Product guide

Humidity and Dew-Point Instruments
© Michell Instruments 2009-2011
92
Uncertainty (accuracy)
The uncertainty of a measurement is defi ned as the
parameter characterising the range in which the ‘true
value’ can be expected to lie. It defi nes a ‘margin of
doubt’ about a reading or estimated value, together
with a level of confi dence (normally 95%) that the ‘true
value’ will lie within this range.
NOTE: Often, the uncertainty for an instrument is
specifi ed for ideal operating conditions at a temperature
of 20°C or 23°C. However, further contributions such
as hysteresis, linearity, reproducibility and temperature
dependence will need to be taken into account when
estimating the overall uncertainty for an instrument. The
method of use may make an important contribution to
the overall uncertainty achieved.
Strictly speaking, ‘accuracy’ is a qualitative term only.
For example, an instrument or measurement might be
described generally as ‘accurate’ or ‘not accurate’. If
accuracy is to be quantifi ed, it should be expressed in
terms of uncertainty, for example, ‘... an uncertainty
of ±5% RH ...’ (not ‘... an accuracy of ±5% RH ...’).
However, the word ‘accuracy’ continues to be used
loosely in specifi cations to refer to the maximum
difference that can be expected between the reading
given by an instrument and the ‘true value’ being
measured.
Repeatability
In general terms, the repeatability of an instrument
is the closeness of agreement of multiple readings
repeated under the same conditions of measurement.
To quantify repeatability, the spread or dispersion of
results may be quoted, e.g. in terms of a standard
deviation of a number of readings.
Reproducibility
In general terms, reproducibility is an instrument’s
capacity to reproduce a previous measurement. This
may be at a later date, or after undergoing signifi cant
changes in conditions, for example, after a change of
operator, or of location. To quantify reproducibility, the
spread or dispersion of results may be quoted, e.g. in
terms of a standard deviation of a number of readings.
Non-linearity
In an ideal world, if a calibration were carried out at
two points in a measurement range (high and low) the
instrument would behave similarly at values in-between.
However, it is usual to fi nd some deviation from this
ideal, due to non-linearity of the sensor. In other words
most sensors exhibit a calibration curve, as opposed to
a straight line.
• range
• resolution
• uncertainty (accuracy)
• repeatability
• reproducibility
• non-linearity
• hysteresis
• response time
• long-term stability
• temperature coeffi cient
This list is not exhaustive, although it indicates the
most important specifi cations to consider when
selecting an instrument. Practical considerations such
as dimensions, power supplies and output signals,
may have little or no signifi cance to the measurement
performance characteristics of the unit. However, these
practical elements should be carefully selected if the
instrument is to meet the full working requirement of
the application.
Each characteristic is explained separately below.
However, although separately defi ned, it is not always
possible in practice to distinguish between some of
them.
Formal defi nitions of these terms are given in the ISO
document ‘International Vocabulary of Basic and General
Terms in Metrology’ (Second edition, 1993).
Measuring range
The stated measuring ranges for temperature and
humidity provide guidance as to the upper and lower
temperature and humidity environments to which
the instruments sensors can be applied. Often an
additional specifi cation: operating range (electronics)
is also stated; this confi rms the conditions in which
the instrument’s electronics (other than the sensor,
or probe) can safely operate. Where a maximum or
minimum humidity is stated, it is often the case that
the humidity sensor will be damaged if these limits are
exceeded.
Resolution
Where an instrument features a display of its measured
values, the indication will have a resolution of, say, 0.1%
RH or 1% RH. Specifi cation of this in the data sheet
simply informs the user to what degree of precision
the measurement can be noted. Resolution may not
signifi cantly affect the uncertainty of measurement,
except in the case of a coarse resolution of, say, 1% RH,
where the display will round the actual value up or down
to the nearest signifi cant fi gure, thereby introducing an
additional error of up to ±1% RH, depending on the
method of rounding.
A Guide to the Measurement of Humidity