IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, VOL. 3, NO. 4, DECEMBER 2017 731 RF-Spectrum Opportunities for Cognitive Radio Networks Operating Over GSM Channels Miguel Luís , Member, IEEE, Rodolfo Oliveira , Senior Member, IEEE, Rui Dinis, Senior Member, IEEE, and Luis Bernardo, Member, IEEE Abstract—In this paper, we characterize the radio frequency spectrum opportunities available in a common global system for mobile communications (GSM) channel to support the opera- tion of a cognitive radio network (CRN). In a first step, we describe the technical details involved to sample the channel using a software defined radio device. Adopting a simple energy-based detector, we identify the two energy regions where the GSM system is active or inactive and evaluate the spectrum sensing accuracy. Based on the output of the detector, we show that the distribution of the durations of busy and idle periods are approximated by geometric distributions. Finally, we validate a theoretical model for the distribution of the service time. The validation results indicate that the service time can be success- fully represented by a discrete generalized Pareto distribution, which is confirmed by the Kolmogorov–Smirnov test. Because the throughput of the CRN is represented by the inverse of the service time, the proposed analysis provides an upper bound for the networks’ throughput, indicating the maximum throughput that can be attained when a single secondary user transmits over a GSM cellular channel. The results presented in this paper are validated with real data, confirming the accuracy of the proposed service time model. Index Terms—Cognitive radio networks, service time, performance evaluation and modeling. I. I NTRODUCTION I N COGNITIVE Radio Networks (CRNs) the transmission channel is licensed to the primary users (PUs), while sec- ondary users (SUs) only access the channel in an opportunistic way when the PUs are inactive, i.e., when the PUs do not use the channel [1], [2]. Because the channel is used by the SUs opportunistically, a SU transmission must be halted whenever a PU becomes active. In a scenario where a SU needs to trans- mit multiple packets (e.g., in a file transmission), or when a packet may be too long, the amount of time required to fin- ish the SU’s service (Service Time) depends on the number Manuscript received April 30, 2017; revised August 25, 2017; accepted October 28, 2017. Date of publication November 8, 2017; date of current version December 22, 2017. This work was supported by the Foundation for Science and Technology of the Portuguese Ministry of Education and Science under Project UID/EEA/50008/2013. The associate editor coordinat- ing the review of this paper and approving it for publication was O. Holland. (Corresponding author: Miguel Luís.) M. Luís is with the Instituto de Telecomunicações, 3810-193 Aveiro, Portugal (e-mail: nmal@av.it.pt). R. Oliveira, R. Dinis, and L. Bernardo are with the Instituto de Telecomunicações, 1049-001 Lisbon, Portugal, and also with the Departamento de Engenharia Electrotécnica, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal (e-mail: rado@fct.unl.pt; rdinis@fct.unl.pt; lflb@fct.unl.pt). Digital Object Identifier 10.1109/TCCN.2017.2771558 and duration of the PUs’ transmissions. By definition, the ser- vice time is the interval of time from the instant when the data arrives at the head of the SU transmitting queue (e.g., a packet or a file, depending on the network stack layer), until the instant when its transmission ends. Service time is an important metric in CRNs because it incorporates the level of activity of the PUs. In this work we characterize the ser- vice time of a cognitive radio network operating in a GSM channel. The rest of this section describes related literature and high- lights the contributions of our work. Section II describes the RF spectrum sensing procedure, by introducing the experi- mental setup and describing the steps followed to determine the sensing accuracy. Section III characterizes the duration of the periods when the GSM channel is found idle or busy due to the GSM operation. Based on the statistics of the exper- imental data, Section IV summarizes the steps involved to numerically compute the distribution of the service time. In Section V we assess the accuracy of the numerical results by comparing them with experimental data. Finally, conclusions are drawn in Section VI. A. Related Work Over the last years, with the introduction of the CR con- cept, a high number of publications have focused on the characterization of transmission opportunities in GSM chan- nels. Recently, the work in [3] evaluated the possibility of using GSM whitespaces in rural areas for dynamic spectrum sharing, therefore supporting the growth of community cel- lular networks, and consequently, improving the rural access to communications services. Gao et al. [4] analyze the prac- tical capacity limits that can be achieved by an opportunistic network when using the 850 MHz GSM uplink frequency. The capacity limits were evaluated under different constraints such as the instantaneous interference power limit to the GSM basestation and the SU transmit power limit. The work in [5] addresses the possibility of exploiting the underutilized GSM channels for secondary communications assuming two distinct communication strategies: i) an interweave communi- cation strategy, where the unlicensed network can only use the channel when it is found vacant, and ii) an underlay strategy, allowing concurrent primary and secondary trans- missions. After comparing the performance of both strategies through simulation the authors focus their work on the access strategies to be adopted by the secondary network under 2332-7731 c  2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.