Introduction
Millimeter-wave applications utilizing high frequencies above 100 GHz show promise in various fields, including high-speed communications, precision detection radars, and non-destructive imaging devices [1]. In particular, millimeter-wave imaging systems can distinguish objects through internal detection and extract dielectric characteristics that are beyond the recognition capabilities of conventional cameras. Many studies have focused on applications such as foreign body detection in food, security imaging systems, micro-crack detection, and drug discrimination [1–4]. The difficulty of implementing a high-power transmitter and high-sensitivity receiver poses limitations on applications utilizing millimeter-wave imaging systems, in which only the magnitude of the millimeter-wave signal transmitted through the target objects can reveal internal information [5, 6]. A phase shifter is a crucial component in a millimeter-wave transmitter that leverages polarization characteristics to achieve resolution beyond the wavelength limit and accurate image detection based on differences in dielectric properties.
A phase shifter based on vector modulators is advantageous for generating a wide range of phase shifts while minimizing millimeter-wave signal losses attributed to the use of passive devices [7]. The variation of the output impedance in the conventional Gilbert-cell structure can be minimized by maintaining a consistent total current consumption through the in-phase (I) and quadrature (Q) signal paths [8]. However, the magnitudes of the I and Q signals in the vector modulator-based phase shifter can fluctuate with wide phase variations at the output. The output impedance can be significantly altered by changes in the current within the phase shifter. Particularly, parasitic components in transistors, which change depending on the phase control voltage, produce large impedance variations in the output of the millimeter-wave band. However, the variation in output impedance resulting from phase change hinders the optimization of the power amplifier performances. This is due to the fixed input impedance characteristics of the amplifier, which are essential for generating high output power in the millimeter-wave transmitter.
In this study, a CMOS phase shifter based on a vector modulator configuration is proposed to achieve output impedance variation insensitivity, which is essential in implementing a high-power transmitter in the millimeter-wave band. The D-band phase shifter, implemented in a 65-nm CMOS process, generates a vector output signal with a power combiner after converting differential I and Q signals into two I/Q single-ended signals by using separate transformers. Section II presents the circuit design of the proposed phase shifter including the output impedance matching network for insensitive variation of the control biasing. Section III provides both small-signal and large-signal measurements and a discussion of the proposed phase shifter and compares its performance with previous studies. The conclusion is presented in Section IV.
Circuit Design
1. Vector Modulator-based Phase Shifter
Fig. 1 shows the overall schematic of the proposed vector modulator-based phase shifter. Although a double Gilbert-cell configuration offers higher phase resolution, the proposed phase shifter employs a structure similar to a single Gilbert-cell configuration for implementation in millimeter-wave band circuits. This choice reduces the number of passive devices at the input, simplifies layout design, and minimizes performance degradation caused by additional parasitic components [9]. The operation of the proposed phase shifter differs from that of the Gilbert-cell configuration, as it utilizes arbitrary control voltages to generate I and Q signals, thereby achieving various phases at the output.
Low amplitude and phase imbalances are required in the generation of quadrature signals in the phase shifter. The RC polyphase filter faces significant losses at millimeter-wave frequencies exceeding 100 GHz. A conventional hybrid coupler necessitates quarter-wavelength transmission lines, thereby contributing to an increase in the circuit’s overall size. The quadrature signals in the proposed phase shifter were generated using a Lange coupler to minimize losses and signal imbalances within a confined area, thereby optimizing the overall performance [10]. The electromagnetic (EM) simulation results of the Lange coupler using the top metal (Metal 9) and the next top metal (Metal 8) in the backend oxide layers of the CMOS process are shown in Fig. 2. The insertion loss across the D-band was simulated to be between −4.2 and −3.4 dB, and the magnitude and phase imbalances are less than 0.43 dB and 3° in the simulation, respectively.
All the transistors in the proposed phase shifter were designed with a gate size of 0.06 μm in length and 40 μm in width. The input impedance at the transconductance (Gm) transistor was implemented with capacitors C1, and C2 and a transmission line TL1. The transmission line TL2 was used between the Gm and switch transistors for gain enhancement through inductive feedback [11]. The body floating effect was implemented in the switch transistors by connecting the body node to the ground through a 5 kΩ resistor. The body floating technique, introduced to mitigate leakages through the body, involves the use of a substantial resistor [12]. Connecting a sufficiently large resistor between the body and the ground causes the nonlinearity due to parasitic components to diminish through the body floating effect, increasing the P1dB of the proposed phase shifter.
2. Output Matching Network Insensitive by Phase Control
The output impedance matching network, which exhibits a small impedance variation with the phase control voltage, consists of two transformers and one power combiner, as shown in Fig. 3. The positive or negative I/Q signals are selected by the phase control biasing and transmitted to the transformer. Both input ports of the transformer are always connected to the drain nodes of the off-state and on-state switch transistors, depending on the magnitude of the applied voltages. This operation prevents the current direction from changing rapidly in the proposed matching network, which helps to keep the impedance changes in the transformer constant. Compared to the conventional structure, where the I/Q signals are first power-coupled and transmitted to the transformer for matching, the proposed matching network minimizes the impedance mismatch condition. This is because there is no scenario in which the current direction supplied to the input terminals of the transformer becomes a common mode, depending on the phase control voltages. The output signal with the desired phase change is generated by combining the I/Q signals, which have passed through the transformer, at the power combiner.
