Just over a year ago I was briefly involved with the final stages in the deployment of a wireless sensor network. During the project I had many thoughts about how the radio system could have been different. The key problem was that requirements for the comms channel were somewhat unknown. At the time I had recently finished a Software defined Radio project based around the TMS320. Since then, my thoughts have often drifted back to the quandary of just how flexible can you make a radio module for a sensor network. For now I am really considering the physical layer, the network topology is an entirely different kettle of fish.
Not only are there unknowns pre-deployment but things change with time. For example, the channel may change with weather – more water, frozen water, snow cover. Or perhaps physical environment – Increase in (wet?) foliage, change in physical location – node attached to a rock in a landslide? Or a moving animal?? So what if the nodes/data sink could change system variables on the fly to negotiate for optimal communications at minimum power usage? This would of course lead into developing algorithms for choosing what actually is the best at any one time, how much overhead should one invest to optimise the channel….
The requirements initially appear quite simple (Of course not the case when you get into things). The key requirement is low power consumption. This off course relates to output power and data rate. For example, a high power, high data-rate burst may be more cost (power) effective than a longer lower power transmission. Most of the radio systems one sees deployed in commercially available WSN’s are fixed frequency, occasionally offering a choice of output power and modulation. So what are the variables that on might be interested in tinkering with?
- Frequency/Bandwidth – Pretty straightforward, generally lower frequencies are attenuated less per distance through substances – rock, ice, water etc. Higher frequency – greater data rate, smaller antennas
- Modulation – More complex modulation schemes verses resilience to noise.
- Data-rate – maximum a result of other variable choices, although features in Bit error rate somewhat.
- Transmit power – self explanatory really.
- Antenna type/direction/matching – how about a choice of an omni antenna then being able to select one from a choice of increased gain (Directional) antennas? Clearly this is linked to frequency and physical node size
What other physical layers are there? Particularly thinking of non free-space situations. Light, Acoustic wave and Inductive for starters.
Induction has two particularly exciting properties. Unlike EM Waves, its attenuation through water containing media (Eg soil/Ice) remains relatively constant Irrespective of the water/particle size. Secondly, the inductive load of the receiver can be detected, providing the transmitting station some indication of a presence of a receiver.
Even more promising for non-free-space environments is acoustic transmission. Commonly used underwater already, there is already a wealth of research. Unfortunately, little appears to show concern for power consumption. Some numbers plucked from a random paper: 500Kbps over 60m and 4Kbps over 4Km, again this is with unknown power usage. My instinct is that its probably massively too high. More investigation needed.
As a brief aside, one area that remains relatively unexplored to me is that of Wide Band Pulse, It looks quite exciting for this application. Apparently good transmissive properties through walls etc More research needed.
