Procedure
(c) Procedure. Use the following linearity verification protocol, or use good engineering judgment to develop a different protocol that satisfies the intent of this section, as described in paragraph (a) of this section: (1) In this paragraph (c), the letter y denotes a generic measured quantity, the superscript over-bar denotes an arithmetic mean (such as y ), and the subscript ref denotes the known or reference quantity being measured. (2) Use good engineering judgment to operate a measurement system at normal operating conditions. This may include any specified adjustment or periodic calibration of the measurement system. (3) If applicable, zero the instrument as you would before an emission test by introducing a zero signal. Depending on the instrument, this may be a zero-concentration gas, a reference signal, a set of reference thermodynamic conditions, or some combination of these. For gas analyzers, use a zero gas that meets the specifications of 1065.750 and introduce it directly at the analyzer port. (4) If applicable, span the instrument as you would before an emission test by introducing a span signal. Depending on the instrument, this may be a span-concentration gas, a reference signal, a set of reference thermodynamic conditions, or some combination of these. For gas analyzers, use a span gas that meets the specifications of 1065.750 and introduce it directly at the analyzer port. (5) If applicable, after spanning the instrument, check zero with the same signal you used in paragraph (c)(3) of this section. Based on the zero reading, use good engineering judgment to determine whether or not to rezero and or re-span the instrument before continuing. (6) For all measured quantities, use the instrument manufacturer's recommendations and good engineering judgment to select reference values, yrefi, that cover a range of values that you expect would prevent extrapolation beyond these values during emission testing. We recommend selecting a zero reference signal as one of the reference values for the linearity verification. For pressure, temperature, dewpoint, power, current, voltage, photoacoustic analyzers, and GC-ECD linearity verifications, we recommend at least three reference values. For all other linearity verifications select at least ten reference values. (7) Use the instrument manufacturer's recommendations and good engineering judgment to select the order in which you will introduce the series of reference values. For example, you may select the reference values randomly to avoid correlation with previous measurements and to avoid hysteresis; you may select reference values in ascending or descending order to avoid long settling times of reference signals; or you may select values to ascend and then descend to incorporate the effects of any instrument hysteresis into the linearity verification. (8) Generate reference quantities as described in paragraph (d) of this section. For gas analyzers, use gas concentrations known to be within the specifications of 1065.750 and introduce them directly at the analyzer port. (9) Introduce a reference signal to the measurement instrument. (10) Allow time for the instrument to stabilize while it measures the value at the reference condition. Stabilization time may include time to purge an instrument and time to account for its response. (11) At a recording frequency of at least f Hz, specified in Table 1 of 1065.205, measure the value at the reference condition for 30 seconds (you may select a longer sampling period if the recording update frequency is less than 0.5 Hz) and record the arithmetic mean of the recorded values, y i. Refer to 1065.602 for an example of calculating an arithmetic mean. (12) Repeat the steps in paragraphs (c)(9) though (11) of this section until measurements are complete at each of the reference conditions. (13) Use the arithmetic means, y i, and reference values, yrefi, to calculate least-squares linear regression parameters and statistical values to compare to the minimum performance criteria specified in Table 1 of this section. Use the calculations for a floating intercept described in 1065.602. Using good engineering judgment, you may weight the results of individual data pairs (i.e., (yrefi, y i)), in the linear regression calculations.