mixer. As shown in Fig. 6, there is little difference in the BLE band. The proposed I/Q mixer can provide quadrature signals with gain and phase mismatches of less than 0.15 dB and 0.65$^{\circ}$ in the BLE band, respectively. These values are sufficient to provide an IRR of more than 21 dB. The I-path conversion gain of the proposed I/Q mixer can be expressed as

The I-path noise factor of the proposed I/Q mixer can be expressed as

where $\textit{V}$$^{2}$$_{\mathrm{MN1}}$, $\textit{V}$$^{2}$$_{\mathrm{MP1}}$, $\textit{V}$$^{2}$$_{\mathrm{MN3}}$, $\textit{V}$$^{2}$$_{\mathrm{MP3}}$ and $\textit{V}$$_{RL}$ represent the output-referred noise voltages generated by $\textit{M}$$_{\mathrm{N1}}$, $\textit{M}$$_{\mathrm{P1}}$, $\textit{M}$$_{\mathrm{N3}}$, $\textit{M}$$_{\mathrm{P3}}$, and $\textit{r}$$_{\mathrm{L}}$, respectively, $\textit{I}$$_{\mathrm{B}}$ denotes the dc current flowing into $\textit{M}$$_{\mathrm{N3,4}}$ of the switching stage, and $\textit{A}$ denotes the amplitude of the LO signal. In $\textit{f}$$_{\mathrm{Mixer}}$, the switching noises of $\textit{M}$$_{\mathrm{N3}}$ and $\textit{M}$$_{\mathrm{N4}}$ are the dominant noise contributions. Because the proposed I/Q mixer with the current-reused quadrature transconductor enhances the overall transconductance, it has a smaller noise factor compared to that of the conventional quadrature transconductor.

The conversion gain and noise factor of the proposed BLE RX front-end can be expressed as

IV. Simulation Results

The proposed low-power 2.4 GHz BLE RX front-end adopting the I/Q mixer with the current-reused quadrature transconductor was designed in a 65-nm CMOS process. Fig. 7 illustrates the layout of the BLE RX front-end. The active area without bond pads is 0.5~mm$^{2}$. It draws a dc bias current of 1 mA from a supply voltage of 1 V. The power breakdown of the designed BLE RX front-end is presented in Table 1. The power consumptions of the LNA and I/Q mixer are 0.6 and 0.4~mW, respectively. The following reported simulation results are based on the layout parasitic extraction. The simulated input return loss ($\textit{S}$$_{11}$) of the BLE RX front-end is depicted in Fig. 8. The $\textit{S}$$_{11}$ is less than -15 dB throughout the BLE operating frequency range of 2.40-2.48 GHz. Fig. 9 depicts the simulated conversion gain of the RX front-end with a 2.44 GHz RF input signal. A simulated conversion gain of 37.6 dB is obtained at an IF frequency of 2 MHz.

The simulated NF of the RX front-end with an IF frequency of 2 MHz is depicted in Fig. 10. The obtained NF is 3.3 dB, which is measured at an RF frequency of 2.44 GHz. The simulated input-referred third-order

intercept point (IIP3) and output-referred third-order intercept point (OIP3) are shown in Fig. 11. The two-tone test conditions for IIP3 are $\textit{f}$$_{1}$ = $\textit{f}$$_{\mathrm{LO}}$ + 5 MHz, and $\textit{f}$$_{2}$ = $\textit{f}$$_{\mathrm{LO}}$ + 8 MHz. The obtained IIP3 and OIP3 values are ${-}$22.73 and 13.26 dBm, respectively.

Table 2 summarizes and compares the performance of the proposed BLE RX front-end to those of the current state-of-the-art approaches. For fair performance comparison of the BLE RX front-ends, the following figure-of-merit (FOM) is used.

where $\textit{f}$ is the noise factor ($\textit{f}$ = 10$^{NF/10}$) and $\textit{P}$$_{dc}$ is the dc power consumption. As presented in Table 2, the proposed RX front-end demonstrates excellent NF performance with low power consumption by employing the I/Q mixer with the current-reused quadrature transconductor.

IV. CONCLUSIONS

In this study, a 2.4 GHz low-power low-IF BLE RX front-end employing the new I/Q down-conversion active mixer with the current-reused quadrature transconductor was designed in a 65 nm CMOS technology. The quadrature signals required for single-quadrature mixing were generated at the transcondcutor of the I/Q mixer to reduce the power consumption and NF. The proposed current-reused current-bleeding circuit of the I/Q mixer improved the overall transconductance of the mixer, provided quadrature signals, and reduced the 1/$\textit{f}$ noise of the switching stage. The designed BLE RX front-end obtained a conversion gain of 37.6 dB, NF of 3.3 dB, and IIP3 of -22.73 dBm.