Abstract
The precision of local oscillator (LO) signal in in-phase and quadrature (IQ) demodulation strongly affects the imaging performance of millimeter wave (mmWave) radars. Therefore, to eliminate the requirement for high-precision LO, a simple yet effective digital IQ demodulation method has been proposed with the aid of a specified sampling scheme in order to eliminate the demand for LO. Based on the bandpass sampling theorem, the characteristic of the intermediate frequency signal of mmWave imaging indicates that the LO is unrequired if the sampling rate is twice of the frequency of the carrier of the intermediate signal. In this way, the in-phase signal would be directly and accurately obtained by performing the Binary-Phase-Shift-Keying (BPSK) modulation on the samples, based on which the IQ demodulation would be completed by using the Hilbert transform. The proposed method does not employ LO and thus simplifies the demodulation process and is suitable for implementation in a Field-Programmable Gate Array (FPGA) with fewer hardware resources. To verify the method, a three-dimensional mmWave radar imaging is carried out at the 30-34 GHz bandwidth, where the sampling and digital IQ demodulation are realized by an ADC (AD9250) and FPGA (XC7K325T), respectively. The results show a simplified transceiver with lower requirements and the prospect of the proposed method being applied in radar imaging and other related fields.
1. Introduction
The millimeter wave (mmWave) technology, with its high resolution and the advantage of penetrability in dealing with insulating materials like paper, clothing, and fog, is now an important topic in radar imaging [1]. It has been applied in various fields, such as medical diagnostics [2, 3], security screening [4–6], military fields [7, 8], and nondestructive testing [9–12]. A vital component of these mmWave imaging systems is the in-phase and quadrature (IQ) demodulation which its performance can affect the imaging quality.
The conventional IQ demodulation in both analog and digital methods requires a local oscillator (LO) signal, whose accuracy and precision strongly affect the IQ demodulation performance [13]. However, the LO signal is inevitably deteriorated by some unfavourable drawbacks in the existing system. In the case of analog demodulation, both the inconsistency of analog devices and the variation of parameters with temperature fluctuation cause the phase imbalance of LO signal and highly limit the demodulation performance [14]. Moreover, the addition of individual analog devices (e.g., mixer and filter) further increases the power consumption and cost of the system. In the case of digital demodulation, the popular way to generate the LO signal is using the numerically controlled oscillator based on direct digital synthesis (DDS) [15, 16]. Currently, a look-up table-based DDS has the problems of relatively large signal spurious and low output frequency range. In order to solve the above problems, several methods have been proposed to achieve a high accuracy in real-time. For instance, [17] presents a direct digital frequency synthesizer based on a Multipartite Table Method. The authors in [18] propose a novel direct digital frequency synthesis architecture based on the differential Coordinate Rotation Digital Computer algorithm. These methods improve the speed or accuracy of DDS to a certain degree. However, with the increasing precision requirement of LO, the DDS-based demodulation requires a large number of hardware resources and is hard to be implemented in FPGA [19]. Therefore, it is necessary to design a LO-free method for IQ demodulation, to reduce the hardware resources and simplify the design of the receiver.
In this paper, LO-free digital IQ demodulation for mmWave imaging has been proposed. Based on the bandpass sampling theorem, the analysis of mmWave echo reveals that setting the sampling rate at twice of the frequency of the carrier of the intermediate frequency (IF) echo, the sampled data reflect the characteristics of in-phase (I) signal. To this end, the proposed digital IQ demodulation has been implemented in two steps. Firstly, the I signal is obtained by directly performing Binary-Phase-Shift-Keying (BPSK) modulation on the sampled data. Secondly, the Hilbert transform is applied on the obtained I signal to obtain the completed IQ demodulated data. The proposed IQ demodulation method eliminates the need for mixers against the traditional methods. Furthermore, this method has simplified digital demodulation and thus is easy to be implemented in FPGA. To verify the method, a three-dimensional (3D) mmWave radar imaging system operating at the 30-34 GHz bandwidth is set up. A 14-bit @ 200MSPS ADC chip (AD9250)-based data acquisition board is designed to sample the IF signal, where the IF carrier frequency is 100 MHz, and an FPGA (XC7K325T) is employed to control the system and achieve the proposed BPSK modulation-based LO-free IQ digital demodulation. The results demonstrate that the system realizes high-precision imaging and that the proposed method, with reduced hardware resources and simplified transceivers, has strong potential in the imaging field.
The structure of this paper is as follows: Section 2 explains the proposed demodulation method and the 3D mmWave radar imaging system. Section 3 presents the results of the experiment and performance analysis. Section 4 draws the conclusions of the paper.
