Mobile QR Code QR CODE

  1. (Department of Electrical Engineering, Pusan National University, Busan, Korea)
  2. (Samsung Electronics, Suwon, Gyeonggi-do, Korea)



Down-conversion mixer, 5G, millimeter-wave, source degeneration inductors

I. INTRODUCTION

The millimeter-wave (mmWave) 28 GHz band is a widely used 5G band. Among the various mmWave 5G bands, the mmWave 28 GHz band is widely used because of the availability of spectrum in the greatest number of countries across the globe, reduced device complexity, and better propagation. Especially, 5G mmWave 28 GHz band n257 is a frequency band defined for short range millimeter-wave communications at high data rates in 5G-new radio networks and covers 26.5 to 29.5 GHz for South Korea, North America, and Japan [1- 4].

A direct-conversion receiver architecture is a suitable candidate for implementing low-power 5G mmWave transceivers to enhance energy efficiency [5]. The linearity performance of the direct-conversion receiver front-end is primarily determined by the down-conversion mixer. Consequently, realizing a linear down-conversion mixer with low power consumption is essential. Among the methods to improve the linearity of a mixer, the derivative superposition technique, which can compensate for the nonlinearity of the main transistor with the nonlinearity of the auxiliary transistor, is often used in mixer circuits [6, 7]. In the early stages of circuit design at the schematic level, when source degeneration inductors are used for the main transistor and the auxiliary transistor, the coupling factor between the inductors is typically neglected during linearity optimization. However, during the actual layout implementation, the mutual coupling between these inductors causes a deviation from the originally optimized linearity point, leading to a need for redesign [8]. To minimize the magnetic coupling effect between the source degeneration inductors used in the main and auxiliary transistors, the inductors must be spaced at least 400 µm apart, as shown in [8]. However, this results in increased silicon area utilization.

In this paper, we propose a source-degeneration inductor structure that shares the inner diameter of the source-degeneration inductors used in the main and auxiliary transistors to reduce the silicon area. Additionally, we verify the derivative superposition linearization technique by considering the coupling coefficient of these inductors, applying it to the design of a 28 GHz down-conversion mixer with a local oscillator buffer in 65-nm CMOS technology.

II. PROPOSED CIRCUIT DESIGN

Fig. 1 shows the simplified schematic of the down-conversion mixer with an LO buffer using the proposed source-degeneration inductor structure which shares the inner diameter of the source-degeneration inductors. When the differential superposition technique, which compensates the nonlinearity of the auxiliary transistor with that of the main transistor, is applied to a mixer circuit, the source inductor for the main transistor and that for the auxiliary transistor are typically implemented separately, as illustrated in Fig. 1. As mentioned in the introduction, the spacing between the inductors should be at least 400 µm to minimize the magnetic coupling effect between the source degeneration inductors used in the main and auxiliary transistors. This leads to an increased utilization of silicon area. Therefore, to reduce the silicon area while taking advantage of the coupling effect between the source inductors used for the main and auxiliary transistors, we propose a shared source inductor design, where the inner diameter of the source inductors is common, as shown in Fig. 1, and apply the differential superposition technique.

Fig. 1. Simplified schematic of the down-conversion mixer with LO buffer using the proposed source-degeneration inductor structure that shares the inner diameter of the source-degeneration inductors.

../../Resources/ieie/JSTS.2026.26.2.159/fig1.png

Volterra series analysis is conducted to rigorously examine the circuit design approach based on the derivative superposition technique, specifically including the impact of inductive coupling between the inductors connected to the main and auxiliary transistors. Fig. 2 illustrates the simplified small-signal model employed to derive the nonlinearity equations of the transconductor shown in Fig. 1. The parasitic gate-drain capacitance $C_{gd}$ is neglected to simplify the nonlinearity analysis. $v_{in}$ denotes the input voltage signal source. $L_1$ and $L_2$ represent the source-degeneration inductors, and $M$ indicates the mutual inductance between them. $C_{gs1}$ and $C_{gs2}$ are the gate-to-source capacitances of M$_1$ and M$_2$, respectively. As shown in Fig. 2, $v_{in}$, $i_1$, and $i_2$ can be expressed using KCL and KVL as follow:

Fig. 2. Simplified small-signal equivalent schematic of the transconductor for nonlinearity analysis.

