I. INTRODUCTION

Recently traffic accidents due to driver's physical abnormalities are increasing. In particular, as the number of elderly drivers is increasing as we enter the aging society, research on a biometric monitoring radar system that monitors the driver's abnormal vital signs more accurately while driving to prevent traffic accidents due to the driver's condition abnormality is currently being conducted ^(1-5). In order to measure the driver's heartbeat or respiration in a non-contact manner, research on a radar sensing transceiver using a millimeter wave (mm-wave) band is in progress. If the millimeter wave band is used, the size of the radar module can be reduced and high resolution characteristics can be obtained.

In order to integrate in a vehicle at low cost, it is necessary to use a direct conversion receiver that can minimize the use of external elements. Since the direct conversion receiver structure is used in the mm-wave band, the balancing characteristic between the I/Q signal paths of the mm-wave receiver is important. Since radar interference signals may exist in the 24 GHz band, a receiver front end with good linearity is required.

Therefore, a 24-GHz receiver front-end is proposed for in-cabin radar systems that can monitor driver’s vital signs such as heart rate and respiration in this paper. It was implemented using a 1-poly 8-metal RF 65-nm CMOS process. In the following sections, the circuit design and measured results of the receiver front-end are described in details.

II. Proposed Circuit Design

Fig. 1 shows the simplified block diagram of the presented 24 GHz direct-conversion receiver front-end. A direct-conversion receiver architecture is suitable for a wireless short-range monitoring systems in a car because it has high integration and low power consumption. The presented 24 GHz receiver front-end consists of a low-noise amplifier, an I(in-phase)/Q(quadrature) down-conversion mixer, and an I/Q LO generator. The RF input signal from the antenna is applied to the input of the low-noise amplifier and amplified by the low-noise amplifier. The amplified RF signal from the low-noise amplifier is converted into the baseband analog I/Q output signal using the LO input signal, which comes from an external LO signal source or an internal frequency synthesizer, by the I/Q down-conversion mixer.

The simplified schematic of a low-noise amplifier (LNA) is shown in Fig. 2. It is composed of a single-ended cascode amplifier and a single-to-differential common-source amplifier. To improve the electro-static discharge (ESD) protection performance of the LNA, the single-ended cascade amplifier with shunt inductance L$_{\mathrm{g1}}$ at input stage is employed. The input impedance Z$_{\mathrm{in}}$ of the LNA is approximately expressed as

where g$_{m}$ and C$_{gs}$ are the transconductance and gate-source capacitance of the transistor M$_{1}$, respectively ⁽⁶⁾. Since the value of the capacitance C$_{1}$ is much larger than that of the capacitance C$_{gs}$, 1/j${\omega}$C$_{1}$ can be ignored in (1). To make the imaginary part of (1) zero, the following (2) must be satisfied at the operating frequency f$_{op}$;

When (2) is satisfied, (1) is can be expressed as

Therefore, in order to match the input impedance Z$_{\mathrm{in}}$ of the LNA to 50 Ω, the LNA was designed by adjusting the design parameters so that (3)becomes 50 Ω.

As shown in Fig. 3, a test bench was set up for the human body model (HBM) ESD simulation ⁽⁷⁾. In order to examine the effect of the metal width of the inductance L$_{\mathrm{g1}}$ on the NF and HBM characteristics of the LNA, simulation was performed by changing the metal width of the inductance L$_{\mathrm{g1}}$ under the same inductance condition. Fig. 4 shows the simulated NF and HBM results of the LNA according to the metal width of the inductance L$_{\mathrm{g1}}$. When the metal width of L$_{\mathrm{g1}}$ is 2 µm, the simulated quality factor of L$_{\mathrm{g1}}$ is 18. When the width is 5~µm and 10 µm, the simulated quality factor of the inductor is 19 and 20, respectively. Regardless of the metal width of L$_{\mathrm{g1}}$, the simulated NF of the LNA is approximately 3.2 dB, and the HBM voltage is about 200~V. Considering the NF and HBM performances of the LNA, the metal width of the inductance L$_{\mathrm{g1}}$ was used as 2 µm in our design.

Fig. 5 shows the simplified of the proposed I/Q down-conversion mixer. An inverter-type transconductor composed of NMOS and PMOS is used at the transconductance stage of the I/Q down-conversion mixer. The inverter-type transconductor has larger transconductance than a conventional NMOS counterpart at the same dc current consumption ^(8,9).

