Mengjie Qin, Zhongmao Li, Pengzhan Liu, Lanlan Liao, Zhiqiang Li, Xin Qiu
©SHUTTERSTOCK.COM/BENSON HE
Nowadays, with the increasing performance requirements of wireless applications, the RF front end needs enough dynamic range to work properly in multiband. The traditional multichannel parallel structure provides outstanding performance, but repeated components consume a large board area and add complexity to the circuit [1]. To solve this problem, tunable architectures have been proposed [2], [3]. At the same time, wireless telecommunication equipment is pushing toward low-cost system-on-chip technology, which integrates the RF front-end module with the back end onto a single chip.
As a key frequency-selective unit of the RF front end, the bandpass filter (BPF) appears at multiple locations in a transceiver with different effects. For example, Figure 1 shows a simplified superheterodyne transceiver architecture. In most cases, a BPF following the antenna is used for frequency band preselection, requiring high frequency selectivity. A BPF before a mixer focuses on image rejection to avoid image frequencies entering the mixer. An intermediate frequency BPF after the mixer is used to filter out-of-band (OOB) clutter and to reduce the requirements for the dynamic range of subsequent modules. A BPF behind the power amplifier is used for harmonic rejection. In addition, as transceiver architectures continue to develop, so too have the requirements for BPFs, such as tunability, reconfigurability, harmonic suppression, and multifunctionality. However, an adjustable BPF meeting specific system requirements is still the most interesting part to be integrated.
Figure 1. The simplified architecture of a superheterodyne transceiver. LNA: low-noise amplifier; VGA: video graphics adapter; LO: local oscillator; LPF: low-pass filter; ADC: analog-to-digital converter; PA: power amplifier; DAC: digital-to-analog converter; IQ: in-phase quadrature.
Several adjustable filter structures in monolithic microwave integrated circuits (MMICs) have been proposed, such as the LC BPF, RC BPF, transconductance-capacitor (gm-C) BPF [4], [5], [6], and N-path BPF [7]. Most of them realize the adjustment by varying filter characteristics, including center frequency, bandwidth (BW), transmission zero (TZ), and quality (Q) factor, using tuning capacitance, inductance, and gm. Each structure has its unique advantages.
This article reviews many excellent works that have been reported, which excel in on-chip tunable BPFs (TBPFs). Three types of adjustable BPFs will be discussed: BPFs with adjustable frequency, BPFs with adjustable BW, and BPFs with adjustable TZs. Each type of filter is categorized according to its tuning method into capacitance tuning, inductance tuning, and switched-based tuning. Finally, the performance of each filter is compared comprehensively, including factors such as tuning range, BW, insertion loss, and power consumption. The development trends and limitations of the on-chip TBPF are summarized, which provide suggestions for the design of on-chip adjustable BPFs.
BPFs with adjustable center frequency are among the most common tunable filters. In this section, the available research on such filters is divided into three types according to tuning method: capacitance, inductance, and switch based.
In addition, the Q-enhanced technique is proposed to optimize the poor frequency response caused by frequency tuning [8]. There are two main technical routes to improve the Q factor. The first is to replace the spiral inductor with an active inductor (AI) to reduce loss. The other is to employ loss compensation techniques (Q-boosting networks), including negative-resistance techniques and transformer-based Q-enhanced techniques [9], [10], [11], [12]. The purpose of negative-resistance technology is to establish a negative-resistance network to offset the series loss resistance and to improve the effective Q factor of on-chip spiral inductors. Cross-coupled CMOS and bipolar transistor pairs are commonly used to exhibit first-order negative impedance at the drain and collector terminals, respectively. Q tuning is mentioned next, but it won’t be classified separately.
Capacitance tuning filters primarily include LC, RC, and gm-C filters. The main components for tuning capacitance are the varactor diode and the switched capacitor. The varactor diode tunes capacitance through a bias voltage, which simplifies the circuit. However, this limits the tuning range. Switched capacitors usually appear in designs as a capacitor array, consuming more area but allowing a wider frequency tuning range.
For microwave frequencies, a varactor-tuned LC filter is a popular choice in TBPF circuit designs, although its limited capacitance tuning range is still a problem.
In 2016, a 5-/60-GHz CMOS receiver with a transformer-coupled Q-enhanced tunable channel-select active filter was proposed in [13]. The partial diagram of the circuit for the 5-GHz receiver path is shown in Figure 2. Two varactors, Ct1 and Ct2 (VT control), tune frequencies from 5.15 to 5.65 GHz. The Q factor is compensated by tuning the tail current (Vq1 control) to get a good frequency response. The results of the tuning frequency and Q are shown in Figure 3. By adjusting the frequency and the Q-value at the same time, the filter maintains a good frequency response in different operating frequencies, as shown in Figure 4. In 2017, a similar receiver architecture using a negative resistance to enhance Q achieved a wider tuning range [14].
Figure 2. A partial diagram of the circuit for the 5-GHz mode in [13].
Figure 3. A measurement of (a) frequency tuning (variable VT with Vq1 = 0.85 V) and (b) Q tuning (variable Vq1 with VT = 0.3 V) of the circuit for the 5-GHz mode in [13].
Figure 4. The measured frequency tuning response of the LC active filter in [13]. ch-A: channel-A; ch-B: channel-B; ch-C: channel-C; ch-D: channel-D; IP1dB: input 1dB compression point; IIP3: input 3rd-order intercept point.
In 2017, Amin et al. [15] demonstrated a dual-band Q-enhanced differential LC BPF. The dual band is generated by two parallel BPFs tuned at different frequency ranges. The two BPFs have the same structure, which is shown in Figure 5. The tunable LC tank consists of Rc, L, and Cv. The loss of the LC tank is compensated by a negative-resistance network, which consists of Q5–8, Rx, and Ry. These elements form a differential degenerated Darlington pair to improve the linearity of the negative resistance, thereby achieving a larger frequency tuning range. Ry is used to linearize the cross-coupled pair and to reduce transistor noise. The measured frequency tuning and Q tuning are shown in Figure 6. The center frequencies in the X and Ku bands can be tuned from 9.7 to 13.9 GHz. The Q of each BPF can be tuned from 20 to 70 independently.