2. Principles and System
2.1. BPSK Modulation-Based LO-Free IQ Digital Demodulation
The schematic diagrams of the traditional IQ demodulation in analog and digital are shown, respectively, in Figures 1 and 2. Both of these methods require an LO signal. In the analog demodulation method, the limitations of the device and the external environment make the LO signal hard to be synchronized with the ideal IF carrier signal and thus affect the demodulation accuracy [20]. While in the traditional digital demodulation, the digital LO signal is indispensable, generating a high-precision LO signal takes up a lot of FPGA resources [21]. In order to alleviate the pressure on FPGA hardware resources, the phase truncation method has been proposed to introduce additional noise and make the spurious suppression capability worse [21]. In short, a high-performance LO signal increases the system complexity and cost but plays an important role in the traditional IQ demodulation methods.
Aiming at LO-free IQ demodulation and simplifying the receiver, the characteristic of the IF signal is analysed, and a new BPSK modulation-based IQ digital demodulation is proposed in this paper. As shown in Equation (1), the IF signal of mmWave radar imaging can be expressed in this form [22]. where is the IF carrier’s frequency, is the scattering echo amplitude, and the is the target-echo phase information. According to bandpass sampling theorem, the sampling rate must meet.
If the sampling rate is twice the frequency of the carrier of the IF signal, i.e., it can be observed that the IQ demodulation would be simplified by incorporating Equation (3) into Equation (1). Therefore, the IF signal can be expressed as: where indicates the sampled data, is the in-phase component, is the quadrature component, and , 1, 2, ….. Bringing the result of into Equation (4), can be represented as:
Equation (5) indicates that I signal demodulation could be easily obtained by performing the BPSK modulation on the sampled data, where BPSK keeps the phase of the even samples and reverses the sign of the odd samples. Then, the IQ demodulation would be completed by applying the Hilbert transform on the obtained I signal, signal is written as:
From the above, the proposed BPSK modulation-based LO-free IQ digital demodulation is achieved. The detailed diagram of the method is given in Figure 3. The proposed method in this paper eliminates the requirement for high-precision LO and also greatly reduces the consumption of hardware resources.
2.2. Proof-of-Principle System
To verify the feasibility of this scheme, a 3D mmWave radar imaging system is set up. The block diagram and the imaging system are, respectively, shown in Figures 4 and 5. The system consists of a mmWave transceiver, an antenna array, and an FPGA-based data acquisition board. The mmWave transceiver transmits a sweep signal to the target through the array antenna, and the echo signal is downconverted to an IF signal, which contains information of the target. The entire system is controlled by the FPGA-based data acquisition board, which samples and processes the IF signal. At last, the processed data is transmitted to a personal computer (PC) for target imaging.
2.2.1. Antenna Array and Signal Parameters
The antenna array includes 64 receive antennas and 64 transmit antennas in this 3D mmWave radar imaging system (as shown in Figure 5). Each transmit antenna radiates linear frequency modulation continuous wave signal. The synthesized aperture imaging is achieved by electrical scanning in horizontal direction and mechanical motion in vertical direction. As shown in Figure 6, this effective electrical scanning technique for synthesized aperture imaging has been incorporated into the system described in [23–26]. This system is configured to operate from 30 to 34 GHz, and the parameters of the mmWave transceiver in this system are shown in Table 1.
The electrical scanning in horizontal antenna array provides focusing only in horizontal direction, and a mechanical scan in the vertical direction is required for obtaining the focusing ability in the vertical direction. In this imaging system, the vertical mechanical scan step is half of the wavelength. Then, the popular range-migration algorithm is used to obtain high-resolution imaging [27].
2.2.2. FPGA-Based Data Acquisition Board
The FPGA-based data acquisition board has a reconfigurable logic in FPGA (XC7K325T), a ADC chip (AD9250), and several I/O interfaces, allowing users to flexibly control the system. The implementation of the proposed demodulation method is shown in Figure 7. (1)Trigger manager module: control of the array antenna and the mmWave transceiver(2)ADC sampling module: sample the IF signal(3)IQ demodulation module: perform digital demodulation processing on the IF signal(4)Ethernet module: transfer the IQ data to the PC through an Ethernet protocol realized in the FPGA
3. Experiments
3.1. Simulation Verification and Resource Consumption
The parameter setting is provided in Table 1, and six targets have been chosen with their IF locations at 105, 110, 115, 120, 125, and 130 MHz. The spectrum after BPSK processing and Hilbert transform are, respectively, shown in Figures 8(a) and 8(b). Through this figure, it can be seen that the proposed method can accurately move the IF signal to its baseband and achieve IQ demodulation without LO.
