Temperature data spikes revisited

[DaveNZ]

Someone on an NZ weather forum had spikes happening on their station for temperature, rainfall, and windspeed. She moved a whole lot of eletrical/wireless stuff away from the console and it seems to be fixed
http://www.nzweather.net/forum/topic/10 ... on-please/

Don't know if it would help you but thought I should mention it

[Synewave]

That 'someone' on the NZ weather forum is JennyLeez (on this forum too). Perhaps she can help Gina?

[Gina]

I've been running the test station with anemometer turning (fan driven) and checked the waveforms at various points with the oscilloscope.

AWS_Wind_Sensor_Cct1_&_Waveforms.png

The waveforms are as before and as expected for the thermistor. The negative end of the timing electrolytic capacitor appears to be connected to the -ve rail - it has a constant 0v on it. So... the first pulse on the common connection reads the vane resistance and the second the thermistor resistance by measuring the time taken to charge the capacitor to about half voltage.

This confirms my earlier idea of how wind direction and temperature are measured. Next test will be with the long cable connected into the feed from the wind sensors.

[Charlie]

Are these really as expected? If so, then perhaps I'm confused about how it works. I was thinking that B and C are floating with A outputting high (3V) then A gets tristated into a comparator input, and B (or C) gets switched low. The amount of time it takes A to ramp down to the comparator reference voltage determines the resistor value. A quick recharge by tristating to an output "1" to recharge then do C (or B), then go back to sleep with A high.

I'd expect B & C to be square on the bottom (not rising) and aligned in time with the pulses on A. I'd also expect the pulse on A to discharge down from 3V in a curve, rather than switch completely low then charge up again.

Maybe I'm getting confused by the reference - what is the scope using as it's reference? Is it the negative rail? It almost looks like it's referenced to +3V and AC coupled...

[Gina]

I've redone the diagram to clarify it. I've also connected the scope "ground" to the -ve battery terminal and rechecked the pulses.

AWS_Wind_Sensor_Cct_&_Waveforms2.png

Are these really as expected? If so, then perhaps I'm confused about how it works. I was thinking that B and C are floating with A outputting high (3V) then A gets tristated into a comparator input, and B (or C) gets switched low. The amount of time it takes A to ramp down to the comparator reference voltage determines the resistor value. A quick recharge by tristating to an output "1" to recharge then do C (or B), then go back to sleep with A high.

I'd expect B & C to be square on the bottom (not rising) and aligned in time with the pulses on A. I'd also expect the pulse on A to discharge down from 3V in a curve, rather than switch completely low then charge up again.

Yes, I see what you mean. The pulses probably do line up as you say - with a "single beam" scope I can only show one input at a time and trigger off that. There isn't even a separate trigger input on this scope (it was relatively cheap). It looks like A, B and C chip connections may be all tristate. I'll see if I can work out what the internal circuit might be.

Maybe I'm getting confused by the reference - what is the scope using as it's reference? Is it the negative rail? It almost looks like it's referenced to +3V and AC coupled...

It was referenced to the +3v rail but the enclosed diagram shows it referenced to 0v. It is definitely DC coupled - it shows a "permanent" level when connected to a supply rail.

[Gina]

I disconnected the wind cable from the transmitter to connect the long cable into the circuit but left the rain plug disconnected while I did other things. I cleared a space on the table to put my main PC (AMD64X2) to fit extra memory and give it a spring clean. PC back in action and just looked at the test plots - still with no wind connection. This shows a much smoother trace in the temperature with nothing connected than with the standard cable to the wind sensors but with no wind.

4hrs-Temp-only3.png

So even the standard setup causes wriggles in the temperature plot - interesting!

Now to connect up again with the long cable (16 metres) in-line.

[Gina]

This plot shows the temperature plot for the test system. Up to point A the standard cable only was used to connect the wind sensors. Between A and B nothing was plugged in to the wind socket on the transmitter unit. From B onwards the wind sensors were connected through 16 metres of the same cable as used for the main station. Towards the end of the plot period a fan was used to rotate the anemometer. This seems to have changed the airflow around the transmitter unit resulting in a slightly higher temperature at that location as can be seen.

Screenshot-1.png

Well!! Where's the spikes? More wiggly plot with the supplied cable only, no wiggles without anything plugged in, smaller wiggles with an extra length of cable inserted between the supplied cable and the transmitter unit. It would seem the extra cable actually improves things"

[Gina]

Ah there it is - a spike!

6hrs-Temp-only-h.png

[Charlie]

Hi Gina,

Thinking about it some more, those curves would be expected if B & C were driven by a constant current source rather than a constant voltage source. Thanks for the updates! I have a couple more questions if you still have patience?
1) What is the time scale on your drawing? (I'm guessing about 10 ms/div? if it's very different then the capacitor value might not be 4.7 uF)
2) When you add the longer cable, how does the waveform change? (The added R & C might narrow the time between finish of the first pulse and beginning of the second, but I would not think by much)
3) It might be interesting to see how the waveform changes as you rotate the vane to different directions. The gap between measurements (seen on A) might change, depending on the software. Certainly the B & A pulse widths should change - with no change in C
The overshoot on A between measurements is a bit suspect for the cause of spikes. If the voltage on A is higher than the reference when the temperature measurement starts, then the value read would be a really big number - resulting in a spike being reported. Adding R & C (by extending the cable) will make the position measurement take longer to complete...