Figure 5. A schematic of the differential Q-enhanced LC BPF with varactor tuning in [15].
Figure 6. (a) The measured frequency tuning response and (b) Q tuning response of the dual bands of the BPF in Figure 5 [15].
Switched-capacitor-based TBPFs have wider capacitance tuning ranges than varactors. In [16], a switched-capacitor-based tunable filter using a GaAs pseudomorphic high-electron mobility transistor (pHEMT) process was demonstrated. The filter consists of series capacitors, parallel resonators, and tunable components. Unlike a distributed structure (which consumes a relatively large area) and a lumped structure (which can cause parasitic effects), the reported filter is designed with a semilumped topology. Figure 7 shows a diagram of the structure. C1–C4 are used to shorten the length of the main transmission. The gate voltage (VC1–VC3) of the field-effect transistor (FET) reconfigures the filter with different center frequencies by controlling the FET state. The presented filter is reconfigurable at 8.8 GHz and 10 GHz. The fractional bandwidth is 13.35% and 24.94%, respectively.
Figure 7. The structure of the switched-capacitor-based TBPF in [16]. GND: ground; MIM: metal–insulator–metal.
In [17], a passive LC filter is demonstrated, which consists of a third-order pi-section structure to meet the OOB attenuation requirement. The filter provides a 300–882-MHz operating frequency range using metal–insulator–metal capacitors and NMOS switches. In [18], a 0.8-/2.4-GHz tunable active-RC BPF is reported. An elliptic filter structure is adopted to realize a large OOB rejection. An operational amplifier offers enough gain to maintain a high Q factor. The tunability of this BPF is achieved by resistor and capacitor banks.
Testi et al. [19] have fabricated a fourth-order tunable Q-enhanced BPF. Figure 8 shows the complete structure of the fourth-order BPF and each stage. As shown in Figure 8(b), an array of differential pairs on the left is used to provide an overall gain in the 72-dB range. The switched-capacitor array on the right is used to tune the center frequency. In addition, the negative-resistance network (a cross-coupled pair, M6 and M7) compensates for the loss of the LC tank during frequency tuning. Figure 9 shows the frequency tuning results. The filter covers a frequency response from 2.35 to 2.48 GHz, and the frequency step is 0.4 MHz.
Figure 8. A schematic of (a) the fourth-order LC TBPF and (b) each of the TBPF stages of the BPF in [19]. Freq: frequency; CTRL: control; VB: voltage bias.
Figure 9. The measured Q-enhanced LC TBPF frequency response of the BPF in [19]. Freq: frequency.
Although spiral on-chip inductors offer a low level of noise and superior linearity, it is not a good choice to use passive inductors at high frequency, because the intrinsic physical properties lead to a large substrate area and a low Q factor. To solve this problem, some passive methods have been proposed, including patterned ground shields and geometry improvements [20]. However, the Q factor is still limited. Unlike spiral inductors, AIs can be demonstrated with less area at high frequencies, without adding any special processing steps.
AIs are active gyrator-C networks that primarily consist of metal–oxide semiconductor (MOS) transistors. When the dc-biasing condition and signal swing reach certain values, these networks present an inductive characteristic. As either the input or the output impedance of a gyrator-C network is finite, AIs are no longer lossless and the networks are inductive only in a specific frequency range.
A gyrator-C network is shown in Figure 10. Go1 and Go2 are the total transconductances at nodes 1 and 2, respectively. This topology can be modeled as a parallel RL tank circuit. The input admittance of this network is given in (1). The elements of the RL tank networks are ${\text{R}}_{\text{p}} = \left({{1} / {{\text{G}}_{\text{o}2}}}\right)$, ${\text{G}}_{\text{p}} = {\text{C}}_{2}$, ${\text{R}}_{\text{s}} = \left({{{\text{G}}_{\text{o}1}} / {{\text{G}}_{\text{m}1}{\text{G}}_{\text{m}2}}}\right)$, and ${\text{L}} = \left({{{\text{C}}_{1}} / {{\text{G}}_{\text{m}1}{\text{G}}_{\text{m}2}}}\right)$. The resonant frequency is given in (2). \begin{align*}{Y} & = \frac{{\text{I}}_{\text{in}}}{{\text{V}}_{2}} \\ & = {\text{sC}}_{2} + {\text{G}}_{\text{o}2} + \frac{1}{{\text{s}}\left({\frac{{\text{C}}_{1}}{{\text{G}}_{\text{m}1}{\text{G}}_{\text{m}2}}}\right) + \frac{{\text{G}}_{\text{o}1}}{{\text{G}}_{\text{m}1}{\text{G}}_{\text{m}2}}} \tag{1} \end{align*} \[{\text{w}}_{\text{o}} = \frac{1}{{\text{LC}}_{\text{p}}} = \sqrt{\frac{{\text{G}}_{\text{m}1}{\text{G}}_{\text{m}2}}{{\text{C}}_{1}{\text{C}}_{2}}} = \sqrt{{\text{w}}_{\text{t}1}{\text{w}}_{\text{t}2}} \tag{2} \]
Figure 10. A gyrator-C network and equivalent model.
where \[{w}_{t1,2} = \frac{{G}_{m1,2}}{{C}_{1,2}}{.}\]
Based on this analysis, the inductance of AIs can be tuned by varying the transconductances of the transistors. The Q factor can be increased by increasing the output resistance of the transconductances. The tunability of AIs is widely exploited in the design of tunable filters.
TBPFs that employ AIs have been widely studied. The most common topology of AIs in a TBPF is a shunt inductor to ground, while an active series inductor is used in some broadband designs [21]. However, AIs have disadvantages, such as higher noise, larger power consumption, and nonlinear behavior, which can be addressed with an optimized topology in the TBPF design.