../../Resources/ieie/JSTS.2026.26.2.159/fig2.png
(1)
$v_{in} = v_{gs1} + v_{s1} = v_{gs1} + j\omega L_1 i_1 - j\omega M i_2$ $= v_{gs2} + v_{s2} = v_{gs2} + j\omega L_2 i_2 - j\omega M i_1,$
(2)
$i_1 = j\omega v_{gs1} C_{gs1} + i_a \text{ and } i_2 = j\omega v_{gs2} C_{gs2} + i_b,$

where $v_{s1}$ and $v_{s2}$ are expressed as $v_{s1} = j\omega L_1 i_1 - j\omega M i_2$ and $v_{s2} = j\omega L_2 i_2 - j\omega M i_1$. Using the Volterra series representation, the gate-to-source voltage $v_{gs1}$ and the drain-source current $i_a$ of the main transistor M$_1$, as well as the gate-to-source voltage $v_{gs2}$ and the drain-source current $i_b$ of the auxiliary transistor M$_2$, can be expressed as follows:

(3)
$v_{gs1} = A_1(j\omega) \circ v_{in} + A_2(j\omega_1, j\omega_2) \circ v_{in}^2$ $+ A_3(j\omega_1, j\omega_2, j\omega_3) \circ v_{in}^3,$
(4)
$i_a = g_{1A} v_{gs1} + g_{2A} v_{gs1}^2 + g_{3A} v_{gs1}^3$ $= B_1(j\omega) \circ v_{in} + B_2(j\omega_1, j\omega_2) \circ v_{in}^2$ $+ B_3(j\omega_1, j\omega_2, j\omega_3) \circ v_{in}^3,$
(5)
$v_{gs2} = C_1(j\omega) \circ v_{in} + C_2(j\omega_1, j\omega_2) \circ v_{in}^2$ $+ C_3(j\omega_1, j\omega_2, j\omega_3) \circ v_{in}^3,$
(6)
$i_b = g_{1B} v_{gs2} + g_{2B} v_{gs2}^2 + g_{3B} v_{gs2}^3$ $= D_1(j\omega) \circ v_{in} + D_2(j\omega_1, j\omega_2) \circ v_{in}^2$ $+ D_3(j\omega_1, j\omega_2, j\omega_3) \circ v_{in}^3,$

where $v_{in}^n$ represents the $n$th power of the voltage source signal, and $A_n(j\omega)$, $B_n(j\omega)$, $C_n(j\omega)$, and $D_n(j\omega)$ denote the Volterra-series coefficients, which are linear functions of $n$ frequencies [9]. The Volterra-series coefficients $A_n(j\omega)$ and $C_n(j\omega)$ are obtained as using Eqs. (1), (2), (3), and (5), as detailed in Appendix A. The corresponding coefficients are given as $g_{1A} = g_{m1}$, $g_{2A} = \frac{1}{2} g_{m1}'$, $g_{3A} = \frac{1}{6} g_{m1}''$, $g_{1B} = g_{m2}$, $g_{2B} = \frac{1}{2} g_{m2}'$, $g_{3B} = \frac{1}{6} g_{m2}''$. The parameters $g_{m1}$ and $g_{m2}$ represent the first-order transconductance of each transistor, while $g_{m1}'$, $g_{m1}''$, $g_{m2}'$, and $g_{m2}''$ denote the first- and second-order derivatives corresponding to the nonlinear characteristics of the devices. Accordingly, $g_{2A}$ and $g_{3A}$ represent the coefficients of the second- and third-order nonlinear terms, respectively, where the factors $\frac{1}{2}$ and $\frac{1}{6}$ are derived from the Taylor series expansion.

From Eqs. (3)-(6), the Volterra-series coefficients $B_n(j\omega)$ and $D_n(j\omega)$ is expressed as

(7)
$B_1(j\omega) = g_{1A} A_1(j\omega),$
(8)
$B_2(j\omega_1, j\omega_2)$ $= g_{1A} A_2(j\omega_1, j\omega_2) + g_{2A} A_1(j\omega_1) A_1(j\omega_2),$
(9)
$B_3(j\omega_1, j\omega_2, j\omega_3)$ $= g_{1A} A_3(j\omega_1, j\omega_2, j\omega_3) + 2g_{2A} A_1(j\omega) A_2(j\omega_1, j\omega_2)$ $+ g_{3A} A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3),$
(10)
$D_1(j\omega) = g_{1B} C_1(j\omega),$
(11)
$D_2(j\omega_1, j\omega_2)$ $= g_{1B} C_2(j\omega_1, j\omega_2) + g_{2B} C_1(j\omega_1) C_1(j\omega_2),$
(12)
$D_3(j\omega_1, j\omega_2, j\omega_3)$ $= g_{1B} C_3(j\omega_1, j\omega_2, j\omega_3) + 2g_{2B} C_1(j\omega) C_2(j\omega_1, j\omega_2)$ $+ g_{3B} C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3),$

where

$\overline{A_1(j\omega) A_2(j\omega_1, j\omega_2)} = \frac{1}{3} [A_1(j\omega_1) A_2(j\omega_2, j\omega_3)$

