Parameter settings for 2.4GHz ISM spectrum measurements

Parameter settings for 2.4GHz ISM spectrum measurements Daniel Denkovski, Mihajlo Pavloski, Vladimir Atanasovski, Liljana Gavrilovska Ss. Cyril and Methodius University Faculty of Electrical Engineering and Information Technologies Skopje, Macedonia {danield, mihajlop, vladimir, liljana}@feit.ukim.edu.mk

Abstract— Spectrum sensing is an important consideration in the future cognitive radio networks. In order to protect communicating users, it is essential to make a reliable and correct occupancy decision. A possible way in providing a more accurate sensing measurement is to adapt measurement parameters to the inspected band and inspected signal. Parameters settings for resolution bandwidth and sweep time are technology dependent and should be different for detecting various signal types in the frequency domain. This paper presents a practical measurement setup using two types of spectrum sensing equipment: a high precision spectrum analyzer and a medium precision USRP2 software defined radio. Different parameter sets are used for the both devices, and results are compared in terms of duty cycle of the occupancy in the measured bands. All the measurements are performed in the Industrial Scientific and Medical (ISM) band. Keywords- spectrum sensing, resolution bandwidth, sweep time, duty cycle, ISM band;

I.

INTRODUCTION

Nowadays the spectrum usage is controlled by the frequency allocation policy bodies. Issues like spectrum access, signal interference, user protection are just few of the covered tasks. With the emergence of new wireless technologies, spectrum organization appears to be a very important matter. Policy bodies, network operators and even users themselves need a greater insight at the situation of used spectrum at any moment of time. Various sensing techniques are developed as enablers of this demand in the last decade. Spectrum sensing represents the process of measuring a piece of spectrum and making a decision of its usage based on measured data. Future cognitive networks envision use of the sensed information for opportunistic access of “secondary” users in a given band. An interesting issue in the field of spectrum sensing measurements is the ability to detect and recognize different signals belonging to specific wireless communication technologies. Moreover, sensing approaches where measurement parameters are adapted to specific signal pattern provide more reliable and accurate results. Spectrum shared bands, like the Industrial Scientific and Medical (ISM) band, appear to be interesting for measurements due to the great number of coexisting technologies.

The scope of this paper is to propose a measurement approach in the 2.4 GHz ISM band. Different measurements performed in this band are using specific parameter sets and two types of equipment: a high precision device – signal analyzer and medium precision devices – USRP2s. Presented results should give directions on optimal settings for the most realistic spectrum occupancy derivation, for both cases. This paper is organized as follows. Section II gives a brief overview and comparison of previous spectrum measurement efforts. Section III illustrates our measurement setup and explains both medium and high precision cases with interest in measurement setup and equipment capabilities. Section IV includes comparison of results by different measurements in terms of duty cycle distribution, threshold selection, measurement parameter settings etc. A brief discussion about lessons learned and conclusions for suitable parameter settings is provided in section V. Finally, section VI concludes the paper. II.

RELATED WORK

Spectrum occupancy measurement is an interesting area of research today. Current measurements usually utilize energy detection as a spectrum sensing technique [1], since it has proven to offer satisfactory results regardless on the simplicity of the referred approach. Duty cycle is the most common metric used to qualify the spectrum occupancy, which represents the percentage of time in which a specific portion of the spectrum is considered as busy. Basically two types of measurement campaigns exist: wideband spectrum measurements which sense the spectrum occupancy in larger bands covering several different technologies; and narrowband measurements focused on specific technologies and bands. Wideband spectrum measurements [2-6] are in the focus in most of the work in this area, observing the spectrum occupancy in larger time periods. Their goals vary from simple evaluation of used spectrum over time, inspecting not occupied spatial and temporal spectrum gaps, proposing spectrum occupancy models etc. Most of the long term measurement campaigns are based on high precision devices such as spectrum analyzers [7] [8]. The narrowband measurements can be rarely found in the literature. Besides the high precision devices they usually employ low and medium precision devices

such as software defined radios (like USRPs [9]), custom made sniffers etc. Mentioned campaigns give insight in many aspects of the measurement process itself. However, there is a lack of works which include full explanation of the measurement methodology, parameter settings etc. Furthermore, the aspect of measurement adaptation to specific technologies and bands should be considered in more details. The work in this paper encompasses couple of different aspects mentioned earlier in this section. The 2.4 GHz ISM band is inspected using energy detector in repeated short term measurements. Adapting sensing parameters to specific signal waveforms provides a more realistic picture of the spectrum thus maximizing the detector’s performances. III.