A differential LC resonator-based filter with high selectivity and low power consumption is proposed in [22]. There are two tuning methods for this design. The first is to introduce an MOS feedback resistor into the AI structure to vary the inductance. The second is to vary the bias current directly [23]. The resulting BPF obtained a tunable center frequency that varies from 145.2 to 436.2 MHz. The power consumption is between 2.34 and 3.45 mW. The 1-dB compression points are 1.77–1.82 and 1.23–2.67 dBm, respectively.
As mentioned earlier, most AIs are realized with the Si-based MOS process. However, in recent years, compound semiconductors have been increasingly used in MMIC design due to the advantages of high-electron mobility and wideband gap. In [24], an AI in GaN is demonstrated—the structure is shown in Figure 11(a)—to achieve a BPF with separate amplitude and Q tuning. Figure 11(b) shows the schematic of the TBPF. The measured frequency tuning range of the active filter is 749 MHz at a center frequency of 3.39 GHz. Figure 12 shows the measured frequency and amplitude tuning ability, respectively. In [25], a proposed TBPF in GaAs pHEMT technology shows high performance. The operating range is from 1,800 to 2,100 MHz.
Figure 11. Structure of the (a) designed AI and (b) TBPF in [24].
Figure 12. Measured (a) frequency tuning ability and (b) Q tuning ability of the circuit in [24].
Switch-based TBPFs are primarily N-path filters. An N-path filter is a popular alternative to an external surface acoustic wave filter in the front of receivers because of its good frequency selectivity, ease of integration, and wide frequency tuning range [26], [27]. It basically consists of a baseband low-pass filter (LPF) and some switches [28]. The center frequency of an N-path filter is controlled by nonoverlapping clocks applied to switches and can be easily tuned for different frequencies. Two common N-path filter forms are shown in Figure 13: one port and two port [29]. The one-port structure reduces the number of clocks, which results in lower power consumption and better noise performance. In addition, most N-path filters work below 3 GHz due to clocking limitations. A recent report of an N-path filter with frequencies up to 12 GHz was published in [30].
Figure 13. The structure of N-path filters. (a) A one port and (b) two port [29]. LO: local oscillator.
The design theory of N-path filters continues to improve, with the goal of improving the frequency selection performance and increasing the filter’s order [31], [32], [33], [34], [35]. For wideband tuning circuits, parasitic capacitance at an N-path filter input causes a passband response, offsetting the expected center frequency [34]. The authors of [33] evaluated an offset-tuning method that achieved a programmable 0.7–2.7-GHz receiver.
An N-path structure equivalent to a series LC coupling network is proposed in [32]. As is shown in Figure 14, quarter-wave transmission lines (T-lines) are added on either side to transform a shunt LC into a series LC. This structure completes the N-path equivalence theory. Moreover, T-lines also eliminate the influence between each stage of an N-path filter. Figure 15 shows the diagram of the proposed sixth-order filter. The quarter-wave T-lines are realized by lumped capacitor–inductor–capacitor T-type circuits. The measured 600–850-MHz tuning results of the filter are shown in Figure 16. The BPF has excellent filter shape and OOB rejection but increases in-band loss.
Figure 14. The conversion for parallel LC and series LC to N paths.
Figure 15. A diagram of the designed sixth-order N-path filter in [32]. CLK: clock.
Figure 16. Measured S21 and S11 of the sixth-order N-path filter in [32].
The filter at the front end of a receiver needs excellent OOB rejection to prevent OOB signal interference and self-interference. A traditional N-path filter has a large insertion loss and poor OOB rejection at high frequencies due to the presence of switch parasitic capacitance and switch resistance [36]. To broaden the bands of the filter response and to improve performance at high frequencies, a T-line N-path filter technology, which absorbed the parasitic capacitance into a T-line, is analyzed in [37]. Figure 17 shows the structure of the T-line N-path filter, and a comparison of S21 with a traditional N-path filter. Figure 18(a) shows a designed four-stage eight-path filter. The coupling between inductors creates a TZ that improves OOB rejection. The measured frequency response is shown in Figure 18(b). The center frequency tuning range is from 0.1 to 1.6 GHz. When the filter turns off, the in-band insertion loss generated by the T-line is roughly 0.5 dB and starts to increase after the cutoff frequency.
Figure 17. (a) The structure of a T-line N-path filter (K is the number of stages) and (b) comparison of the T-line N-path filter and a traditional N-path filter(K = 4) [37]. LO: local oscillator.
Figure 18. (a) A schematic of the four-stage eight-path filter with digital logic divider and (b) the measured response of the filter in [37]. LOs: local oscillators; clock gen: clock generator.
In [38], a fourth-order filter is demonstrated using voltage-in, voltage-out and voltage-in, current-out conversion to increase the filter roll-off. Figure 19 shows a simplified diagram of the proposed filter. The two-stage BPF exhibits different impedances with different frequency signals, which suppresses the OOB signal. In addition, unlike switches in traditional N-path filters using top-plate mixing, a bottom-plate mixing structure is used to improve linearity. Figure 20 shows the measured gain and S11 in three center frequencies. The fabricated filter achieves an OOB IIP3 of 44 dBm. The center frequency tuning range is 0.1–2 GHz.
Figure 19. A simplified architecture of the fourth-order N-path filter in [38]. V-V: voltage in, voltage out; V-I: voltage in, current out.
Figure 20. Measured gain and S11 for three center frequencies of the filter in [38].
In a frequency-division duplex system, a filter at the front of the receive signal path needs to suppress interference caused by the transmitted signal, which means self-interference. In [39], a 0.2–3.6-GHz cascading TBPF and bandstop filter (BSF) is proposed to solve this problem. Figure 21 shows the structure. A fully differential structure is used to reduce local oscillator (LO) leakage. The BPF and BSF can be tuned independently. Figure 22 shows the tuning results. The OOB rejection reaches 15 dB over the tunable frequency range.
Figure 21. A diagram of the differential N-path filter in [39]. Clk: clock; LO: local oscillator; Tx: transmitter; Rx: receiver.
