RF Technologies and Challenges for Future MBR Systems in Cellular Base Stations

1 Introduction
    3G commercialization and evolution have been advanced and promoted by 3GPP and this has resulted in higher data rates and higher quality wireless communication services. With the development of the Internet of Things (IoT) and cloud computing, full convergence and resource sharing have become the trend of future networks (Fig. 1). In the future, mobile communication networks will be one of the primary carriers of data traffic.  

    With global warming and energy crises, green, energy-efficient network construction is also a key issue. This raises challenges for network operators because in a climate of competitive pressure they must optimize their investment and operating cost in order to lower CAPEX and OPEX. This is especially true for base stations. Therefore, future radio access networks (RANs) need to be smart and have advanced features such as dynamic spectrum allocation (DSA) for higher spectrum efficiency, higher performance, higher power efficiency, higher end-to-end efficiency, and all resource (active) sharing. Cognitive radio (CR) satisfies these features and has been the subject of many recent IEEE papers. At its core, a CR system can sense, adapt, and learn from its surroundings. With computational intelligence, it can also change communication parameters in response to changes in application needs or changes in the radio frequency landscape. A CR system must be wideband. Therefore, future cellular base station systems (CR systems) should be multimode, multiband, multifunctional, flexible, and smoothly upgradable. They should also be cost and power efficient, smart, and capable of coexisting and sharing equipment. Several key technologies for achieving these features are being researched [1]-[4].

    Multicarrier multimode base stations are currently in commercial use and run on a software defined radio (SDR) platform. SDR is defined as a radio in which the radio frequency (RF) operating parameters of frequency range, modulation type, and/or output power can be set or altered by software, or the technique by which this is achieved [1]. An ideal SDR system allows all signals to be processed digitally except those of the analog antenna [1]. However, current SDR systems can only partly implement functions through programmable software. Bandwidth is limited by hardware, and power efficiency is not ideal.

    SDR will evolve into multiband radio (MBR), which is a wideband system with frequency-agile devices [1]. An MBR system can change how and where devices operate within the radio spectrum, moving between a set of frequency bands in response to interference or other constraints. MBR can be called an advanced level of SDR. CR will be a more advanced technology than MBR. It will not only satisfy MBR requirements but will also be smart and have computational intelligence. CR can be seen as an expansion of SDR. Currently, radio technology is moving from SDR to MBR. As shown in Fig. 2, all SDR, MBR and CR systems run on the SDR technology platform. However, CR is smart and has computational intelligence [1]. MBR plays an important role in the development of radio technology.

2 Architecture and Features of an MBR System
    An MBR system with frequency-agile characteristics can be multiband, multimode, multifunctional, flexible, and smoothly upgradable. It can also be cost and power efficient and capable of sharing equipment. Therefore, MBR can be applied to many scenarios to help operators reduce CAPEX and OPEX.

    Fig. 3 shows an MBR system architecture for cellular base station. Most RF devices in this system are required to support wideband, and the duplexer or analog filter should be tunable within a frequency range of, for example, 300 MHz. It is a key point for MBR and CR to have frequency-agile characteristics. Based on the SDR platform, many MBR system functions could be performed or partly performed by a software program to reduce the difficulty in designing hardware such as RF devices.

    MBR can be classified into two levels according to the range of frequency agility. One level is reconfigurable within a frequency band range of 300 MHz; that is, 700-1000 MHz. The other (higher) level is reconfigurable within a frequency band range of no less than 1 GHz; that is, 1-2 GHz. The system design for the latter is more difficult than for the former. On every level, subsystems except the duplexer allow wideband from 0 Hz to the maximum bandwidth of, for example, 300 MHz or 1 GHz. The duplexer comprises multiband filters, and every sub-band filter satisfies the protocol requirements of the communication mode in that sub-band.
MBR is a whole solution that caters for the advancement of network. However, in implementing key MBR technologies such as multiband/broadband antenna, multiband duplexer, broadband/multiband filter, high efficiency broadband power amplifier (PA), broadband synthesizer, high speed ADC/DAC converters, and high performance field-programmable gate array (FPGA) or digital signal processor (DSP), there are challenges. The cost of implementing MBR should also be carefully considered.

