Recent advancement in the design of mixers for software‐defined radios

The implementation of software‐defined radio necessitates a versatile radio frequency (RF) front‐end that can support multiple standards in different frequency bands. This article presents the state‐of‐the‐art multiband and wideband mixers in the context of different receiver architectures suitable for software‐defined radio (SDR). Wideband and multiband mixer designs reported in the literature are categorized on the basis of their circuit architecture. This study tabulates the results of mixer designs with an emphasis on noise figure (NF), conversion gain (CG), linearity, and image rejection ratio. The best parameters reported in the comparison are 68–84 dB of CG, 3.2 dB of NF, 2.5 mW of power consumption, 12.5 dBm of third‐order input intercept point, and 0.015 mm2 of area at different frequencies. The study also discusses the design challenges and techniques to overcome those issues. The article concludes with the summary and prospective developments for SDR architectures.


| INTRODUCTION
Wireless communications have been around since the early 1990s, which subsequently brought up many standards and application areas such as in civil, military, and medical industries. With technological advancement, the number of wireless standards has been increasing. 1 Consequently, it is desired to design a radio front-end that can support multiple wireless standards. [2][3][4][5] This was difficult or even impossible for the traditional radios because of their hardware dependency. On the contrary, softwaredefined radio (SDR)s have the ability to reconfigure their operation. [6][7][8][9][10][11] In a conventional SDR receiver, after analog-to-digital conversion (ADC), most signal processing was carried out within the digital domain through field-programmable gate arrays (FPGAs). 12,13 This approach maximally enabled the software-controlled feature of SDRs. However, it led to high power consumption, which formed the bottleneck that prevented the wide adoption of SDRs. Additionally, some hardware components (mixers, amplifiers, and antenna) are also required that are not completely replaceable by their software counterparts.
An alternative approach to implement SDR is to use discrete-time and mixed-signal systems where the mixing process occurs before ADC. Besides, a power-scalable ADC is utilized, which is capable of providing tunable resolution and bandwidth. This approach was proven to be successful as they provide good reconfigurability with lowpower consumption. 10 Mixers are used to alter the frequency of the electromagnetic signal while maintaining all other characteristics of the original signal such as phase or amplitude. The main purpose of frequency conversion is to enable amplification of the received signal at a frequency other than radio frequency (RF). 14,15 As mixers are the important blocks of the receiver circuit, the overall performance of the receiver gets affected by the mixer's performance. Therefore, for good SDR receiver performance, a mixer is required to have low noise figure (NF), high conversion gain (CG), good linearity, and high image rejection ratio (IRR), while consuming small die/chip areas.
This survey aims to provide a timely summary of the recent development of mixers for SDRs. Section 2 discusses the mixer fundamentals, followed by the design topologies in Section 3. Section 4 focuses on the mixer specifications, followed by the wideband and multiband mixers for the SDR receivers in Section 5. Section 6 discusses the impedance matching techniques used in multiband mixers for SDR and Section 7 provides a comparative study of different mixers. Section 8 discusses the design challenges, and finally, Section 9 concludes the survey with future developments.

| MIXER FUNDAMENTALS
A mixer is one of the key components in a transceiver, and it can shift the signal from one frequency to another. Figure 1 illustrates the operation of a mixer. Upon mixing, the frequency of the output signal is in the form of a sum or difference of those of the input signals due to the presence of two input signals. In transmitters, mixers can perform frequency upconversion to shift intermediate frequency (IF) to RF. Similarly, in receivers, mixers can perform frequency downconversion to shift RF to IF. 17 For good performance with low NF, high linearity, polarity switching through local oscillator (LO) input is required. Such a mixer will have RF signal divided into in-phase and out-of-phase parts; the conversion switch operated by LO signal can alternatively choose in-phase and out-of-phase signals. Ideally, mixers will introduce the minimum amount of noise and have good linearity. 18 Moreover, they should be independent of LO amplitude and intermodulation products. However, practical mixers have the following limitations: non-negligible NF, limited CG, and linearity. 19 For the sake of convenience, we focus on conversion mixers in this survey.