Figure 22. Measured results of S11 and S21 of the filter in [39].
Harmonic foldback (HFB) is a drawback of N-path filters. During the clock-switching process, the filter shifts higher harmonics (NK ± 1) fc (where K = 1,2,3 … and fc is center frequency) into its passband, resulting in harmonic aliasing. This is a result of the circuit and sampling being a linear periodically time-varying system. In [40], a foldback cancellation technology is proposed. This method adds an additional N-path filter, which has the same clock frequency but different clock phases. The proposed BPF achieves HFB attenuation of 25 to 45 dB, at the expense of higher power consumption and larger area. The tuning frequency range is 0.25–1 GHz. A harmonic-free filter, which is achieved by adding additional paths, is illustrated in [41]. Unlike the aforementioned method, this filter reduces power consumption by filtering higher-order harmonics of the LO clock signal. The frequency tuning range of the fabricated filter is 0.1–5 GHz. The total power consumption is less than 8.5 mW. Its area is 0.2 mm2 and OOB IIP3 is +6 dBm.
In [42], a differential 16-path filter was fabricated in CMOS SOI, which realized a wide tuning range. A differential structure is used to reduce harmonic aliasing at the even harmonics of the center frequency. The proposed filter, shown in Figure 23, can be tuned from 0.25 to 2.25 GHz. The 3-dB BW is roughly 20 MHz. The measured results are shown in Figure 24. This BPF has a 0.9-dB insertion loss.
Figure 23 A schematic of the differential 16-path filter in [42]. Clk: clock.
Figure 24 Measured S21 and S11 of the 16-path filter in [42].
As shown in Figure 25, a 0.1–1.2-GHz tunable sixth-order BPF was fabricated [43]. Eight phases are chosen to reduce the HFB. A differential structure is chosen to combat common-mode disturbance. Moreover, a Miller compensation method is used to optimize the passband shape. Figure 26 shows the measured S21 in the 0.1–1.2-GHz tuning range and details of the bandpass response in some of the frequency bands. This filter has a 59-dB stopband rejection. The passband ripple is less than 0.6 dB.
Figure 25. The differential circuit of the sixth-order TBPF in [43].
Figure 26. Measured S21 in tuning range of the sixth-order TBPF in [43].
BW tuning provides more flexible filtering characteristics for a front-end filter. When the OOB blockage is close to the target signal, a narrow-band filter with high rejection performance is more suitable, but it will bring a large insertion loss. When the OOB blockage is far from the target signal, a broadband filter can be used to reduce the insertion loss. A BPF with adjustable BW can adaptively balance the insertion loss and OOB rejection performance according to different requirements. BW adjustment is more complicated than frequency adjustment. Most of the filters change BW by adjusting the coupling of the resonator. For multistage filters, the BW can be adjusted by adjusting the interstage coupling of the resonators.
In [44], a Q-enhanced, LC-based programmable four-stage filter is presented. Its center frequency can be tuned from 1.3 to 1.7 GHz (a 50-MHz tuning step), and the signal passband can be tuned from 25 to 1,000 MHz (a 50-MHz tuning step). The structure of the proposed Q-enhanced LC tank is illustrated in Figure 27. An LC tank circuit is in the middle of the structure. C1 and R1 are tunable to tune center frequency and the Q factor of the BPF. The Gm and Gmp array are 6-bit tunable gain and Q-boosting transconductance stages, respectively, to adapt to different frequency responses. Cascading four-filter stages provide flexibility for BW tuning. The buffers between four LC tanks are used to provide BW and frequency programming flexibility. Figure 28 shows the measured results of center frequency programmability and signal BW programmability, respectively. This filter achieves the widest fractional BW in this review.
Figure 27. A schematic of the Q-enhanced LC TBPF in [44]. CMFB: common-mode feedback.
Figure 28. Measured results of (a) center frequency (constant 350-MHz BW) and (b) BW (center frequency is 1.5 GHz) programmability of the filter in [44].
As shown in Figure 29, a Q-enhanced differential fourth-order LC filter is designed in [45]. Two second-order filters are in parallel to improve frequency selectivity. The LC tank circuit consists of an inductor and a switched-varactor array for frequency tuning. The measured results of the frequency tuning and BW tuning are shown in Figure 30. The fractional BW tuning range is 2–25%. The frequency tuning range is 4–8 GHz.
Figure 29. (a) Structure of a differential fourth-order filter. (b) Schematic detail of one path of the differential BPF and signal subtractor in [45].
Figure 30. Measured (a) frequency response with fractional BW tuning at 6 GHz and (b) frequency tuning with constant absolute BW of 200 MHz of the filter in [45].
In [46], an active eight-path filter is proposed and shown in Figure 31. The low-pass gm-C filter is implemented by a fully differential Miller-compensated Butterworth filter, which provides BW tuning. A 3-dB BW is tuned from 2.5 to 51 MHz by six digital bits. A set of resistors is controlled to change the value of gm. A capacitor array is controlled to realize another degree of freedom in BW tuning. Figure 32 shows the tuning results with different digital bit combinations. The center frequency is tuned from 100 MHz to 2 GHz by different clock frequencies.
Figure 31. Schematic of the eight-path filter in [46].
Figure 32. Tuning capability of the filter in [46] for different digital bit combinations.
Based on capacitively coupled parallel resonators, a BPF is realized with passive switched-capacitor resonators with high linearity in [47]. The fourth-order TBPF is shown in Figure 33. The center frequency and BW are controlled by the clock frequency and coupling capacitors, respectively. The coupling capacitors C1–3 are designed as a switched-capacitor array to tune the BW. Switches in switched-capacitor resonators are driven by nonoverlapping clocks.
Figure 33. A diagram of the fourth-order N-path BPF with switched-capacitor resonators in [47]. CLK: clock.