[Gina]

Hi Charlie

Just a quick reply (while I've stopped for a coffee break) - I'll write a detailed reply later. The temperature pulse was about 40ms wide and the wind vane about 70ms. I'll check in more detail later and post the results.

[mcrossley]

I may have this completely wrong (my electronics knowledge is basic to say the least), but from a few posts back the 'reference' cct posted on another site used a mechanism to calibrate the measuring cct by charging through a standard resistance, then measuring the time taken through the thermistor? Are you saying the FO uses the same mechanism?

If that is the case - what happens if the wind vane moves between/during these measurements. Could that affect the total resistance in the cct and thus throw out the calibration against the measurement.

The wind vane resistance changes up and down depending on direction, which could account for spikes in either direction.

This may be a load of bovine steam, so feel free to shoot it down

Edit: just thought, if that were the case then everyone should see the effect

[Gina]

OK... More detailed reply as promised.

Thinking about it some more, those curves would be expected if B & C were driven by a constant current source rather than a constant voltage source. Thanks for the updates! I have a couple more questions if you still have patience?

No problem, I'm considered to have a lot of patience

I agree about a constant current source, and yet the waveforms would indicate an RC charging characteristic. Whilst the -ve end of the 4.7uF capacitor seems too show 0v, it goes to the chip. If it were going to the 0v rail I'd expect a direct connection to 0v on a nearby track. I think it could be going into the neg input of an op-amp, measuring or setting the current.

1) What is the time scale on your drawing? (I'm guessing about 10 ms/div? if it's very different then the capacitor value might not be 4.7 uF)

Time scale was 20ms/div

2) When you add the longer cable, how does the waveform change? (The added R & C might narrow the time between finish of the first pulse and beginning of the second, but I would not think by much)

Didn't seem to make any difference.

3) It might be interesting to see how the waveform changes as you rotate the vane to different directions. The gap between measurements (seen on A) might change, depending on the software. Certainly the B & A pulse widths should change - with no change in C

Just checked that (turned fan off so that the vane would stay where I put it) and strangely, the waveform does not change! I'm wondering if I have a bad connection or broken wire. I'll unplug it and check with ohmmeter

The overshoot on A between measurements is a bit suspect for the cause of spikes. If the voltage on A is higher than the reference when the temperature measurement starts, then the value read would be a really big number - resulting in a spike being reported. Adding R & C (by extending the cable) will make the position measurement take longer to complete...

Curiously, there seems to be a smaller overshoot now, just a couple of ms and less amplitude.

I'm off now to check the wind vane resistance. LATER... Vane resistance values are exactly as per marked values for the four cardinal points. Circuit is complete right up to the RJ11 plug. Gremlins!!

Testing with scope suspended due to flat batteries

[Gina]

I may have this completely wrong (my electronics knowledge is basic to say the least), but from a few posts back the 'reference' cct posted on another site used a mechanism to calibrate the measuring cct by charging through a standard resistance, then measuring the time taken through the thermistor? Are you saying the FO uses the same mechanism?

Could be. But earlier tests would indicate that they didn't bother with a standard resistance - just measured thermistor resistance and wind vane circuit resistance.

If that is the case - what happens if the wind vane moves between/during these measurements. Could that affect the total resistance in the cct and thus throw out the calibration against the measurement.

If the vane moves during its measurement it would give an erroneous value (which might explain why the wind direction is so variable, but that's another issue).

The wind vane resistance changes up and down depending on direction, which could account for spikes in either direction.

I don't think that would cause the trouble.

This may be a load of bovine steam, so feel free to shoot it down

Edit: just thought, if that were the case then everyone should see the effect

Probably, yes. But any suggestion can be valuable however silly it may sound. We so called "experts" can easily miss the obvious!

[Gina]

Had a thought about the wind vane - looked at the console Yes, the station is indicating the correct wind direction. Which makes the constant pulse width puzzling. I'll sort out some batteries for the scope tomorrow - or recharge some NiCDs. Then I can check again.

[Gina]

Right... Have scope back in operation with new batteries. It seems to be triggering better now - the batteries were probably near exhausted before.

Please ignore all previous results except that the pulse shapes are still similar. There are 3 pulses on A with the 3rd one varying with vane direction. Only thing is, I seem to have lost the X/Y axis scales and not yet found the right combination of buttons to get them back. The 1st and 2nd pulses are about 40ms and the 3rd about 20ms for south and 60ms for west.

I'll post some more detailed results shortly.