| MIXER DESIGN TOPOLOGIES
Mixers are broadly classified as passive mixers and active mixers as shown in Figure 2. For passive mixers, they will introduce signal attenuation. The mixing is achieved through passive switches. Therefore, the switches are turned on and off depending on the LO signal, which is compared to a reference voltage and mixing is achieved through the multiplication operation of RF and LO frequency signals (in terms of square wave or sinewave). Passive mixers are widely used because of their simplicity, zero power consumption, high third order input intercept point (IIP3), and good NF at the expense of port isolation. However, their main drawback is the high LO power requirement. 17 Different from passive mixers, active mixers can provide high CG, good port isolation, low NF, and low LO (A) power requirements. However, it is difficult for them to achieve good linearity without the use of additional circuitry.

| Passive mixers
Passive mixers can be implemented using nonlinear diodes or field-effect transistors (FETs).

| Passive diode mixers
The mixers adopt nonlinear diodes and filtering elements. As shown in Figure 3, they can be subdivided into single-ended, single-balanced, and double-balanced mixers, respectively. Figure 3A, single ended mixers do not require DC power supply and are easy to implement. However, they do require filtering elements to improve the port isolation. 17 2. Single-balanced mixers employ two devices, which are often implemented as two single-device mixers linked by 180 or 90 couplers. If a 90 coupler is used within the mixer circuitry, matching networks must be employed, but port isolation will be reduced. The 180 couplers, on the other hand, provides good port isolation but matching will be poor. Based on this concept, a single-balanced mixer circuit is proposed as shown in Figure 3B. 20 The circuitry contains a 90 hybrid coupler, two Schottky diodes, matching circuitry, and a bandpass filter. The LO input signal is divided within two Schottky diodes. Similarly, the RF input signal is equally divided at the Schottky diode with a phase difference. After mixing of RF and LO signals within these diodes, the outputs obtained are combined at the IF. Thus, to filter out the desired frequency component at the output, the bandpass filters are used. Balanced mixers can provide improved port isolation compared to single-ended mixers. However, they require larger LO input power. 3. As shown in Figure 3C, double-balanced mixers contain differential inputs and outputs. The switching operation of mixers depend on the LO signal behavior (positive and negative cycles). During this operation, the lefthand and right-hand pair diodes alternatively switch on and off and the nodes "a" and "c" refer to the virtual ground for RF. The points "b" and "d" refer to the balanced RF signal, which connects to other points alternatively. Such mixers can cover a wideband and attain high linearity at the expense of elevated NF. 21

| Passive FET mixers
As shown in Figure 4, passive FET mixers employ FETs as switches. In passive FET mixers, drain-source resistance acts like a voltage-controlled resistor, where the channel resistance is a function of gate-source voltage. During positive half-cycles, two FETs M 1 , M 3 are enabled and M 2 , M 4 are disabled. Thus, IF balun secondary winding gets connected to that of RF balun through switched on FETs. The reverse operation occurs during the Passive FET quad mixer. 22 FET, field-effect transistor negative half cycle and the reverse polarity RF input reaches the IF output. The frequency of FET switching operation is determined by the LO signal frequency. 22 Passive FET mixers outperform diode mixers in terms of linearity with similar CGs. To conclude the section, a comparative summary of passive mixers is provided in Table 1. Table 1 shows that for high linearity performance, a double balanced passive diode mixer or passive FET mixer can be used. Likewise, single-ended mixers can be used for simple architecture, and for wideband operation double balanced passive mixers can be used. 17

| Active mixers
Active mixers can be classified as single device active mixers and balanced active mixers.

| Single device active mixers
Single-device mixers are further subdivided into single gate and dual gate mixers, as shown in Figure 5A,B, respectively.
1. Single gate mixers use a single transistor to take RF and LO inputs through the gate. Therefore, diplexing is needed to distinguish both inputs. 17 2. Dual gate mixers use dual-gated transistors to take RF and LO signals through separate gate terminals, resulting in good port isolation and IIP3. 17

| Balanced mixers
Balanced mixers are further grouped into single-balanced and double-balanced (Gilbert) mixers, as shown in Figure 6A,B, respectively.
1. Single-balanced mixers adopt single-ended RF input and differential LO inputs, leading to a differential IF output signal as shown in Figure 6A. These mixers attain good CG, low NF, and high port isolation. Moreover, they consume less power due to small area as compared to double-balanced mixers. 23 2. Gilbert mixers adopt differential inputs and output as shown in Figure 6B. These mixers combine two single-balanced mixers to achieve better performance in terms of IIP3, port isolation, and CG. 17 RF stage transistors (operating in saturation mode) are used to convert RF input voltage to current and hence amplify it. The transistors (operating in triode mode) connected to the LO stage are used for switching operation. The LO signal voltage must be higher than the RF signal voltage for proper operation.