In [48], a passive N-path BPF with variable TZs realized high selectivity and wide BW. Figure 34 shows the schematic of the designed BPF. Unlike traditional N-path filters, spiral inductors are added in the filter’s series circuits to decouple the resonators from each other, avoiding unnecessary charge sharing and improving frequency selectivity. All the capacitors are realized by a switched-capacitor array, which adjusts the BW. Figure 35 shows the measured filter response while the BW is 30, 40, and 50 MHz. The transition band roll-off slope is larger than 100 dB/100 MHz.
Figure 34. Schematic of the differential N-path TBPF in [48].
Figure 35. Measured filter response (BW is 30, 40, and 50 MHz) of the filter in [48].
An adaptive N-path filter that achieved tunable center frequency, tunable BW, and reconfigured filter in-band shape is reported in [49] and shown in Figure 36. A differential structure is employed to suppress common-mode disturbance. Each stage uses the same resonator. The first stage is decoupled to reduce noise and to increase the gain and the dynamic range of the filter. BW tuning primarily depends on adjusting resonators and external couplings. A switched-capacitor array Cb varies from 3.18 to 42.1 pF to adjust external couplings and to maintain a suitable Q at different frequencies. Figure 37 shows the 0.2–1.2-GHz tuning results of the filter for different in-band responses.
Figure 36. The full structure of the differential N-path filter in [49].
Figure 37. Measured response of the fabricated filter in [49] configured as (a) Butterworth type (BW is 20 MHz) and (b) Chebyshev type (BW is 40 MHz).
Creating TZs is a popular strategy for suppressing OOB interference. The filter can obtain sharp roll-off and proper stopband rejection by creating TZs in a transfer function [50], [51], [52]. The following BPFs with adjustable TZs are introduced in terms of capacitance tuning and inductance tuning.
As shown in Figure 38, an octagonal wideband BPF with a controllable TZ was demonstrated in 0.18-µm CMOS technology [53]. Based on specific metal layers provided by the CMOS process, two ring resonators of the BPF are routed in an octagonal interwinding shape to provide stronger coupling. In addition, the TZ is determined by two open stubs loaded at the input and output ports. An MOS varactor is used at one end of the open stub to tune the TZ. The simulated and measured S-parameters are shown in Figure 39. The TZ keeps moving by varying the bias voltage.
Figure 38. Structure of the BPF with a tunable TZ in CMOS technology in [53].
Figure 39. Simulated and measured S-parameters of the TBPF in [53].
A tunable blocker-tolerant filter with dual adaptive TZs close to the passband was reported in [54]. As shown in Figure 40, the filter consists of a TBPF and two TBSFs in parallel. Having the BPF and BSFs in parallel allows the filter to create adjustable TZs without loss of passband gain. Two BSFs allow more freedom to adjust each TZ. Figure 41 shows the measured results of TZ tuning while the center frequency is 1 GHz. Zeros on both sides can be tuned to roughly 20 MHz.
Figure 40. Schematic of the BPF with two tunable TZs in [54].
Figure 41. Measured performance of the BPF with TZs tuning in [54].
In 2018, the authors of [55] proposed a BPF with two tunable TZs based on a lumped L-C ring. The prototype of the employed stepped-impedance ring is shown in Figure 42(a). It is well known that the ring resonator has two TZs above and below the passband, which can improve stopband rejection. Figure 42(b) shows a lumped L-C model equivalent of the ring resonator. A cascode AI with feedback structure is employed to realize L1 and L2, as shown in Figure 43, which achieves low insertion loss and a small chip area. The inductance of the AI can be tuned by adjusting the bias current of AIs. Figure 44 shows the variable BW with TZ tuning. The measured insertion loss is 2.9 dB at 2.4-GHz center frequency, and the two TZs are located at 2.1 GHz and 3.1 GHz, respectively. A low-noise amplifier (LNA) is placed before the BPF to reduce noise. The noise factor of LNA+BPF is 6 dB.
Figure 42. (a) A prototype of the stepped-impedance ring filter and (b) the equivalent lumped L-C model in [55].
Figure 43. A complete schematic of the active LC TBPF in [55].
Figure 44. Measured S-parameters of the LC TBPF with TZ tuning in [55].
To summarize, Table 1 compares performance parameters of filters that follow different tuning methods. It can be seen from the table that research on on-chip tunable filters is primarily focused on frequency tuning using LC filters and N-path filters. For frequency tuning, the tuning method using a varactor diode requires fewer tuning components. However, the tuning range of a tunable filter based on a varactor diode is narrow, which is restricted by the tuning range of the varactor diode itself. Switched-capacitor-based tunable filters have a switched-capacitor array that enables wider-range frequency switching. Switch-based tunable filters mainly employ an N-path structure. For this type of filter, the structures of LPFs are diverse and include gm-C and RC. In addition, an N-path structure is affected by the switching speed, which lowers the operating frequency. However, the frequency tuning range becomes wider. Moreover, an N-path filter can be designed in a multistage form, and its design flexibility is higher than that of the LC filter.
Table 1. A comparison of tuning performance for different structures of on-chip adjustable RF BPFs.
It’s clear that active filters entail high power consumption and worse noise performance but with excellent frequency selectivity, while passive filters have lower noise. Moreover, some tradeoffs between parameters cannot be avoided, such as the tradeoff between insertion loss and frequency selectivity. Finally, the OOB interference faced by a filter at the front end of the receiving chain and the problem of HFB suppression faced by a filter connected after a linear RF component also require better technical solutions.
In recent years, on-chip tunable filters have been developed to be multitarget tunable. For example, frequency tuning, BW tuning, and in-band response reconfigurability can now be achieved in one filter. Most researchers prefer using multistage filters to add more flexibility to the design tuning, even though that requires more area. Furthermore, on-chip filters are mainly composed of lumped components, and their design flexibility is not as high as that of planar filters (such as printed circuit boards). Moreover, due to the cost of chip design, the performance of on-chip TBPFs is more in line with market demand, and the design experience at commercial communication frequencies is more abundant. On-chip TBPFs have great research value in terms of tuning theory and design flexibility.