MEASUREMENT SETUP

The measurement setup employed in this study focuses on spectrum measurements in the 2.4GHz ISM bands. The main target of the setup is to test different parameters settings for the measurements and observe how they reflect in the spectrum occupancy results for different technologies. The second target is to check whether a medium precision device can be used to perform this type of measurements with acceptable precision. Several parameters characterize device sensing capabilities. The sensitivity of a spectrum sensing device is an important feature, since it defines the lowest power level that can be detected. A Spurious Free Dynamic Range (SFDR) is a metric that specifies how low power signals are detected in presence of strong ones. Furthermore the frequency and time resolutions are also crucial for the performance of a sensing device. The measurements results presented in this paper are based on a high precision spectrum analyzer Anritsu MS2690A [10] and USRP2 [11] (Universal Software Radio Peripheral 2) software defined radio hardware custom made sniffer solution. The spectrum analyzer is capable of spectrum measurements up to 6 GHz and gives options to choose one of the two types of sensing modes, sweeping and real-time. Sweeping mode includes inspecting of each frequency point successively, while the real-time mode performs the measurement on a set of frequency points at the same time. However, the real-time sensing mode can handle a bandwidth of 25 MHz at the most, what makes this mode inapplicable for the 2.4 GHz ISM band measurements. This spectrum analyzer has high frequency and time resolutions. It has a sensitivity value of -155 dBm and SFDR value of 88.5 dBm, which ranks it in the middle class of spectrum analyzers. In the measurement setup the spectrum analyzer uses transceiver VERT2450 ISM antenna in the range 2400 MHz – 2480 MHz. It is connected to a standard laptop via Ethernet cable, and Matlab is used to remotely setup the device and to request and store the measurement results. USRP2 is the successor of USRP (Universal Software Radio Peripheral), a software radio hardware enabled to use different daughterboards covering different frequency ranges. RFX2400 daughterboard is a transceiver daughterboard that operates in the 2.3 – 2.9 GHz band and is used in this setup. This daughterboard has noise figure in the limits 6-10 dB and AGC receiver gain range 0 – 70 dB. The precise values for the spurious free dynamic range of the USRP2 are daughterboard

specific and need further examination. GNU Radio [12] is the open source software package used by the USPR2 offering the possibilities to implement various types of spectrum sensing methods. The USRP2 device is connected to the same type of antenna as the analyzer. Using Ethernet connection to a second laptop, the USRP2 receives instructions and saves data. Both the described setups are placed in an indoor location at the faculty in the vicinity of couple WLAN access points. The USRP2 based sniffer is developed as a C++ application that includes the GNU Radio USPR2 specific blocks. It enables the USRP2 to work as a wideband energy detector sweeping over the full band on interest. The sweeping mode of operation of the sniffer enables various discrete values for the resolution bandwidth, from 190 KHz up to 25 MHz, or decimation and interpolation factors of the USRP2 in the range of 4–512. Furthermore, this sniffer solution can perform joint FFT analyzes of the received samples, in that manner multiplicatively increasing the resolution bandwidth of the USPR2 hardware. It can handle various sizes of FFT (multiples of 2), however, FFT with more than 16 bins is not recommended since the time needed for calculation increases exponentially. When FFT is used the time samples are windowed with Hamming Window to reduce the spectral leakage. Additionally, FFT overlapping of 25 % is used to reduce the non linear response of the digital down converter (DDC) at the edges of the FFT analyzes. This FFT feature can be wisely used to enable the USRP2 to work with smaller sweep times, since the time required to shift between frequency points is relatively high – 2 ms, and consequently when handling a large number of frequency points the sweep time cannot be kept low. Before the actual measurements of the spectrum occupancy in the 2.4GHz ISM bands, measurements were performed in order to calculate the noise floor of the both setups for several parameters of interest. For the both devices, the spectrum analyzer and the USRP2 hardware, load matching was performed with 50 ohms terminator and sufficient amount of noise data was captured. The post-processing of the noise data for the both devices included the calculation of the mean noise level, as well as the signal decision thresholds – probability of false alarm threshold, several dBs above mean noise level threshold. The calculation of the probability of false alarm threshold implies the estimation of a value that is being exceeded from certain percentage of the noise samples. IV.