As shown in Fig. 6(a), if a small signal input voltage v$_{i}$ is applied to the inverter transconductor of the down-conversion mixer, the small signal output current I$_{o}$ can be expressed as

where g$_{m}$$^{\prime}$ is the first derivative and g$_{m}$$^{\prime}$$^{\prime}$ is the second derivative of the transconductance g$_{m}$ with respect to the small signal input voltage v$_{i}$, respectively. Since the third-order nonlinearity of CMOS circuit is dominated by g$_{m}$$^{\prime}$$^{\prime}$ nonlinearity ⁽¹⁰⁾, g$_{m,n1}$$^{\prime}$$^{\prime}$ + g$_{m,p1}$$^{\prime}$$^{\prime}$ should be zero to achieve high linearity. By adjusting the proper size and bias point of PMOS M$_{\mathrm{p1}}$, the g$_{m,n1}$$^{\prime}$$^{\prime}$ term of NMOS M$_{\mathrm{n1}}$ can be cancelled by the g$_{m,p1}$$^{\prime}$$^{\prime}$ term of PMOS M$_{\mathrm{p1}}$. As shown in Fig. 6(b), when the width of PMOS M$_{\mathrm{p1}}$ with a channel length of 65 nm is approximately 32 µm, the second derivative g$_{m}$$^{\prime}$$^{\prime}$ of the transconductance g$_{m}$ with respect to the small signal input voltage v$_{i}$ is zero.

Fig. 7 shows the simulated output-referred third-order intercept point (OIP3) of the proposed down-conversion mixer according to the width of PMOS M$_{\mathrm{p1}}$ with a channel length of 65 nm. The largest simulated OIP3 result of the down-conversion mixer is obtained when the width of PMOS M$_{\mathrm{P1}}$ with a channel length of 65 nm is approximately 35 µm. The proposed I/Q down-conversion mixer improves more than 3 dB OIP3 performance in comparison to the conventional I/Q down-conversion mixer with an NMOS transconductor.

The balancing accuracy between the I and Q paths is critical in achieving the targeted signal-to-noise ratio in the direct-conversion receiver. Especially, it is important to obtain balanced I/Q LO signals in the millimeter-wave band ⁽¹¹⁾. Fig. 8 shows the simplified schematic of the I/Q LO generation circuits. The I/Q LO generation circuit consists of a single-ended cascode amplifier, a single-to-differential common-source amplifier, two-stage RC polyphase filter (PPF), and two differential common-source amplifiers. These amplifiers are used to compensate for the signal attenuation by two-stage PPF composed of passive components and to drive the switching pairs of the down-conversion mixer. The values of the resistance and capacitance of the two-stage PPF are destermined by considering the parasitic inductance value from equation in ⁽¹²⁾.

III. Experimental Results

The proposed 24-GHz direct-conversion receiver front-end was implemented using a 65-nm CMOS process as part of an in-cabin radar system. Fig. 9 shows the chip photograph of the receiver front-end. The silicon area of the receiver front-end is 1250 ${\mathrm{\mu}}$m ${\times}$ 550 ${\mathrm{\mu}}$m excluding ESD protection circuits and PADs. It draws 21-mA from a 1.2-V supply voltage. For the measurement, an external signal generation instrument is used to provide the LO input signal. The measured return losses of the RF input and LO input are less than -10 dB in the frequency range from 24 GHz to 24.5 GHz.

Fig. 10 presents the measured RF input return loss of the RF front-end, which is less than -10 dB from 24 GHz to 24.5 GHz. Fig. 11 shows the measured conversion gain and phase mismatch between the I-path and Q-path at the IF output with an IF of 10 MHz of the RF front-end versus RF input frequency. The RF front-end shows a conversion gain of greater than 30 dB in the frequency range of 24 - 24.5 GHz. The gain and phase mismatch

between the I- and Q-paths at the IF output are less than 0.2 dB and 0.5$^{\circ}$ in the frequency range from 24 GHz to 24.5 GHz, respectively.

$^{(1)}$ OIP3 ${\approx}$ Gain + IIP3 (= P$_{\mathrm{1dB}}$ + 9.6dB),

$^{(2)}$ $^{\mathrm{F.O.M.=\frac{Gain[abs]\cdot IIP3[mW]}{P_{DC}[mW]}\cdot \frac{1}{(NF- 1)[abs]}\cdot f[GHz]}}$

Fig. 12 shows the measured IIP3 and double-side band (DSB) noise figure (NF) results of the proposed RF front-end, respectively. Two-tone spacing is 2 MHz for the linearity test. The RF front-end has an IIP3 of greater than -10 dBm over the entire operating frequency bands. The RF front-end has a DSB NF of less than 4.5 dB over the entire operating frequency bands.

Table 1 summarized and compares the performances of the proposed RF front-end with those of other published mm-wave RF front-ends. The proposed RF front-end has good figure of merit (FOM) compared with the recently published millimeter-wave RF front-ends.

IV. CONCLUSIONS

A 24-GHz direct-conversion receiver front-end was presented for in-cabin applications. The proposed RF receiver is composed of a low-noise amplifier, an I/Q down-conversion mixer, and an I/Q LO generator circuit. An inverter transconductor with third-order nonlinearity cancellation is applied to the I/Q down-conversion mixer to improve the linearity of the I/Q down-conversion mixer. By adopting a linear I/Q down-conversion mixer and a balanced I/Q LO generator, the 24 GHz receiver front-end obtains high linearity and good I/Q balancing performance. The mm-wave receiver front-end can be utilized to in-cabin radar systems that can monitor driver’s vital signs such as heart rate and respiration.