[1] P. Bahramzy et al., “A tunable RF front-end with narrowband antennas for mobile devices,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 10, pp. 3300–3310, Oct. 2015, doi: 10.1109/TMTT.2015.2470237.
[2] H. Shao, G. Qi, P.-I. Mak, and R. P. Martins, “A low-power multiband blocker-tolerant receiver with a steep filtering slope using an N-path LNA with feedforward OB blocker cancellation and filtering-by-aliasing baseband amplifiers,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 69, no. 1, pp. 220–231, Jan. 2022, doi: 10.1109/TCSI.2021.3098810.
[3] D. Lee and K. Kwon, “CMOS channel-selection LNA with a feedforward N-path filter and calibrated blocker cancellation path for FEM-less cellular transceivers,” IEEE Trans. Microw. Theory Techn., vol. 70, no. 3, pp. 1810–1820, Mar. 2022, doi: 10.1109/TMTT.2022.3142140.
[4] H. Le Vu, H. T. Thi Luu, L. Dinh Tran, and H. V. Tran, “Implementation of CMOS tunable on-chip Gm-C IF filter in RF front-end IC for SDR transceiver,” in Proc. 7th Int. Conf. Integr. Circuits, Des., Verification (ICDV), 2017, pp. 46–51, doi: 10.1109/ICDV.2017.8188636.
[5] M. Darvishi, R. Van der Zee, E. Klumperink, and B. Nauta, “A 0.3-to-1.2GHz tunable 4th-order switched Gm-C bandpass filter with >55dB Ultimate rejection and out-of-band IIP3 of +29dBm,” in Proc. IEEE Int. Solid-State Circuits Conf., 2012, pp. 358–360, doi: 10.1109/ISSCC.2012.6177050.
[6] Z. Y. Chang, D. Haspeslagh, and J. Verfaillie, “A highly linear CMOS Gm-C bandpass filter with on-chip frequency tuning,” IEEE J. Solid-State Circuits, vol. 32, no. 3, pp. 388–397, Mar. 1997, doi: 10.1109/4.557637.
[7] M. N. Hasan, Q. J. Gu, and X. Liu, “Tunable blocker-tolerant RF front-end filter with dual adaptive notches for reconfigurable receivers,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), May 2016, pp. 1–4, doi: 10.1109/MWSYM.2016.7540332.
[8] W. B. Kuhn, N. K. Yanduru, and A. S. Wyszynski, “Q-enhanced LC bandpass filters for integrated wireless applications,” IEEE Trans. Microw. Theory Techn., vol. 46, no. 12, pp. 2577–2586, Dec. 1998, doi: 10.1109/22.739250.
[9] J. Kulyk and J. Haslett, “A monolithic CMOS 2368±30 MHz transformer based Q-enhanced series-C coupled resonator bandpass filter,” IEEE J. Solid-State Circuits, vol. 41, no. 2, pp. 362–374, Feb. 2006, doi: 10.1109/JSSC.2005.862348.
[10] B. Georgescu, H. Pekau, J. Haslett, and J. McRory, “Tunable coupled inductor Q-enhancement for parallel resonant LC tanks,” IEEE Trans. Circuits Syst. II. Analog Digit. Signal Process., vol. 50, no. 10, pp. 705–713, Oct. 2003, doi: 10.1109/TCSII.2003.818366.
[11] B. Georgescu, I. G. Finvers, and F. Ghannouchi, “2 GHz Q-enhanced active filter with low passband distortion and high dynamic range,” IEEE J. Solid-State Circuits, vol. 41, no. 9, pp. 2029–2039, Sep. 2006, doi: 10.1109/JSSC.2006.880618.
[12] T. Soorapanth and S. S. Wong, “A 0-dB IL 2140±30 MHz bandpass filter utilizing Q-enhanced spiral inductors in standard CMOS,” IEEE J. Solid-State Circuits, vol. 37, no. 5, pp. 579–586, May 2002, doi: 10.1109/4.997850.
[13] Y.-C. Hsiao, C. Meng, and S.-T. Yang, “5/60 GHz 0.18 μm CMOS dual-mode dual- conversion receiver using a tunable active filter for 5-GHz channel selection,” IEEE Microw. Wireless Compon. Lett., vol. 26, no. 11, pp. 951–953, Nov. 2016, doi: 10.1109/LMWC.2016.2615012.
[14] W. L. Chang, C. Meng, S.-D. Yang, and G.-W. Huang, “0.18 μm SiGe BiCMOS microwave/millimeter-wave dual-mode dual-conversion receiver architecture with a tunable RF channel selection at low-flicker-noise microwave mode,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), Jun. 2017, pp. 1778–1780, doi: 10.1109/MWSYM.2017.8058992.
[15] F. Amin, S. Raman, and K.-J. Koh, “An integrated microwave tunable dual-band Q-enhanced LC band-pass filter in 0.13 μm SiGe BiCMOS,” in Proc 47th Eur. Microw. Conf. (EuMC), 2017, pp. 33–36, doi: 10.23919/EuMC.2017.8230792.
[16] H.-R. Zhu, X.-Y. Ning, Z.-X. Huang, and X.-L. Wu, “An ultra-compact on-chip reconfigurable bandpass filter with semi-lumped topology by using GaAs pHEMT technology,” IEEE Access, vol. 8, pp. 31,606–31,613, Feb. 2020, doi: 10.1109/ACCESS.2020.2972932.
[17] D. Im, H. Kim, and K. Lee, “A broadband CMOS RF front-end for universal tuners supporting multi-standard terrestrial and cable broadcasts,” IEEE J. Solid-State Circuits, vol. 47, no. 2, pp. 392–406, Feb. 2012, doi: 10.1109/JSSC.2011.2168650.
[18] Z. Xu et al., “A 0.8/2.4 GHz tunable active band pass filter in InP/Si BiCMOS technology,” IEEE Microw. Wireless Compon. Lett., vol. 24, no. 1, pp. 47–49, Jan. 2014, doi: 10.1109/LMWC.2013.2287236.
