Instrument calibration

Calibration compares what an instrument reads with a reference of known, better accuracy, records the difference, and, if the difference is outside an agreed tolerance, adjusts the instrument and checks it again. Done well, it tells you not only that an instrument is right today, but how far it had drifted and whether that drift affected the process.

What calibration is, and what it is not

People use "calibration" for several different activities, and the difference matters when you read a service report:

• Calibration is the comparison against a reference and the record of the error. It does not necessarily include adjustment.
• Adjustment (often called trimming) changes the instrument so it reads closer to the reference. It should always be preceded by an as-found check and followed by an as-left check.
• Re-ranging a smart transmitter, by changing the values it reports at 4 and 20 mA, is a configuration change, not a calibration. It does not correct the sensor's error.
• A functional or bump check confirms a device responds, without measuring its accuracy. Gas detectors are a familiar example; see gas detector bump testing and calibration.

The reference must itself be calibrated, with a certificate that traces its accuracy back through an unbroken chain to national measurement standards. In Canada, the National Research Council maintains those standards. A common rule of thumb is that the reference should be several times more accurate than the tolerance you are checking, often four times or better, so that the reference's own error does not hide the instrument's.

How a calibration is done

For a pressure transmitter, the technician isolates it from the process, connects a pressure source and a reference standard to its process side, and reads its output with a calibrated milliamp meter or a documenting calibrator. The same pattern applies to temperature (a dry block or bath with a reference probe) and to other measurements with the appropriate source.

A typical procedure, with what to record at each step:

1. Identify and plan. Confirm the instrument tag, range and tolerance from the instrument index. Record the reference equipment used and its certificate due dates.
2. Make it safe. Arrange with operations, put the loop in manual or bypass any trips it feeds under the site's permit process, and isolate and depressurise the instrument following the site's lockout procedure. Record what was bypassed.
3. As-found test. Apply test points across the range, commonly 0, 25, 50, 75 and 100 percent, going up and then down so that any hysteresis shows. Record the applied value, the output and the error at each point, before touching anything.
4. Decide. If every error is within tolerance, no adjustment is needed. If not, adjust zero and span (or perform a sensor trim on a smart transmitter) following the manufacturer's procedure.
5. As-left test. Repeat the test points and record the results.
6. Restore. Reconnect, reopen isolation valves, check the reading against the process and the controller display, and remove bypasses. Record who returned the loop to service.

Setting tolerances and intervals

A tolerance should come from what the measurement is used for, not simply from the manufacturer's best-case accuracy. A transmitter feeding a safety trip, a custody or billing measurement, a control loop and an operator's trend each need a different tolerance. Set it tight enough to protect the process and loose enough that a healthy instrument passes.

The interval between calibrations depends on how critical the measurement is, the manufacturer's recommendations, any regulatory or quality requirements, and, most usefully, the instrument's own history. That is why the as-found result matters so much.

If an instrument is repeatedly found well within tolerance, its interval can often be extended. If it is found out of tolerance, the interval should be shortened, and someone should ask what that error meant for the process since the last good calibration: product made, energy billed, or a trip that might not have worked.

What a calibration record should include

• Instrument tag, description, make, model and serial number.
• Calibrated range, units and the tolerance applied.
• Reference equipment used, with serial numbers and calibration due dates.
• As-found readings at each test point, up and down, with the error.
• What was adjusted, if anything.
• As-left readings at each test point.
• Pass or fail, the date, the technician's name, and any comments on condition.
• Bypasses applied and confirmation that they were removed.

A record that shows only "calibrated OK" with no readings cannot be used to trend drift or to defend a measurement if it is ever questioned.

In place or on the bench

Calibrating in place tests the instrument in its real installation, including its impulse lines and environment, and avoids disturbing connections. Bench calibration offers better control and is often easier for temperature sensors and small devices. Either way, a loop check afterwards confirms that the controller reads the same value as the instrument. If the controller's scaling does not match the transmitter's range, a perfectly calibrated instrument still produces a wrong number on screen; our guide to 4–20 mA current loops covers scaling.

A typical case

A plant has two level transmitters on the same tank, one for control and one for a high-level trip. At the annual check, the trip transmitter is found reading low by more than its tolerance across the whole range, a classic zero shift. The technician records the as-found values, adjusts, and records the as-left. The as-found result then triggers two questions. Could the trip have operated late since the last good check? Engineering compares the error with the margin between the trip setpoint and overflow and records the answer. And why did it drift? A look at the history shows the same transmitter drifting at each of the last three checks, so its interval is shortened and the sensor is put on the replacement list.

Common mistakes

• Adjusting before recording the as-found values, which destroys the evidence of drift.
• Using a reference that is out of its own calibration, or not accurate enough for the tolerance.
• Testing at one point only, which misses span and linearity errors.
• Re-ranging a smart transmitter and calling it a calibration.
• Setting every interval the same, regardless of criticality and history.
• Forgetting to remove a bypass or return a loop to automatic.

Some errors are not the instrument's fault at all. Plugged impulse lines, a sensor in the wrong place or a short thermowell will survive any number of calibrations. Our guide to sensors and transmitters covers those installation problems.

How ehZee can help

We set up instrument indexes and calibration programs, define tolerances and intervals based on what each measurement is used for, review calibration records and plan the loop checks that go with them. Hands-on calibration is carried out by qualified instrumentation technicians with traceable reference equipment. See our instrumentation and controls service.

Common questions