| PERFORMANCE METRICS OF MIXERS
Mixers play an important role in wireless communication systems, especially in superheterodyne receivers. [24][25][26][27][28][29][30][31][32][33] Therefore, to characterize mixer performance, metrics should be selected carefully. But, there exists a trade-off among different performance parameters like CG, NF, IIP3, and power consumption. High CG mixers experience high NF, whereas high IIP3 can be achieved by sacrificing CG and power consumption. 23 Port isolation can be accomplished using mixer balance and hybrid junctions. Thus, it is important to maximize the isolation as the undesired signal dissipates RF power and may affect the required output signal. It may also cause electromagnetic interference. 34 Image rejection must be considered while designing any mixer as the unwanted signal present at the same input port may generate the false output signal. This will lead to wasted power consumption. 35 Impedance matching refers to the extent of matching between source and load impedance. Improper matching may lead to reflection and filters are required to avoid distortion at the receiving end. 36 To measure the performance of a mixer, vector network analyzers (VNAs) with calibration advancements can be used, including scalar mixer calibration (SMC) and vector mixer calibration (VMC). To perform SMC, the frequency offset introduced by mixer must be taken into account. For this purpose, VNA will provide input IF or RF that will be upconverted or downconverted to the desired frequency at the mixer output. As the output frequency at VNA Port 2 differs from the input frequency at Port 1, Port 2 must be set to frequency-offset mode. SMC uses conventional Short-Open-Load-Thru (SOLT) calibration, with the condition that the calibration must account for all frequencies of interest. SOLT calibration Active balanced mixers measurements can be done over both supplied frequency ranges when the frequency-offset mode is enabled. Although this approach is time consuming, it provides accurate results. Another approach is to perform SOLT measurement by disabling frequency-offset mode. This approach quickly generates the calibration results. However, it ends up in interpolation when actual measurements are taken. In addition to the mixer-under-test, VMC also requires another calibration mixer and filter. VMC calibration depends on the frequency conversion process. Although SMC setup is simpler compared to VMC, VMC calibration is important in phase and group delay measurement. 38 With the technology advancement, SDR mixers emerged to handle multiple standards with different frequencies in a concurrent or discrete manner. 39 Table 2 summarizes the specifications of these mixers as per SDR applications. SDR mixers should satisfy these specifications for all supported bands 40 without affecting circuit size, power consumption compared to narrowband mixer designs, making them suitable for mobile applications.

| DESIGN TECHNIQUES
As shown in Figure 7, SDR mixers can be broadly categorized as wideband and multiband mixers. Wideband mixers face the difficulty to attenuate interfering signals (especially for harmonics). 42 Therefore, multiband mixers are needed for the SDR. One of the easiest solutions for multiband is to use several mixers in parallel, each operating at different bands. However, this approach leads to complex mixer designs with high power consumption and large circuit size. Thus, the preferred solution is to design a single mixer, which is robust enough to operate at different standards without significantly affecting the performance.

| Wideband mixers
Wideband mixers demonstrate flat gain response across a wideband for different applications. Additionally, they require tuning circuits for frequency reconfiguration. Some of the common wideband mixer topologies include cross-coupled common gate, 37 inductive peaking, 41,43,44 bulk cross-coupling, and current bleeding 45 and current reuse with the current mirror approach. 2 Figure 8 shows cross-coupled common gate transconductance stage with a wide LO-based Gilbert mixer. It can achieve flat CG with good matching and low-power consumption. However, it suffers from high NF and low IIP3. Alternatively, an inductive peaking technique can be used. Conventional approaches use an inductor with cross-coupled PMOS transistors (CCPT), diode connected NMOS transistors (DC-NT) dependent IF load to enhance the load impedance, while maintaining high transconductance and hence CG. The inductive peaking will balance the CG drop at higher frequencies while improving the bandwidth. 43,46 Figure 9 shows the use of an inductive peaking topology at the transconductance stage to provide wideband impedance matching. 41 The technique is employed to resonate with the parasitic capacitors leading to a large and flat gain. In the design, the load stage consists of large resistances followed by the transistors acting like diode loads to provide a virtual AC ground at the gates of the transistors, leading to high voltage headroom affecting the overall NF and IIP3 performance. Yang et al. utilized a resistive double-balanced structure that attained good IIP3 and CG as shown in Figure 10. 44 In the design, on-chip baluns are developed to meet single to differential transformation requirements. Besides, inductors are placed in the RF signal path between the balun and mixer core for optimizing input matching. Likewise, for obtaining high and wideband gain, the optimum gate and source inductive feedback technique is employed.