MEASUREMENT RESULTS

This section will provide measurement results on the spectrum occupancy in the 2.4GHz ISM band conducted using the previously elaborated measurement setups. All measurements are performed in several sessions of 30 minutes each. Every session includes simultaneous measurements with the spectrum analyzer and the USRP2 hardware with similar measurement parameter settings. At the end of each session, a Bluetooth exchange is initiated to test its impact on the parameters selection. Measurements were conducted for different pairs of resolution bandwidth and sweep time. All duty cycle results in the 30 minutes time periods are calculated after the actual measurements using several decision thresholds (represented through the probability of false alarm).

Figure 1. Spectrum analyzer duty cycle results for a) RBW=300KHz, ST=1s; b) RBW=300KHz, ST=4s; c) RBW=1MHz, ST=1s; d) RBW=1MHz, ST=4s; e) RBW=300KHz, ST=4s, last 5 min; f) RBW=300KHz, ST=4s, last 5 min

A.

Signal analyzer measurements

1) Resolution bandwidth Fig. 1 depicts duty cycle results (obtained with the spectrum analyzer) for resolution bandwidths of 300 KHz and 1 MHz and sweep times of 1s and 4s. Each graph in the figure contains 3 duty cycle curves calculated for different probability of false alarm, 2%, 10% and 20%, respectively. It is evident that the value of 1MHz for the resolution bandwidth (the graphs on Fig. 1c) and Fig. 1d)) allows more suitable detection of wideband signals (such as WLAN signals). The rationale behind is that the measurement process in this case needs a smaller number of points (i.e. 22) in order to sweep the entire WLAN bandwidth catching most of the transmissions and offering acceptable resolution. Additionally, higher resolution bandwidths yield increased duty cycles for the most utilized channels (in this case WLAN channel 11). On the other hand, the usage of 300KHz resolution bandwidth accommodates better signals with narrower bands (i.e. Bluetooth signals). This is obvious on the graphs in Fig. 1e) and Fig. 1f) that plot the last 5 minutes of the sessions (the time when the Bluetooth communication is initiated). The comparison between the 300KHz and 1MHz resolution bandwidths shows that the lower resolution bandwidth results

Figure 2. 3D representation of the moments of Bluetooth transmissions for RBW=300KHz and ST=4s

in duty cycle curves that can distinguish Bluetooth transmissions in WLAN networks surroundings. Fig. 2 depicts the moments during the actual Bluetooth communication for resolution bandwidth of 300KHz and sweep time of 4s. It is clear that the Bluetooth transmissions are distributed as thin peaks along frequency and time, while the shape of the WLAN transmission is preserved and can be also noticed. 2) Sweep time The influence of the different sweep time values on the signal detection in the 2.4GHz ISM band is also given in Fig. 1. The higher sweep time of 4s, Fig. 1b) and Fig. 1d), results in non-realistic duty cycles (approaching 100% for all frequencies) for higher probabilities of false alarms (Pfa = 10% or 20%). Namely, as the sweep time increases, the noise has more time to pass the threshold. If this threshold is low (allowing statistically more noise samples to be detected as signal), then the false alarms would be more frequent. Otherwise, using lower sweep time of 1s, Fig. 1a) and Fig. 1c), the duty cycle results for higher probability of false alarm thresholds are acceptable. Also, the higher sweep time of 4s results in more realistic duty cycle results for the more traffic overloaded channels (channel 11 in the case on Fig. 1) since low sweep time yields higher probability of missing a WLAN transmission in the most utilized channels.

