RF spectrum sensing technique for cognitive UWB radio network

RF Spectrum Sensing Technique for Cognitive UWB Radio Network M. Anis1 , M. Ortmanns1 , N. Wehn2 Institute of Microelectronics, University of Ulm, 89081 Ulm, Germany 2 Microelectronic Systems Design Research Group, University of Kaiserslautern, 67663 Kaiserslautern, Germany 1

Abstract—This paper proposes the architecture of an RF spectrum sensing scheme for cognitive UWB radio network. The design architecture consists of multiple super-regenerative bandpass filters which have the ability to configure for wide and narrow bandwidths. Twelve digitally controlled superregenerative oscillators are implemented to cover the UWB spectrum. The oscillators are tuned for specific ranges of frequencies during certain periods of time. Multiple quench cycles with different bias currents of oscillators are applied to adjust the selectivity and sensitivity in discrete time intervals. Occupied and free spaces within the UWB spectrum are identified by digital correlation between the output patterns of all on-chip super-regenerative bandpass filters. Index Terms — Low power, cognitive radio, UWB, superregenerative filters, LC oscillator.

I. INTRODUCTION Nowadays most of the devices are adopting wireless communication. This is not difficult to imagine that future technologies will face the crisis of spectrum availability. On the other hand, most of the allocated spectrums are inefficiently utilized. This situation necessitates the use of the frequency spectrum in a more flexible way, such as a cognitive radio. Cognitive radio has the ability to sense the spectral environment over a wide frequency range and adjust its operating parameters accordingly to achieve a high spectrum efficiency compared to existing systems [1]. It has the potential to build a new methodology for the future of wireless communication systems. There are two frequency bands where the cognitive radios might operate in a near future: 400-800MHz (UHF TV bands) and 3-10GHz for long and short range applications respectively [2]. But there are number of complex requirements for the implementation of a cognitive radio network like no interference with licensed systems, adjustable pulse shapes, bandwidths, transmit power, providing multiple access and ensuring the security of information. UWB technology is considered as attractive candidate because it has an inherent potential to fulfill the major requirements for cognitive radios [3]. An efficient RF spectrum sensing technique is required to identify the free and occupied transmission channels in order to build the cognitive radio network. This is one of the severe problems in implementing a cognitive radio. The previously reported RF sensing schemes in sub-GHz regime require complex and

978-1-4244-6664-1/10/$26.00 ©2010 IEEE

power consuming architectures which are based on reconfigurable LNAs, mixers, variable notch filter, wideband tunable oscillators, IF stages and ADCs [4-5]. Many research papers have been published but no test chip currently offers UWB spectrum sensing scheme. This works presents the concept of using super-regenerative bandpass (SR-BP) filters in RF spectrum sensing scheme for low power, low cost and short range medium data rate cognitive UWB radio networks. The application of super-regenerative filters for UWB [6] and narrowband [7-9] radio technologies has been previously reported. In this work, reconfigurable SR-BP-filters are used in both wide and narrowband detecting mode to sense the spectrum in fine and coarse manner. The multiple superregenerative filters are required to cover the whole UWB (3.1-10.6GHz) spectrum. These oscillators are digitally tuned for a specific range of frequencies during certain periods of time to sense the spectrum in discrete time intervals. After the digitally correlating the response of all SR-BP-filters during multiple quench cycles, the free and occupied places in the UWB spectrum are detected. II.

SYSTEN OVERVIEW

The proposed architecture of RF spectrum sensing for low power cognitive UWB radio networks is shown in figure 1. SR-Filter 3.25GHz

UWB Impulses NB Radios

+

SR-Filter 3.8GHz

Digital Correlation Block (DCB) & Control Unit

E0-11

SR-Filter 9.7GHz Q0-15

Fig.1. The conceptual diagram of the proposed RF spectrum sensing technique for cognitive UWB radio networks. It is assumed that multiple numbers of narrowband radios and impulses of around 500-800MHz bandwidth within the UWB

