RF and Baseband Filtering Techniques for Cognitive Radio

Jerzy Dąbrowski


Objective

The objective of this project is to develop practical filtering techniques for Cognitive Radio (CR) considered one fundamental technology for the future wireless communications. The CR systems require flexible and hardware efficient RF filters. Such filters based on impedance transformation technique have emerged, recently. Getting mature they would play a central role in operation of CR, especially due to the need of spectrum sensing and spectrum mobility that makes a difference to wireless systems with fixed band assignment. The respective CR functions need efficient control scenarios implemented at baseband along with specialized filtering techniques, specifically to distinguish between noise and weak signals of spread spectrum origin. In this project we aim to demonstrate the effectiveness of the proposed filtering techniques for CR by a system composed of a wideband RF front-end CMOS chip with an integrated tunable RF filter and spectrum sensor based on RF power detection and dedicated baseband processing. The key role to control the filtering process would be a software implemented on FPGA.

Field overview

The idea of Cognitive Radio (CR) dates back to the paper by Joseph Mitola [Mitola99] published in 1999. It is assumed that a radio in a CR system is capable of observing whether a particular frequency band is being occupied and, if not, it can use the band without interfering with other authorized systems. If the primary system of the band restarts transmission, the CR jumps to another frequency band or stays in the same band, altering its transmission power level or modulation scheme to avoid interference [Walko05], [Akyildiz08]. The available frequency bands, in particular, can be the licensed frequency bands of other operators, the ISM bands or the TV White Space (WS).

In principle, CR is a method of improving spectrum efficiency [Kitsunezu12]. Over the last two decades, the demands on wireless services have been increasing rapidly. The data rates in wireless communications have been experiencing an exponential growth leading to a demand for more transmission spectrum. Fortunately, the occupation of different radio frequency bands largely varies. While some bands are fully occupied, others are largely underutilized. The reported utilization of spectrum between 30 MHz and 3 GHz is about 5.2%, and it is less than 0.5% between 3 GHz and 5 GHz [SharSpect10], [Cabric05]. In this situation one of the most promising solutions is the deployment of CR, which enables spectrum sharing by dynamic and opportunistic spectrum access [Akyildiz08]. It is believed that with CR, the efficiency of spectrum usage can be significantly enhanced in the future.

A major bottleneck for the deployment of CR is the enabling hardware, i.e. spectrum-agile radio transceivers. Since the CR transceivers are required to operate in any unoccupied band within a wide range of frequencies, in practice, all the involved RF components, beginning from the transmit/receive antenna and RF filters, are challenged by the wideband operation requirements [Cabric05], [Laskar06], [E2R07], [Kaltioka11], [Parssinen11]. Another key challenge is in the spectrum sensing which is essential for CR, and which makes difference with the Software Defined Radio (SDR). CR is more demanding in terms of flexible filtering compared to the SDR. In this regard, the ubiquitous interference and the sensitivity levels place very high demands not only on the spectrum sensor dynamic range but also on the linearity [Park08], [Razavi10], [Svensson10], [Kallioinen11], [Alink12], [Axel12].

Different spectrum management scenarios and sensing techniques that have been proposed reflect tradeoffs among the necessary processing time (transmission latency), the scanned spectrum range, and the reliability in detection of spectrum holes which might be difficult to distinguish from spread-spectrum signals of very low power. One critical factor in this regard is spectrum filtering. To facilitate the spectrum scan process, especially in terms of spectrum mobility [Akyildiz08], flexible RF filters have been highly sought. Tunability over a wide range of frequencies such as 400 MHz - 4 GHz and bandwidth control could largely enhance both the spectrum sensing and the CR performance during regular transmission in various bands. The existing solutions, like frequency -agile radios (ideally SDR), possess limited RF filtering based on hardware inefficient banks of fixed SAW filters or filters with a low tunability range.

Many researchers are currently engaged in developing the communication technologies and protocols required for CR/SDR systems. However, the existing prototypes are still insufficient in terms of flexibility, bandwidth, performance, degree of integration, input/output power range, and power efficiency. The analog/digital co-design as well as the application of wideband and tunable components are still very challenging, especially for mobile terminals where the size, power consumption, and price are important factors.