Figure 3. Decision threshold criteria selection for USPR2 sniffer

Figure 4. Results comparison between USRP2 hardware (RBW=400KHz) and spectrum analyzer (RBW=300 KHz)

B. USRP2 measurements Same measurements were performed with the USRP2 equipment. Fig. 3 depicts the duty cycle criteria selection for this type of measurement devices. It is evident that the probability of false alarm criteria is not suitable for USRP2 because of the accuracy of the RF part of the USRP2 hardware (higher noise uncertainty of the RFX2400 daughterboard of the USRP2). Therefore, appropriate criteria for duty cycle calculation in this case would be to qualify the samples as real signals if they are 10 dB above the noise level of the system. This approach significantly improves the duty cycle results of the USRP2 (Fig. 3). C. Comparison of both setups The comparison of the results between the spectrum analyzer and the USRP2 (Fig. 4) shows that the USRP2 based sniffer provides duty cycle curves comparable to the spectrum analyzer ones. The used resolution bandwidth of the USRP2 is 400KHz, higher than the spectrum analyzer’s 300KHz with the same number of measurement points, i.e. 200. The settings are different for both devices as they operate with discrete values for the bandwidth sizes and an existing match for lower bandwidth sizes cannot be clearly identified. Fig. 4 shows that the USRP2 offers acceptable results for the duty cycle compared to the precision of the spectrum analyzer. However, in the case of lower sweep time (i.e. 1s) the USRP2 is outperformed since it needs half of the sweep time (0.5s full, 2ms actual tune delay) for switching between frequency points and the other half for the actual measurements. This may result in an inability for the USRP2 to handle very low sweep times when using low resolution bandwidths. This drawback can be overcome with the FFT option of the USRP2 increasing the resolution of the USRP2 device. The performances of the USRP2 using this feature compared to the spectrum analyzer in terms of duty cycle are depicted at Fig. 5. The figure shows 3 curves, the first one is spectrum analyzer duty cycle for resolution bandwidth of 100KHz and sweep time 1s, whereas the other two are the USRP2 based duty cycle curves for resolution bandwidth of 400KHz and FFT of size 4 and resolution bandwidth of 800KHz and FFT of size 8, respectively. Both USRP2 settings offer effective resolution bandwidths of 100 KHz and sweep times of 1s. Fig. 5 shows

Figure 5. Duty cycle results of the USPR2 sniffer using FFT feature in comparison with spectrum analyzer for RBW=100KHz and ST=1s

that both USRP2 FFT cases outperform the spectrum analyzer in terms of the duty cycle measurement. The reason lies in the lower need of number of points to cover the full span of 80 MHz for the USRP2. Therefore, the USRP2 spends more time on each point catching more of the transmission. The best duty cycle results are therefore gathered in the case of FFT size 8. V.

LEASONS LEARNED

Section IV presented preliminary spectrum measurements results in the 2.4GHz ISM band in terms of duty cycle of the spectrum usage. Measurements were conducted using both high precision and medium precision spectrum sensing devices. Based on these measurements, Table I provides a brief summary of parameter settings impact on the duty cycle performance. The comparison between the high precision spectrum analyzer Anritsu MS2690A and the USRP2 hardware shows that the medium precision USRP2 hardware can be used to perform spectrum measurements with acceptable precision. Moreover, the FFT feature of the USRP2 sniffer allows for even better duty cycle results than the spectrum analyzer. However, possible usage of various spectrum analyzer enhancements, such as amplifiers, filters etc., should significantly increase its sensitivity and the precision of the spectrum measurement results. TABLE I.

PARAMETER SETTINGS SUMMARY Small ST

• • Small RBW

•

• • Large RBW

•

larger Pfa values for decision criteria better for Bluetooth signal detection smoother duty cycle for less utilized WLAN channel larger Pfa values for decision criteria better for WLAN signal detection better for highly utilized WLAN channels

Large ST

• • •

• •

smaller Pfa values for decision criteria better for Bluetooth signal detection better for highly utilized WLAN channels smaller Pfa values for decision criteria better for WLAN signal detection

VI.