506

spectrum are present at the antenna of cognitive radio. The front end consists of multiple SR-BP-filters tuned within 3.1 to 10.6GHz. These filters can extract the energy from UWB impulses and narrowband received signals. The unoccupied places in the spectrum are called white spaces. These spaces act as noise for super regenerative filters. Noise is uncorrelated in all SR-BP-filter output patterns during multiple quench cycles. While the narrow band signals and UWB impulses generate the correlated SR-BP-filter output patterns during multiple quench cycles, the digital control unit enables each filter for a certain period of time and provides the different bias levels, frequencies and shapes of the quenching patterns. The presence and absence of any radio transmission in the time and frequency domain are detected by digital correlation between the response of each SR-BP-filter during multiple numbers of quench cycles. Twelve SR-filters with integrated LC tanks are implemented to cover the complete UWB spectrum. III.

SUPER-REGENERATIVE BANDPASS FILTER

The super-regenerative principle has been widely used since it has advantages of extraordinary gain, simplicity and low power consumption. However, the poor frequency selectivity of traditional super-regenerators limits its usage in narrow band applications, and makes them more attractive for UWB applications. The frequency selectivity for narrowband applications can be improved by Q-enhancement technique, external control circuitry, off chip resonant tanks or autocalibrated current control [7-9]. Thus the SR-BP-filters have the ability to configure for wide and narrow bandwidths to sense the RF spectrum in coarse and fine manners. The architecture of a reconfigurable SR-BP-filter for an RF spectrum sensing technique is shown in figure 2. L

oscillatory nodes, the oscillation builds up slowly by amplifying noise present near to the tuned frequency. With an RF signal, the oscillator starts up more quickly and strongly. The start-up time depends on the bias current of the oscillator, the quench signal, the strength of the injected RF signal and the frequency gap between the tuned frequency of oscillator and the frequency of injected RF signal. In this architecture, all super-regenerative oscillators are digitally controlled. The tuning frequency of each oscillator can be varied by selecting the bank of capacitors in parallel to the inductor. 16-Bit control voltages, Vc0-15 are used to change the tuned frequency of each oscillator in order to detect the narrowband radios within the frequency range of each SR-filter. The sensitivity and bandwidth of the super-regenerative filters are mostly depended on the bias current of the oscillator, and the frequency and the shape of the quench patterns [10]. Due to this reason the bias current of the oscillator, the quenching frequency and the slopes of the sawtooth quenching patterns are varied in discrete steps to detect the radio signals of different power levels and bandwidths in discrete intervals of time.

EN0

t1

t11

t12

Critical Bias levels B2 B1 B0 t11 u1 u2 Quench Frequencies Q2 Q1 Q0 u1 u11 u12 Quench Slopes S2 S1 S0

CLK

M1

Counter

Env. Det.

E

DAC VQ IQ

EN11 t2

Control Voltages VC2 VC1 VC0

VC

-A

EN1 t1

0-15

IA

Quench Shaper

S0-15 EN

RST

(a)

Enable signals

DCB & Control Unit

u11

v1

v2

t11

EN0 t12

t

VC15

(b) VC0

t15

t2

t

B15

(c) B0

u15

t12

t

Q15

(d) Q0

u15

u2

t

S15

(e) S0

v15

u12

t

Fig. 3. The control flow for RF spectrum sensing mechanism

B0-15 Q0-15 Quench VCO

Fig. 2. The block diagram of a super-regenerative bandpass filter for RF spectrum sensing. The isolation amplifier between the antenna and oscillator is used to inject the RF input into the oscillator tank. It reduces the oscillation leakage back to the antenna. The oscillator is the core of the super-regenerative filter, and it is periodically driven in and out of oscillation by the periodic quench signal. The principle of a super-regenerative filter is based on the observation of the difference in the start-up time in each quench cycle. If there is no RF signal injected onto the

