Millimeter Wave Active Bandpass Filters
A review of millimeter wave (mmWave) active bandpass filters (BPF) is presented to motivate research and development of high performance mmWave BPFs, especially above 60 GHz. Active BPFs exhibit the merits of low loss, good out-of-band rejection, good selectivity and a high integration level. The details of various design approaches and realization techniques are provided. Their strengths, weaknesses and design challenges are discussed.
The mmWave spectrum spans the frequency range from 30 to 300 GHz. It offers various unlicensed frequency bands and supports gigabits/second (Gbps) wireless communication. At around 60 GHz, the unlicensed frequency bands of 57 to 64 GHz in the U.S. and Korea, 59 to 66 GHz in Japan and Europe, 59.4 to 62.9 GHz in Australia and 59 to 64 GHz in South Africa are available. Free-space path loss, mmWave absorption by atmospheric oxygen and water vapour, rain attenuation, losses in building materials and absorption by floors and walls affect and limit propagation. These effects make the 60 GHz band more suitable for short distance (less than 1 km) communications.
Electronic filters are used to allow the desirable frequency components and to reject the unwanted components of a signal. A bandpass filter (BPF) passes the frequencies within a specified range and attenuates other frequency components. A BPF is located between the antenna and the low noise amplifier (LNA) in the RF receiver of a communication system.
BPFs can be divided into three types, including purely active, purely passive and active (active + passive or semi-passive).1 In purely passive (passive) filters, only passive components, i.e., resistors, capacitors and inductors are used. These filters do not have the capability of signal amplification. A power supply is not required for the operation of passive filters, therefore, they do not dissipate power. Since these filters do not consist of active devices, there is no active noise. They generate only thermal noise because of the resistive components. Consequently, passive filters demonstrate better noise performance. In purely active and active (semi-passive) filters, transistors are used as active components, therefore, these filters exhibit operating frequency limitations, poor linearity and high noise figure (NF) and power dissipation. They do, however, require less area than that of passive filters.
Passive filters suffer from high loss. They also have the shortcomings of incompatibility with tunable elements, they require trade-offs between bandwidth (BW) and insertion loss (IL) and they have low out-of-band rejection levels. No complementary metal oxide semiconductor (CMOS) passive BPF with 3 dB fractional bandwidth (FBW) below 10 percent or above 65 percent has been reported.
Active filters are realized by applying loss compensation (Q-enhancement) techniques, and can be used in transceiver modules to reduce loss and size.2 CMOS active BPFs that have 3 dB FBWs less than 10 percent with low IL have been reported.3,4 Loss compensation techniques can reduce IL to 0 dB.3,5,6 Moreover, good out-of-band rejection can be achieved at low as well as high frequencies. Besides the advantages of small size and low loss, active monolithic filters also have the qualities of good selectivity, a high level of integration with other circuits, and an electronic tuning capability.
Traditionally, gallium arsenide (GaAs) and indium phosphide (InP) (III-V technologies) have been assumed to be the only appropriate options for mmWave circuit implementation because of the high speed of devices using these technologies. III-V heterojunction bipolar transistor (HBT) and high electron mobility transistor (HEMT) technologies offer high cutoff frequencies (fT). GaAs and InP substrates show semi-insulating characteristics, with high resistivity in the order of 107 to 109 Ω-cm. However, III-V technologies exhibit the drawbacks of low thermal conductivity (0.46 W/cm-K for GaAs and 0.68 for InP at 300K, while silicon (Si) is 1.41), high leakage current, device reliability issues, a low integration level and high cost.
The CMOS process demonstrates low cost, a level of high integration and high reliability. MOS field effect transistor (MOSFET) scaling has supported the production of high speed transistors in CMOS technology. Nevertheless, low Q-factors of on-chip passives and poor noise performance are the major drawbacks of CMOS process.
Besides CMOS, silicon germanium bipolar CMOS (SiGe BiCMOS) is another prominent Si technology used in RF and mmWave circuit design. A SiGe HBT offers higher gain and better noise and power performance than a Si bipolar junction transistor (BJT). The performance of the SiGe BiCMOS process is competitive with GaAs and InP processes. The strength of SiGe technology is the ability to use the CMOS process for the fabrication of digital, analog and RF modules on a single chip. This is not available with any other technologies (e.g., GaAs, InP).