3 Key RF Technologies and Challenges for an MBR System
    An MBR system is wideband and based on the open SDR platform. Therefore, the following discussion also applies to SDR systems.

3.1 Broadband/Reconfigurable Antenna
    The antenna in an MBR system must be broadband and multiband for port sharing, providing broadband coverage, and providing flexibility. Most currently used multiband antennas are multiband and multiport (or a port with a combiner inside). Although the antenna provides dual/triple/multiband performance, every sub-band corresponds to a port.

    The broadband and multiband antenna design of cellular base station antennas is based on microstrip antenna technology. Microstrip technology has a planar electric dipole and a shorted patch antenna (equivalent to a magnetic dipole) for achieving a wide range of voltage standing wave ratio (VSWR) performances and operating bandwidth with excellent electrical characteristics [2], [3]. Broadband load matching with excellent electrical characteristics is a challenge to antenna design. Moreover, cost, size, and weight of antennas are also of concern to operators.

    Broadband and multiband antennas covering about 300 MHz have been released by vendors such as Andrew. However, implementing such antennas in outdoor base stations covering about 1GHz and achieving excellent electrical performance is a challenge with present technology. Mobile Mark Inc. has released an antenna called the Surface Mounted Multiband (SMW) antenna. It is a small distributed system antenna that has many different indoor wireless applications. It has two broadband antenna elements: an 800-2700 GHz element that can be used for Cellular 850/1900/2100 MHz, WiMAX 2.5 GHz, or a second Wi-Fi, and a 1700-2700 MHz element that can be used for Advanced Wireless Services (AWS-1) band 1.7-2.1 GHz, GSM 1.9 GHz, Wi-Fi 2.4 GHz, and WiMAX 2.5 GHz.

    One approach to this challenge is to use reconfigurable antenna technology based on RF Micro-Electro-Mechanical System (MEMS) [5]-[9]. This technology is intelligent and state of the art. A reconfigurable antenna design using a network of MEMS switches can change its operating frequency and radiation/polarization characteristics, which are the goals of MBR. Barriers to implementing RF MEMS reconfigurable antennas in cellular base stations are performance, reliability, power limit of RF MEMS switches, design of switch bias networks, and control algorithms [8], [9].

3.2 Multiband/Tunable Duplexer
    In a current SDR system, the duplexer works in a single band. A 900 MHz frequency band duplexer for GSM\UMTS\LTE has RX of 890-915 MHz, TX of 935-960 MHz, filter bandwidth of 25 MHz, and total bandwidth of 70 MHz. The maximum bandwidth of the duplexer for GSM\UMTS\LTE 1800 MHz is about 170 MHz (with 75 MHz filter bandwidth). However, a multiband duplexer should be tunable to no less than 300 MHz with sub-band filter covering no less than 40 MHz for an SDR base station. Therefore, the multiband duplexer may be the most difficult component to design in an MBR system.

    For wireless duplexer applications, TX and RX filters are required to have extremely sharp roll-offs. Current base stations mostly use high-Q air cavity duplexers in which multiple cavities are in series with TX OUT or RX OUT for high performance. A duplexer cavity comprises resonant cavities which can simply be two carefully tuned resonant circuits. One circuit sets the bandpass frequency the cavity is resonate on, and the other is for coupling energy into the cavity. Respectively, these circuits have a bolt for tuning the bandpass frequency and for tuning the frequency of the cavity notch.

    The difficulty in designing a reconfigurable multiband duplexer for an MBR system lies in mechanical requirements and achieving excellent electrical characteristics with frequency shift and variable sub-bandwidths. Currently solutions only use a motor to tune the bolt depth or move coppers in the top covers of all cavities for resonating on the desired bandpass frequencies. Also, they only cover 200 MHz bandwidth for narrower sub-band application (Fig. 4). High cost is also a key obstacle for applying multiband duplexers.