The voltage controlled and current controlled quenching methods are used to configure these filters in wideband and narrowband detection mode across the tuned frequency of oscillators. In a voltage controlled quenching scheme, the oscillator is turned on and off by a change in voltage-VQ across the threshold level of transistor M1. In a current controlled quenching scheme, the oscillator is turned on and off by a change in quenching current-IQ across the critical level, which is adjusted by a digital to analog current converter. A sawtooth quench shape is selected in order to build a slow turn-on and fast turn-off pattern for each oscillator. Sixteen different slopes, S0-15 of the sawtooth quenching voltage VQ and current IQ are generated by switched controlled parallel current sources in the quench shaper. Sixteen different quenching rates, Q0-15 are generated

507

Vc5

Vc6

Vc7

Vc8

Vc9

Tuned Freq.

5.6GHz 5.5GHz 5.4GHz 5.2GHz

5.3GHz

Quench Signals

The control flow for the RF spectrum sensing mechanism is based on five major steps in one complete cycle, as shown in figure 3. In a first step, the control unit activates each filter for a certain period of time slots. Within the time slot 0-to-t1 the 1st filter, t1-to-t2 the 2nd filter and similarly t11-to-t12 the 12th filter is activated, as shown in fig 3a. In the second step, the control voltage of the switched capacitor bank is varied in sixteen small steps during each time slot in order to sense the frequency spectrum across the tuned frequency of the selected oscillator, as shown in fig 3b. In the third step, within the time slot for each step of the control voltage, the bias current of the selected oscillator is varied in sixteen small steps across the critical bias current level of the oscillator, as shown in fig. 3c. In a forth step, within the time slot for each step of the bias current, the quenching frequency is varied in sixteen small steps, as shown in fig. 3d. In step five, within the time slot for each step of the quench rate, the slopes for the quenching voltage VQ and the quenching current IQ are varied in sixteen small steps, as shown in fig. 3e. Complex DSP algorithms will be required in the future for the implementation of cognitive UWB radio networks within highly dense RF environments. In our measurement setup, the free and occupied spaces in the UWB spectrum are detected by using a simple logic comparison and redundancy checks in digital correlation block. The countered bit patterns are analyzed in multiple quench cycles with all binary combinations of control voltages, bias currents, quench slopes and frequencies. The random counter bit patterns in multiple quench cycles of enabled SR-BP-filter shows that the SR-BPfilter is triggered by noise and spectrum is free across the tuned frequency of oscillator to build cognitive UWB radio network. The redundant counter bit patterns in multiple quench cycles of enabled SR-BP-filter shows that spectrum is occupied near to the tuned frequency of selected of oscillator and not available to build cognitive UWB radio networks. The operation principle of the SR- BP-filter tuned across 5.4GHz is shown in figure 4. The injected RF signal is the combination of three narrowband signals of frequencies

Control Bits

RF SPECTRUM SENSING MECHANISM

Oscillator Response

IV.

5.25GHz, 5.45GHz and 5.65GHz. The control voltage of the SR-oscillator is increased step by step, and the tuned frequency of the oscillator rises consequently in discrete time steps with the selected bias current level, quenching frequency and slopes. The tuned frequency of the oscillator is 5.2GHz in the first step, 5.4 GHz in the third step and 5.6GHz in the fifth step. During the first step, the oscillator is triggered by the RF signals of 5.25GHz, which is closed to the tuned frequency of 5.2GHz. Similarly, the 5.45GHz and 5.65GHz frequency component triggers the oscillator in the third and fifth step, respectively. This shows that the free and occupied narrowband RF spectrums can be detected in discrete intervals of time by changing the tuned frequency of the super-regenerative oscillators. RF Signals

by a low frequency quench VCO. The high quenching rate is suitable to detect UWB pulses, and low quenching patterns for detecting narrowband radios [6]. Sixteen different bias current levels, B0-15 are provided in order to adjust the critical bias level. The enable bit, EN is used for selecting the SRfilter in discrete time intervals. The filter is disabled by providing voltage VQ higher than threshold level of transistor M1, and the current IQ lower than the critical level. The envelop detector is used to sense the amplitude and width of the oscillations. The output of the envelope detector is compared to an adjustable reference level. The output of the comparator is used to enable and disable the high speed counter. The counter is getting reset in each quench cycle. The digital representation of the oscillation width is achieved by digitally counted values, which are processed by the digital correlation block (DCB).

