Photos of the insides of Fine Offset sensors.

[Charlie]

The way this sensor works is that a magnet attached to the rotating vane turns on one or two of the reed switches. It is meant to detect wind directions between the 8 compass points by relying on the magnet turning on two adjacent reed switches. That is thought to explain the rather weird sequence of resistor values associated with the corresponding switches. The chip in the main transmitter unit converts the resulting resistance into a digital value which is then turned into a number between 0 and 15 in the firmware.

I've been trying to puzzle out how this works with a spreadsheet. When adjacent switches close, the resistor value between the terminals varies in some cases by as little as 200 ohms, ans in other cases by as much as 42K! By my calculations, as the vane turns, the value at the terminals jumps up and down, sometimes a little, sometimes a lot. I honestly don't see how you can acurately measure it, at least not within a couple ms.

Mine is way up on the roof, so I can't double check the diagram Gina posted, although I'm sure she has it right. Any guesses how this works in real life?

[AllyCat]

Hi Charlie,

The resistor (or two in parallel) has one end switched onto the supply rail (at the start of each A/D conversion cycle) which charges up a capacitor (4.7uF). Then the microcontroller measures the time it takes for the voltage to reach half the supply rail. This takes between a few and about 500 milliseconds. There is a "timeout" if this voltage is not reached and the sensor/console reports a disconnected/faulty vane (no wind direction shown).

Presumably the micro uses a lookup table to relate wind directions to A/D time delays. Note that there are also some intermediate "dead spaces" between certain resistances (or time delays) when again the console displays "no direction". The precise value of the electrolytic capacitor and voltage threshold are not important because the micro pre-calibrates them against a (more accurately) known resitance value.

IMHO the FO resistor values may well have been "designed" by trial and error/guesswork. But it's an interesting challenge (because the values must wrap around in a circle) to design the most efficient or tolerant set of resitor values for any particular number of directions and resistor tolerances.

Cheers, Alan.

[Charlie]

Hi Alan,
I understand THAT part of how it works - my issue is with the resistor values and the possible acuracy of the sampling.
For example, the difference between E and W is 1K to 120K, so very easy to detect. However, ENE=891, E=1000, ESE=688. Distinguishing between these is a LOT harder, especially if you need a range that also goes to 120,000!
Maybe some of the Fine Offset direction stability does not have to do with the vane at all, but rather the ability to discern which way it's pointing...

[Philip]

Here is a quick table of values I did when I was looking at this subject,
values are in ohms.


SW1 R1 33K N 33000 N NE 6567.96
SW2 R2 8K2 NE 8200 E NE 891.30
SW3 R3 1K E 1000 E SE 687.50
SW4 R4 2K2 SE 2200 S SE 1406.56
SW5 R5 3K9 S 3900 S SW 3135.68
SW6 R6 16K SW 16000 W SW 14117.65
SW7 R7 120K W 120000 W NW 42120.06
SW8 R8 64K9 NW 64900 N NW 21876.40


Philip

[AllyCat]

ENE=891, E=1000, ESE=688. Distinguishing between these is a LOT harder, especially if you need a range that also goes to 120,000

Hi Charlie,

The closest difference is about 12%, most resistors now are better than +/- 5% tolerance and +/- 1% are readily available if needed. The difference is about 0.1% of the full-scale value so it is going to need at least a 10 bit timer/counter. All micros have at least an 8-bit counter, often with an optional pre-scaler and some (as I suspect here) support 16-bits (65k or about 0.0015% resolution) or more.

The A/D converter needs to achieve this resolution anyway for the temperature sensing. The resolution is 0.1 degree C and the range over 100 degrees (-40 to +65), so better than 0.1%. But thermistors normally have a logarithmic resistance characteristic, so better than 0.01% resolution may well be needed.

When I was testing my magnetic modification to "equalise" the 16 wind direction sectors, I did notice that one sensor occasionally confused NW for W. That's the highest resistance, so I suspect that there was some current leakage or maybe the additional capacitance of my long extension cable was upsetting the timing. Otherwise the 16 directions on several units which I've tested have been decoded perfectly.

Cheers, Alan.

[aussiewmr]

For those of you that have WH3081 Weather Stations as used most probably in Australia, NZ, UK (EG Non USA) here are some photos of the Outdoor and Solar / Lux Sensors.

From what I can see the Wind Sensors are the same as those posted earlier in this thread.

The solar sensor photo here is not the one I modified as per a post here (Solar and UV Sensors going offline - fix?). There is little point posting that because the mod did not fix the issue I was having.

Cheers
Phil

[AllyCat]

Hi Phil,

Thanks for those very detailed pictures; from another thread it seems that you're aware of most of the components, but perhaps it's worth adding the details here.

The Lux sensor is under the white diffuser dome on the right, the microcontroller under the black blob nearby and the UV sensor is a tiny chip at lower left. The 8-pin package near the centre is a small serial flash memory, presumably for Lux/UV lookup tables and perhaps calibration for the specific components.

Personally, I suspect that the UV sensor is a standard UVA+UVB type (the "UV index" is fundamentally UVB) and should be mounted under a UVB-pass filter (which may explain the poor accuracy). Recently, I saw a patent which claimed that UVB could be inferred (calculated) from individual UVA and Visible light measurements but I don't know if this is correct or if FO attempt it.

Just as this solar module first "saw the light of day" purely as a PV cell to charge the transmitter batteries, FO appear have plans to extend its functionality further. The unpopulated area beside the L-shaped slot is obviously for a radio transmitter and I guess some of the pads on the right will accept a daughter-board, perhaps for temperature and humidity sensors. This, together with the pads for two tag-ended batteries (or a holder), could make a complete "external unit" (which might suffer less problems than the present "two-processor" arrangement). A socket for the rain sensor is already fitted so the main unknown is whether the wind vane and/or anemometer would be omitted, or perhaps attached by a "captive" cable from some of the other "spare" pads on the PCB.