    The novel duplexer architecture being researched in [10] may be an ideal solution for reconfigurable multiband duplexers in MBR system. This solution can be used to replace the duplexer or reduce its strict performance so that it can be tunable easily. The architecture combines a low isolation device with an adaptive loop canceling scheme and is same as feedforward fashion. It provides the required transmitter leakage and transmitter noise isolation over wideband by using a delay element and an adjustable vector attenuator in cancellation path (Fig. 5) [10]. The feasibility of wideband cancellation depends mainly on the delays in the main path and cancellation paths. These delays are restricted by attenuation coefficients in the cancellation paths. The achievable cancellation bandwidth also depends on the duplex frequency; smaller delay differences and smaller duplex frequencies give higher cancellation bandwidths. Wider bandwidth cancellation can also be achieved by employing two or more loops. Arithmetic and noise floor from RF devices may be a more important factor that others affecting cancellation. The test results are not very ideal and further study is needed [10].

    A more advanced technology for application in multiband duplexers is metamaterial duplexer technology.  This is currently in the theoretical and lab testing stage. When applying this technology in multiband duplexers, the left-hand artificial structure metamaterial with negative refractive index exhibits unusual properties. If RF signals input it with different injection angle, different pass band and stop band which is the want of designing multiband duplexer will appear [11].

3.3 Wideband PA Technologies
    For high efficiency and broadband connectivity in an MBR system, the PA in the system should cover wide bandwidth and have a high linearity for amplifying signals without distortion. These performances require transistors with higher impedance and allow for easier and lower loss matching networks in amplifiers. Purely real impedances can theoretically be matched to a 50 Ω system over any bandwidth by using an infinite number of matching elements. However, actual devices have optimum impedances with a reactive component. Complex loads can be matched only over a limited bandwidth as defined by Fano’s limit [4]. A suitable figure of merit for high power broadband capability in a device technology is a low pF/W gate and drain capacitance. Therefore, a wide bandgap (WBG) semiconductor such as gallium nitride (GaN)—which can be operated at high drain voltage and has low parasitic capacitances per watt of output power—is a favorable choice for use in frequency-agile pulsed applications such as military radar, air traffic control radar, and communications jamming [4].

    In current SDR systems, laterally diffused metal oxide semiconductor (LDMOS) is the most important technology for the PA. Its large device parasitic capacitances per watt of output power lead to low device input/output impedances. However, because of the narrow instantaneous bandwidth (about 40 MHz) and gain characteristic of LDMOS devices, it is no longer fit for MBR systems. GaN technology has been developing fast and will gradually be commercialized. The bandwidth of GaN RF PA transistors is generally more than 300 MHz, even GHz in microwave frequency band. There are several circuit architectures for GaN PA applications, including Class AB, Doherty, and very-high-efficiency Class D/E/F/F-1. The leading vendors of GaN PA transistors include RFMD and CREE. Performance, cost, and reliability are the main challenges to GaN PA application [4].

    Digital predistortion (DPD) plays an important role in an SDR system; it improves PA efficiency and reduces the difficulty of designing hardware devices. A accurate non-linear RF PA transistor model ensures high performance of DPD, and LDMOS device models are rich and perfect. However, GaN RF is an emerging technology, and research on non-linear modeling is still challenging.

3.4 Wideband Transceiver Technologies
    In an MBR system, a transceiver should support wideband and should have reconfigurable frequency band.  There are conflicting requirements on the transceiver, including wideband, multimode, high dynamic range, high power efficiency, cost and size. In current SDR systems for cellular base stations, the transceiver architecture is mainly based on a direct or single conversion solution with wideband performance at zero or high intermediate frequency. As shown in Fig. 3, the challenges for this MBR system architecture lie mainly in the design of the ADC\DAC, multiband synthesizer, and image filter. High dynamic range and high resolutions of ADCs\DACs (such as 14-bit ADC and 16-bit DAC) can be used in the direct or single conversion transceiver architecture in order to satisfy the systemic requirements. Several multiband synthesizers have been released, including the AD4350 (which covers 137.5-4400 MHz), and the HMC22 (which divides RF output into three bands: 665-825 MHz, 650-1330 MHz, and 2660-3300 MHz). The frequency band in the image filter is variable within the required range. Combining multiband filters with switches is a current solution for the image filter in multimode and multiband handsets. However, this solution is large and expensive and is not flexible and advanced.

