Analytical Industries GPR-35 Oxygen Analyzer - Manual (Page 12)

Advanced Instruments Inc.
12
Accuracy & Calibration
Single Point Calibration: As previously
described the galvanic oxygen sensor generates
an electrical current proportional to the oxygen
concentration in the sample gas.
Absolute Zero: In the absence of oxygen the
sensor exhibits an absolute zero, e.g. the
sensor does not generate a current output in
the absence of oxygen. Given these linearity
and absolute zero properties, single point
calibration is possible.
Pressure: Because sensors are sensitive to the
partial pressure of oxygen in the sample gas
their output is a function of the number of
molecules of oxygen 'per unit volume'.
Readouts in percent are permissible only when
the total pressure of the sample gas being
analyzed remains constant. The pressure of the
sample gas and that of the calibration gas(es)
must be the same (reality < 1-2 psi).
Temperature: The rate oxygen molecules diffuse into the sensor is controlled by a Teflon membrane otherwise known as an
'oxygen diffusion limiting barrier' and all diffusion processes are temperature sensitive, the fact the sensor's electrical output
will vary with temperature is normal. This variation is relatively constant 2.5% per ºC.
A temperature compensation circuit employing a thermistor offsets this effect with an accuracy of better than +5% (over the
entire Operating Range of the analyzer) and generates an output function that is independent of temperature. There is no error
if the calibration and sampling are performed at the same temperature or if the measurement is made immediately after
calibration. Lastly, small temperature variations of 10-15º produce < 1% error.
Accuracy:
In light of the above parameters, the overall accuracy of an analyzer is affected by two types of errors: 1) those
producing 'percent of reading errors', illustrated by Graph A below, such as +5% temperature compensation circuit, tolerances
of range resistors and the 'play' in the potentiometer used to make span adjustments and 2) those producing 'percent of full
scale errors', illustrated by Graph B, such as +1-2% linearity errors in readout devices, which are really minimal due to today's
technology and the fact that other errors are 'spanned out' during calibration. Graph C illustrates these 'worse case'
specifications that are typically used to develop an transmitter's overall accuracy statement of < 1% of full scale at constant
temperature or < 5% over the operating temperature range. QC testing is typically < 0.5% prior to shipment.
Example: As illustrated by Graph A any error, play in the multi-turn span pot or the temperature compensation circuit, during
a span adjustment at 20.9% (air) of full scale range would be multiplied by a factor of 4.78 (100/20.9) if used for
measurements of 95-100% oxygen concentrations. Conversely, an error during a span adjustment at 100% of full scale range is
reduced proportionately for measurements of lower oxygen concentrations.