$\quad\quad\quad\quad\quad\quad\quad\quad\quad+ A_1(j\omega_2) A_2(j\omega_1, j\omega_3)$

$\quad\quad\quad\quad\quad\quad\quad\quad\quad+ A_1(j\omega_3) A_2(j\omega_1, j\omega_2)],$

and

$\overline{C_1(j\omega) C_2(j\omega_1, j\omega_2)} = \frac{1}{3} [C_1(j\omega_1) C_2(j\omega_2, j\omega_3)$

$\quad\quad\quad\quad\quad\quad\quad\quad\quad+ C_1(j\omega_2) C_2(j\omega_1, j\omega_3)$

$\quad\quad\quad\quad\quad\quad\quad\quad\quad+ C_1(j\omega_3) C_2(j\omega_1, j\omega_2)].$

In Fig. 2, the total drain current $i_t$ is expressed as

$i_t = i_a + i_b$

$= H_1(j\omega) \circ v_{in} + H_2(j\omega_1, j\omega_2) \circ v_{in}^2$

$+ H_3(j\omega_1, j\omega_2, j\omega_3) \circ v_{in}^3,$

where Volterra-series coefficient $H_1(j\omega)$ is given by the sum of $B_1(j\omega)$ and $D_1(j\omega)$, while $H_3(j\omega_1, j\omega_2, j\omega_3)$ is obtained as the summation of $B_3(j\omega_1, j\omega_2, j\omega_3)$ and $D_3(j\omega_1, j\omega_2, j\omega_3)$.

From Eqs. (9) and (12), $H_3(j\omega_1, j\omega_2, j\omega_3)$ can be expressed as follows:

(13)

$H_3(j\omega_1, j\omega_2, j\omega_3)$

$= B_3(j\omega_1, j\omega_2, j\omega_3) + D_3(j\omega_1, j\omega_2, j\omega_3)$

$= \left[ 1 + \frac{g_{1A} \left( \begin{array}{l} g_{1B}(\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs2} \\ - j(\omega_1 + \omega_2 + \omega_3) L_1 \end{array} \right)}{V(j(\omega_1 + \omega_2 + \omega_3))} \right.$

$\left. + \frac{j(\omega_1 + \omega_2 + \omega_3) M g_{1B}}{V(j(\omega_1 + \omega_2 + \omega_3))} \right]$

$\times \left( \begin{array}{l} 2 \overline{g_{2A} A_1(j\omega_1) A_2(j\omega_1, j\omega_2)} \\ + g_{3A} A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3) \end{array} \right)$

$+ \left[ 1 + \frac{g_{1B} \left( \begin{array}{l} g_{1A}(\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs1} \\ - j(\omega_1 + \omega_2 + \omega_3) L_2 \end{array} \right)}{V(j(\omega_1 + \omega_2 + \omega_3))} \right.$

$\left. + \frac{j(\omega_1 + \omega_2 + \omega_3) M g_{1A}}{V(j(\omega_1 + \omega_2 + \omega_3))} \right]$

$\times \left( \begin{array}{l} 2 \overline{g_{2B} C_1(j\omega_1) C_2(j\omega_1, j\omega_2)} \\ + g_{3B} C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3) \end{array} \right).$

To reduce the design degrees of freedom, if the terms inside the brackets of Eq. (13) are designed to be identical, the following condition is satisfied:

(14)
$C_{gs1}(L_1 + M) = C_{gs2}(L_2 + M).$

Moreover, if Eq. (14) is satisfied, $A_1(j\omega)$ and $C_1(j\omega)$ in Appendix A become identical. Therefore, under the condition given in Eq. (14), Eq. (13) can be expressed as follows.

(15)

$H_3(j\omega_1, j\omega_2, j\omega_3)$

$= \left[ 1 + \frac{g_{1A} \left( \begin{array}{l} g_{1B}(\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs2} \\ - j(\omega_1 + \omega_2 + \omega_3) L_1 \end{array} \right)}{V(j(\omega_1 + \omega_2 + \omega_3))} \right.$

$\left. + \frac{j(\omega_1 + \omega_2 + \omega_3) M g_{1B}}{V(j(\omega_1 + \omega_2 + \omega_3))} \right]$

$\times \left( \begin{array}{l} 2 \overline{g_{2A} A_1(j\omega_1) A_2(j\omega_1, j\omega_2)} \\ + g_{3A} A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3) \\ + 2 \overline{g_{2B} C_1(j\omega_1) C_2(j\omega_1, j\omega_2)} \\ + g_{3B} C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3) \end{array} \right).$

The third-order intermodulation intercept point (IIP3) at the frequency $2\omega_b - \omega_a$ can be calculated using the Volterra-series coefficient $H_3(j\omega_1, j\omega_2, j\omega_3)$, with the condition that $\omega_1 = \omega_b$, $\omega_2 = \omega_b$, and $\omega_3 = -\omega_a$. When the difference between the adjacent frequencies $\omega_a$ and $\omega_b$ is small, it can be assumed that $\omega \approx \omega_a \approx \omega_b$ [10]. The input tone amplitude at the intercept point of the IMD3 response at $2\omega_b - \omega_a$ with the fundamental response at $\omega_a$ is given by [7].