Fig. 4. Operation principle of SR-BP-filter tuned across 5.4GHz

V. EXPERIMENTAL RESULTS The test structure of the proposed RF UWB spectrum sensing mechanism has been implemented in a 0.18um CMOS technology with an active area of 2.5mm2. The die photo is shown in figure 5. 8-Bit serial to parallel interfacing (SPI) registers are used to transfer the control bits to configure the SR-BP-filter, quenching slopes, tuned frequencies, and bias currents of the oscillators. The quench VCO, counter, digital correlation and control unit are implemented in an external microcontroller. The selectivity and sensitivity of each filter is measured by comparing the BER of injected data patterns modulated by sinusoidal signals. Different behavior for each filter is observed at a fixed level of bias current, quench frequency and slope due to different Q-factor of on-chip LC tanks. Nearly similar results of bandwidth and selectivity are achieved by adjusting the bias current, quenching frequency and slopes for all filters. In the current controlled quenching scheme, SR-BP-filters have relatively small bandwidth. The bandwidth of 25-to-80MHz across the tuned frequency of 7GHz is measured for for input signals having power levels from -90dBm to -20dBm at an average level of all

508

combinations of bias currents, quenching slopes and frequencies, as shown in figure 6(a). In the voltage controlled quenching scheme, the bandwidth of 180-to-250MHz is measured for input signals having power levels from -90dBm to -20dBm at an average level of all combinations of bias currents, quenching slopes and frequencies, as shown in figure 6(b). The changes in bandwidths and sensitivity levels are observed by varying the quenching frequency with a fixed bias current and quenching slope.

stages and ADCs for a cognitive UWB radio network, multiple SR-BP-filters are proposed in this work to sense the complete RF spectrum. Digitally controlled tuned oscillators with adjustable bias currents, slopes and frequencies of sawtooth quenching patterns are provided to adjust the selectivity and sensitivity of each SR-BP-filter. Occupied and free spectrum spaces are identified after manipulating the digital response of each filter by digital correlation block. 9090 80

Bandwidth (MHz)

Bandwidth (MHz)

270270

(a)

80

70 70

6060 5050 4040 3030 2020 -90 -80 -70 -60 -50 -40 -30 -20 -90 -80 -70 -60 -50 -40 -30 -20

250250 230230 210210 190190 170170

150150 -90 -80 -70 -60 -50 -40 -30 -20 -90 -80 -70 -60 -50 -40 -30 -20 RF Input Power (dBm)

RF Input Power (dBm)

Fig.5. The die photograph of test structure having twelve SR-BP-filters with an active area of 2.5mm2

Fig. 6. Measured bandwidth of SR-BP-filter tuned at 7GHz with current (a) and voltage (b) controlled quench schemes

In the current controlled quenching scheme, the bandwidth declines from 100-to-15MHz for -90dBm signals and from 125-to-30MHz for -20dBm signals by decreasing the sawtooth quenching current rate form 100MHz to 500 KHz. This is due to the improved quality factor in the filtering mode by small variations of the quench current across the critical level [11]. In the voltage controlled quenching scheme, the bandwidth is increased from 250-to-500MHz for -90dBm signals and 460-to-700MHz for -20dBm signals by decreasing the rate of the sawtooth quenching voltage from 100MHz to 500KHz. This is due to increased time duration available for changing the equivalent conductance from high to low level in the filtering mode. The minimum sensitivity increased to -60dBm with increasing the quenching current frequency to 2MHz and the quenching voltage frequency to 4MHz. These results shows that SR-BP-filters are suitable for the detection of narrowband and wideband signals in current controlled and voltage controlled quenching schemes. In order to detect the free and occupied spectrum place within the UWB spectrum, the measurement setup requires a bank of multiple wideband and narrowband radio transmitters within the spectrum of 3 to 10GHz, with adjustable power levels, bandwidths, and configurable modulation schemes. Eight narrowband radio signals with ASK and FSK modulation schemes are injected with UWB impulses of around 1GHz bandwidth, presented in our previously reported UWB receiver design [12]. The digital correlation unit detects the narrowband radios and UWB impulses with increased bandwidth of 8MHz and 25MHz across the center frequency of the injected signals respectively. More efficient and complex DSP algorithms can be implemented in the digital correlation block to achieve more precise results for detecting free and occupied places in RF UWB spectrum. VI.

