Sandaysoft

Hi Scott

I've just spent 3 months developing a station that will accept data from a number of sensors, including one wire, and then display that data in Cumulus. This is based on a microcontroller to gather the data, process it into usable form, then present it by the USB interface to Cumulus. It probably has a couple weeks effort yet before it's ready for prime time, mostly because I received a bad one wire device and I'm waiting for a replacement. There is curently no wireless link, although the architecture allows for adding one in future if I ever think it's worth the effort. It does have onboard sensors for the indoor measurements, and a built in data logger the same size as the Fine Offset. It has been happily displaying data in Cumulus for a couple weeks now.

I started the project after my own Fine Offset frustrations, and with discussion and support from "41South". If fast and easy is what you are looking for, then rather than trying to build your own, send me a private message and we can discuss getting a copy of it constructed for you.

I can tell you that I wouldn't want a weather station I can't use with Cumulus, and it's unlikely Steve will ever support home made stations unless they use an existing interface. I can also tell you the project was non-trivial with plenty of surprises.

Final point, re my original comment: one wire only brings transport layer to the table - you still need to code for all the sensors. If you use WIFI for the transport layer, then what does one wire add?

I like one-wire because I can add and drop sensors with minimal wiring, without touching more processor pins. (I'm prone to frying whatever I touch, so I want to touch very little.) It's purely convenience; I want to get two pins working and be done with messing with the processor forever after, except throwing code on it.

Alright, so... having priced one wire components for everything, I think I'm giving up and moving away from one-wire for windspeed, rainfall, and wind direction. Windspeed and rainfall are just digital pulses, which the processor I have (netburner 5270) won't have any problem debouncing and counting.

Wind direction looks dicier. It's an array of 8 different resistances, up to around 100k. The cheap way is to use the resistances to charge a cap, and measure the time it takes the cap to get high enough in voltage to look like a logic 1 to the processor. The problem with that approach is that caps drift as the temperature changes... this won't work.

So instead I'm considering using the resistance from the wind wave to control the duty cycle of a 555 timer. Messing with some numbers, I can probably get a measurement in 15ms, worst case. (Wait for the 555 to go high, measure the time high, measure the time low, divide and look up the answer.) Since duty cycle is independent of the capacitor value, it doesn't matter much if the cap value varies with temp.

Has anyone tried this? Is there a better way? The one-wire solution in this case is around $58 - too much.

Hi,

If you're considering the FO wind vane, that's exactly what they do. But they pre-calibrate the capacitance against a known resistance first.

The issue is that the resistances don't run in sequence (and there are actually 16 values because two reeds can operate at the same time) so some form of "lookup table" is required. A small microcontroller is the obvious solution, it's exactly what I'm planning to do (with a PICaxe) in my current project, maybe soon to be revealed in the Homebuilt section.

Cheers, Alan.

>If you're considering the FO wind vane, that's exactly what they do.

I already have the FO weather vane, so I guess I'm using it.

How do they account for cap drift? reed switch debounce? The vane moving in the middle of the measurement? Or do they just hope for the best? Do they use a 0.1uf cap (Which means the worst case read time is probably around 11ms?)

I suppose I'll just remeasure until I get 3 equal values in a row. I was hoping for something better.

Hi Scott,

Most of my information came from about page 22 of Gina's mammoth thread on spikes in the external data caused by Electric Fences.

I believe the capacitor is 4.7uF, the micro first charges this through a fixed resistor (10k?) taking about 50ms for calibration, then through the external temperature thermistor (also about 50ms at normal temperature) and finally through the wind vane, which takes from about 3ms up to 500ms. If the vane is disconnected it times out after about 1 second. The vane is so "lively" that reed switch bounce must be irrelevant, but if the micro reads a value that it doesn't recognise, then it (or maybe the console) appears to hold the previous reading for one cycle and then displays "--" subsequently.

But with a PIX(axe), for example, you can just feed the vane from the supply via a resistor of about 3k - 4k ohms and, even with only 8 bits A/D conversion, easily differentiates the values. My application indeed involves "processing" the raw vane data and re-encoding it into FO's A/D converter.

Cheers, Alan.