ACTIVE BPF DESIGN APPROACHES
Various approaches to active BPF design have been reported, some of which include transversal, recursive, cascade connections of passive filters with amplifiers, active matching, active inverter, negative capacitance and negative resistance. Of all these approaches, the negative resistance technique is the simplest and most appropriate for mmWave active filter design.2 This topology can be used in both lumped and distributed resonators. Negative resistance is used to increase the Q-factor of the passive network through resistive loss compensation. This approach makes active filters more stable, while also exhibiting good rejection ratios.7 The negative resistance topologies include transformer feedback, tapped-inductor feedback, active capacitance, active device reduction, cross-coupled pair, and others.
Figure 1 illustrates a resonator made of a quarter-wavelength (λ/4) transmission line (TL). The short circuited port of the resonator is replaced with a negative resistance (-RN) for resonator loss compensation. The expression for loss as a function of attenuation constant (α) of the TL is represented by Equation 1. Equation 2 expresses the gain generated by the negative resistance in terms of line characteristic impedance (Z0) and RN. When Equation 3 is satisfied, loss of the resonator is completely compensated, and its Q-factor becomes infinite. For this condition, RN is represented by Equation 4. If the value of RN is greater than that provided by Equation 4, the resonator holds a loop-gain, and oscillation arises.2
Besides satisfying Equation 4, negative resistance should be constant over a broad frequency range to compensate the loss of a passive resonator without causing oscillation and instability. Figure 2 shows the negative resistance circuit with a FET as an active component, and Figure 3 is a schematic of a two-stage active BPF using the negative resistance technique.
ACTIVE BPF REALIZATION TECHNIQUES
Active BPFs can be broadly classified into two categories: Q-enhanced non-TL based and Q-enhanced TL based. The Q-enhanced non-TL based technique can be further divided into two parts: LC based and active capacitance based.
Q-enhanced LC based active BPFs use mainly transformer feedback8,9 and tapped-inductor feedback10 architectures. The tapped-inductor feedback technique provides high inductance, low power dissipation and small size compared to the conventional transformer feedback topology. Nonetheless, active BPFs realized with both of these topologies have the disadvantages of relatively high NF and high power consumption. These topologies generally use common-source or common-gate series feedback structures, which are mostly employed in oscillator circuit design. Owing to the series feedback structure, the noise performance of these active BPFs is degraded.1
In low GHz (less than 10 GHz) RF integrated circuit (IC) design, the main challenge is the low Q-factors of on-chip passive inductors. The IL of active BPFs at these frequencies is mostly governed by the low Q of these inductors. At mmWave frequencies, passive inductors show desirable Q-factors of 15 or more. The Q-factors of on-chip passive inductors increase with increasing frequency, while those of on-chip passive capacitors tend to decrease. Therefore, in mmWave filters, the IL is affected mainly by the Q-factors of the capacitors. The NF also depends on IL, so both IL and NF can be reduced by using high-Q capacitors in active filter design.11 Furthermore, with increasing frequency, capacitances of on-chip passive capacitors also increase. Low Q-factors and deviation in capacitance values with frequency are serious drawbacks. Consequently, the overall performance of mmWave ICs is significantly degraded.
On-chip active capacitors are better substitutes for on-chip passive capacitors. Active capacitors are constructed with transistors and possess high Q-factors and tunable capabilities. These high-Q capacitors can be used for loss and noise performance improvement of mmWave active BPFs. Q-enhanced LC based active BPFs suffer from poor noise and power performance, while BPFs realized with active capacitors do not.1,11
Active BPFs have been reported that use Q-enhanced TL based techniques, such as cross-coupled pair,3-5,12 and coupled negative resistance.13 Different cross-coupled pair architectures are applied to the resonator to compensate for the resistive losses, including nMOS5,12,14 and complementary3,4 cross-coupled pair.