(16)
$AIP3(2\omega_b - \omega_a) = \sqrt{\frac{4}{3} \left| \frac{H_1(j\omega_b)}{H_3(j\omega_b, j\omega_b, -j\omega_a)} \right|}.$

Complex mathematical analyses using Eq. (15) and (A.3)-(A.6) lead to the following expression for $H_3(j\omega_b, j\omega_b, -j\omega_a)$.

(17)

$H_3(j\omega_b, j\omega_b, -j\omega_a)$

$\approx H_3(j\omega, j\omega, -j\omega)$

$= A_1(j\omega) |A_1(j\omega)|^2$

$\times \left[ 1 + \frac{g_{1A} \left\{ \begin{array}{l} g_{1B} \omega^2 (L_1 L_2 - M^2) \\ + j\omega^3 (L_1 L_2 - M^2) C_{gs2} \\ - j\omega L_1 \end{array} \right\} + j\omega M g_{1B}}{V(j\omega)} \right]$

$\times \left[ \begin{array}{l} (g_{3A} + g_{3B}) \\ + \frac{2}{3} \frac{\left[ \begin{array}{l} 4\omega^2 (L_1 L_2 - M^2) (g_{2A}^2 g_{1B} + g_{1A} g_{2B}^2) \\ + j \left\{ \begin{array}{l} 8\omega^2 (L_1 L_2 - M^2) \\ \times (C_{gs2} g_{2A}^2 + C_{gs1} g_{2B}^2) \\ - 2\omega (L_1 g_{2A}^2 - 2M g_{2A} g_{2B} \\ + L_2 g_{2B}^2) \end{array} \right\} \end{array} \right]}{V(j2\omega)} \end{array} \right].$

As shown in Eq. (17), the coupling effect between source degeneration inductors influences the linearity of the transconductance stage. Maintaining the relationship expressed in Eq. (14), as shown in Eq. (16), the design was carried out to enhance linearity by minimizing the final term in Eq. (17), such that it approaches zero as closely as possible.

Fig. 3(a) shows the layout of inductors L$_1$ and L$_2$ used for the main and auxiliary transistors, respectively. Fig. 3(b) presents the inductance and coupling coefficient val-

Fig. 3. (a) Layout of the source degeneration inductors L$_1$ and L$_2$, (b) inductance and coupling coefficient of the source degeneration inductors L$_1$ and L$_2$.

../../Resources/ieie/JSTS.2026.26.2.159/fig3.png

ues obtained through EM simulation for the inductor layout shown in Fig. 3(a). At a frequency of 28 GHz, the coupling coefficient between the two inductors is 0.193, and the inductance values of L$_1$ and L$_2$ are 590 pH and 380 pH, respectively. The bias voltages (V$_{B1}$ and V$_{B2}$) for the main and auxiliary transistors are provided by the current mirror bias circuits [5]. To enhance linearity at a low supply voltage without stacking, the transconductance stage and switching stage are magnetically coupled via the on-chip $2:1$ transformer T$_1$.

III. MEASUREMENT RESULTS

The proposed 28- GHz down-conversion mixer employing magnetically coupled source degeneration inductors with an LO buffer was implemented using a 65-nm CMOS process for verification. Fig. 4 shows the chip photograph of the mixer with an LO buffer, of which silicon area is 700 µm $\times$ 500 µm excluding and PADs. The

Fig. 4. Chip photograph of the down-conversion mixer with an LO buffer.

../../Resources/ieie/JSTS.2026.26.2.159/fig4.png

down-conversion mixer draws 11.7-mA from a 1-V supply voltage. For the measurements, a $-10$ dBm power level was applied to the LO input via an external signal generator for the measurements.

