Our sensors


Figure 1: The MicaZ motes
from Xbow
We are using one of two types of devices (called “motes”) in the sensor patches: the MicaZ from Xbow and the Telos mote from MoteIV. The MicaZ mote consists of a 7.3 MHz ATmega128L processor, 128KB of code memory, 4KB of data memory, and a Chipcon CC2420 radio capable of 250 Kbps and an outdoor transmission range of approximately 30m at ground level. The device measures 5.8cm×3.2cm×1.8cm and is typically powered by 2 AA batteries with an expected lifetime of days to months, depending on application duty cycle. The Telos mote has a similar configuration. The motes run TinyOS [TinyOS], a small-footprint operating system written in nesC [nesC], developed by UC Berkeley. While the range is about 30 meters at ground level, through a multi-hop routing algorithm they can pass on information along a chain of sensors until it is received by the patch gateway.

Figure 2: The Xbow StarGate SBC
The motes come with a set of on board sensors such as ambient temperature, light, and humidity sensors. While these are useful, they are not specifically designed around our needs: we are also interested in soil moisture content that cannot be measured with the existing sensors.

For the patch gateways, we are going to use the Stargate node [Stargate]. This is a small embedded device using the X-scale processor with 64MB of RAM and 32MB of Flash memory (Figure 2). The Stargates will be powered from an external power supply. These nodes will connect to the sensor network through a locally connected mote. Furthermore, Stargates have a 802.11 radio (WiFi) that will be used to relay collected measurements to the patch's Internet gateway.

Energy Consumption

We power the motes with inexpensive AA Alkaline batteries which store about 2100mAh of energy. During a 2 minutes cycle a node keeps the radio on for 1 second (drawing 23mA), samples the sensors twice (once per minute with each sample lasting half a second and drawing 0.6mA of current) and stays in low power mode for the rest of the time (0.16mA). The average current draw is therefore 0.35mA.

After 70 days operation we noticed that the batteries' voltage dropped to 2.6V, a drop of 0.4V from 3V. If we consider a battery cutoff voltage of 0.8V and a linear discharge model to approximate the amount of energy in the battery, as suggested in Energizer's website, a drop of 0.4V corresponds to consumption of 600mAh. This is very close to the 595mAh computed using the average current drawn by the mote. The difference is due to the power consumed during data downloads, a factor not included in our analysis. Our calculation holds even for smaller scales: the voltage drop during one week is almost 0.02V (as observed through our online monitoring tools corresponding to an expense of 60mAh using the linear battery model.

For the same period, our average current model estimates energy consumption of 59.47mAh, a 0.8% difference. The high accuracy of of this model indicates that it can be used as a planning tool for estimating the lifetime of a network for a given sampling frequency, or reversely for determining the highest sampling frequency given a network lifetime goal.

The MicaZ specification recommends operation voltage range from 2.7V to 3.6V but from our tests we determined that motes can reliably operate down to 2.2V when the flash stops responding. We also noticed that the mote and the radio still works till about 2.17-2.10V. Because the current set of batteries was installed at the end of November we expect the nodes to stop recording the data in mid April and to stop sending status messages one to four weeks later.

To increase the deployment lifetime we can either increase the sleep time and consequently lower the sampling rate; this will decrease the number of status messages, and will incur a bigger delay for the start of a download, or we can provide the system with more energy by using batteries with bigger capacity e.g. Alkaline D batteries which are rated at 21500 mAh.

The reception at the basestation located at the first floor of the nearby building was highly variable but remain very good and constant for long periods of time (days and weeks) with most of the nodes (6-7 out of 10). For the ones which we didn't have good radio connectivity we went nearby the building and download the data using a laptop.

Figure 3: The voltage drop over our motes in operation from Nov 2005 to Feb 2006.