The realization of passive resonators using synthetic quasi-transverse electromagnetic (quasi-TEM) TLs, also known as complementary conducting strip TLs (CCS TL), has been reported.3,4 Compared to conventional thin-film microstrip line, CCS TL can provide more parameters for synthesis without any changes in the process and material constants. In addition, a CCS TL can be meandered in a 2-D plane with fewer coupling effects. Efficiently meandered CCS TLs can provide compact layouts, size miniaturization, and a high degree of integration. All of these attractive features of CCS TL support its application in monolithic microwave integrated circuits (MMIC) and system on chip (SOC) implementation.
A modified form of the CCS TL, known as a condensed complementary conducting strip TL (C-CCS TL) has been described for passive resonator realization.12 The C-CCS TL has a significant capability for further area reduction. Both the CCS TL and C-CCS TL facilitate size reduction and Q-factor improvement simultaneously and may, therefore, be selected for TL based compact mmWave BPF design.
SELECTIVITY AND MULTIMODE
The selectivity of low order on-chip BPFs is unsatisfactory.10 Conventional second-order BPFs show low out-of-band rejection levels. The selectivity of a filter can be improved by increasing its order, at the expense of increased IL and chip area. Various highly selective active BPFs are reported using the concept of transmission zeros. These transmission zeros are placed on either side of the filter passband.
Most on-chip BPFs have been designed using singlemode resonators. The size and loss of a filter can be reduced by modifying the conventional resonator to generate additional modes for the realization of multimode operation.15 A SiGe BiCMOS on-chip multimode passive BPF based on a multimode resonator (MMR) is reported by Ma et al.15 It exhibits low IL in a small chip area. Dualmode BPFs have achieved great importance in advanced wireless communication systems because of their high Q-factors, good selectivity and narrowband performance. The dualmode active BPF was first reported by Karacaoglu et al.16 A ring resonator based dualmode CMOS active BPF using CCS TLs is presented by Su and Tzuang.3
The performance parameters of previously reported active BPFs1,3,5,8,10,17 are compared in Table 1. N, fc, PDC, P1dB and RL are the order (number of poles), center frequency, DC power dissipation, in-band input 1 dB compression point, and return loss, respectively, of a BPF. Figure of merit (FOM) is a parameter used to compare active RF BPFs and is defined as:8,10
where PDC and P1dB are in watts, fc is in Hz, and NF and FBW are actual values (not converted to dB and percentage, respectively).
WEAKNESSES OF ACTIVE BPFs
Weaknesses of active filters include high NF, operating frequency limitations and sensitivity to process variations as well as environmental conditions. Very few active BPFs are implemented at mmWave frequencies. The center frequencies of those reported include 31.8 GHz in GaAs,18 34.2 GHz in CMOS process,5 40 GHz in CMOS11 and 65 GHz in GaAs.2 Reducing the NF and increasing the operating frequency of active BPFs are major design challenges.
This article provides a comprehensive review of mmWave active BPFs. On-chip active BPFs have the main disadvantages of high NF and moderate operating frequency. To the best of the authors’ knowledge, only one research article2 has reported active BPFs with center frequencies around and above 60 GHz, and these filters are implemented in GaAs process. The performance and feasibility of the CMOS process for the design of such high frequency active filters need further investigation. While CMOS RF circuits face the problem of poor noise performance, the SiGe BiCMOS process may be an appropriate alternative.
- S. Wang and W. J. Lin, “C-Band Complementary Metal-Oxide-Semiconductor Bandpass Filter Using Active Capacitance Circuit,” IET Microwave, Antennas and Propagation, Vol. 8, No. 15, December 2014, pp. 1416–1422.
- M. Ito, K. Maruhashi, S. Kishimoto and K. Ohata, “60-GHz-Band Coplanar MMIC Active Filters,” IEEE Transactions on Microwave Theory and Techniques, Vol. 52, No. 3, March 2004, pp. 743–750.
- L. Su and C. K. C. Tzuang, “A Narrowband CMOS Ring Resonator Dual-Mode Active Bandpass Filter with Edge Periphery of 2% Free-Space Wavelength,” IEEE Transactions on Microwave Theory and Techniques, Vol. 60, No. 6, June 2012, pp. 1605–1616.
- K. K. Huang, M. J. Chiang and C. K. C. Tzuang, “A 3.3 mW K-Band 0.18-µm 1P6M CMOS Active Bandpass Filter Using Complementary Current-Reuse Pair,” IEEE Microwave and Wireless Components Letters, Vol. 18, No. 2, February 2008, pp. 94–96.
