Replacement wind vane

[Charlie]

I've been running my 1-wire weather station with Cumulus for about a month now and it's great to have eliminated most of the Fine Offset frustrations. I say "most" because I'm still using the Fine Offset wind and rain sensors with new electronics.

Yesterday, while doing a little gardening, I noticed the Fine Offset wind vane actually going round in circles. By broadening it's tail a couple years ago, I've damped the wild swings if the wind is over a couple KPH, but at really slow speeds, this does not help.

So can anybody recommend a relatively low cost replacement wind vane? It does not necessarily have to plug into the Fine Offset station as I'd be comfortable with building an interface for it, it just has to point in the right direction

[philcdav]

Hi Charlie.

I tried the 'extended tail' solution using pvc tape (curled in sun) and a solid sheet of thin plastic clipped on (blew off in wind)

Increasing the mass should be useful, Bluetack or lead weight riveted on maybe.

Some folk have very 'steady' values but I just ignore the randomness for now.

[Charlie]

Hi Phil,

I did a few mods that improved it dramatically, but honestly, with the 8-point/16-point issues, sticking bearing issues, stability issues, it's really trying to make a silk purse out of a sow's ear. I was hoping someone had found a cheap alternative that solved the world's problems, but based on the number of posts in this thread I'm starting to believe no such beastie exists. Back to the tool shed to do my own, I guess. I think it's harder than it looks to do it well, though.

[AllyCat]

Hi Charlie,

....and the FO choice of resistor values in the vane aren't exactly optimum either. But I have largely overcome the 16-point issue by adding a couple of magnets internally (reported elsewhere here - I'll look if you can't find it) and it might be worth checking/lubricating/replacing the bearing at the same time.

However, for another reason, I am planning to use a (PICaxe) micro to decode the wind direction (resistance) and then re-encode it for the transmitter (it's the re-encoding for the transmitter with its battery power limitations that is the hard part). But as part of that process it will be quite easy to read the wind direction more frequently and "filter" it by keeping a rolling average. This should overcome the unstable direction issue, which might then actually be an advantage, in using a kind of "Pulse Width Modulation" to resolve a more accurate direction between the major 8 (unmodified) or 16 reported directions.

Since I believe you already are using a micro, such a "software" solution might be an easy "fix".

Cheers, Alan.

[Charlie]

I'll certainly watch how you make out with interest, however, once I saw it actually going around in circles... well, "filter me that, Batman" LOL. Besides, there often is some real variation, and Steve's graph provide a nice "visual filter".

As to the resistor choices, they are not as bad as they first appear. I built a spreadsheet to play with a voltage divider from 5V supply to GND, using the wind vane to 5V and a resistor of various values to GND, and graphed the voltage for all directions. It seems a value of 5.7K for the reference is best, although anything from about 2K to 10K should work. Monte Carlo analysis shows with tolerancing, virtually all manufactured units will work and give enough difference to write software to use an A-D converter. I'm guessing whoever did the original calculation put a lot of effort into their choice, to enable using standard resistor values, but still distinguish direction in the worst case.

At any rate, there isn't a single suggestion for an alternative piece of hardware on this thread, which indicates there are none, so I might try to build my own. I do know it's harder than it looks - balance, centering, symmetry, all need to be nearly perfect - but I've got a bag of Hall Effect sensors in the corner that I got on sale...

[AllyCat]

Hi Charlie,

I'm not convinced that there is any fundamental difference between filtering with hardware (damping) or filtering in software (integration/averaging). The problem with the FO implementation is that it only reads an instananeous sample once every 48 seconds, so there are bound to be some "rogue" values logged.

I'm guessing whoever did the original calculation put a lot of effort into their choice, to enable using standard resistor values, but still distinguish direction in the worst case.

I don't disagree that the values selected are just about acceptable using standard (5%) resistor tolerances, but I don't believe they were designed well. I think it was Steve who pointed out that they're basically the nearest "preferred" (E12) values to 1k, 2k, 4k , 8k ... 128k. Some of those don't convert well and there are a couple of values (including the paralelled pairs) that differ by only 12 - 13%. Conversely there, are six gaps of over 50%, whilst generally only the two associated with the highest resitor value are essential (because it must be paired with one of two lower values).

I couldn't devise a "formula" to calculate optimum values (for given resistor tolerances) but in less than an hour with a spreadsheet came up with a better series based directly on E12 values. The smallest differential resistance is 20% (cf. 12%) but the highest to lowest resistance ratio is only about 80:1 compared with 175:1 for FO.

Being E12 values, the whole series is easily "scaleable", but starting with the same low value as FO, I got the following series (in k ohms) with the actual resitors in bold: 1.0 , 0.82 , 4.7 , 2.99 , 8.2 , 6.8 , 39 , 23.0 , 56 , 9.9 , 12 , 3.8 , 5.6 , 1.58 , 2.2 , 0.69 . However, I would probably have started with a 1k5 resistor (or higher) giving a lowest value still above 1k.

But yes, I came to the same conclusion that the optimum driving source resistance for A/D conversion of the FO vane is about 4k - 5k (I've used 4k7).

Cheers, Alan.