    With low power consumption, high isolation, high density, and high integration advantages, RF MEMS is an emerging area in tunable filter design. By changing the values of MEMS switches or varactors within RF filters, tunable characteristics can be achieved [5]. Film bulk acoustic wave resonator (FBAR) filter techniques are also a current area of research interest. FBAR devices have lower loss (high Q value), better power-handling, and better robustness with the most demanding specifications. Because they are more expensive to manufacture than other solutions, there is still much research to be done [12].

    In Fig. 3, the transmitter and feedback receiver have no filters because the DPD bandwidth is too wide, and the algorithms such as DPD and Auto Quadrature Error Correction (AQEC) in baseband or digital intermediate frequency help reduce spurious signals. Accordingly, the difficulty in designing analog devices is reduced.

    The architecture for a zero intermediate frequency (ZIF) receiver has a pathway for full on-chip integration of the receiver because the signal is directly demodulated to baseband I and Q signals. Because the intermediate frequency (IF) is zero, there is no need for an external IF surface acoustic wave (SAW) filter for channel selection, and there is no need for an additional IF synthesizer section or an image reject RF filter (required in a high IF receiver). The basic RF filter in the duplexer rejects out-of-band blockers and transmitter leakage (Fig.3). However, an external bandpass RF filter might still be required after a low noise amplifier (LNA) in order to further reject out-of-band blockers and transmitter leakage at the demodulator input caused by limited finite duplexer TX-RX isolation. Channels are selected at baseband by on-chip low-pass filters that filter channels and reject close-in blockers. The bandwidth of on-chip baseband filters can be programmed on-chip so that the filters can operate the receiver in multiband and multimode applications. After channel filtering in a ZIF receiver, I/Q signals at baseband are amplified by variable gain amplifiers before they are digitized in the analog baseband section. Challenges to designing a ZIF receiver include high IIP2 with the second-order distortion, and the algorithm on current AQEC. The available IF bandwidth of this solution is about 20 MHz and cannot satisfy the requirements of multicarrier GSM because of the dynamic ADC range [13].

    Using flexible radio architecture on an SDR platform in order to support an MBR system and multiple wireless standards has attracted much interest. The digital design flow enables a higher level of system integration and higher bandwidth, simplifies testability, and provides reduced power, size, and cost. Some advanced transceiver architectures that have not been put into commercial use will be introduced here. These include RF sampling receiver and all digital transmitter [13]-[23].

    The digital transmitter with switching mode PA (SMPA) is an attractive choice for current SDR systems and next generation MBR systems because of its high power efficiency, linearity, low complexity, flexibility, reconfigurablity, and wideband. Much research has been focused on all-digital transmitter architectures such as low-pass delta-sigma modulator (DSM), bandpass DSM, and direct-pulsewidth/position-modulation (PWPM) [13]-[18]. Theoretically, all-digital transmitter uses a quantizer at the modulator’s output to generate a pulse-shaped signal. The quantization noise is spread over a wide band and is shaped outside the useful band of the signal using interpolation and the delta-sigma transfer function. This architecture generates an entirely digital two-level signal at RF, so it is configurable and suitable for multistandard and multiband applications. Fig. 6 shows novel bandpass DSM transmitter architecture [18].

    The DSM transmitter for wireless RF applications has the following two drawbacks:

    (1) RF signals are centered at several gigahertz, and oversampling of the carrier requires enormous digital clock rates. This limits the signal bandwidth to 25 MHz based on currently available technologies [13]-[18].

    (2) The architecture is complex. The modulator needs to work at several gigahertz and requires high-speed computational capability for digital signal processor or FPGA. This greatly increases the cost and power consumption of the designed circuits. These challenges give direction to future research.