Project description

We propose to develop flexible RF filtering technique based on impedance transformation that has been introduced recently [Mirzaei09], [Ghaffari10], [Mirzaei11a]. A transformed baseband impedance in combination with a transconductance amplifier (LNTA) can make a tunable high-Q band-pass filter (BPF) with a gain in stop-band proportional to the impedance value at the corresponding offset frequency as shown in Fig.1. The filter performance depends on the quality of the multiphase RF clock, the LNTA output impedance, and the parasitics of the transistor switches used for the transformation. The reported filter rejection is usually around 15 dB which is not sufficient to replace a standard SAW filter but it can be improved by increasing the output impedance of LNA. Additionally, more than one such a filter can be used in the frontend. Unlike [Mirzaei11b] we propose to keep the selective RF impedances well separated from each other to achieve a rejection suitable for the CR application, i.e. double or triple as compared to one filter stage as illustrated in Fig.2. Such a rejection is necessary to reduce intermodulation distortions evoked by out-of the scan-band blockers that can obscure useful spectrum holes hampering the CR operation. In other words, we propose to scan the spectrum over multiple sub-bands, which can be selected using the spectrum history (statistical data) and then corrected dynamically to avoid temporary in-band blockers.

In our preliminary design work we have achieved > 40 dB rejection using a cascade of two highly linear LNTAs [Duong11] with selective loads provided by transformation of a capacitive impedance for frequencies < 1 GHz, and > 30 dB rejection for 2-3 GHz while sacrificing the noise figure only by 1 dB. The out-of-band IIP3 > 20 dBm has been met that is a promising result for CR applications [Qazi13a].

Fig.1. Example of selective receiver frontend, tuned in frequency by 4-phase clock and programmed for bandwidth by impedance ZBB(w). Ideally, transformed impedance Zin(w) ~ ZBB(w -ws), wS = 2*pi/Ts .

Fig.2. Architecture of two-stage SAW-less receiver frontend where both Mixers serve as high-Q RF filters (impedance transformers). LNA2 with high input impedance is used to support filtering in first stage. Mixer2 also serves as zero-IF downconverter. Elevated supply voltage in LNA1 is used to tolerate 0 dBm blockers.

One disadvantage of the impedance transformation filtering are spurious bands around odd harmonic frequencies of the clock. To avoid interference from those bands, band limitation by fixed SAW filters can be used with cut-off frequencies below 3xflow (where flow is the minimum frequency of a signal to be sensed). This approach, however, requires using a bank of fixed filters when the range of tunability exceeds (flow, 3xflow). For example, if the lowest frequency of interest is 400 MHz, one SAW filter can support spectrum sensing up to 1200 MHz (with a reserve) and when higher RF frequencies are aimed at, another fixed filter with cut-off frequency below 3600 MHz is necessary, and so on.

In this project, we propose to combine impedance transformation filtering with the harmonic frequency cancellation technique. This technique was first proposed for signal transmission, but it appeared suited for reception as well [Ru09]. To this end, additional impedance transformation paths can be used in a parallel setup. When well balanced, they ideally should cancel the spurious bands and thereby enable spectrum sensing over a wide frequency range without employing SAW filters. Additionally, we consider replacement of the baseband capacitive load of the RF filter by a low-pass or band-pass impedance to increase the filter roll-off to possibly approach the selectivity of SAW filters. This is not a trivial task since a typical low-pass transfer function such as Butterworth- or Chebyshev-type according to Brune criterion cannot be implemented as a one-port impedance composed of RLC elements only [Wanh11, p.89]. Instead, active implementations [Darv12] might be considered carefully in terms of extra noise and nonlinear contribution (Fig.3). In this case the Flicker noise appears critical but when the filter is adapted to low-IF operation this effect can be mitigated and also a partial image rejection can be achieved at RF [Qazi13b].

Fig.3. Model of 4-path transformation of virtual LC impedance achieved with baseband Gm-C cells.

Alternatively, we propose to synthesize a dedicated LC impedance for the best possible performance in terms of rejection, flatness of amplitude characteristics, and selectivity. In this case we have to depart from the classical synthesis techniques which follow Butterworth, Chebyshev or other common frameworks.