[19] N. Testi, R. Berenguer, X. Zhang, S. Munoz, and Y. Xu, “A 2.4GHz 72dB-variable-gain 100dB-DR 7.8mW 4th-order tunable Q-enhanced LC band-pass filter,” in Proc. IEEE Radio Frequency Integr. Circuits Symp. (RFIC), 2015, pp. 87–90, doi: 10.1109/RFIC.2015.7337711.
[20] C. P. Yue and S. S. Wong, “On-chip spiral inductors with patterned ground shields for Si-based RF ICs,” IEEE J. Solid-State Circuits, vol. 33, no. 5, pp. 743–752, May 1998, doi: 10.1109/4.668989.
[21] F. Hu and K. Mouthaan, “L-band bandpass filter with high out-of-band rejection based on CMOS active series and shunt inductors,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), Jun. 2014, pp. 1–3, doi: 10.1109/MWSYM.2014.6848409.
[22] M. Mhiri, A. Ben Hammadi, F. Haddad, S. Saad, and K. Besbes, “VHF, UHF voltage-controlled bandpass filter employing tunable active inductor,” in Proc. 15th Int. Multi-Conf. Syst., Signals Devices (SSD), Mar. 2018, pp. 435–440, doi: 10.1109/SSD.2018.8570592.
[23] A. B. Hammadi, F. Haddad, M. Mhiri, S. Saad, and K. Besbes, “Compact tunable bandpass filter for RF and microwave applications,” in Proc. 18th Mediterranean Microw. Symp. (MMS), 2018, pp. 160–163, doi: 10.1109/MMS.2018.8611979.
[24] T. B. Herbert, J. S. Hyland, S. Abdullah, J. Wight, and R. E. Amaya, “An active bandpass filter for LTE/WLAN applications using robust active inductors in gallium nitride,” IEEE Trans. Circuits Syst., II, Exp. Briefs, vol. 68, no. 7, pp. 2252–2256, Jul. 2021, doi: 10.1109/TCSII.2021.3054739.
[25] L. Pantoli, V. Stornelli, G. Leuzzi, L. Hongjun, and H. Zhifu, “GaAs MMIC tunable active filter,” in Proc. Integr. Nonlinear Microw. Millimetre-Wave Circuits Workshop (INMMiC), Apr. 2017, pp. 1–3, doi: 10.1109/INMMIC.2017.7927304.
[26] H. Noori, R. Jiang, and F. F. Dai, “A 1.8 GHz–2.4 GHz SAW-less reconfigurable receiver frontend RFIC in 65nm CMOS RF SOI,” in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), 2018, pp. 1–4, doi: 10.1109/ISCAS.2018.8351847.
[27] A. Ghaffari, E. A. M. Klumperink, and B. Nauta, “Tunable N-path notch filters for blocker suppression: Modeling and verification,” IEEE J. Solid-State Circuits, vol. 48, no. 6, pp. 1370–1382, Jun. 2013, doi: 10.1109/JSSC.2013.2252521.
[28] L. E. Franks and I. W. Sandberg, “An alternative approach to the realization of network transfer functions: The N-path filter,” Bell Syst. Tech. J., vol. 39, no. 5, pp. 1321–1350, Sep. 1960, doi: 10.1002/j.1538-7305.1960.tb03962.x.
[29] N. Reiskarimian, J. Zhou, T.-H. Chuang, and H. Krishnaswamy, “Analysis and design of two-port N-path bandpass filters with embedded phase shifting,” IEEE Trans. Circuits Syst., II, Exp. Briefs, vol. 63, no. 8, pp. 728–732, Aug. 2016, doi: 10.1109/TCSII.2016.2530338.
[30] K. Kibaroglu and G. M. Rebeiz, “An N-path bandpass filter with a tuning range of 0.1–12 GHz and stopband rejection > 20 dB in 32 nm SOI CMOS,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), 2016, pp. 1–3, doi: 10.1109/MWSYM.2016.7539982.
[31] Z. Shang, Y. Zhao, and Y. Lian, “A low power frequency tunable FSK receiver based on the N-path filter,” IEEE Trans. Circuits Syst., II, Exp. Briefs, vol. 66, no. 10, pp. 1708–1712, Oct. 2019, doi: 10.1109/TCSII.2019.2931840.
[32] N. Reiskarimian and H. Krishnaswamy, “Design of all-passive higher-order CMOS N-path filters,” in Proc. IEEE Radio Frequency Integr. Circuits Symp. (RFIC), 2015, pp. 83–86, doi: 10.1109/RFIC.2015.7337710.
[33] K. B. Östman et al., “Analysis and design of N-path filter offset tuning in a 0.7–2.7-GHz receiver front-end,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 62, no. 1, pp. 234–243, Jan. 2015, doi: 10.1109/TCSI.2014.2358331.
[34] K. B. Östman et al., “A 2.5-GHz receiver front-end with Q-boosted post-LNA N-path filtering in 40-nm CMOS,” IEEE Trans. Microw. Theory Techn., vol. 62, no. 9, pp. 2071–2083, Sep. 2014, doi: 10.1109/TMTT.2014.2333714.
[35] A. Ghaffari, E. A. M. Klumperink, M. C. M. Soer, and B. Nauta, “Tunable high-Q N-path band-pass filters: Modeling and verification,” IEEE J. Solid-State Circuits, vol. 46, no. 5, pp. 998–1010, May 2011, doi: 10.1109/JSSC.2011.2117010.
[36] C. M. Thomas and L. E. Larson, “A CMOS broadband distributed N-path tunable bandpass filter,” IEEE Microw. Wireless Compon. Lett., vol. 24, no. 8, pp. 542–544, Aug. 2014, doi: 10.1109/LMWC.2014.2321254.
[37] C. M. Thomas and L. E. Larson, “Broadband synthetic transmission-line N-path filter design,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 10, pp. 3525–3536, Oct. 2015, doi: 10.1109/TMTT.2015.2473161.
