http://tao.ndbc.noaa.gov/proj_overview/pubs/mil96paper.html
> "The board draws approximately 1 mW when in the sleep state, and around 250 mW when sampling sensors. The entire system averages 10 to 15 mW including power for all sensors, sampling, and Argos telemetry. By using switching power supplies where possible, the same battery space available in the original ATLAS tube will provide a conservative 18-month deployment life with a battery pack comprised of 84 D-cell alkaline batteries. The low power system allows the use of inexpensive batteries instead of solar cells or wind generators. These would attract vandals and add significantly to the acquisition and maintenance cost of each mooring."
This bouy is fascinating. It has a sub-surface set of sensors and check out the description of how those underwater sensors communicate with the main board of the bouy:
> "The sensors transmit and receive data from the buoy with an inductive coupling technique."
[...]
> "The tube drives the cable while the sensor drives the secondary coil on the inductive element. An FSK method of modulating the signal onto the cable is used with 2 kHz for a logic one, and 4 kHz for a logic zero. With a 2-kbaud data transfer rate, a one bit is 1 cycle and a zero bit is 2 cycles. By having data frequencies that are exact integers of the baud rate, the bits can be phase coherent at their edges, thus eliminating a source of jitter. The data uses standard ASCII hex format with one start and one stop bit."