An additional important aspect to consider is the phase linearity of the filters. For spectrum sensing based on power measurements, the phase distortion (group delay variation) can be neglected. In that case, digital correction of the linear distortion is required only for the regular reception, not during the sensing. On the other hand, for frequency bands which are occupied by spread spectrum systems, in which case it is difficult to distinguish the signal from noise, more sophisticated sensing techniques based on signal statistics and correlation measures may have to be adopted [Axel12], [Alink12], [Kosunen13]. In such cases, one has to make sure that the phase distortion does not destroy the sensing capability. A task here is to derive appropriate filter design techniques for this purpose.

Flexibility in frequency and bandwidth of the CR RF filter can largely enhance the spectrum mobility management also referred to as the spectrum hand-off [Akyildiz08]. As this process should be time effective to limit CR latency, one possible scenario is to scan the spectrum only over a relatively narrow band(s) which according to statistical data is likely to be unoccupied. Then a quick decision can be taken upon the measured RF power provided the signal can be distinguished from the noise level. For bands which are used by spread spectrum systems, more sophisticated are required as noted above. In either case, including white spaces, flexible RF filtering plays a central role to prevent contamination of the sensed spectrum by intermodulation products of possible blockers (strong interferers). Securing very high linearity of the front-end might not be sufficient in this case. For example, two moderate in-band blockers of -20 dBm power each, can provide spurious IM3 products of -80 dBm falling in the examined band when the receiver IIP3 = +10 dBm, while satisfactory power levels within this band should be less than -100 dBm. In effect the respective fragment of the band looks occupied and this problem can be resolved by reducing the filter bandwidth appropriately. Alternatively, the IIP3 can be measured independently and compared against power of the actual blockers and the IM3 products. However, when more blockers are involved and other intermodulation effects come to play the latter approach is rather problematic.

This example also shows that insufficient blocker rejection can hamper the spectrum sensing process equally. A similar in-band IM3 power can be expected when two out-of-band blockers of 0 dBm (not unusual power) are applied at the input and are attenuated by 20 dB. To scan a band the filter center frequency can be precisely tuned by the multi-phase clock while the bandwidth can be programmed using a bank of baseband impedances (capacitors). Different programming scenarios can be considered to optimize the scan process. For spectrum sensing by RF power measurement an RF detector of high sensitivity is necessary [Duong12]. In fact, such a detector should include a variable gain amplifier and A/D converter. As already mentioned the programmability in frequency and bandwidth should make the scan process fast and immune to out-of-band interference. Whether the scanned band is occupied or not can be seen from the enhanced DC signal level at the detector output, or additionally from low-frequency intermodulation components produced by the detector itself due to its nonlinear characteristics. In such a case either the band should be abandoned (considered occupied), or scanning can be repeated for a narrower bandwidth and depending on the scan strategy also a correction of the center frequency can be introduced based e.g. on the band history.

Because of manufacturing, temperature, and supply voltage variations (PTV) the detector measurement chain (detector+VGA+ADC) should be calibrated. In this project, calibration of the integrated RF detector chain can be considered a complementary task. We have shown that it can be attained using a statistical Monte-Carlo model and a multivariate regression technique [Ramzan08]. Based on this the actual calibration is achieved by on-chip DC measurements and mapping of the DC- into RF characteristics. For bands which are exploited by spread spectrum systems, signal downconversion followed by A/D conversion would be necessary [Johans13]. No extra mixer is needed in this case since the RF filter can equally serve this purpose on its low frequency port. Nyquist filtering can also be avoided provided the ADC sampling frequency follows the RF bandwidth with some reserve.

Ultimately, we want to demonstrate the effectiveness of the proposed techniques by designing a wideband RF frontend chip for frequency range 400 MHz – 4 GHz using 65 or 40 nm CMOS technology. The chip supported by a baseband processor (DSP/FPGA with ADC and DAC) is supposed to make a CR receiver suitable both for intelligent spectrum sensing and reception in normal operation mode. The respective architecture is shown in Fig.4. By using dedicated software implemented in DSP/FPGA, various CR scenarios (frequency and band selection, RF power level sensing, spread spectrum signal sensing) can be verified in the proposed setup.

Fig.4. Possible architecture of CR receiver. Impedance transformer serves both RF filtering and downconversion necessary during normal receive mode or when sensing bands with weak signals. The preliminary band sensing is achieved with the RF detector, calibrated by DC signal path.