| Active wideband mixers
Gilbert topology is well known for providing high CG, good port isolation. [49][50][51][52][53] Adiseno et al. proposed a Gilbert mixer with flat CG, but the design suffers from high NF and low IIP3. 54 Although the design is suitable for compact, low IF receivers, the overall power consumption is high.
To maintain high CG and high IIP3 simultaneously, an active and resistive combined topology can be used, which is convenient for direct receiver architectures with low-power consumption. 55 Solati et al. employed derivative superposition and noise cancelation approach for the mixer circuitry. 56 For linearity enhancement purposes, auxiliary transistors have been used. Additionally, a novel linear path with multiple gated transistor topology has been introduced for the thermal noise cancelation of the input transistors. The design achieved not only high CG and high IIP3, but also low NF.
Active Another design challenge is to simultaneously maintain large bandwidth and high CG. Blaakmeer et al. proposed a blixer circuitry, which is basically introduced as a combination of low noise amplifier (LNA), Balun, and I/ Q mixer. 57 The proposed design is compact with lowpower consumption. Hamed et al. adopted a novel passive balun for a mixer that uses a multi-dielectric layered structure for a broadband operation within coplanar structure. 58 The design attained high return loss, while maintaining high CG and bandwidth as well.

| Passive wideband mixers
Passive mixers can be used for low voltage operation as they can provide high IIP3 because of linear channel resistance. [59][60][61][62] Murphy et al. proposed a noise-canceling receiver with an eight-phase passive mixer. 63 With the employment of passive mixers and high-gain baseband operational amplifiers, the virtual ground behaves ideally and also suppresses voltage swing prior to baseband filtering. As with the rise in mixer phases, the noise folded from high-order harmonics was suppressed, leading to better NF. In addition, in the presence of fixed transimpedance amplifier outputs, desired signal I/Q components can be attained and the undesired signal components are canceled. Namsoo et al. proposed a resistively degenerated mixer, which achieved low NF, high IIP3, and low-power consumption. 64

| Dual topology mixer
Active mixers are good for attaining high CG and passive mixers can attain high IIP3. Therefore, to attain high performance in CG and IIP3 simultaneously, one solution is to combine active and passive mixers in one design. 65 The design uses an active RF stage and passive core stage while utilizing the current mirror technique 2 without degrading the overall performance.
SDR mixers based on dual topology mixers require compact mixers that provide high performance and consume low power while operating within a wideband. Existing mixers with on-chip inductors suffer from problems like large chip area, low-quality factor, and susceptible to electromagnetic interference within unified systems. For wideband SDR mixers, capacitive parasitics associated with CMOS transistors need extra attention because they affect the CG and NF performance, especially within complex designs. Although there is a possibility to tune them with on-chip inductors, off-chip inductors prove to be a better solution that maintains low area requirement without performance degradation. 66

| Multiband mixers
Multiband mixers can handle multiple bands simultaneously with continuous or discrete tuning. 67 They can be further classified as concurrent mixers and reconfigurable mixers.

