Calibration
Core concepts
Calibration is the documented comparison of an instrument's output against a traceable reference standard. Distinguish three activities that questions deliberately blur: verification (an as-found check with no adjustment), adjustment/trim (changing the instrument so output matches the reference), and calibration (the whole process, including recording as-found and as-left data).
Know these terms cold: accuracy, precision, resolution, repeatability, hysteresis, deadband, span (URV minus LRV), live zero (4 mA = LRV, not zero process), and error expressed as percent of span versus percent of reading. Traceability and test uncertainty ratio (TUR) matter: the reference standard should be about four times more accurate than the device under test (3:1 is the usual minimum).
For a 4-20 mA transmitter the standard check is five points (0, 25, 50, 75, 100 percent) upscale then downscale, recording input, ideal output, actual output, and error. Understand HART trims: a sensor trim corrects the digital process value, an output (D/A) trim corrects the 4-20 mA conversion. Trimming the wrong one leaves the loop wrong.
Traps
- Recording error as percent of reading when the spec is percent of span.
- Adjusting before capturing as-found data, destroying the evidence.
- Simulating a 3-wire RTD with a 2-wire decade box and ignoring lead resistance.
- Treating 4 mA as zero process value on a live-zero loop.
- Calibrating in ambient conditions when the specification is at reference conditions.
Worked example
A 0-100 psi, 4-20 mA transmitter reads 11.80 mA at 50 psi. Ideal output = 4 + 0.5(16) = 12.00 mA. Error = (11.80 - 12.00) / 16 x 100 = -1.25 percent of span.
Rule of thumb
No adjustment until as-found is written down; no calibration without a traceable standard at least 4x better than the device.