Temperature data spikes revisited

[Charlie]

Hi Gina,
Since the issues only started with extending the cable, it may be the design makes some assumption about the cable characteristics. I'm not sure if you can put the original cable on your spare unit and see if it then produces spikes or not.
If the original cable looks good, then we can look up the original cable electrical characteristics and try to match those. i.e. a larger gauge will have smaller "R", 2 @ 2 conductors will have less "C" than the 4 conductor, etc.
Unless you have a method to measure the frequency components of that fast edge switch on (FFT?) it's going to be hard to do something about the "L", but that might not matter.
(The shielding in the replacement cable you have now will certainly have higher "C" than the original.)
This also would explain how many of us see different behaviour, as no two custom extensions will be the same. I've been running with 10 nF across the anemometer and 10 nF on each of it's leads to the negative terminal for 2 weeks now with reasonable results (occasional spikes less than 0.5 degrees) - of course, by posting, I might be causing that to change. I'm reminded of a certain cat in a box...

[Gina]

Hi Gina,
Since the issues only started with extending the cable, it may be the design makes some assumption about the cable characteristics. I'm not sure if you can put the original cable on your spare unit and see if it then produces spikes or not.

Yes, I should be able to do that - if I can find one of the original cables again. And yes, it all started when extending the original cable either by using an RJ11 extension cable or replacing the cable altogether. In fact my replacement cable is actually twice as long as it needs to be to reach the Stevenson screen but I left it at it's original length to reach the house. Also, (if cable length is the problem) I could move the screen box nearer the mast to reduce the cable length further.

If the original cable looks good, then we can look up the original cable electrical characteristics and try to match those. i.e. a larger gauge will have smaller "R", 2 @ 2 conductors will have less "C" than the 4 conductor, etc.

The cable I'm using has two twisted pairs, screened from each other. The cable capacitance will certainly be a lot more than the original However, I don't think it's the capacitance between the pair that's the trouble as adding up to 470nF seems to have little affect. That much capacitance to unit earth has little effect from one connection but causes noise on the other. I guess it could be that a wave (edge pulse) is being reflected back from the far end of the improperly terminated cable and the longer cable gives more delay. I have some small chokes (somewhere) that I might try in series. If the problem is high frequency reflections that might help.

Unless you have a method to measure the frequency components of that fast edge switch on (FFT?) it's going to be hard to do something about the "L", but that might not matter.
(The shielding in the replacement cable you have now will certainly have higher "C" than the original.)
This also would explain how many of us see different behaviour, as no two custom extensions will be the same. I've been running with 10 nF across the anemometer and 10 nF on each of it's leads to the negative terminal for 2 weeks now with reasonable results (occasional spikes less than 0.5 degrees)

I don't think my little scope will show such high frequency components of the short delay of the cable but I'll check. I'll also try 10nF Cs like you're using.

- of course, by posting, I might be causing that to change. I'm reminded of a certain cat in a box...

I know what you mean! Only too well!!

[Gina]

I'm thinking now that there's little left to test and the remaining spikes can be removed by software. I feel I have discovered quite a lot about the workings of the F.O. weather stations though and that in itself has been a worthwhile result of the exercise.

Connecting the cable screen to either battery terminal cures a lot of interference picked up by the cable but with that well screened I don't think RFI is the problem now. I agree with Charlie that it would seem to be cable characteristics. Apart from shortening the cable it might be possible to adjust it's characteristics but I haven't got the equipment to measure it and anything in that line would have to be trial and error.

I think I shall move the Stevenson screen box and attach it to the post the mast is attached to, shortening the cable and avoiding the exposed length from mast to box making it less vulnerable to possible attack. Meanwhile, I'll write code to remove temperature spikes as Steve has done in Cumulus.

[Gina]

As of 19:40 this evening I've added an LC filter into one leg of the anemometer connection as shown in the diagram below :-

AWS_Anemometer_LC.png

[Gina]

I don't think that helped very much so have gone back to capacitor only as from 08:30 this morning. Here's the overnight plot.