Fig. 5 presents the measured RF input |S$_{11}$|, conversion gain, noise figure, and IIP3 performance. The measured RF input |S$_{11}$| is less than -10 dB from 26.5 GHz to 29.5 GHz. The conversion gain is greater than 10.6 dB over the 26.5–29.5 GHz frequency range and reaches 11.8 dB at 28 GHz frequency. As shown in Fig. 5(b), a noise figure of 9.7 dB was achieved at a 50 MHz IF frequency, with a gain of 11.8 dB at an RF operating frequency of 28 GHz. Fig. 5(c) shows the measured IIP3 under two tone excitation at 28 and 28.01 GHz with an LO frequency of 27.99 GHz. The proposed mixer achieves an IIP3 that is 5 dB higher than that of a conventional mixer without the linearization technique. Table 1 compares the performances of the proposed 28 GHz down-conversion mixer with those of other mmWave ones. The proposed mixer has good figure of merit (FOM) compared with the previously published mmWave ones.

Fig. 5. (a) Measured conversion gain and RF input return loss, (b) double side-band NF, (c) IIP3 of the proposed mixer.

../../Resources/ieie/JSTS.2026.26.2.159/fig5.png

IV. CONCLUSIONS

This paper has presented a 5G millimeter-wave CMOS down-conversion mixer employing inner-diameter-shared source degeneration inductors with magnetic coupling for 5G wireless communication systems. By taking the coupling coefficient of the inductors into account, a derivative superposition linearization technique was applied to enhance the linearity of the proposed mixer. The proposed mixer demonstrates promising linearity performance and a competitive figure of merit, indicating its suitability for 28- GHz 5G applications.

Table 1. Summary and comparison of performance.

[3] [11] [12] [13] This work
Operating frequency ( GHz) 26.5-29.5 31 26.5-39 24-40 26.5-29.5
Gain (dB) 11 3.4 3.87$^{**}$ $-4.1$-1.2 11.8
OIP3 (dBm) 13.5 21.4 4.54$^{**}$ 4-6.8 18.8
NF (dB) 8.9 9.5 13.62$^{**}$ 12.4-15.3 (SSB NF) 9.74 @ 50 MHz IF
LO input power (dBm) -10$^*$ 3 5 N.A. $-10^*$
Power consumption (mW) 13.7/23.8$^*$ @ 1 V (Mixer + LO buffer) 21.2 @ 1.5 V (Mixer) 9.75 @ 1 V (Mixer) 7.9/17.6$^*$ @ 1.1 V (Mixer + IF buffer + LO buffer) 11.7/21.7$^*$ @ 1 V (Mixer + LO buffer$^*$)
Technology 65 nm CMOS 45 nm SOI CMOS 65 nm CMOS 40 nm CMOS 65 nm CMOS
Area 0.39 mm$^2$ 0.8 mm$^2$ 0.4 mm$^2$ 0.654 mm$^2$ 0.35 mm$^2$
F.O.M$^{(1)}$ [8] 1.9/1.1$^*$ @ 28 GHz 17.2 @ 31 GHz 0.24 @ 28 GHz 0.61/0.27 @ 28 GHz$^{(2)}$ 5.5/2.99$^*$ @ 28 GHz

$^*$ Value including LO buffer, $^{**}$ Value at 28 GHz

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

$^{(2)}$ Value calculated based on data estimated in [13]

Appendices

APPENDIX A

In the main transistor of Fig. 2, the following equation can be derived:

(A.1)

$v_{in} = v_{gs1} + v_{s1}$

$= v_{gs1} + j\omega L_1 i_1 - j\omega M i_2$

$= v_{gs1} + j\omega L_1 (j\omega v_{gs1} C_{gs1} + i_a)$

$- j\omega M (j\omega v_{gs2} C_{gs2} + i_b)$

$= (1 - \omega^2 L_1 C_{gs1} + j\omega L_1 g_{1A}) v_{gs1} + j\omega L_1 g_{2A} v_{gs1}^2$

$+ j\omega L_1 g_{3A} v_{gs1}^3 + (\omega^2 M C_{gs2} - j\omega M g_{1B}) v_{gs2}$

$- j\omega M g_{2B} v_{gs2}^2 - j\omega M g_{3B} v_{gs2}^3$

$= (1 - \omega^2 L_1 C_{gs1} + j\omega L_1 g_{1A})$

$\times [A_1(j\omega) \cdot v_{in} + A_2(j\omega_1, j\omega_2) \cdot v_{in}^2$

$+ A_3(j\omega_1, j\omega_2, j\omega_3) \cdot v_{in}^3]$

$+ j\omega L_1 g_{2A} [A_1(j\omega_1) A_1(j\omega_2) \cdot v_{in}^2$

$+ 2A_1(j\omega) A_2(j\omega_1, j\omega_2) \cdot v_{in}^3 + \cdots]$

$+ j\omega L_1 g_{3A} [A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3) \cdot v_{in}^3 + \cdots]$