(a)
(b)
Figure 9 shows the FPGA resources consumed by the proposed method under different phase accumulator word lengths. With the increase of the word length of the phase accumulator, the slice LUTs and slice registers of DDS demodulation method increase linearly as shown Figures 9(a) and 9(b). Compared with DDS demodulation, the number of slice LUTs and slice registers in the proposed demodulation has been reduced by approximately 20 and 70%, respectively. As shown in Figure 9(c), the RAM increases exponentially in the DDS demodulation, but it is not consumed in the proposed demodulation. Compared with the DDS demodulation, the proposed demodulation only uses 70% of the DSP block, as seen from Figure 9(d). In the DDS demodulation, mixing will lead to the increase of data width, and the use of data truncation, for achieving fixed-point data processing, will cause spectrum spurious [28]. However, the proposed demodulation only needs symbol transformation and is not affected by the word length of phase accumulator. It does not produce the spectrum spurious caused by data bit fixed-point processing. At the same time, the required hardware resource is also reduced significantly.
(a)
(b)
(c)
(d)
3.2. Imaging Verification and Analysis
In this section, the proposed method was applied to the actual mmWave radar for 3D imaging, and an experimental target was imaged for verification. In the imaging experiment, the IQ demodulation part can run well under a 200 MHz system clock.
As shown in Figure 10(a), complex steel plate, divided into “A,” “B,” “C,” and “D” parts and partially inserted into a foam, is used to test the imaging performance. The length/width of the A, B, and C areas is, respectively, 120/24, 80/16, and 40/8 mm. The width of area “D” is shorter than 4 mm which is beyond the system-resolution limitation as the mmWave transceiver has a frequency bandwidth of 30-34 GHz.
(a)
(b)
The target is located at 10 cm from the antenna plane in the experiment (as shown in Figure 10(b)). The IF frequency of the target can be calculated as 0.3 MHz; however, a frequency offset would be added, due to the nonignorable line-loss delay of the system. The frequency offset is tested as 13.6 MHz. Therefore, the output frequency of the IF signal would be .
The I signal spectrum obtained after sampling and BPSK processing of the signal is shown in Figure 11, in which the “A” signal is a direct-current (DC) signal, and the “B” signal is a target signal. It can be seen from Figure 11 that after BPSK modulation, the IF carrier signal with a frequency of 100 MHz can be removed accurately to complete I channel demodulation. Therefore, the target signal frequency is 13.9 MHz. Then, the Hilbert transform is used to obtain the IQ signal and IQ complex signal spectrum as shown in Figure 12. The image rejection ratio [29] of BPSK-based IQ demodulation reaches 55 dB, which can meet the requirements of the actual millimeter wave radar imaging system [30].
Figure 13 is the imaging result, which indicates that parts “A,” “B,” and “C” are imaged clearly (the image of “D” area was blurred due to its size beyond the system-resolution limitation). The results are consistent with the theoretical situation, where the theoretical resolutions of azimuth dimension, height dimension, and range dimension are, respectively, 4.9, 5.2, and 3.75 mm. Furthermore, the bottom part of the target inserted into the foam can be imaged well and shows the good penetrability of mmWave.
4. Conclusions
Future mmWave imaging radar receivers require faster processing speed and higher accuracy. However, the existing digital IQ demodulation methods rely on the LO signal which is typically generated using the DDS, resulting in problems of relatively large signal spurious and low output frequency range, thereby restricting the radar speed or accuracy. This paper has presented a simple yet effective method to achieve a digital IQ demodulation in mmWave radar system which eliminates the LO signal and simplified the design of radar transceivers. Using a 3D mmWave radar imaging system, the proposed method has been validated with a good performance. The findings indicate that the proposed method has strong application potentials in the radar and imaging fields.
Data Availability
The test data, simulation data, and the proposed method used to support the findings of this study are available from the corresponding author upon request. The algorithm proposed in this paper does not involve any patent problems.
Conflicts of Interest
The authors declare no conflict of interest.
Acknowledgments
This work was supported in part by the National Key R&D Program of China (2018YFF01013003), in part by the National Natural Science Foundation of China (61731020 and 12105177), in part by the 111 Project (D18014), in part by the International Joint Lab Program supported by Science and Technology Commission of Shanghai Municipality (17590750300), and in part by the Key Project Supported by Science and Technology Commission of Shanghai Municipality (YDZX20193100004960).