12hrs-Temp-only1.png

To mix metaphors I feel I'm groping about in the dark, clutching at straws and generally getting nowhere in getting rid of the spikes. Spikes seem to appear in groups, mainly unrelated to anything I do with the sensor connections themselves.

Must get back to the science I think and stop messing about! If there was interference with either of the wind sensor data I could understand it but NO it's the temperature which would seem totally unrelated. If it's not large pulses of interference getting in which it doesn't seem to be. the only thing I can think of is that the extra capacitance of the longer cable is loading the microchip and causing a malfunction. I have already tried decoupling the battery supply with a 220uF electrolytic and 470nF ceramic capacitor in parallel.

Anyway, I won't be doing anything for the rest of the morning - I'm off out shortly and back lunchtime.

[hans]

have you tried winding your cables into a choke balun?
usually done with coax,but worth to try.
examples:

[Gina]

Yes, I tried it with some ferrite cores from Maplin - made no difference.

[hans]

makes me think more and more it might be the tempsensor itself.
you have practically done everything to eliminate the other sensors.
maybe a small condensator(a few pF,non polarised) directly over the tempsensor.to smooth it a litle.

[Gina]

makes me think more and more it might be the tempsensor itself.
you have practically done everything to eliminate the other sensors.
maybe a small condensator(a few pF,non polarised) directly over the tempsensor.to smooth it a litle.

I may try that another time but I've stopped testing now and put the transmitter unit back outdoors.

I've moved the Stevenson screen box up close to the mast with the wind sensors on. And having put the transmitter unit back into it, have cut the cable and joined it up so that there is no extra length - it goes straight from the mast and into the box to the test setup and thence to the transmitter. Length of cable is now just under 6m rather than over twice that (I'll measure the piece I've cut out later).

Weather station is now fully connected as from 16:00 today (15/11/10) and I've just re-started the main weather station computer which uploads to the main web site.

[hans]

hope that works,

but if not (sorry,couldn't resist) here is some food for thought:

''''RTD and thermistor output signals are typically in the millivolt range, making them susceptible to noise. Lowpass filters are commonly used in RTD and thermistor data acquisition systems to effectively eliminate high frequency noise in RTD and thermistor measurements. For instance, lowpass filters are useful for removing the 60 Hz power line noise that is prevalent in most laboratory and plant settings.''''

taken from here
http://zone.ni.com/devzone/cda/tut/p/id/3643

so the wires may not be the culprits,but the sensor itself.
also the bit about selfheating is very informative.

this all is probably not the issue,but i myself prefer to be proved wrong before dismissing it.
could be as simple as reheating the solderjoints of the thermistor(which is the litle thingy near the cableconnectors)

fascinating stuff....drives you crazy sometimes..

edit;
when you look at this picture;

you can see that the tempsensor is very close to the cableconnection wires.
so interference picked up by the cables(which are huge antennas in a way) could very well have influenced the the sensor so close to it

[Gina]

Thank you Hans, that is a very good point And thank you for the link to the tutorial on using thermistors. If the spikes come back I'll definitely look at that approach. So far, after nearly 3 hours there have been no spikes (since I shortened the cable) but there has also been no wind. Unusually, we have an absolute flat calm - the anemometer isn't moving at all. I imagine the breeze will come up again tomorrow and I'll see what result I get then.

[hans]

wind(or a small breeze) can build up static on an antenna,dependent on atmosferic conditions,like humidity,static charge of the air,nearby elements(trees,poles) so wether the spikes ar build ups or discharges,i don't know.

yet another possibility

a piece of tin between the tempsensor and the cable connector and soldered to the neg battery lead could work in that case

[Super-T]

Isn't it more to the point that Gina has replaced the Transmitter and still gets the problem? It seems to me that nobody else using Fine Offset Stations is getting the horrendous spikes that Gina gets. I get the occasional spike of perhaps 1.5 degrees and that is all.
Is there anyone else getting these large spikes as often as Gina?

It's almost as if someone lights a match under the sensor at random times

[daj]

Is there anyone else getting these large spikes as often as Gina?

Never for me!

[hans]

just an occasional small one.

what if it is the base unit all that time,sometimes it't right in front of you and you don't see it........