Preliminary results

The proposed project has its roots in our recent work aimed at a blocker tolerant wideband RF receiver design [Ahsan12] and flexible receiver front-end design based on impedance transformation technique [Qazi11, Qazi13a,Qazi13b]. The latter has been supported by high-performance LNTA design [Duong11]. Also the research related to on-chip RF detectors with calibration [Ramzan08, Duong12] considerably contributes to this project.

References

[Ahsan12] Ahsan N., Svensson C., Ramzan R., Dabrowski J., and Samuelsson C., "1.1 V 6.2 mW, wideband RF Front-end for 0 dBm Blocker Tolerant Receivers in 90 nm CMOS," in Analog Integrated Circuits and Signal Processing, vol. 70, No. 1, 2012, pp. 79-90.

[Akyildiz02] F. Akyildiz, et al., "A survey on sensor networks," IEEE Commun. Mag., vol. 40, pp. 102-114, Aug. 2002.

[Akyildiz08] F. Akyildiz, et al.,"A Survey on Spectrum Management in Cognitive Radio Networks," IEEE Communications Magazine, April 2008, pp. 40-48.

[Alink12] M. S. Alink, et al., "A CMOS-Compatible Spectrum Analyzer for Cognitive Radio Exploiting Crosscorrelation to Improve Linearity and Noise Performance," IEEE TCAS1, 2012 pp. 479–492.

[Axel12] E. Axell, G. Leus,et al., "State-of-the-art and recent Advances Spectrum Sensing for Cognitive Radio State-of-the-art and recent advances," 2012, IEEE Signal Processing Magazine, (29), 3, 101-116.

[Cabric05] D. Cabric , R. W. Brodersen, "Physical layer design issues unique to cognitive radio systems," Intl. Symposium PIMRC, 2005, vol.2, pp 759 - 763.

[Darv12] M. Darvishi, et al., "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," ISSCC Dig. Tech. papers, pp. 358-359, 2012.

[Duong11] Q-T. Duong and J. Dabrowski, "Low Noise Transconductance Amplifier Design for Continuous-Time Delta Sigma Wideband Frontend," in European Conference on Circuit Theory and Design (ECCTD), pp. 825 - 828, 2011.

[Duong12] Q-T. Duong and J. Dabrowski, “Wideband RF Detector Design for High Performance On-Chip Test,” in Proc. IEEE Norchip, pp. 1-4, November 2012.

[E2R07] E2R II white paper, "RF transceiver architecture for cognitive radio user equipment," https://ict-e3.eu/project/white_papers/e2r/3.E2RII_TRXArchitecture_White_Paper.pdf

[Ghaffari10] A. Ghaffari, et al., "A Differential 4-Path Highly Linear Widely Tunable On-Chip Band-Pass Filter," IEEE RFIC Symp., pp. 299-302, 2010.

[Ghaffari11] Ghaffari A. et al., "Tunable High-Q N-Path Band-Pass Filters: Modeling and Verification," J. Solid-State Circuits, Vol.46, No.5, 2011, pp. 998-1010.

[Harada05] H. Harade, "Software defined radio prototype toward cognitive radio communication systems," DYSPAN 2005, p. 539.

[Heskamp05] M. Heskamp, et al. (2005), "A Proof-of-concept demonstrator for Cognitive Radio," ProRISC 2005.

[Johans13] H. Johansson, “Sampling and quantization," in Academic Press Library in Signal Processing: Signal Processing Theory and Machine Learning, vol. 1, pp. 169–244, 2013.

[Kallioinen11] S. Kallioinen, et al., "Multi-Mode, Multi-Band Spectrum Sensor for Cognitive Radios Embedded to a Mobile Phone," 6th Int. Conf. on Cognitive Radio Oriented Wireless Networks and Comm. (CrownCom), June 2011.

[Kaltioka11] M. Kaltiokallio, et al., "Wideband 2 to 6GHz RF Front-End With Blocker Filtering,.” in Proc. European Solid-State Circuits Conference ESSCIRC, Sep. 2011.

[Kitsunezu12] M. Kitsunezuka, et al., "Efficient Use of the Spectrum", IEEE Microwave Magazine, p.55, Jan/Feb 2012.

[Kosunen13] M. Kosunen, et al., "Survey and Analysis of Cyclostationary Signal Detector Implementations on FPGA," IEEE JETCAS, Vol.3, No.4, 2013.

[Laskar06] J. Laskar, et al., "Reconfigurable RFICs and modules for cognitive radio," SMIC05.