    It has recently been popular to use RF-sampling techniques for implementing more flexible front-ends in terms of frequency tuneability, filtering, and easing of requirements on the ADC [19]-[23]. A receiver with this RF-sampling ADC is most compact and has a very simple design, no LO and mixer, and a relaxed filter. But its drawbacks are evident—it requires more digital processing, is has the highest sampling rate, it is most sensitive to jitter, require a higher power converter, and its overall linearity is limited by ADC. Fig. 7 illustrates a direct RF sampling receiver architecture with wide frequency or multiband coverage and arbitrary tuning. This architecture provides a high degree of reconfigurability of tuning range and bandwidth by using a tunable or selectable anti-aliasing filter before the stage of RF conversion [19]-[23]. The sampling rate of ADC for direct RF bandpass conversion receiver is defined by a tunable bandwidth of, for example, 300 MHz. This is at least double the bandwidth for meeting the Nyquist law, so it is not extremely high. The difficulty with this type of ADC application is that, currently, its dynamic range cannot meet the requirements of cellular base stations. Companies such as ADI and TI have been researching and developing this technology. It is conceivable that this RF-sampling technique will be in commercial deployment soon. Another type of ADC architecture is low pass sampling and oversampling. This has a very high sampling clock of about several gigahertz if applied in RF band, for example, level two MBR scenes. This architecture is now in a technologically ideal state.

4 Conclusions and Prospects
    MBR with frequency-agile characteristics can be considered the advanced stage of SDR and the basis of CR. MBR is a wideband and frequency-agile system that is implemented on the SDR platform. Based on MBR for cellular base stations, GSM, UMTS, LTE, and other communication systems can be smoothly upgraded, can coexist, and can share a number of network elements. However, there are still several technical challenges to MBR implementation and commercial deployment. These will require further study. 

 
    Because of limited hardware bandwidth, power inefficiency, size and cost of radio systems, as well as improvements in DSP and FPGA, analog processing is being turned into digital processing using digital compensation and algorithms. The compensation solutions presented in this paper allow for an easing of analog requirements for small, low cost, flexible and highly reconfigurable radios in broadband communication systems. This trend is inevitable and attractive.

Acknowledgement:
Thanks to Wei He from the Shanghai RRU department of ZTE Corporation for his contribution to this work.