| Concurrent mixers
Concurrent mixers can receive different frequency bands concurrently. However, to satisfy the miniaturization and low-cost needs, low-power consumption should be emphasized as most of the power consumption comes from mixers and amplifiers within the receiver circuit. Likewise, the design complexity should be reduced. Therefore, a single mixer with multiple operational bands will be beneficial. 68,69 Similar work proposed mixers operating in dual bands by utilizing capacitive tuning for multiband operation and attaining a high IRR. 69 Noise-canceling designs were proposed by utilizing current bleeding and transistor-based attenuation 30 to attain high CG and low NF. Likewise, the mixer switching approach in 70 can also be used for dual-band concurrent operation. Chen et al. proposed a concurrent dual-band LNA with a dual-band subharmonic mixer. 71 This approach eliminated the need for multiple frequency synthesizers for the receiver and hence reduced the overall manufacturing cost. The overall CG and NF were good at the expense of linearity.
Abdelrheem et al. presented a dual-band impedance matching network for concurrent narrowband matching at two frequencies and a resistive degeneration approach for performance improvement. 47 The design achieved a small chip area because of the utilization of off-chip F I G U R E 1 1 Dual-band mixer 47 passive components. As shown in Figure 11, Q 1 , Q 2 transistors form an RF stage and Q 3 -Q 6 form transistors of the core stage and Figure 12 shows the return loss which is >10 dB within the entire band. The design also attains high CG within dual bands. It was found that transconductance is large for the frequencies smaller than unity-gain frequency f T , which makes it easier for a mixer to operate at more than one frequency. Therefore, the obtained transconductance, G m can be defined by: where G m refers to the driving stage, g m is the smallsignal transconductance, and R d refers to the degeneration resistance, respectively. It is not necessary to use off-chip components to reduce the overall area of a mixer design. On-chip filters/ resonators/components can be used to provide high performance while reducing the overall cost of the circuitry. Wang et al. have proposed the design that adopted both on-chip baluns and LC series-parallel resonator. 72 The resonator circuitry helps to operate within dual bands. The proposed mixer consumed low power at the expense of degraded CG and NF performance. Nevertheless, it attained high port isolation.
Hsu et al. proposed a novel concurrent dual-band receiver employing dual-band LNA, stacked mixer, voltage-controlled oscillator, and variable gain amplifiers. 48 The design adopted the current-reuse approach. Thereafter, for power reduction purposes, it adopted a common gain stage and component stacking. The design attained high CG and low NF with low-power consumption. Figure 13 shows the concurrent dual-band receiver circuitry, where a parallel-series combination-matching network was proposed for concurrent matching. The complete design covered a moderate chip area, including matching networks and pads. Its return loss is >10 dB in both bands as shown in Figure 14. The design also attained good CG at the desired concurrent bands.

| Reconfigurable mixers
High-performance reconfigurable mixers are desired to support different bands of operation. The conventional way of designing reconfigurable mixers for multiband applications is to integrate mixers in parallel, each operating within a different band. However, they are limited to lower frequency bands. The reconfigurable mixers can be classified as switchable mixers and tunable mixers.
F I G U R E 1 3 Concurrent dual-band receiver 48 F I G U R E 1 2 Performance metrics of dual-band mixer 47 Switchable mixers Switchable mixers can operate within different bands using switches. The reconfiguration can be possible at different frequencies by toggling the switches, which are discrete in nature. The resultant frequency bands have the highest return loss. One of the switchable mixers is shown in Figure 15, which is based on the Gilbert cell approach. The parasitic capacitor of the switching stage (LO stage) is the main factor for the small dynamic range that is tuned with spiral inductors, programmable inductors developed using transformers, or small spirals at different frequencies. The switches presented in the filter section allow switching at different frequencies. However, to control the programmable inductors, NMOS varactors are also employed within the filter circuitry. The programmable inductor section is shown in Figure 15. The input impedance is inductive and the effective impedance can be expressed as L 1 , L 2 , and M are the primary, secondary, and mutual inductance, respectively. k is the magnetic coupling coefficient. With the variation in C 2 , the current ratio varies and hence the effective inductance. Generally, the programmable inductor is employed for tuning the parasitic capacitors to enhance the loading impedance to improve the noise performance and the second-order intermodulation distortion. The overall design covered a large chip area while attaining good performance in terms of CG, NF, and IIP3.

F I G U R E 1 4 Performance metrics of concurrent dual-band receiver 48
F I G U R E 1 5 Reconfigurable mixer 73 Similarly, Cao et al. proposed current driven passive mixer and switchable transconductance array to attain low flicker noise and good IIP3. 74 The buffer section is designed using the Tow-Thomas transimpedance biquad amplifier as a filter. The switchable amplifier is responsible for providing high wideband gain. The mixer section as shown in Figure 16 is designed using a passive mixer because of its ability to provide high linearity and low flicker noise. To further enhance the performance, threshold variation is taken into consideration with the help of the DC bias voltage controlling process for every transistor gate within the circuitry, resulting in high DC offset. Therefore, to overcome this issue, digital to analog converters are added at the IF end that helps in the DC offset calibration process.