Fig. 4 shows the output impedance variation as a function of phase control voltages, with the minimum, maximum, and intermediate states represented. Considering the input impedance characteristics of the power amplifier, which is the next stage of the millimeter-wave transmitter, it is designed to have an inductive impedance characteristic for complex conjugate matching. The impedance variation with the phase-controlled voltage was simulated to be within 18.9%, confirming the achievement of a phase-controlled insensitive output matching network.
3. Simulation Results of the Proposed Phase Shifter
By tuning the bias voltages applied to the Gm and switch transistors, the simulated S21 of the proposed phase shifter at 130 GHz is depicted as a polar plot in Fig. 5. Both the Gm and switch transistors were biased 0.4 V above Vth in the on state and 0 V in the off state. As the nonlinearity of the gain increases with the bias voltage in the Gm transistor, the voltage ranges for the Gm and switch transistors were set to 0.4–0.8 V with a 50-mV step and 0.4–1.2 V with a 100-mV step, respectively. The simulation results of the proposed phase shifter demonstrate a phase shift range exceeding 180°, facilitating the necessary polarization variation for millimeter-wave imaging systems.
Measurement Results and Discussion
The proposed phase shifter was fabricated on an area of 0.4 mm × 0.46 mm, inclusive of all pads, employing the 65-nm CMOS process with 1-poly and 9-metals in the back-end oxide layers, as shown in Fig. 6. The core area of the proposed phase shifter was implemented in an area of 185 μm × 285 μm.
The performance of the proposed phase shifter was evaluated through small-signal measurements using a network analyzer to obtain gains, input/output reflection coefficients, and phase shifts based on control biasing.
A part of the measurement data shown in Fig. 5 illustrates the gain and phase shift characteristics of the proposed circuit. The range of the phase shift was measured to be over 180°, but the gain and phase shift were not uniformly distributed depending on the bias voltage control of the transistors, as precise control by the digital-to-analog converters (DACs) was not included. Unlike in the simulation, achieving a phase shift between 90° and 250° was difficult because the phase shift characteristics due to the bias of the M5 transistor showed a large error in the gain and phase shift. These characteristics can be seen in detail in Fig. 7, which represents the small-signal measurement results as a function of frequency. The small-signal measurements were performed with on-wafer probing using a vector network analyzer (Keysight N5222B) and a frequency extension module (VDI WR8.0 VNAX). The results show good input and output impedance matching achieved with |S11| and |S22| below −7 dB in the frequency band of 115–135 GHz. In particular, the output impedance variation according to the phase control signals was measured to be within 2 dB, as depicted in Fig. 7(a), demonstrating the insensitive output impedance matching characteristic controlled by the signals. The power gains of the proposed phase shifter were measured in the range from −11.0 dB to −30.0 dB, exhibiting a significant variation depending on the phase control. This variation may result from the imbalance of signals going to each switch transistor. In addition, a DAC, typically employed for calibrating magnitude and phase variations in a phase shifter, was not integrated into the proposed phase shifter. This absence is considered to have contributed to the larger gain deviation. The phase transition of the phase shifter shows that a phase shift of more than 180° was achievable, consistent with the simulation. Significant nonlinearity in the phase transition was observed at the phase control voltage, confirming a small power gain. The root-mean-square (RMS) gain errors, indicating the deviation from the average gain depending on the bias control, and the RMS phase errors, indicating the deviation from the ideal phase shift characteristics, are plotted in Fig. 8 [9, 13].
Large-signal measurements were conducted using a signal generator and analyzer to determine the maximum saturation output power at the center frequency of the phase shifter. The frequency of 125 GHz, corresponding to the measured bandwidth, was generated using a signal source (Keysight N5183B) and a frequency multiplier (VDI WR6.5 SGX-M) for large-signal measurements. As the input power to the phase shifter was constrained by the equipment performance, the input-referred P1dB, shown in Fig. 9, was determined to be approximately over 7 dBm in the proposed phase shifter using a power meter (VDI PM5).
A performance comparison with previously published phase shifters in similar frequency bands is presented in Table 1 [9, 10, 14]. Despite the low operating frequency imposed by fabrication process limitations, the proposed phase shifter demonstrates the lowest variation in the magnitude of S22 with phase control voltage and the highest input-referred P1dB. This demonstrates that the design successfully meets the specified requirements.
Conclusion
A vector modulator-based phase shifter is proposed for a millimeter-wave transmitter to mitigate the impact of output impedance changes caused by phase control signals. The proposed D-band phase shifter, fabricated in a 65-nm bulk CMOS process, exhibited output return loss variation within 2 dB as a function of phase control voltages and an input-referred P1dB of more than 7 dBm at 125 GHz. Compared to previously published studies implemented with advanced and expensive semiconductor processes, the output impedance stability and high P1dB, despite generating lower output magnitudes, indicate that the proposed phase shifters can be utilized in polarization-variable and beamforming image detection systems within the transmitter of millimeter-wave imaging systems.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
Print