(b)

CONCLUSION

An RF spectrum sensing mechanism is proposed for low power, low cost and medium data rate cognitive UWB radio networks. Instead of using reconfigurable LNAs, mixers, IF

REFERENCES [1]

Federal Communication Commission, “Facilitating Opportunities for Flexible, Efficient, and Reliable Spectrum Use Employing Cognitive Radio Technologies”, NPRM & Order, ET Docket No. 03–108, (Dec.

30, 2003). [2]

D. Cabric, S. M. Mishra and R. W. Brodersen, “Implementation issues in spectrum sensing for cognitive radios”, Proc of Asilomar conference on signals, systems and computers, Nov 2004, Vol 1, pp 772-776. [3] H. Arsalan, M.E. Sahin, “Cognitive Radio, Software Defined Radio, and Adaptive Wireless Systems”, pages 355-381, Springer 2007. [4] N.A Moseley, E. Klumperink, B. Nauta, "A Spectrum Sensing Technique for Cognitive Radios in the Presence of Harmonic Images," New Frontiers in Dynamic Spectrum Access Networks, 2008. DySPAN 2008. 3rd IEEE Symposium on , vol., no., pp.1-10, 14-17 Oct. 2008 [5] G. Cafaro, N. Correal, D. Taubenheim, and J.Orlando, “A 100MHz– 2.5GHz CMOS tranceiver in an experimental cognitive radio system”, SDR Forum Technical Conference, 2007. [6] M. Pelissier, D. Morche, P. Vincent, "Super-Regenerative Architecture for UWB Pulse Detection: From Theory to RF Front-End Design," Circuits and Systems I: Regular Papers, IEEE Transactions on , vol.56, no.7, pp.1500-1512, July 2009 [7] P. Thoppay, C. Dehollain, and M. Declercq, “A 7.5mA 500 MHz UWB receiver based on super-regenerative principle,” Solid-State Circuits Conference, 2008. ESSCIRC 2008. 34th European, pp. 382–385, sept. 2008. [8] J.-Y.Chen, M.Flynn, andJ.Hayes, “A fully integrated auto-calibrated super-regenerative receiver in 0.13μm cmos,” Solid-State Circuits, IEEE Journal of, vol.42, no. 9, pp.1976–1985,sept. 2007. [9] B. Otis, Y. Chee, and J. Rabaey, “A 400μW-RX, 1.6mW-TX superregenerative transceiver for wireless sensor networks,” Solid-State Circuits Conference, 2005. Digest of Technical Papers. ISSCC. 2005 IEEE International, pp. 396–606 Vol.1, feb. 2005. [10] F.X .Moncunill-Geniz, P. Pala-Schonwalder, O .Mas-Casals, "A generic approach to the theory of superregenerative reception," Circuits and Systems I: Regular Papers, IEEE Transactions on , vol.52, no.1, pp. 54- 70, Jan. 2005 [11] M. Anis, R. Tielert, N. Wehn, "Fully integrated super-regenerative bandpass filters for 3.1-to-7GHz multiband UWB system," VLSI Design, Automation and Test, 2008. VLSI-DAT 2008. IEEE International Symposium on , vol., no., pp.204-207, 23-25 April 2008 [12] M. Anis, R. Tielert, "Low power UWB pulse radio transceiver frontend," Solid State Circuits Conference, 2007. ESSCIRC 2007. 33rd European , vol., no., pp.131-134, 11-13 Sept. 2007

509