Tunable mixers
Tunable mixers can be tuned at different frequencies within a band, continuously or discretely. In general, digitally tunable SDR receivers employ multiband mixers for broadband frequency range coverage. 75 Channelization and waveform detection is possible with different encoding schemes along with frequency hopping that makes use of digital signal processing algorithms and FPGAs. The work refers to digital receiver capability to perform ADC sampling and FPGA-based digital downconversion (DDC) process. 76 A frequency-tuning algorithm was discussed by Ibrahim et al. 77 The proposed algorithm in the work utilized a platform consisting of universal software radio peripheral and RFX2400 daughterboard. This FPGAbased DDC approach allows lossless signal processing. For example, FPGA devices are commonly used for digital signal processing and communication with the advancement in SDR technology. This is mainly because of their reconfiguration capability. 78,79 For SDR receivers, RF front-ends operate in a wideband. During the selective filtering process, out-of-band blockers (strong carriers beyond the digital receiving band) and harmonic interference are amplified as the desired signal, which degrades the SNR. Some SDR receivers rely on the voltage-mode passive mixer to provide filtering for the suppression of out-of-band blockers, but they are not effective to suppress harmonic interference. Fortunately, this can be done with harmonic rejection mixers. However, the gain and phase mismatch leads to the degradation of the harmonic rejection ratio (HRR). Therefore, the gain and phase must be calibrated if high HRR is required, although the calibration is a tedious task.
Different techniques can be used to improve the HRR at the expense of CG and phase mismatching. Harmonic rejection approaches consume high power due to complicated circuit and large die area. 80 For a high IRR, good CG, and low-power consumption, the mixer proposed by Zhang et al. can be chosen. 81 It featured a zero/low IF variable SDR structure that uses digital calibration for both narrowband and wideband purposes. Three modes were considered. For low NF performance, a currentmode mixer was used. Additionally, for high linearity performance, a voltage passive mixer was utilized specifically for out-of-band rejection. Then, the harmonic rejection passage was used for vector gain calibration. However, it only supports a limited number of bands with degraded CG, NF, and linearity. 81 As such, continuous tuning is preferred over discrete tuning, as the former is less sensitive to process variations.

| Active multiband mixers
Several approaches have been discussed in the literature for attaining high CG and NF performance. Different mixers were developed and proposed by Douss et al., based on current bleeding and charge injection approaches. 3 However, they may lead to high power consumption and degraded CG. 65 Additionally, the mixer was proposed by Fan et al. that used Tow-Thomas topology as a filter to achieve high performance without degrading CG, NF, and IIP3. 28 The mixer proposed by Chen et al. is a good example to adopt the current reused G m stage, source degeneration, and multi-gated transistor approaches. 59 The design attained high IIP3 at the expense of CG and NF. Similarly, Qin et al. proposed a mixer with an improved complimented common gate pair as the RF stage. 82 The proposed design also used a transformer for chip area reduction and PMOS with inductive peaking better for current bleeding. This improved the NF, CG, and bandwidth performance at the expense of IIP3.
Other than the above-mentioned approaches, the use of RLC tunable resonators 16 helps in attaining high CG and low NF, as shown in Figure 17. The resonance frequency is known as The current driven passive mixer 74 The design achieved low NF and good CG as shown in Figure 18. Nevertheless, there is always a trade-off among CG, NF, and IIP3. Yang et al. showed that passive mixers dependent on power amplifier degeneration can be used to lower the NF without additional switching requirements. 83 Although the proposed design was fully reconfigurable, the power consumption was quite high. Thus, this design is not suitable for low-power SDR applications. Likewise, Sowlati et al. proposed a transceiver to include a passive mixer while maintaining low phase noise and better port isolation. 84 The design also covers low area and attains high CG. Both inductor-based and inductorless designs can cover large chip area. An inductorless I/Q mixer was proposed by Poobuapheun et al., which employs an operational amplifier-based output section with a passive core stage. 85 The proposed mixer attained high CG, high IIP3, and good NF at the expense of chip area.

| Dual-mode multiband mixers
Passive mixers are highly linear with low-power consumption. However, they suffer from high conversion loss and need a large LO power. On the contrary, active mixers can provide better CG and require small LO power. However, they suffer from low linearity and high NF. 86 To the best of our knowledge, the mixer proposed by Nguyen et al. is the first dual-mode mixer designed using an active balun coupled ring mixer that can alternatively work as an active and passive mixer. 87 In their design, the CG control was possible by varying the control voltage. The design of active balun was using a DC-coupled differential pair approach to eliminate off-chip IF balun. In high CG mode, the design attained moderate IIP3, while, in low CG mode the linearity performance is better.