- M. J. Chiang, H. S. Wu and C. K. C. Tzuang, “A 3.7-mW Zero-dB Fully Integrated Active Bandpass Filter at Ka-Band in 0.18-µm CMOS,” IEEE MTT-S International Microwave Symposium Digest, June 2008, pp. 1043–1046.
- T. Soorapanth and S. S. Wong, “A 0-dB IL 2140 ± 30 MHz Bandpass Filter Utilizing Q-Enhanced Spiral Inductors in Standard CMOS,” IEEE Journal of Solid-State Circuits, Vol. 37, No. 5, May 2002, pp. 579–586.
- K. W. Fan, C. C Weng, Z. M. Tsai, H. Wang and S. K. Jeng, “K-Band MMIC Active Band-Pass Filters,” IEEE Microwave and Wireless Components Letters, Vol. 15, No. 1, January 2005, pp. 19–21.
- B. Georgescu, I. G. Finvers and F. Ghannouchi, “2 GHz Q-Enhanced Active Filter with Low Passband Distortion and High Dynamic Range,” IEEE Journal of Solid-State Circuits, Vol. 41, No. 9, September 2006, pp. 2029–2039.
- J. Kulyk and J. Haslett, “A Monolithic CMOS 2368 ± 30 MHz Transformer Based Q-Enhanced Series-C Coupled Resonator Bandpass Filter,” IEEE Journal of Solid-State Circuits, Vol. 41, No. 2, February 2006, pp. 362–374.
- S. Wang and B. Z. Huang, “Design of Low-Loss and Highly-Selective CMOS Active Bandpass Filter at K-Band,” Progress in Electromagnetics Research, Vol. 128, June 2012, pp. 331–346.
- A. Ghadiri and K. Moez, “High-Quality-Factor Active Capacitors for Millimeter- Wave Applications,” IEEE Transactions on Microwave Theory and Techniques, Vol. 60, No.12, December 2012, pp. 3710–3718.
- M. L. Lee, H. S. Wu and C. K. C. Tzuang, “1.58-GHz Third-Order CMOS Active Bandpass Filter with Improved Passband Flatness,” IEEE Transactions on Microwave Theory and Techniques, Vol. 59, No. 9, September 2011, pp. 2275–2284.
- C. Y. Chang and T. Itoh, “Microwave Active Filters Based on Coupled Negative Resistance Method,” IEEE Transactions on Microwave Theory and Techniques, Vol. 38, No. 12, December 1990, pp. 1879–1884.
- C. K. C. Tzuang, H. H. Wu, H. S. Wu, and J. Chen, “CMOS Active Bandpass Filter Using Compacted Synthetic Quasi-TEM Lines at C-Band,” IEEE Transactions on Microwave Theory and Techniques, Vol. 54, No. 12, December 2006, pp. 4548–4555.
- K. Ma, S. Mou and K. S. Yeo, “Miniaturized 60-GHz On-Chip Multimode Quasi-Elliptical Bandpass Filter,” IEEE Electron Device Letters, Vol. 34, No. 8, August 2013, pp. 945–947.
- U. Karacaoglu, I. D. Robertson and M. Guglielmi, “A Dual-Mode Microstrip Ring Resonator Filter with Active Devices for Loss Compensation,” IEEE MTT-S International Microwave Symposium, Vol. 1, June 1993, pp. 189–192.
- Z. Gao, J. Ma, M. Yu and Y. Ye, “A Fully Integrated CMOS Active Bandpass Filter for Multiband RF Front-Ends,” IEEE Transactions on Circuits and Systems-II: Express Briefs, Vol. 55, No. 8, August 2008, pp. 718–722.
- W. Mouzannar, H. Ezzedine, L. Billonnet, B. Jarry and P. Guillon, “Millimeter-Wave Band-Pass Filter Using Active Matching Principles,” Proceedings of the IEEE-Russia Conference on High Power Microwave Electronics, September 1999, pp. 11–14.
Editor’s note: The authors’ biographies were printed with their prior article, “Millimeter Wave Passive Bandpass Filters,” published in the January 2017 issue.