$+ (\omega^2 M C_{gs2} - j\omega M g_{1B})$

$\times [C_1(j\omega) \cdot v_{in} + C_2(j\omega_1, j\omega_2) \cdot v_{in}^2$

$+ C_3(j\omega_1, j\omega_2, j\omega_3) \cdot v_{in}^3]$

$- j\omega M g_{2B} [C_1(j\omega_1) C_1(j\omega_2) \cdot v_{in}^2$

$+ 2C_1(j\omega) C_2(j\omega_1, j\omega_2) \cdot v_{in}^3 + \cdots]$

$- j\omega M g_{3B} [C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3) \cdot v_{in}^3 + \cdots].$

Similiarly, in the auxiliary transistor, the following equation can be derived:

(A.2)

$v_{in} = v_{gs2} + v_{s2}$

$= v_{gs2} + j\omega L_2 i_2 - j\omega M i_1$

$= v_{gs2} + j\omega L_2 (j\omega v_{gs2} C_{gs2} + i_b)$

$- j\omega M (j\omega v_{gs1} C_{gs1} + i_a)$

$= (1 - \omega^2 L_2 C_{gs2} + j\omega L_2 g_{1B}) v_{gs2} + j\omega L_2 g_{2B} v_{gs2}^2$

$+ j\omega L_2 g_{3B} v_{gs2}^3 + (\omega^2 M C_{gs1} - j\omega M g_{1A}) v_{gs1}$

$- j\omega M g_{2A} v_{gs1}^2 - j\omega M g_{3A} v_{gs1}^3$

$= (1 - \omega^2 L_2 C_{gs2} + j\omega L_2 g_{1B}) [C_1(j\omega) \cdot v_{in}$

$+ C_2(j\omega_1, j\omega_2) \cdot v_{in}^2 + C_3(j\omega_1, j\omega_2, j\omega_3) \cdot v_{in}^3]$

$+ j\omega L_2 g_{2B} [C_1(j\omega_1) C_1(j\omega_2) \cdot v_{in}^2$

$+ 2C_1(j\omega) C_2(j\omega_1, j\omega_2) \cdot v_{in}^3 + \cdots]$

$+ j\omega L_2 g_{3B} [C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3) \cdot v_{in}^3 + \cdots]$

$+ (\omega^2 M C_{gs1} - j\omega M g_{1A}) [A_1(j\omega) \cdot v_{in}$

$+ A_2(j\omega_1, j\omega_2) \cdot v_{in}^2 + A_3(j\omega_1, j\omega_2, j\omega_3) \cdot v_{in}^3]$

$- j\omega M g_{2A} [A_1(j\omega_1) A_1(j\omega_2) \cdot v_{in}^2$

$+ 2A_1(j\omega) A_2(j\omega_1, j\omega_2) \cdot v_{in}^3 + \cdots]$

$- j\omega M g_{3A} [A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3) \cdot v_{in}^3 + \cdots].$

The conditions that satisfy Eqs. (A.1) and (A.2) are as follows:

(A.3)
$A_1(j\omega) = \frac{1 - \omega^2 (L_2 + M) C_{gs2} + j\omega (L_2 + M) g_{1B}}{V(j\omega)},$
(A.4)
$C_1(j\omega) = \frac{1 - \omega^2 (L_1 + M) C_{gs1} + j\omega (L_1 + M) g_{1A}}{V(j\omega)},$
(A.5)

$A_2(j\omega_1, j\omega_2)$

$= \left[ \begin{array}{l} (\omega_1 + \omega_2)^2 (L_1 L_2 - M^2) g_{1B} g_{2A} \\ + j(\omega_1 + \omega_2)^3 (L_1 L_2 - M^2) C_{gs2} g_{2A} \\ - j(\omega_1 + \omega_2) L_1 g_{2A} \end{array} \right]$

$\times A_1(j\omega_1) A_1(j\omega_2)$

$+ \frac{j(\omega_1 + \omega_2) M g_{2B} C_1(j\omega_1) C_1(j\omega_2)}{V(j(\omega_1 + \omega_2))},$
(A.6)

$C_2(j\omega_1, j\omega_2)$

$= \left[ \begin{array}{l} (\omega_1 + \omega_2)^2 (L_1 L_2 - M^2) g_{1A} g_{2B} \\ + j(\omega_1 + \omega_2)^3 (L_1 L_2 - M^2) C_{gs1} g_{2B} \\ - j(\omega_1 + \omega_2) L_2 g_{2B} \end{array} \right]$

$\times C_1(j\omega_1) C_1(j\omega_2)$

$+ \frac{j(\omega_1 + \omega_2) M g_{2A} A_1(j\omega_1) A_1(j\omega_2)}{V(j(\omega_1 + \omega_2))},$
(A.7)

