On the Benefits of Random FDMA Schemes in Ultra Narrow Band Networks Minh-Tien DO ∗† , Claire GOURSAUD ∗ , and Jean-Marie GORCE ∗ ∗ Centre of Innovation in Telecommunications and Integration of service, INSA-Lyon, Lyon, F-69621, France † Sigfox Wireless, Bˆ atiment E-volution, 425 rue Jean Rostand, Lab` ege, F-31670, France Email: minhtien.do@sigfox.com; claire.goursaud@insa-lyon.fr; jean-marie.gorce@insa-lyon.fr Abstract—Ultra narrow band transmission (UNB) systems have already been deployed and have proved to be ultra-efficient for point-to-point communications. This paper presents this technology and gives some insights on the scalability of UNB for a multi-point to point network. This configuration corresponds to an uplink scenario where multiple nodes compete to send their packets, with neither coordination nor feedback from the sink. In particular, we present and analyze two multiple access schemes based on random frequency selection: discrete random FDMA (DR-FDMA) and the new continuous random FDMA (CR- FDMA). An ideal system where the carrier frequencies are exactly obtained is first considered and extended to a more realistic case, with rough carrier frequencies. We analyze the system performance in terms of bit error rate and outage probability. The presented results clearly show that, even if in the ideal case, the DR-FDMA scheme outperforms the CR-FDMA scheme; in the realistic case, both schemes lead to similar performance. Thus, this paper highlights the fact that the use of CR-FDMA is very relevant in a realistic network as it bypasses the need of an accurate carrier frequency control, and thus permits the use of even the cheapest transmitters without loss of performance. I. I NTRODUCTION Wireless sensor networks (WSNs) are increasingly being used in a wide field of applications in various domains, from health-care to network control and monitoring [1] [2]. In such networks, each node has a small amount of data to transmit (e.g. in applications such as temperature monitoring, electrical metering etc.). Thus, the main issue in WSN is not the indi- vidual capacity but rather finding an optimal resource sharing approach to either reduce the energy consumption or maximize the global capacity. In the case of a very large amount of nodes which compete to transmit their data to a common sink, many research works have been devoted with multi- hop and cooperative transmissions. In this paper, we present a different approach referred to as ultra narrow band (UNB). This technique is based on a highly asymmetric system, where the transmitters used UNB individual transmissions over a relatively total large band. The central base station (BS) uses a wide-band receiver to gather and decode all UNB signals transmitted by the distributed sources. UNB technology is not a new concept. However, to ob- tain a very narrow band, H.R. Walker proposed in 1997 to use Very Minimum Sideband Keying (VMSK) modulation [3] and claimed to satisfy extremely high spectral efficiency (beyond Shannon’s theorem). However, it was demonstrated that the claimed performances can only be obtained for perfect components [4]. So, we do not consider VMSK, but use more realistic modulation: BPSK for UNB technology to sat- isfy cost-effective and bandwidth-efficiency for low-throughput network. One of the main advantages of UNB technique is the reduced occupied bandwidth that induces a reduced noise contribution. Thus, the reception power sensitivity is very low, producing a very large coverage area using a single sink or base-station (BS) (more than 50 km in open field). With such an extended coverage, a large amount of source nodes can be served. Thus, the medium access procedure is probably the most critical issue. Usual approaches based on reservation techniques are not efficient in regard of the low quantity of information to be transferred and would lead to a waste of time for protocols or synchronization issues [5]. Therefore, random access protocols are more appealing, as they present more flexibility to manage bursty and random transmissions. Furthermore, in UNB networks, we aim at reducing the cost and the complexity of the source nodes even at the price of an increasing complexity of the receiver. Therefore, random access methods are interesting since they do not require a feedback loop to trigger the transmission. The well-known drawback of random channel access is the collisions. Interference might take place when several nodes are transmitting at the same time in the same frequency band. Commonly used protocols consider the transmission time as the random variable [5]–[8]. In this case, the transmission frequency is fixed, and the nodes either try to send their packets at randomly chosen times and re-transmit in case of transmission failure (ALOHA based protocols) [5], or obtain information before transmitting with CSMA (Carrier Sense Multiple Access) or ISMA (Inhibit Sense Multiple Access). This approach is not viable in our setting. However, it is also possible to see the transmission frequency carrier as a random variable, which comes in complement to the transmission time. In this case, the nodes can perform their transmission at any randomly chosen time and frequency. To the best of our knowledge, only few recent works considered a random access protocol based simultaneously on both time and frequency selection [9]–[12]. However, all these works focused on a discrete set of frequencies which may be compared to our DR-FDMA approach. By extension, we could also consider [13] where the authors proposed to split a given bandwidth into sub-channels, and where users are randomly affected to each sub-band. Our contribution is thus an extension of these recent works, adapted to UNB networks, with no feedback loop and where the individual transmission bandwidth is negligible with respect to the total band. Besides in our work, the individual bandwidth is not by default constrained by the spacing between the carriers. Furthermore, the CR-FDMA has never been studied and present in the proposed context where tight frequency synchronization is hardly exactly obtained (due 978-3-901882-63-0/2014 - Copyright is with IFIP WNC3 2014: International Workshop on Wireless Networks: Communication, Cooperation and Competition 672