CONCLUSION

This paper presents 2.4 GHz ISM band spectrum occupancy measurements conducted with a high end spectrum analyzer and a medium precision USRP2 device for several various parameter settings. The results compare the performance of USRP2 custom designed spectrum sniffer with the benchmark ones obtained with the spectrum analyzer. It is evident that a suitable choice of measurement parameter settings may provide sufficient accuracy of the medium precision USRP2 based device. Moreover, the results, in terms of duty cycle, give guidelines on different measurement parameters selection in different surroundings. Future work will include expanding the methodology in this paper for different spectrum bands (both licensed and unlicensed), medium and long term wideband measurements, possible enhancements of the USRP2 sniffing solution in order to optimize its performances etc. ACKNOWLEDGMENT Parts of this work were funded by the EC through the FP7 project FARAMIR (248351) [13]. The authors would like to thank everyone involved.

[4]

[5]

[6]

[7]

[8]

[9]

REFERENCES [1]

[2]

[3]

T. Ycek and H. Arslan, “A survey of spectrum sensing algorithms for cognitive radio applications,” IEEE Communications Surveys & Tutorials, vol. 11, no. 1, pp. 116–160, 2009. M.Wellens and P.Mähönen, “Lessons Learned from an Extensive Spectrum Occupancy Measurement Campaign and a Stochastic Duty Cycle Model”, Springer Mobile Networks and Applications, 2009. M. Islam, C. Koh, S. Oh, X. Qing, Y. Lai, C. Wang, Y.-C. Liang, B. Toh, F. Chin, G. Tan, and W. Toh, “Spectrum Survey in Singapore: Occupancy Measurements and Analyses” in Proc. of International Conference on Cognitive Radio Oriented Wireless Networks and Communications (CROWNCOM), Singapore, May 2008.

View publication stats

[10] [11] [12]

[13]

M. López-Benítez, F. Casadevall, A. Umbert, J. Pérez-Romero, J. Palicot, C. Moy, and R. Hachemani, “Spectral occupation measurements and blind standard recognition sensor for cognitive radio networks,” in Proc. 4th Int’l Conf. on Cognitive Radio Oriented Wireless Networks and Communications (CrownCom 2009), Jun. 2009, pp. 1–9. A. Petrin and P. G. Steffes, “Analysis and Comparison of Spectrum Measurements performed in Urban and Rural Areas to Determine the Total Amount of Spectrum Usage,” in Proc. of International Symposium on Advanced Radio Technologies (ISART), Boulder, CO, USA, March 2005, pp. 9–12. McHenry, M. A., Tenhula, P. A., McCloskey, D., Roberson, D. A., and Hood, C. S., “Chicago spectrum occupancy measurements & analysis and a long-term studies proposal,” in Proc. of Workshop on Technology and Policy for Accessing Spectrum (TAPAS), Boston, USA, Aug 2006. M. López-Benítez, F. Casadevall, A. Umbert, J. Pérez-Romero, R. Hachemani, J. Palicot and C. Moy, “Spectral occupation measurements and blind standard recognition sensor for cognitive radio networks,” in Proceedings of the 4th International Conference on Cognitive Radio Oriented Wireless Networks and Communications (CrownCom 2009), Special session on Flexible and opportunistic wireless access, Hannover, Germany, June 22-24, 2009, pp. 1-9 M. López-Benítez and F. Casadevall, “Methodological aspects of spectrum occupancy evaluation in the context of cognitive radio,” in Proceedings of the 15th European Wireless Conference (EW 2009), Aalborg, Denmark, May 17-20, 2009, pp. 199-204. M. López-Benítez, F. Casadevall and C. Martella, “Performance of spectrum sensing for cognitive radio based on field measurements of various radio technologies,” in Proceedings of the 16th European Wireless Conference (EW 2010), Special session on Cognitive Radio, Lucca, Italy, April 12-15, 2010 MS2690A Signal Analyzer Product Brochure, Anritsu, Anritsu EMEA Limited, Luton, Beds, LU1 3LU, UK Universal Software Radio Peripheral 2 (USRP2). Information available at: www.ettus.com. S. Valentin, H. von Malm, and H. Karl. “Evaluating the GNU software radio platform for wireless testbeds” in Technical Report TR-RT-06273, 2006 EC FP7 project FARAMIR, Information available at: http://www.ictfaramir.eu.]