[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] Open Source optical networking

... "it's probably the cheapest wireless system ever" ?

Acquiring a digital data sample from many sensors requires on the order of 1 nJ.  Threshold detection at discrete time periods will require substantially less energy in most cases.  Higher performance sensors will require more energy per sample, but the nJ/sample number is applicable to, for example: whisper-to-chainsaw acoustic, sub-degree accuracy temperature, milli- to kilo- gravity acceleration sensing (Which also provides tilt and vibration information), magnetic field to 0.1% of earth's field, barometric pressure to 5m, wind flow to 1 m/s, relative humidity to 2%, ambient light level and spectrum.

Transmitting a bit of data over 10-100 meters by RF today takes approximately 100nJ with Bluetooth, Wavelan, and other local area RF networks.  Transmitting a kilometer takes at 10 to 100 microJoules.  These numbers are not likely to fall much, since they are often pushed up close to the fundamental physical limits.  Another order of magnitude may be available by sacrificing immunity to unintentional jamming from nearby transmitters. If the dynamic range of the radio receivers is reduced, substantial improvements inpower can be realized.

Collimated line-of-site optical communication systems will transmit 10m with an energy cost of 10pJ/bit, more than 10,000 times lower than existing radio technology.  We have demonstrated 1nJ/bit in the lab already.  This incredible gain over RF is due entirely to an antenna gain of roughly 7 orders of magnitude when going from an isotropic radiator to a 1 mrad divergence beam.

32 bit computation currently costs around 1nJ/instruction on power-optimized microprocessors.  Engineering limits in the next 5 years or so are approximately 1pJ/instruction for dedicated hardware.

Good batteries provide roughly 1 J/mm^3 .  Solar cells provide approximately 100uW/mm^2 in full sunlight, more than 100nW/mm^2 in average room lighting.  Vibrational energy availabe in an office setting is in the nW/mm^3 range.  RF power in a simple antenna is generally not useful, unless there is a cell phone in use in the room, or a dedicated RF power source, in which case microWatts can easily be generated.  Conversion is difficult, but feasible.

Assuming a simple task of sampling a sensor, performing some relatively simple processing (threshold, FIR/IIR filtering, statistical analysis, or FFT), listening for incoming messages, and transmitting a simple outgoing message, the energy cost will be a few nanoJoules.

Combining this with the power source information, a cubic millimeter battery will provide enough power to perform such a simple task once a second for 10 years.  A cubic millimeter vibrational energy rectifyer will operate at that rate forever. Indoors a square millimeter solar cell will provide enough power to perform 100 tasks/second, or in full sunlight 100,000 tasks/second.

For indoor optical line of sight communication, a cubic millimeter battery will provide enough energy to transmit 50 billion bits (roughly half a dozen full-length movies).
< http://robotics.EECS.Berkeley.EDU/~pister//SmartDust/in2010 >

An open-source CITRIS R&D project based on "Smart Dust" and a TinyOS
http://www-bsac.EECS.Berkeley.EDU/~mattlast/research/index.html >

Video Semaphore Decoding for Free-Space Optical Communication
http://www-bsac.EECS.Berkeley.EDU/~mattlast/papers/spie_ei_2000.ppt >

Jack Park wrote:


Ronja is an Open-Hardware optical datalink that connects two PC's
point-to-point. Ronja's design is licensed under the GNU Public License:
you get all the necessary documentations and construction guides free. The
construction costs are minimal, it's probably the cheapest wireless system
ever. The operation is very reliable and immune to interference.

10 mbs, full duplex 0.5 to 1 km distance

I might use one of these myself!