It doesn't sound like this idea ever went anywhere, but it sounds like it might be the only way to get around the FO station lockup issues.

Has anyone made any progress on this? I know nothing about electronics, but am interested in this project. I don't even care about the base station, really - I just want the data to be logged by Cumulus.

I've just recently restarted my weather station project. I'm using an Arduino, a wifi shield, and some one wire sensors, plus the rain gauge, wind vane and anemometer from the FO station. Not the cheapest solution, but I have the wifi and temperature working and it's been simple so far. I expect the wind vane to be the fussy part; the Arduino has an analog input and I'm hoping to use a simple voltage divider to measure it. We'll see...

I'm building an Arduino plus 1-wire based station. Arduino handles 1-wire with a library - I have another thread on it - though been too busy to progress it much lately.

_________________
Gina

Sorry, no banner - weather station out of action. Hoping to be up and running with a new home-made one soon.

I've had success with the one wire library, though very occasionally one of the sensors decides not to report back. (It's possible the poor Arduino can't quite handle everything I'm throwing at it, but it's more likely my prototype's wiring isn't perfect, and things will improve when I use a real circuit board.) The wifi shield works well at the distance I have in mind, and the nice thing about real wifi is real TCP - so both the station and the receiving computer can tell when data isn't flowing, and take steps to restart the connection.

My plan is to put the rain gauge on one interrupt (debounced at 100ms - the rain gauge can't possibly tip faster than that), the anemometer on the other (debounced at 15ms? I guess - experimentation will tell me more), and the wind vane on A0, using the vane as a voltage divider against some fixed resistance. Everything else (light meter, temperature, humidity) is one wire. I don't have a solution for pressure yet, but then there's no reason why the pressure sensor can't be inside and a completely separate solution.

When I get a solution I like, I'll post schematics. There won't be much to them; it's just a few resistors needed to get lines pulled up and down as needed.

I have everything "working" in the sense of electrical signals and software to process them; but I'm unhappy with the way I'm doing wind direction. I have a voltage divider with a constant 45k pulling up to +5v, and the wind vane's varying resistance pulling to ground. The Arduino only sees a 2 out of 1023 difference (18/1023, 20/1023) between 2 adjacent directions, in the worst case. That's too close. Is anyone using a value better than 45k?

Also, while I get windspeed numbers from a spinning anenometer, I don't believe the number I get.

Hi Scott,

45k is far too high to use with the FO wind vane. The resistance values (note that they do not run in sequence) vary from about 690 ohms up to 120 kohms. The "optimum" pull-up (or pull-down, as appropriate) resistor is about 4k (because there are three values of 0.69k, 0.89k and 1k at the "low" end, but only 65k and 120k to resolve at the "top" end).

In what way don't you believe/like the windspeed readings? Have you tried any "controlled" tests?

Cheers, Alan.

I don't have any controlled tests available. In my ad hoc test, the peak windspeed and average windspeed seem too different, but I'll experiment more and check my math.

I'm trying a 5k resistor, but ENE and E are still very close together for a ten bit A/D, even with that. Not that big a deal, really; just annoying.

Hi Scott,

Yes, FO didn't do a very good job in "designing" the Wind Vane resistor values. There are two positions where the nominal difference between adjacent positions is only about 12% (one is where the 8k2 comes in parallel with 1k to make 0.89k, and diametrically opposite 120k comes in parallel with 16k to make 14.1k). If 5% tolerance resistors are used then the differential may be very small. So the "optimum" series resistance could be as low as about 3k.

For amusement, I designed my own set of 8 preferred (E12) values which spans a smaller resistance range but still gives a nominal minimum 20% separation between the 16 adjacent positions, e.g. 1k5, 6.8, 12, 56, 82, 18, 8.2 and 3.3 (but easily scaled up or down as required). However, not much use unless you're building/modifying your own vane (and decoder).

Cheers, Alan.

The F.O. resistors are 1% tolerance and my measurements show them as actually being much closer than that. But my measuring equipment is only accurate to 0.1% so can't say closer than that.

_________________
Gina

Sorry, no banner - weather station out of action. Hoping to be up and running with a new home-made one soon.