| IMPEDANCE MATCHING TECHNIQUES
Conventional mixer circuits were developed with off-chip surface mount components, making it flexible for designers to use different impedance matching techniques, like stubs or quarter-wave transformers. Additionally, the characterization can be done first before the actual implementation. Moreover, off-chip components have higher quality factors than their counterparts. However, they are not widely used because it is difficult to integrate them in circuits. Modern integrated circuit technology makes it easier for designers to use on-chip components. Nevertheless, these components are limited in size and band of operation. 88 Lumped element, transformer-based, or microstrip line matching are impedance matching techniques using on-chip components. For different applications, they have varied performance. 28,89,90

| Lumped elements matching
Matching using lumped elements is widely used in any multiband mixer for SDR. Frequency tuning is possible by varying the resonator capacitance employing F I G U R E 1 7 Tunable Gilbert mixer 16 F I G U R E 1 8 Performance metrics of tunable mixer 16 electrostatic, piezoelectric, or thermal micro-actuators using the electrostatic technique. 91 Wang et al. proposed a matching network for a mixer operating in a high band to compensate for the effect of the parasitic capacitances seen at its IF output. 72 Likewise, peaking inductor was employed for the low band mixer to extend the IF port 3-dB bandwidth. The technique allows concurrent matching within 6-18 GHz bands. Similar work was discussed by Hashemi et al. that allows concurrent matching at 2.45 and 5.25 GHz, respectively. 92 The parallel network was proposed to resonate with both frequencies of interest. To achieve concurrent multiband gain, the drain load network should have high impedance only at frequencies of interest. This can be fulfilled by connecting a series LC branch to parallel LC tank. Each series LC branch inserts a zero into the design's gain transfer function at its series resonant frequency, which sets the frequency of the transfer function notches. This notch is intended to improve the receiver's image rejection. Likewise, Kia et al. proposed a tunable differential active inductor to tune with the parasitic capacitances. 93 The reconfiguration is possible by using a variable feedback resistor employed within the inductor. A similar approach was also discussed by Tseng et al., where lumped element-based couplers are employed to obtain different inputs. 94 The overall design consumed low power and required small die area. The design also achieved good port isolation.

| Transformer-based matching
Transformer matching can be used in multiband frontends, among which multitap transformers proved as a better solution compared to other transformer-based approaches. Although the design is quite flexible, it may degrade the quality factor. 95 Yu et al. employed a transformer is at the output end for differential to single-ended signal conversion for wideband matching. 96 The design also employs coupling capacitors for amplitude and phase matching of the transformer. Likewise, Hermann et al. employed an on-chip resonating transformer to attain high current gain. 97 The proposed design featured a small die area and achieved good port isolation. Additionally, Tiebout et al. discussed a single-balanced transformer-based mixer topology and maintained good port isolation with small die area. 98

| Microstrip line matching
Microstrip line-based matching is another potential approach for reconfigurable multiband mixers. Wu et al.
used a T-shaped transmission lines. 99 To enhance the bandwidth, a modified Marchand balun is proposed. The proposed design covered a small area and better port isolation. However, it has limited application scenarios. Likewise, Lin et al. used a thin film microstrip line for matching networks within the circuitry and the design attained high port isolation. 100 Seyedhosseinzadeh et al. used series transmission lines to form a -shaped network in both RF and LO stages of the mixer for wideband operation. 101 The design covered a small area and achieved high CG. Table 3 summarizes the different mixer designs related to SDRs. Each mixer topology operates within a wideband. From Table 3, most of the works have been implemented in 180 nm CMOS technology. High performance was achieved by using switching 109,116 and multi-gated transistor technique. 107 In Table 3, the highest NF is as less as 1.5 dB and the maximum IIP3 is 12.5 dBm. Besides, reconfigurable mixers are more linear than concurrent mixers. As CG increases, the overall power dissipation increases, which degrades the IIP3 performance. The overall mixer circuit area depends on the design complexity.

| DESIGN CHALLENGES AND TECHNIQUES
Multiband mixer design is complex and requires a large number of passive components. Thus, it often leads to high power consumption and large chip area due to their size. 25,110 Improper mixer design may cause an imbalance among different performance metrics. 103,110,120 Therefore, different techniques have been proposed in the literature while considering the flexibility within the mixer and their wideband operation. 121 This section discusses some of the common challenges in designing such mixers and the corresponding techniques used to overcome these problems.

