On 23 May, 21:09, Neil wrote:
This thread does throw up an interesting question - What is the
recommended calibration procedure for temperature and RH sensors.

There are plenty of methods and procedures available for commercial and
pharmaceutical instruments but I've never looked into the standards for
calibrating meteorological instruments.

I presume they would be published and/or available from the MetO but a
quick look at the website revealed nothing.

Anyone got any insight on this??

To be technically correct, formal temperature calibration would
involve a true calibration check against known standards at a variety
of spot temperatures. This is rarely feasible or affordable for the
majority of amateur-run sites, not least because losing the
instruments for 3-4 weeks while they were off being calibrated would
likely leave a gap in records.

However, the next-best method is to use a travelling reference
instrument as a benchmark i.e. one which has been calibrated under
laboratory conditions. If your instruments are in a Stevenson Screen
or similar, it's easy enough to expose the travelling standard
alongside existing sensors and compare the readings over a period of
time (and ideally a wide range of temperatures). If the instruments
are not in a screen it's a little more troublesome getting the two
sensors close enough together to be reasonably sure they are measuring
the same thing. Best to compare under cloudy, breezy, dry conditions,
best of all at night and when the temperature is changing only slowly
(otherwise differential thermal inertia will skew results).

I've used such a method with a calibrated TinyTag logging
thermohygrometer (http://www.geminidataloggers.com/data-loggers/
tinytag-hand-held) over the last six months, and after even just a few
days of logged data it's surprisingly easy to pin down an accurate
calibration against existing instruments (both logged sensors and
mercury-in-glass thermometry). I found that the manufacturer's
calibration on my expensive Vaisala temp/RH sensor had drifted by 0.4
degC in 18 months. After 4 months comparisons using the new
calibration coefficients the mean error was 0.05 degC under all
conditions, and at night under slow rates of temp change the adjusted
calibration of the Vaisala sensor was within 0.2 degC of the benchmark
99.4% of the time. That's good enough for me!

The logger I used is quite expensive, and calibration is another £60
or so, but it's proved an easy and accurate method of checking the
calibration of my sensors. Sending it back to Gemini for recalibration
every couple of years isn't too expensive either, and of course does
not lose the record from the 'primary' sensor. A repeat of the cross-
calibration every 2-3 years should be sufficient for all but the most
exacting of meteorological purposes.

I will be writing up the method for COL soon, with more detailed
instructions and a template spreadsheet to generate calibration
coefficients. I hope to persuade COL to buy a few of these calibrated
loggers for loan to members, which will make the whole process cheaper
and easy to carry out.

I also had my thermohygrometer humidity sensor calibrated, but I have
to say I've been less impressed with the accuracy of this RH sensor
and its calibration than I have with the temperature sensor. Not a
particular problem, as temperature was the element I was seeking a
calibration method for, but RH is altogether more difficult to pin
down to within 3-5 per cent at best, despite the claims of the
manufacturers.

I can highly recommend the TinyTag logger range btw: I have since
bought another smaller temperature logger without a display but with a
trailing thermistor on a 600 mm lead (ideal for comparing inside, say,
a Davis AWS screen, which is where mine currently is). They are about
half the price of the logging thermohygrometer with display. I
previously compared the new unit against my existing calibrated
standard over 2-3 weeks and 'out of the box' the mean error was
0.05 degC and 98% of readings were within 0.2 degC of the calibrated
unit. You can also buy these smaller units with calibration
certificates as new and they can be recalibrated for a relatively
small charge as and when required. Don't buy the small 'self-
contained' temp logging units though, at least not for accurate
meteorological purposes, as their response time is slow (~ 20 minutes
is quoted, vs 10 sec for the trailing thermistor sensor).

HTH.

--
Stephen Burt
Stratfield Mortimer, Berkshire