Calibrating dataloggers

In the recent exchanges about the advantages and disadvantages of
different makes of dataloggers, several people mentioned the
difficulty or expense of calibrating them.  I thought my experience
of calibrating dataloggers might be of interest.  In my previous job
at English Heritage, I used several types of logger: ACR (old and
new versions), Humbugs, Tinytags, and I always calibrated them
myself, using saturated salt solutions.  I had three desiccator
cabinets with the solutions in the trays in the bottom.  I chose
salts whose RH change very little with temperature: potassium
acetate (approximately 23%), magnesium nitrate (approximately 54%)
and ammonium chloride (approximately 79%), the data being taken from
an NBS compilation.  The cabinets were kept in a room where the
temperature did not vary too much, though an insulated cabinet can
be improvised from an old refrigerator or freezer (not switched
on!).

After setting the calibration factors to zero, I placed the loggers
in the lowest RH cabinet for at least three days: I found it takes
two or three days for the sensors to respond fully to low RH, then
in the medium RH cabinet for three days, and finally in the high RH
cabinet for three days. I then downloaded the loggers and picked
points on the three regions of the graph where the temperature and
RH had been steady for at least 12 hours. These points were then
used to calculate the calibration factors.

For ACR loggers, I used a simple spreadsheet to calculate the
calibration factors.  I then used the "Modify file" function to
insert these calibration factors into the calibration graphs.  If
the calibration is correct the graph should now show the theoretical
RH for each of the saturated salts.  I checked this by using the
"Compound equation" function to calculate the theoretical RH for
each saturated salt solution, using the measured temperature data
and the temperature dependence given in the NBS paper and overlaying
them on the calibration graph.  It is useful to do this because you
can then ensure that the whole calibration graph is a reasonable fit
to the theoretical values, rather than just at the points you have
chosen.

Using this method I have usually got the calibration graph to fit
the theoretical values to within 2-3%, which is as good as can be
expected for general-purpose RH measurements.

If anybody would like further details of the spreadsheet or compound
equations, please contact me off-list.

Dr Barry Knight
Head of Conservation Research
The British Library
96 Euston Road
London NW1 2DB
+44 20 7412 7229
Fax: +44 20 7412 7658


                                  ***
                  Conservation DistList Instance 17:4
                   Distributed: Monday, June 23, 2003
                        Message Id: cdl-17-4-002
                                  ***