[38] Y.-C. Lien, E. A. M. Klumperink, B. Tenbroek, J. Strange, and B. Nauta, “High-linearity bottom-plate mixing technique with switch sharing for N-path filters/mixers,” IEEE J. Solid-State Circuits, vol. 54, no. 2, pp. 323–335, Feb. 2019, doi: 10.1109/JSSC.2018.2878812.
[39] C. Luo, P. S. Gudem, and J. F. Buckwalter, “A 0.2–3.6-GHz 10-dBm B1dB 29-dBm IIP3 tunable filter for transmit leakage suppression in SAW-less 3G/4G FDD receivers,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 10, pp. 3514–3524, Oct. 2015, doi: 10.1109/TMTT.2015.2460733.
[40] A. Mohammadpour, B. Behmanesh, and S. M. Atarodi, “An N-path enhanced-Q tunable filter with reduced harmonic fold back effects,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 60, no. 11, pp. 2867–2877, Nov. 2013, doi: 10.1109/TCSI.2013.2256238.
[41] P. Karami, A. Banaeikashani, B. Behmanesh, and S. M. Atarodi, “An N-path filter design methodology with harmonic rejection, power reduction, foldback elimination, and spectrum shaping,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 67, no. 12, pp. 4494–4506, Dec. 2020, doi: 10.1109/TCSI.2020.3009191.
[42] C. Luo and J. F. Buckwalter, “A 0.25-to-2.25 GHz, 27 dBm IIP3, 16-path tunable bandpass filter,” IEEE Microw. Wireless Compon. Lett., vol. 24, no. 12, pp. 866–868, Dec. 2014, doi: 10.1109/LMWC.2014.2352972.
[43] M. Darvishi, R. van der Zee, and B. Nauta, “Design of active N-path filters,” IEEE J. Solid-State Circuits, vol. 48, no. 12, pp. 2962–2976, Dec. 2013, doi: 10.1109/JSSC.2013.2285852.
[44] H. Nie et al., “A CMOS 1.3-1.7GHz Q-enhanced LC band-pass RF FILTER with 1.5-67% tunable fractional bandwidth,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), Jun. 2021, pp. 503–506, doi: 10.1109/IMS19712.2021.9574792.
[45] F. Amin, S. Raman, and K.-J. Koh, “Integrated synthetic fourth-order Q-enhanced bandpass filter with high dynamic range, tunable frequency, and fractional bandwidth control,” IEEE J. Solid-State Circuits, vol. 54, no. 3, pp. 768–784, Mar. 2019, doi: 10.1109/JSSC.2018.2882266.
[46] B. Behmanesh and S. M. Atarodi, “Active eight-path filter and LNA with wide channel bandwidth and center frequency tunability,” IEEE Trans. Microw. Theory Techn., vol. 65, no. 11, pp. 4715–4723, Nov. 2017, doi: 10.1109/TMTT.2017.2698466.
[47] R. Chen and H. Hashemi, “Passive coupled-switched-capacitor-resonator-based reconfigurable RF front-end filters and duplexers,” in Proc. IEEE Radio Frequency Integr. Circuits Symp. (RFIC), 2016, pp. 138–141, doi: 10.1109/RFIC.2016.7508270.
[48] P. Song and H. Hashemi, “RF filter synthesis based on passively coupled N-path resonators,” IEEE J. Solid-State Circuits, vol. 54, no. 9, pp. 2475–2486, Sep. 2019, doi: 10.1109/JSSC.2019.2923561.
[49] M. N. Hasan, S. Saeedi, Q. J. Gu, H. H. Sigmarsson, and X. Liu, “Design methodology of N-path filters with adjustable frequency, bandwidth, and filter shape,” IEEE Trans. Microw. Theory Techn., vol. 66, no. 6, pp. 2775–2790, Jun. 2018, doi: 10.1109/TMTT.2018.2809573.
[50] H. Shaman and J.-S. Hong, “A novel ultra-wideband (UWB) bandpass filter (BPF) with pairs of transmission zeroes,” IEEE Microw. Wireless Compon. Lett., vol. 17, no. 2, pp. 121–123, Feb. 2007, doi: 10.1109/LMWC.2006.890335.
[51] P.-H. Deng and J.-T. Tsai, “Design of microstrip cross-coupled bandpass filter with multiple independent designable transmission zeros using branch-line resonators,” IEEE Microw. Wireless Compon. Lett., vol. 23, no. 5, pp. 249–251, May 2013, doi: 10.1109/LMWC.2013.2253601.
[52] T. Yang and G. M. Rebeiz, “Tunable 1.25–2.1-GHz 4-pole bandpass filter with intrinsic transmission zero tuning,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 5, pp. 1569–1578, May 2015, doi: 10.1109/TMTT.2015.2409061.
[53] Y. Wang, T. Huang, W. Wu, and Y. Li, “Octagonal on-chip wideband bandpass filter with a tunable transmission zero in 0.18-μm (Bi)-CMOS technology,” in Proc. IEEE Asia-Pacific Microw. Conf. (APMC), 2020, pp. 1003–1005, doi: 10.1109/APMC47863.2020.9331664.
[54] M. N. Hasan, Q. J. Gu, and X. Liu, “Tunable blocker-tolerant on-chip radio-frequency front-end filter with dual adaptive transmission zeros for software-defined radio applications,” IEEE Trans. Microw. Theory Techn., vol. 64, no. 12, pp. 4419–4433, Dec. 2016, doi: 10.1109/TMTT.2016.2623707.
[55] Y.-C. Hsiao, C. Meng, H.-H. C. Chien, and G.-W. Huang, “2.4-GHz tunable miniature CMOS active bandpass filter with two transmission zeros using lumped Stepped-impedance ring resonator,” in Proc. IEEE/MTT-S Int. Microw. Symp. (IMS), Jun. 2018, pp. 405–408, doi: 10.1109/MWSYM.2018.8439312.
Digital Object Identifier 10.1109/MMM.2023.3321226