Generally, the overall quality of the PCB assembly (soldering, etc.) seems far higher than previous FO units and it must be disappointing (for all concerned) that the current performance and reliability is so poor.

Cheers, Alan.

[aussiewmr]

Hi Alan,

Here's the detail of the light sensor with the diffuser removed. I think it's some sort of Photo Diode rather than the LDR I expected.

Sorry about the focus - its a bit hard to get closeups with the equipment I have.

Phil

[Andyman]

Hi all. The wires in my wind direction sensor have come off. Does anyone have any info on which colour wire goes where?

regards
Andy

[AllyCat]

Hi Andy,

Yes, I encountered the same problem, so I'm not sure of the original connections either. However, as the cable is a ribbon style there are really only two (logical) ways to reconnect the wires. Electrically, it shouldn't matter which way they are connected because the two inner wires go to the anemometer and the two outer ones to the vane. However, the A/D converter driven by the vane resistance is rather sensitive so there might be a subtle difference. Perhaps somebody else knows?

Sometime I'm going to dismantle another vane and will photograph it, but as you (and Gina in her photo earlier in the thread) have discovered the wires are quite fragile (and the adhesion of the tracks on the PCB not too good either) so I'm not going to hurry.

Cheers, Alan.

[ScottM]

This seems to be the Thread of Fine Offset Goodness, so I'm certain someone here knows the answer -

Are the rain gauge, wind speed and wind direction sensors all "Dallas One Wire" based?

I ask because I'm completely and totally fed up with my Ambient WS-2080 - the receiver loses contact with the sensor array routinely, and it's not like the distance between them is great. It's under 100' and there's a single pane of glass between them. I've had Ambient replace the display/receiver - no real improvement.

But now I'm thinking I can buy an inexpensive microprocessor board, a one-wire/USB converter, and write my own code to read wind and rain data. Add a one wire temperature and humidity sensor, and done. Give the microprocessor a USB wifi adapter, and now I have a sensor array data transmitted to my house wifi, which is where I want it anyway.

Feasible? And if the electrical interfaces to the various sensors aren't one-wire, what are they? Many thanks.

[AllyCat]

Hi,

No none of the FO sensors are one-wire.

The cabled sensors are based on magnetic reed switches, one pulse per tip of the rain see-saw, 2 pulses per anemometer rotation and an analogue value (resistance) derived from the rotational position of the vane. The external temperature is obtained from the resistance of a thermistor (or the 3080/1 may use a dedicated chip) and I believe humidity is from a variable-frequency oscillator - all measured by the microcontroller in the "transmitter" unit (which may also have a calibration table in a memory chip).

Most of the sensors work reasonably well (temperature spikes excluded) so IMHO the main problem is with the very low cost wireless modules. If you're lucky, even those can work moderately well, but many don't, certainly not near the maximum specification range.

Cheers, Alan.

[ksangeelee]

... thinking I can buy an inexpensive microprocessor board, a one-wire/USB converter, and write my own code to read wind and rain data. Add a one wire temperature and humidity sensor, and done. Give the microprocessor a USB wifi adapter, and now I have a sensor array data transmitted to my house wifi, which is where I want it anyway.

Feasible? And if the electrical interfaces to the various sensors aren't one-wire, what are they? Many thanks.

Rather than try to intercept at the sensors, you might consider doing it at the RF stage. On my 1080 transmitter, the board was laid out into two halves - sensor-reader and RF transmitter. There were only a few tracks leading to the RF transmitter section of the circuit, so I put a logic analyser on the wires, and sure enough there were signals being sent as regular long and short pulses (hi/lo) on these tracks - 88 bits of data as I recall.

I suspect you could hook up (either directly or via a PIC chip if necessary) something like a Hope-RF module to send similar signals at the frequency and modulation of your receiver (though I don't know how the RF on these units is modulated). Perhaps you can do some digging on the forums or inside your base-station to figure out what the transmitter is speaking to in order to figure out your next step.

As for a microprocessor board to WiF?, A Raspberry-Pi with a WiFi dongle would do the job - again I'd suggest hooking into the RF wires, but you'd need to spend some time decoding the low-level sensor-data format. Raspberry Pi boards should be on general sale within a couple of months.

All good fun, but in my experience, any hacking of devices like this usually takes a lot more effort than I first envisage.

[ScottM]

All good fun, but in my experience, any hacking of devices like this usually takes a lot more effort than I first envisage.

Hrm. Interesting, but it means voiding the warranty in a fundamental way. I want to keep the option of sending this junk back and getting my refund (to help defray the cost of better solutions.)

To reduce clutter on the board, let me move the discussion to
https://cumulus.hosiene.co.uk/viewtopic.p ... 83&e=62283
in Homebuilt.

I'll keep your suggestion in mind, though.

[Darrel]

Gina, very nice photos.
My experience with the vane directions is exactly the same as yours.

To emphasize Gina's comment about the intermediate directions (NNW, WNW ...) being poorly represented, I've attached a plot of the relative probabilities of different directions measured over several days. This is a wind vane from the Ambient WS-2080, which I believe uses the same Fine Offset sensor. As you can see, the chance of a 3-letter cardinal point being indicated is very small - about 3% - compared to the other directions. The system of using 2 adjacent reed switches being closed simultaneously, as an indicator of the intermediate direction, is not working very well. If it were possible to use a stronger magnet, or to move the magnet closer to the circuit board, it might be possible to equalize the direction probabilities.
Thanks,
Darrel.