$A_3(j\omega_1, j\omega_2, j\omega_3)$

$= 2 \overline{A_1(j\omega) A_2(j\omega_1, j\omega_2)}$

$\times \left[ \begin{array}{l} (\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) g_{1B} g_{2A} \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs2} g_{2A} \\ - j(\omega_1 + \omega_2 + \omega_3) L_1 g_{2A} \end{array} \right]$

$+ A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3)$

$\times \left[ \begin{array}{l} (\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) g_{1B} g_{3A} \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs2} g_{3A} \\ - j(\omega_1 + \omega_2 + \omega_3) L_1 g_{3A} \end{array} \right]$

$+ 2 \overline{C_1(j\omega) C_2(j\omega_1, j\omega_2)} [j(\omega_1 + \omega_2 + \omega_3) M g_{2B}]$

$+ \frac{C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3) [j(\omega_1 + \omega_2 + \omega_3) M g_{3B}]}{V(j(\omega_1 + \omega_2 + \omega_3))},$

(A.8)

$C_3(j\omega_1, j\omega_2, j\omega_3)$

$= 2 \overline{C_1(j\omega) C_2(j\omega_1, j\omega_2)}$

$\times \left[ \begin{array}{l} (\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) g_{1A} g_{2B} \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs1} g_{2B} \\ - j(\omega_1 + \omega_2 + \omega_3) L_2 g_{2B} \end{array} \right]$

$+ C_1(j\omega_1) C_1(j\omega_2) C_1(j\omega_3)$

$\times \left[ \begin{array}{l} (\omega_1 + \omega_2 + \omega_3)^2 (L_1 L_2 - M^2) g_{1A} g_{3B} \\ + j(\omega_1 + \omega_2 + \omega_3)^3 (L_1 L_2 - M^2) C_{gs1} g_{3B} \\ - j(\omega_1 + \omega_2 + \omega_3) L_2 g_{3B} \end{array} \right]$

$+ 2 \overline{A_1(j\omega) A_2(j\omega_1, j\omega_2)} [j(\omega_1 + \omega_2 + \omega_3) M g_{2A}]$

$+ \frac{A_1(j\omega_1) A_1(j\omega_2) A_1(j\omega_3) [j(\omega_1 + \omega_2 + \omega_3) M g_{3A}]}{V(j(\omega_1 + \omega_2 + \omega_3))},$

where $V(j\omega)$ is given by the following:

(A.9)

$V(j\omega) = (1 - \omega^2 L_1 C_{gs1} + j\omega L_1 g_{1A})$

$\times (1 - \omega^2 L_2 C_{gs2} + j\omega L_2 g_{1B})$

$- (\omega^2 M C_{gs2} - j\omega M g_{1B})$

$\times (\omega^2 M C_{gs1} - j\omega M g_{1A}).$

ACKNOWLEDGEMENT

This work was supported by the Financial Supporting Project of Long-term Overseas Dispatch of PNU's Tenure-track Faculty, 2022.

