Poster abstract:Outdoors Range Measurements with Zolertia Z1 Motes and Contiki

Host Publication: Proceedings of the 5th International Workshop, REALWSN 2013

Authors: M. Paule Uwase, L. Nguyen, J. Tiberghien, K. Steenhaut and J. Dricot

Publication Date: Sep. 2013

ISBN: 978-3-319-03070-8

Abstract:

1. Motivation To study experimentally the interactions of the Radio Duty Cycling (RDC), MAC and Routing over Low power and Lossy Networks (RPL) protocols for wireless sensor and actuator networks (WSAN) a test facility is being built in the garden of a countryside house because observations will be easier to interpret far from potentially disturbing radio transmitters and from reflecting structures causing multi-path interferences. The facility consists in a WSAN, with a dozen static motes, positioned all over a 1000 m2 lawn, one mobile mote, affixed on a robotic lawn mower, and one gateway mote, located near a veranda, where computers and test equipment are housed. The position and the transmission power of the static motes should force the RPL routing algorithm to select as many as possible multi-hop routes. As a survey of the relevant literature [1,2,3,4,5,6] did not provide sufficient practical data to implement this last requirement, we undertook themselves a practical study of radio links , inspired by the experimental work reported in [5] and [6], but using modern motes available on today's market. We believe some of our observations can be useful for anybody designing a WSAN. 2. Experimental set-up Most of the experiments with WSANs in our environment are done with Zolertia Z1 motes [8] running Contiki [9]. As there was no good reason to change this habit, we decided to use the same motes for this project. The Z1 exists in several versions: the simplest one has a built-in antenna, while another has an external antenna. Both were investigated. The quality of a radio link can be characterized by three indicators. Two of them are evaluated by the radio receiver itself: the analog Radio Signal Strength Indicator (RSSI) and the digital Link Quality Indicator (LQI) [7] estimated by the demodulator. The third is directly relevant to the applications: the Ratio of Packets correctly Delivered (PDR). Studies showing that these three indicators do not correlate very well can be found in [8] and [9]. Therefore, whenever relevant, the values of the three will be given. To evaluate these indicators, a point to point unicast link was set up, using the communication functions provided by Contiki rime. The motes were mounted on 1m wooden poles by means of nylon straps. The battery powered transmitter could freely be positioned anywhere in the garden and given any orientation. The receiver was permanently positioned at approximately 1m from the veranda and was connected by a USB cable to a PC recording the received messages. All range measurements were done along one straight line, with no obstacles between sender and receiver, but some additional measurements were done along two other lines for checking influences of unsuspected causes. For avoiding interference with possible WiFi networks all tests were performed in channel 26, the channel which is least susceptible to WiFi interference. The absence of interfering radio fields was monitored during the experiments by means of a SPECTRAN HF񮶬 portable spectrum analyzer from Aaronia. Radio traffic was monitored by a Texas Instruments packet Sniffer with a CC2531DK dongle. 3. The measurements To test a link, the sender transmitted every second a 45 bytes packet. For 50 consecutively received packets the values of RSSI, LQI and the sequence number of the packet were recorded by the PC connected to the receiver. The global values of RSSI and LQI were obtained by averaging the recorded values, while the PDR value was computed by dividing the number of received packets by the number of sent packets. The latter was obtained from the sequence number included by the sender in each packet. For these tests, the radio was continuously left on. 3.1. Directivity measurements The first test aimed at comparing the motes with built-in antenna and those with an external vertical ?/4 whip antenna. Preliminary tests showed that for links exceeding 1m, the motes with built-in antenna needed to be held vertically. For evaluating the horizontal directivity, the properties of the radio link were measured with 8 different orientations of the sender with respect to the receiver. This clearly showed that the internal antenna of the Z1 motes is far from omnidirectional (20 dB difference between the best and worst direction) and that transmission in front of the mote is especially poor. It was suspected that the presence of batteries in front of the internal antenna was the cause of this poor performance, but removing the batteries proved this hypothesis wrong. As the internal antenna would be a source of unwanted and hard to explain artifacts, it was decided to use exclusively motes with an external vertical antenna. As these antennas appeared to be horizontally omnidirectional and to have a main vertical radiation pattern covering approximately 20 degrees, they seemed to be the best choice for a versatile test facility. 