| Image rejection
Desired and image signals enter together at the input and degrade the performance of a mixer. Therefore, it is important to reject the image signals. Recently, it was found that RF sampling can be used as downconverters, where a discrete-time structure is more advantageous compared to the continuous-time structure. However, samplers in wideband SDRs will affect their performance. For example, the charge sampling approach shows the inverse CG and frequency relation. Although this issue was resolved with voltage samplers, wideband noise folding occurs with this sampling process. Therefore, prefiltering is required in both cases. 122 However, this approach not only is difficult to implement, but also increases the overall cost of the receiver as well as the design complexity. 123 Hence, the best way is to design an image rejected mixer that will reduce the overall receiver complexity, while enhancing its reliability and performance. [124][125][126][127] Gong et al. proposed mage rejection mixers, which is a combination of a single-balanced mixer and the orthogonal bridge. 125 The design has attained an acceptable IRR, but the conversion loss is increased.

| Conversion gain-linearity balance
Modern transceivers are developed using CMOS process technologies due to their ability to provide low-cost design and low-power consumption. 128

| Power consumption
The recent development of mobile communication devices has witnessed an ever-increasing level of circuit integration. This advancement would have been incomprehensible without a device downscaling process that further leads to performance improvement. The latest research focuses on the system-on-chip. The CMOS technology faces challenges in achieving continued downscaling of power consumption while satisfying the development trend of software-defined radios for modern wireless communications. 132 It has been found that CMOS wideband mixers consume low power. 133,134 However, it degrades other performance metrics. 60,115,129,135 Moreover, CCPT, inductive peaking, and current bleeding approaches were used for high CG, low-power utilization, but these approaches also degraded the linearity. 136 Likewise, Bazrafshan et al. showed that the use of cascode common source approach is useful for low-power consumption. 10 Similar works were discussed by researchers that consumed low power but at the expense of other performance metrics. 10

| Noise performance
Several techniques are available to reduce the NF of mixers. One of the most common approaches is to adopt filters. For example, Hu et al. proposed a wideband tunable Gilbert mixer using an appropriate filter that attained a moderate NF with low-power consumption at the expense of overall die area. 16 Cordova et al. used a dynamic current injection approach along with the filters attaining a moderate NF. 39 Likewise, Gladson et al. proposed a mixer that operates in two modes, that is, high gain mode and high linearity mode for low rate wireless personal area network applications. 66 The proposed design in high gain mode attained high CG, low IIP3, and moderate NF. On the contrary, its high linearity mode attained high IIP3,

| SUMMARY AND PROSPECTIVE DEVELOPMENT
Through the detailed literature survey, we notice that techniques may vary depending on the applications and requirements. Indeed, it is quite challenging to balance the trade-offs among different parameters when designing mixers for SDRs. Common mixer design techniques are current bleeding, 120,139-141 active balun, 87 current mirror, 142 gain boosting, 129 current reuse, 135 and current injection, 102 double linearizations. 143 In general, the selection of a particular technique depends on the performance optimization goal for the mixer design as every technique has its benefits, however, there are some associated drawbacks as well. Great care must be taken while opting for a particular technique. This information together with the aforementioned literature study provides ample opportunity for further investigations in the design of mixers. This section not only summarizes the survey but also discusses the prospects of mixer design based on the design problems as summarized in Table 4. Based on this survey, we take the liberty to provide some general advice for implementing multiband or wideband mixers as follows: 1. Wideband or multiband mixers can be used as a part of SDR receiver circuitry to enclose the complete desired frequency band compared to narrowband mixers to overcome the need of several receivers capable of covering different frequency bands. Fully balanced mixer structures can be used to eliminate evenorder distortion within the receiver circuitry. 2. Impedance matching plays an important role for simultaneous reconfiguration while proposing reconfigurable multiband mixers. Lumped element-based matching is suitable for low-frequency bands, and microstrip line-based matching is required for highfrequency bands, which may increase the overall chip area and power consumption. 3. Most works in the literature with low-power consumption and chip area operate within single frequency bands. However, it is difficult for multiband mixers to achieve low-power operation as they require complex matching circuits, which unavoidably increases the power consumption and circuit area of the design. Additionally, extra care is required for the parasitics handling as they might affect the input impedance, in turn affecting the overall CG and NF performance.