Project History

The design started around 2006  with  the need to provide instant and autonomous weather reporting for remote airfields which don’t have an official reporting system.

An initial design goal was to have a high resolution wind sensor that can properly capture short burst wind gusts without integrating the peaks away yet still have enough resolution to be able to display wind speeds down to 0.5 miles/h at a 1 second measure interval.   The wind direction needed to have a resolution of  better than 2 degrees, not just the rough guess-o-meter approach  some commercial products provide . Also it needed to resolve the entire 360 degrees not have a dead zone around north  like some systems have.   Furthermore,  some sort of sky cover detection device was desired to gauge if there was severe blue sky or a cloud cover present, and of course the data needed to be instantly accessible as well as recorded  to see which way the weather is trending.

CIMG2920

Second prototype, still running

It was clear from the beginning that using the Weather Underground system was the way to go for  reporting, logging and playback .  Their system,  first off , being open to anyone who wants to send them data, and second being free of cost, was a perfect fit.  Except for the fact that you needed a windows PC running at all times as a gateway between the station and  the  internet to reformat the readings and transmit the data to Weather Underground.  A major hassle considering all the issues with seemingly never ending Windows and Virus  updates, not to mention the needless power consumption of  it basically just sitting there doing barely anything.

None of the existing weather stations  quite fit the bill , either because the wind data resolution was not there or  the cost was more than we wanted to spend.   Most stations had some wireless sensors that would report at a slow interval to conserve battery power and had a replaceable battery somewhere up and out of reach.  Having to chase empty batteries up on top of the building in the middle of winter was not appealing.   We needed a system that was hard wired, would automatically come back online after a power or network outage and be relatively cheap.

The first station was designed with these concerns in mind. I used a Microchip 8-bit PIC   micro controller board from an Australian outfit which had an 10BaseT Ethernet controller on it.  The initial sensors where purely analog and lacked factory calibration.

IMG_1106

RC motor based wind speed sensor

The wind speed sensor was constructed from the guts of an electric motor as used in  remote control airplanes.  The motor provided the bearing for the anemometer and 16 magnets that would provide 8 pulses per revolution. To build an anemometer I would buy a new motor, take it apart, strip it of any armature and mount the hall effect sensor in it’s place.  Quite a labor intensive undertaking, and none of the sensors where mechanically quite the same in the end .

CIMG2922
After initial experimenting with dual (expensive!) potentiometer and not being satisfied with the friction and the longevity of it I found a slick device from Austrian Micro-Systems that provides absolute rotation information while being totally friction less. It is based on a quad arrangement of hall-effect sensors over which a rotating magnet is arranged.  Reading the magnetic field strength and direction of each sensor the orientation of the magnet can be calculated with a bit of trigonometry.  Quite a slick device, albeit a bit pricey, but well worth it.

The original system is still up and running as of spring 2013. Much was learned from the installations and some software bugs were fixed. A couple of other installations were done to get feedback and the consensus was that the wired Ethernet connection was a pain when having to string a long wire to the corner of the house from the room the router was in. Someone pointed out the potential for disaster should the wind tree get hit by lightning.

Hard to argue with that, so I started the next incarnation based on a WiFi-Rabbit board from Digi International. I had to rewrite the firmware to run in this environment because the programing language they use is Not Quite C  and not quite anything else for that matter. Lack luster product support from Digi, a doubling in price and an eventual end of life announcement  prompted me to abandon that path after much frustration.

Being tired of paying someone else money for poorly supported boards that didn’t quite do what I had in mind I decided to bite the bullet and design my own board.  Going back to the Microchip devices  I could use the same TCP/IP stack I used in the first series and choosing a newer member of the PIC family meant I could get out  from under a slow 8 bit CPU with limited memory.

The 3rd try was a charm ( it almost always is)! I started with the rework  fall 2012 and had proto boards in hand by the end of the year.  Having 256Kb of Flash means that after I added all the initial features from my list I still have more than 50% space unused.  All sensors were reworked as well to be pre-calibrated kind and the magnetic rotor wind speed sensor was replaced by an optical encoder.  The absolute rotary encoder for the wind vane is the only sensor that is still the same since the first prototype.