Unusual UV readings

[marcopolo]

Happy New Year !

Hi I'm getting unusual readings twice a day at regular intervals. I've attached the graph to show. At regular intervals both morning and afternoon im getting UV reading gaps. Not sure what's causing it. UV sensor is fully exposed to sun light. I'm trying to find the data log for the UV sensor through cumulus and cant find it to see what data is being recorded no luck or I'm just blind? Has anyone had similar issues? I checked through the forum threads couldn't see anything similar. Thanks Regards marcopolo


Ellenbrook Weather
http://www.ellenbrookweather.id.au/
pwsweather - ELBRK01
wunderground - IWESTERN246

[steve]

That's solar radiation, not UV. Cumulus derives it from the 'Light' figures from the station. The figures are logged to the 'monthly' log files, e.g. the current one is Jan14log.txt. See the help or the wiki for the description, or you can view the log from within Cumulus using View -> Data logs.

The solar radiation graph in Cumulus looks similar - http://www.ellenbrookweather.id.au/images/solar.png - so it looks like your sensor is giving spurious readings at the start and end of the day. Those extra lines just shouldn't be there at all. The rest of the day looks pretty good, though, and follows the theoretical max very well.

[AllyCat]

Hi,

Yes, as Steve says, it's Solar Radiation, which Cumulus calculates from the "Lux" value supplied and displayed by the Console.

Those extra lines just shouldn't be there at all.

It looks very much as if the "low" light levels are being displayed with an incorrect scale factor, perhaps 100 times too large (e.g. 14 being displayed as 1400). I know that the "Solar Pod" transmits the Lux value as a "pure" integer (i.e. potentially any value from 0 and 1 up to several hundred thousand) but don't know if the actual A/D conversion, from the light sensor itself, uses some form of high/low scaling.

So, either the Solar Pod is faulty (in which case the Lux value displayed by the console will be equally weird), or somewhere in the later processing/display of the data, a decimal point (or exponent) is getting misplaced.

Cheers, Alan.

[marcopolo]

Cheers Alan and Steve!

[steve]

It would be useful to know what happens to the Lux figures on the console. It could be that the representation of low Lux figures in the console memory is different to higher figures (and hence is being misinterpreted by Cumulus), but this isn't mentioned in any versions of the spec that I've seen. And you would expect others to have experienced this if that were the case, and I don't think I've seen it before.

Are you using the standard conversion in Cumulus from Lux to W/m2 (0.0079) or have you changed it?

[mcrossley]

Looking at this mornings readings it doesn't seem to be a factor of 10 thing? Looks like it is roughly 30x (EDIT: looking at this evenings data as well...) 40x too high. Once the value drops below 50 W/m^2

0 watts/m^2
43 watts/m^2
166 watts/m^2
345 watts/m^2
578 watts/m^2
824 watts/m^2
1164 watts/m^2
1523 watts/m^2
1749 watts/m^2
2012 watts/m^2
2354 watts/m^2
83 watts/m^2
95 watts/m^2
110 watts/m^2
126 watts/m^2
145 watts/m^2

[marcopolo]

Thanks guys I'm going through the data and checking settings. Ill give you feedback as soon as I can.

[marcopolo]

I've gone through the data. The peaks occur and start roughly at regular intervals around 05:20 (sunrise?) and drop off at 07:30 then it seems to normalise itself after that then pick up again at dusk around 16:50 and normalizing around 19:00. Each is around 2 hours in duration. I've attached a graph with 3 days of data. Overall the light profile is good as Steve pointed out. You can see clearly on day 2 when it was cloudy. Apart from these weird spikes it looks ok. I've contacted the supplier and provided him with the info. I'm waiting for a response. I'm climbing the roof today to have a closer look at the sensor and the cover to observe any anomalies.

I extracted about 856 data points over the three days and ran the data through excel. Nice pwetty graph. Thanks guys. Regards Marco

[steve]

If you turn on the debug log in Cumulus a few minutes before the expected jump occurs this evening (coming up in the next hour or so, at the time I'm writing this, I think), then Cumulus will log the raw data from the station to the debug.log file, and we'll be able to see the actual Lux values.

[marcopolo]

Hi Steve I missed the evening window to use the debug option. Ill try and run it first thing in the morning. Cheers Marco

[marcopolo]

Hi Steve I've attached 2 files one is the raw txt log file and a marked up word version. I've trimmed the data (lots of it). The word version has the areas marked blue and red to mark the spots. 1st reading for lux was at 05:21 reference number 7973.6401 where the values peaked at 06:16 reference number 8302.9161. This is where it get interesting the solar values disappear at this point and don't appear again until 6:20am reference point 8326.9355. This might explain the gaps in the graphs. Does the WS provide the solar data values? Or does cumulus fetch the data from the WS log files? Cheers Marco

[steve]

The Lux readings increase from zero rapidly until they exceed the stated maximum value for the station (300,000) so Cumulus discards the values at that point (i.e. it doesn't convert to W/m2). When they start behaving sensibly, it starts processing again.

It looks like the station is just logging inflated readings at the start and end of the day. Perhaps someone can spot a pattern in the values?

[marcopolo]

Thanks Steve. The data/readings from WS goes off scale and then the WS resets itself (if that is the right description) and starts recording/processing normal readings during the course of day until late evening the spikes start occurring. Sounds more like a sensor fault or circuit? The pattern is very unusual and consistent as it how it occurs. I was looking into the values and couldn't interpret o much from. I'm going to experiment with sensor to see if I can change the behaviour. My gut feeling it the sensor as it been behaving like this straight out of the box when I purchased it. Cheers Marco

[AllyCat]

Hi,

Perhaps someone can spot a pattern in the values?

For this particular data, the "Lux" value appears to hit a limit of 21 bits, i.e. a value of "31" in the third byte. Of course the FO sensor (or its D/A converter) doesn't actually work to that resolution - the Least Significant Bit appears to represent 0.1 Lux. The data coming from the Solar Pod increases in much larger steps, IIRC very approximately 10 Lux, so it is presumably applying some (pre-programmed) calibration multiplier to the measured A/D binary values.

My concern is the lack of a pattern beteween the two tables shown earlier in the thread. In the first table the W/m2 steps from 2281 down to 80 (i.e. ~1/30x), but in the second from 518 up to 2298 (i.e ~5x). Of course the measurement conditions are very different: In the first case the light level is rising, in the second falling (a hysteresis effect?), the sun will be in a completely different location, and the temperature is very different, ~15 degrees in the morning, at least 35 degrees in the evening (probably higher in the Solar Pod , because even FO's "rubbish" sunscreen must give some protection from direct sunlight).

However, as Steve mentioned earlier, the Lux display on the Console (or a log from the Easyweather-Plus software) at the time of the "crossover" should at least give us a definitive answer whether this is a "Fine Offset" problem, (i.e. a faulty Solar Pod, my guess) or a strange issue with the later data manipulation.

Cheers, Alan.

[marcopolo]

Cheers Alan!!