References
[1] T. Weingart, “A method for dynamic reconfiguration of a cognitive radio system,” Ph.D dissertation, Dept. Comp. Sci., Uni. Colorado, Boulder, 2006.
[2] T.G. Spence and D.H. Werner, “A novel miniature broadband/multiband antenna based on an end-loaded planar open-sleeve dipole,” IEEE Trans. Antennas Propag., vol.54, no12, pp.3614-20, 2006.
[3] Zhi Ning Chen and Kwai-Man Luk, “Antennas for Base Station in Wireless Communications”, New York: McGraw Hill, 2009.
[4] S. Azam, C. Svensson and Q. Wahab, “Comparison of two GaN transistor technologies in broadband power amplifiers”, Microwave Journal, 53(4), p.184, 2010.
[5] R. J. Richards and H. J. De Los Santos “MEMS for RF/Microwave Wireless Applications: The Next Wave,” Microwave Journal, March 2001.
[6] B. Cetiner, H. Jafarkhani et al, “Multifunctional reconfigurable MEMS integrated antennas for adaptive MIMO systems,” IEEE Commun Mag, vol. 42, no. 12, pp. 62-70, 2004.
[7] T. Ativanichayaphong, Ying Cai, Jianqun Wang, Mu Chiao and J.Chiao, “Design considerations of reconfigurable antennas using MEMS switches,” in Proc. SPIE Microelectronics, MEMS, and Nanotechnology Symp., Brisbane, Australia, 2005.
[8] J. Bernhard, “Reconfigurable antennas and apertures: state of the art and future outlook,” in Proc. SPIE Conf. Smart Electronics, MEMS, BioMEMS and Nanotechnology, San Diego, CA, 2003, pp.1-9. 
[9] B. Cetiner, H. Jafarkhani, J. Qian et. al, “Multifunctional reconfigurable MEMS integrated antennas for adaptive MIMO systems,” IEEE Commun. Mag., vol.42, pp.62-70, Dec. 2004.
[10] T. O'Sullivan, R. York, B. Noren, P. Asbeck, “Adaptive duplexer implemented using feedforward technique with a BST phase shifter,” IEEE Trans. Microw.Theory Techniques, vol. 53, no 1, pp. 106-114, Jan. 2005.
[11] S.Geelani, “High Q-factor metamaterial duplex filters in suspended stripline technology,” Ph.D dissertation, Technischen Fakult?t, Universit?t Erlangen-Nürnberg zur Erlangung des Grades, Erlangen, M?rz, Germany, 2009.
[12] C. H. Tai, T. K. Shing, Y. D. Lee and C. C. Tien, “A novel thin film bulk acoustic resonator (FBAR) duplexer for wireless applications,” Tamkang J. Sci. Eng., vol. 7, no. 2, pp. 67-71, 2004.
[13] W.Y Ali-Ahmad, “Radio transceiver architectures and design issues for wideband cellular systems,” Proc. IEEE Int. Workshop Radio Freq. Integration Tech., Singapore, p.21, 2005.
[14] M.Helaoui, S. Hatami, R. Negra and F. Ghannouchi, “A novel architecture of delta-sigma modulator enabling all-digital multiband multistandard RF transmitters design,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol.55, no.11, pp.1129-1133, Nov. 2008.
[15] V. Parikh, P. Balsara and O. Eliezer, “All digital-quadrature-modulator based wideband wireless transmitters,” IEEE Trans.Circuits Syst. I, Reg. Papers, vol.56, no.11, pp.2487-2497, Nov. 2009.
[16] Sung-Rok Yoon and Sin-Chong Park, “All-digital transmitter architecture based on bandpass delta-sigma modulator,” Int. Symp.Commun.Inf. Tech (ISCIT), Incheon, Korea, pp.703-706, 2009.
[17] B. Thiel, S. Dietrich, N. Zimmerman and R. Negra, “System architecture of an all-digital GHz transmitter using pulse-width/position-modulation for switching-mode Pas,” Asia Pacific Microwave Conf. (APMC), Singapore, pp. 2340-2343,2009.
[18] A. Jayaraman, P. F. Chen, G. Hanington, L. Larson, and P. Asbeck, "Linear high efficiency microwave power amplifiers using bandpass delta-sigma modulators," IEEE Microwave and Guided Wave Letters, vol. 8, pp. 121-123, Mar. 1998.
[19] Johann-Friedric Luy, T Mueller, T. Mack, A Terzis, “Configurable RF receiver architectures,” IEEE Microwave Magazine, pp. 75-82, March 2004.
[20] D. M. Akos, M. Stockmaster, J.B.Y. Tsui, J. Caschera, “Direct bandpass sampling of multiple distinct RF signals,” IEEE Transactions on Communications, pp. 983-988, July 1999.
[21] Gerald L. Fudge, “Reconfigurable direct RF bandpass sampling receiver and related methods,” United States Patent Application # 20070081617, 12 April 2007.
[22] R. Barrak, A. Ghazel, F. Ghannouchi, “Design of sampling-based downconversion stage for multistandard RF subsampling receiver,” IEEE Int. Conf. on Electronics, Circuits and Systems, pp. 577-580, 10-13 Dec. 2006.
[23] M. L. Psiaki, S. P. Powell, H. Jung and P. M. Kintner, “Design and Practical Implementation of Multi-frequency RF Front Ends Using Direct RF Sampling,” IEEE Trans. on Microwave Theory and Techniques, vol. 53, Issue. 10, pp. 3082-3089, Oct. 2005.

[Abstract] This paper describes the advances and features of future cellular base stations. Software defined radio (SDR) evolves to cognitive radio (CR), which is smart and has wideband, and multiband radio (MBR) with reconfigurable wideband can be regarded as the basis of CR and an advanced level of SDR. Based on the SDR platform, several radio frequency (RF) solutions for implementing MBR systems are proposed, and some challenges to MBR implementation are discussed.

[Keywords] future cellular base station; SDR; MBR; RF; challenges