3.2. RSSI calibration The next test consisted in verifying the accuracy of the RSSI indications by the receivers. One by one, all motes, under identical conditions, received 50 packets from a single sender. The average values of RSSI for each mote were compared. They all were within a 2 dB margin, except for one mote that gave a 20 dB lower value. The 2 dB margin complies with the CC2420 specifications [7] but the 20 dB were considered an indication of malfunctioning. The antenna or its cable was suspected but substitution proved the problem was elsewhere. 3.3. Range measurements First, the maximal range was evaluated at 0 dBm in an open field. It was found to be 160 m. Thereafter the quality of radio links was measured several times in three different directions, over distances ranging between 3 and 21m with steps of 3 m and with four different levels for the transmitted power (0 dBm, ᆞ dBm, ᆣ dBm and ᆫ dBm). An almost monotonous, but far from predictable, decrease of the RSSI values in function of the distance could be observed, but large differences between measurements done in different directions and/or weather conditions confirmed the affirmation that RSSI values cannot be used to estimate distances [10]. The LQI values did not show a similar pattern: often they were better at medium distances (12m) than nearby (3m). However, most observed values were above 90 which corresponds to a good digital link. The quality of the digital link was confirmed by the PDR values, which were 100% for all links, except for one 21 m link with ᆫ dB sending power were RSSI (ᇱ dBm), LQI (60) and PDR (<10%) clearly showed that this was out of range. 3.4. The influence of Radio Duty Cycling When range measurements were started in 2012, naively the default configurations of the motes and their Contiki operating system (power = 0 dBm, MAC = CSMA, RDC = contikimac, 8Hz [13]) were used. Despite good RSSI and LQI figures, abnormally low PDR values were obtained. Disabling the power saving radio duty cycling (RDC) by replacing the contikimac driver by the nullrdc driver brought the PDR to 100% in almost all experiments, as reported above. The very low RDC was found to be due to a timing bug in the Z1 version of the contikimac driver. The bug was reported in the appropriate forum [14] and corrected. This resulted in a PDR close to 100% for RSSI levels better than ᇟ dBm, but degrading rapidly beyond that limit (with a RSSI of ᇤ dBm, PDR is as low as 16%). This degradation is probably due to the Clear Channel Assessments threshold set at ᇤ dBm used to detect arriving packets. The degradation due to power savings was explored further by adding to the range measurements described above, tests with the three different RDC protocols readily implemented in Contiki (contikimac, Xmac [15] and Low Power Probing (LPP) [16]. With Xmac, no degradation of the PDR could be observed, even with an RSSI as low as ᇨ dBm. The case of LPP is quite different: below ᇆ dBm, messages are sometimes duplicated or triplicated because the LPP driver in the sender ignores some acknowledgments received and forces the CSMA MAC layer to retry or to drop the packet. This is clearly due to a timing issue, because the problem disappears when the DEBUG option is enabled in the LPP driver. 4. Conclusions The Zolertia Z1 motes can ensure reliable radio communications over links with attenuations that do not exceed 90 dB. In an open dry grass field, this corresponds to a distance of some 180m when the vertical whip antennas provided with the motes are used. Wet grass or moderate rain can cost up to 15 dB and a single leafy bush on the path costs 15 dB , meaning that radio links exceeding 150m should be avoided. The Z1 mote is also available with an internal antenna. Compared with the whip antenna, this costs between 25 and 45 dB, depending on the relative orientation of the motes, practically restricting the motes with an internal antenna to desktop experiments. The above measurements were conducted with the radios always on, and resulted in PDRs very close to 100%. Enabling energy saving RDC protocols changed this figure, especially over long links, but we suspect this is due to some inadequate tunings of parameters built into the Contiki implementation of these protocols.