12th International Symposium on District Heating and Cooling

The <strong>12th</strong> <strong>International</strong> <strong>Symposium</strong> on District Heating and Cooling,September 5 th to September 7 th , 2010, Tallinn, EstoniaTo obtain IPv6/6LoWPAN functionality in the Mulles,the lightweight operating systems Contiki [14] andTinyOS [15] have been successfully ported to the Mulleplatform. Both operating systems were specificallydesigned to be compatible with resource limitedembedded systems such as Mulle. Moreover, Contikiand TinyOS both support IPv6 and 6LoWPAN.However, TinyOS was selected for this study becausestability issues due to edge-routing problems withContiki.Sensor platform energy usageObtaining an acceptable life expectancy is one of thebiggest challenges to battery powered, wirelessdevices. In Sweden, heat meters are inspected every 5to 10 years, depending on the size of the meter. Thelife expectancy of wireless devices should beequivalent to the inspection period to avoid frequentand expensive battery replacements. All sensor nodesdo however not need to be battery powered. In thecase of available electric power in close proximity, e.g.for platforms mounted in pumps or valves there is noexplicit need for batteries since there are electricityavailable. At other sensor platforms, battery power isthe only feasible solution, for instance outdoortemperature sensors.To determine the amount of energy used by a wirelesssensing device, the current at the sensor platformassociated with IPv6/6LoWPAN communication wasmeasured. To measure the current used by the device,a 1 ohm high precision resistor was connected in seriesto the Mulle power connector. The voltage dropgenerated across the resistor was amplified 100 timeswith a MAX4372H amplifier circuit. Using an analogacquisition card, the amplified signal was measuredand stored in an ordinary PC. Due to poor precision atvery low current, complementary measurements wereperformed with a high precision ampere-meter todetermine the current usage of the Mulle, when it wasin deep sleep mode.To evaluate the energy cost of transmitting datapackets with UDP on IPv6/6LoWPAN, packets withpayload sizes between 1 and 100 bytes weretransmitted, and the expected lifetime of the sensorwas calculated. Fig. 8 displays the expected lifetime ofa sensor with a 500 mA battery and a 15 minutetransmission interval. Out of curiosity, both TinyOS andContiki were programmed to transmit UDP packets ofdifferent sizes at consecutive time intervals to observeany differences in energy usage between the two. Theresults indicated that the energy usage of 50 to 80-bytepayloads in Contiki and Tiny OS were significantlydifferent. The observed difference between operatingsystems is most likely related to the method of headercompression. Specifically, Contiki uses HC1, whileTinyOS is based on HC01. However, both methods area part of the 6LoWPAN standard. Additionally, TinyOSuses short addressing, while Contiki employs longaddressing. The type of addressing and headercompression used by the OS can be changed, but inthis particular test, default settings were used.For payload sizes greater than 60/90 bytes, the IPpacket had to be divided into two separate 802.15.4frames because the maximum frame size of IEEE805.15.4 is 127 bytes. The separation of IP packetsincreased energy usage and decreased the expectedlifetime of the sensor. Thus, software developersshould consider the maximum frame size if absolutemaximization of sensor lifetime targeted. Howeverincreased payload sizes can of course becompensated with a larger battery.As shown in Fig. 8, the fixed transmission interval wasset to 15 minutes, and the effect of transmissioninterval on the expected lifetime of the sensor wasanalyzed. Additionally, sensor lifetime was evaluated atvarious transmission frequencies and a fixed payloadof 80 bytes, as shown in Fig. 9. In accordance to theorythe results indicated that a low transmission frequencyhas a positive effect on sensor lifetime. In the case ofcontext aware sensors, which only transmit data whenrequired e.g. when a measured temperature exceeds aset threshold, sensor life expectancy will in most casesbe increased. However, the impact of thesleep/standby energy usage will make up a largerpercentage of the total energy usage, which hence willmean that the importance of keeping the sleep currentlow will be even bigger.Fig. 8. The effect of payload size on the expected lifetimeof a sensor platform at a transmission rate of4 transmissions per hour (1 to 100 bytes).9