[Mirzaei09] A. Mirzaei, et al., "A Frequency Translation Technique for SAW-Less 3G Receivers," VLSI Circ. Symp. Dig. Tech. Papers, 2009.

[Mirzaei11a] A. Mirzaei, H. Darabi, "Analysis of Imperferctions on Performance of 4-Phase Passive-Mixer-Based High-Q Bandpass Filters in SAW-Less Receivers," TCAS I, Vol.58, No.5, pp. 879-892, 2011.

[Mirzaei11b] A. Mirzaei, et al., "A 65 nm CMOS Quad-Band SAW-Less Receiver SoC for GSM/GPRS /EDGE," J. Solid-State Circuits, Vol. 46, No. 4, 2011, pp. 950-964.

[Mitola99] J. Mitola III and G. Q. Maguire, Jr., "Cognitive radio: Making software radios more personal," IEEE Pers. Commun., vol. 6, no. 4, pp. 13-18, Aug. 1999.

[Oude13] Oude Alink M.S., et al., "Spectrum Sensing With High Sensitivity and Interferer Robustness Using Cross- Correlation Energy Detection," IEEE JETCAS, Vol.3, No.4, 2013.

[Park08] J. Park, et al., "A fully integrated UHF-band CMOS receiver with multi-resolution spectrum sensing (MRSS) functionality for IEEE 802.22 cognitive radio applications," ISSCC08, p. 526.

[Parssinen11] A. Parssinen, "Multimode-Multiband Transceivers for Next Generation of Wireless Communications," Solid-State Device Research Conf. (ESSDERC), pp.42-53, 2011.

[Qazi11] F. Qazi, Q-T Duong, and J. Dabrowski,"Wideband RF Frontend Design for Flexible Radio Receiver, " International Symposium on Integrated Circuits (ISIC), 2011, (4 pp).

[Qazi13a] F. Qazi, Q-T Duong, and J. Dabrowski, "Two-Stage Highly Selective Receiver Front-End Based on Impedance Transformation Filtering," TCAS II Vol.62, No.5 pp. (421-425), May 2015.

[Qazi13b] F. Qazi, Q-T. Duong, and J. Dabrowski, "Blocker and Image Reject Wideband Low-IF Frontend," ECCTD-2013, 4 pp.

[Ramzan08] R. Ramzan, J. Dąbrowski, “On-chip Calibration of RF Detectors by DC Stimuli and Artificial Neural Networks," RFIC Symposium, (4 pp), 2008.

[Razavi10] B. Razavi, ."Cognitive Radio Design Challenges and Techniques," IEEE J. Solid-State Circ., vol. 45, no. 8, pp. 1542-1553, Aug. 2010.

[Ru09] Ru, Z., Klumperink, E. A. M., Wienk, G. J. M., & Nauta, B. (2009). "A software-defined radio receiver architecture robust to out-of-band interference," in IEEE International Solid-State Circuits Conference, Digest of Technical Papers (pp. 230-231), Feb. 2009.

[ShaSpect10] "General Survey of Radio Frequency Bands (30 MHz to 3 GHz): Vienna, Virginia, Sept. 1-5, 2009," online: www.sharespectrum.com, http://www.sharedspectrum.com/wp-content/uploads/2010_0923-General-Band-Survey-30MHz-to-3GHz.pdf

[Svensson10] C. Svensson, "The blocker challenge when implementing software defined radio receiver RF frontends," Analog Integrated Circuits and Signal Processing, 64(2): 81-89, 2010.

[Walko05] J. Walko, "Cognitive radio," IEE Rev., vol. 51, no. 5, pp. 34-37, May 2005.

[Wanh11] L. Wanhammar, "Analog Filters Using Matlab," Springer, 2009.

[Wanh13] L. Wanhammar and H. Johansson, Digital Filters Using Matlab, 650 pages, Linkoping University, 2013.

[Zhang07] C. Zhang, et al., "Built-In Test of RF Mixers Using RF Amplitude Detectors," International Symposium on Quality Electronic Design (ISQED'07) 2007, pp. 404-409.

[Zhang11] H. Zhang, E. Sanchez-Sinencio, "Linearization Techniques for CMOS Low Noise Amplifiers: A Tutorial," TCASI, vol. 58, no. 1, 2011.