REFERENCES

1 
, https://www.qualcomm.com/news/releases/2018/07/qualcomm-delivers-breakthrough-5g-nr-mmwave-and-sub-6-ghz-rf-modules-mobile.URL
2 
Han Y. , Lee S. , Lee D.-H. , Jeong G. , 2024, A 28- GHz CMOS power amplifier with ribbon cell, Journal of Integrated Circuits and System, Vol. 10, No. 3DOI
3 
Myung J. , Park G. , Kim H. , Kwak Y. , Nam I. , 2024, A 28- GHz down-converter for 5G direct-conversion receivers in 65-nm CMOS technology, Journal of Integrated Circuits and System, Vol. 10, No. 3DOI
4 
Chen F. , Wu T. , 2024, Contraband identification algorithm for intelligent millimter-wave security screening device based on regional-convolution neural network algorithm for civil aviation, IEIE Transactions on Smart Prcocessing and Computing, Vol. 13, No. 2, pp. 186-195DOI
5 
Kim H. T. , Park B.S. , Song S.S. , Moon T.S. , Kim S.H. , Kim J.M. , Chang J.Y. , Ho Y.C. , 2018, A 28- GHz CMOS direct conversion transceiver with packaged 2$\times$4 antenna array for 5G cellular system, IEEE Journal of Solid-State Circuits, Vol. 53, No. 5, pp. 1245-1259DOI
6 
Kim B. , Ko J.-S. , Lee K. , 2000, A new linearization technique for MOSFET RF amplifier using mulitple gated transistors, IEEE Microwave and Guided Wave Letters, Vol. 10, No. 9, pp. 371-373DOI
7 
Aparin V. , Larson L. E. , 2005, Modified derivative superposition method for linearizaing FET low-noise amplifiers, IEEE Transactions on Microwave Theory and Techniques, Vol. 53, No. 2, pp. 571-581DOI
8 
Kim C. , Kim M. , Jeon Y. , Lee O. , Son J. H. , Nam I. , 2019, A 28- GHz CMOS down-conversion mixer with low-magnetic-coupled source degeneration inductors for 5G applications, Journal of Semiconductor Technology and Science, Vol. 19, No. 4, pp. 373-377DOI
9 
Pedro J. C. , Carvalho N. B. , 2003, Intermodulation Distortion in Microwave and Wireless CircuitsGoogle Search
10 
Kwon K. , Nam I. , 2013, A linearization technique for a transconductor using vertical bipolar junction trassistors in a CMOS process, IEEE Transactions on Microwave Theory and Techniques, Vol. 61, No. 1, pp. 195-203DOI
11 
Wu C.-L. , Yu C. , O K. K. , 2015, Amplification of nonlinearity in multiple gate transistor millimetre wave mixer for improvement of linearity and noise figure, IEEE Microwave and Wireless Components Letters, Vol. 25, No. 5, pp. 310-312DOI
12 
Bae B. S. , Han J. H. , 2022, A wideband CMOS dwon-conversion mixer for 5G FR2 applications, Journal of Integrated Circuits and System, Vol. 8, No. 3DOI
13 
Lee D. , Lee M. , Park B. , Song E. , Lee K. , Lee J. , Han J. , Kwon K. , 2022, 24-40 GHz mmWave down-conversion mixer with broadband capacitor-tuned coupled resonators for 5G new radio cellular applications, IEEE Access, Vol. 10, pp. 16782-16792DOI
Jinman Myung
../../Resources/ieie/JSTS.2026.26.2.159/au1.png

Jinman Myung received his B.S. degree in electronic engineering from Dong-A University, Busan, Korea, in 2019, and he is currently pursuing a combined M.S/Ph.D. degree in electrical engineering at Pusan National University, Busan, Korea. His main interests are CMOS RF/mmWave/PMIC/ analog circuits for wireless communications.

Geonwoo Park
../../Resources/ieie/JSTS.2026.26.2.159/au2.png

Geonwoo Park received his B.S degree in information and communication engineering from Kyungsung University, Busan, Korea, in 2019 and and an M.S degree in electrical engineering from Pusan National University, Busan, Korea, in 2021, and is currently working toward a Ph.D. degree in electrical and electronic engineering at Pusan National University, Busan, Korea. His current research interests are CMOS RF/PMIC/analog circuits for wireless communications.

Ho Kim
../../Resources/ieie/JSTS.2026.26.2.159/au3.png

Ho Kim received his B.S. degree in electronics engineering from Kumoh National Institute of Technology, Gumi, Korea, in 2019, and an M.S. degree in electrical and electronics Engineering from Pusan National University, Busan, Korea, in 2024. His research interests include CMOS RF, millimeter-wave (mmWave), and analog circuits for wireless communications.

Suyeon Lee
../../Resources/ieie/JSTS.2026.26.2.159/au4.png

Suyeon Lee received her B.S. degree in electrical engineering from Pusan National University, Busan, Korea, in 2020 and an M.S. degree in electrical and electronics engineering from Pusan National University, Busan, Korea, in 2022. Since 2022, she has been with Samsung Electronics. Her main interests are CMOS RF/mmWave/analog circuits for wireless communications.

Ilku Nam
../../Resources/ieie/JSTS.2026.26.2.159/au5.png

Ilku Nam received his B.S. degree in electronics engineering from Yonsei University, in Seoul, Korea, in 1999 and his MS. and Ph.D. degrees in electrical engineering and computer science from Korea Advanced Institute of Science and Technology (KAIST), in Daejeon, Korea, in 2001 and 2005, respectively. His Ph.D. work at KAIST was related to low-power CMOS RF transceiver integrated circuits (IC) design for low-power IoT applications. From 2005 to 2007, he was a Senior Engineer with RF development team, Samsung Electronics, Korea, where he was involved in developing world first multistandard multiband mobile digital TV tuner ICs supporting DVB-H/T, T-DMB, and ISDB-T/H. In 2007, he joined the Department of Electrical and Electronics Engineering, Pusan National University, Busan, Korea, where he is currently a Professor. From 2013 to 2014, he was an advisory professor with communication solution team, Samsung Electronics, Korea, where he was involved in the design of the 60 GHz WiGig circuits. His research interests include CMOS RF/mmWave/analog integrated circuits and RF security system for wireless communications such as IoT and 5G mobile systems.