Calibration is the process of comparing an instrument against a more accurate, traceable reference standard, and documenting how far it's drifted. Every measurement instrument drifts over time — some more, some less — and calibration is how you find out how much, before that drift turns into a bad measurement, a failed batch, or a safety incident. This guide covers the practical side of instrumentation calibration: why it matters, how RTD/Pt100 temperature sensors work, what calibration uncertainty actually means, how dry block calibration is done properly, and the software and tools we use to manage it all.

Why calibration matters

The basic reason to calibrate any process instrument is simple: to confirm it's still measuring correctly. But in practice, several things force the issue:

Drift. Every instrument drifts over time, faster in extreme temperatures, high humidity, or heavy use. Left unchecked, an instrument keeps reporting numbers with growing error and nobody notices until something downstream goes wrong.
Product quality. In processes like thermal treatment, food and beverage, and pharmaceutical manufacturing, the finished product often can't be tested directly — the only proof of quality is that every in-process measurement stayed within tolerance, which only calibration can confirm.
Efficiency and profit. Accurate measurement lets a plant run closer to its real operating limits instead of padding for error, and even small accuracy gains often pay for the entire calibration program many times over.
Regulatory compliance. ISO 9001, ISO 14000, FDA 21 CFR Part 11/211, and industry-specific standards all place calibration requirements on regulated processes. Failing an audit here can shut down a line or cost a license to operate.
Safety. In oil and gas, chemical, and power industries, safety-system calibration is heavily regulated because the consequences of a bad reading can be an explosion. In food, beverage, and pharma, the same logic applies to consumer safety.
Money. When a measurement is directly tied to invoicing or custody transfer, a small calibration error becomes a direct financial error.

The cost of running a calibration program is easy to calculate. The cost of not running one — failed audits, safety incidents, product recalls, unplanned downtime — is much harder to estimate, and in almost every industry it's far higher than just doing the calibrations on schedule.

How often should you calibrate?

There's no single correct interval. The right period depends on the manufacturer's stability specification, how critical the measurement location is, how harsh the operating conditions are, and the instrument's own track record — a device with a long history of staying stable can safely go longer between calibrations, while one that keeps drifting or failing needs tightening up. Where a regulation or standard sets the interval, that overrides everything else.

RTD and Pt100 temperature sensors

The Pt100 is the most common temperature sensor in the process industry, and it's worth understanding properly since it shows up everywhere in instrumentation work.

RTD stands for Resistance Temperature Detector — a sensor whose electrical resistance changes predictably with temperature. Most RTDs are made from platinum, and a platinum RTD is called a PRT (Platinum Resistance Thermometer). The Pt100 is a PRT with exactly 100 ohms of resistance at 0°C, which is where the name comes from.

PRT vs. thermocouple

Thermocouples measure much higher temperatures, are cheap and rugged, and need no external power — but they're less accurate and need cold-junction compensation. PRTs are more accurate, linear, and stable, don't need cold-junction compensation, and can use plain copper extension wire — but they're more fragile, more expensive, and need a known excitation current. In short: thermocouples for high-temperature work, PRTs where accuracy matters more.

Measuring an RTD correctly

Since an RTD's resistance is what you're actually measuring, the wiring matters. A 2-wire connection is only good enough for rough troubleshooting, because wire and connection resistance adds directly to the error. Any real process measurement should use a 3-wire or 4-wire connection, which cancels that lead resistance out. The measurement current is small — typically under 1 mA — because too much current causes self-heating, which warms the sensor slightly and introduces its own error.

Temperature coefficient (alpha) and accuracy classes

The temperature coefficient (α) describes how much a Pt100's resistance changes per degree. The global standard is "385," meaning 0.003851 Ω/Ω/°C — if a sensor doesn't specify its coefficient, assume 385. Using the wrong coefficient in your measuring device causes a real error: mixing up a 385 sensor with a 391 sensor, for example, produces roughly 1.5°C of error at 100°C.

Pt100 sensors are also graded into accuracy (tolerance) classes under IEC 60751 — AA, A, B, and C, from tightest to loosest tolerance — plus older 1/3 and 1/10 DIN classes. Since an RTD sensor can't be adjusted to correct its own error, you pick a sensor with an accuracy class that already fits the application, or you correct the error mathematically using coefficients (Callendar-van Dusen or ITS-90 are the two most common equations) derived from an accurate lab calibration.

What a calibration curve actually shows

A calibration curve is a graph of an instrument's measured output plotted against a known, accurate input. It's how you visually confirm an instrument tracks the true value across its whole range, not just at one point — and it's how errors that only show up partway through the range get caught. Modern calibration software like Beamex CMX and LOGiCAL can generate these curves automatically from test data instead of by hand.

Calibration uncertainty, in plain terms

Uncertainty is the "doubt" in a measurement — how much you don't know, even after you've made the measurement. It's not the same as error: error is the difference between your reading and the reference, but that error means nothing until you know how uncertain the whole process is.

A classic way to picture this: give the same piece of string to three people. One measures once with a 10cm ruler and says "about 60cm." Another checks a few times with a tape measure and says "70cm." The third measures multiple times with a proper tape, calculates an average and standard deviation, accounts for how much the string stretches, and reports "67.5cm ± 0.5cm." All three technically "measured the string" — only the third result is actually usable.

Uncertainty is built from two types of components:

Type A — calculated directly, usually by repeating a measurement several times and computing the standard deviation of the results.
Type B — estimated, and includes things like the reference standard's own uncertainty, environmental conditions, the operator, display resolution, and other process-specific factors.

Independent uncertainty components don't simply add together — that would be overly pessimistic. Instead they combine using the root-sum-of-squares method: square each component, add the squares, then take the square root. The result is then multiplied by a coverage factor (commonly k=2, giving a 95% confidence level) to produce the "expanded uncertainty" that gets reported alongside a result.

This matters most at pass/fail time. Two measurements can show the exact same reading, but if one has a small uncertainty and the other a large one, only the first can honestly be called a "pass." Ignoring uncertainty when judging a result is how instruments get signed off as compliant when they're actually failing.

Dry block calibration: where the real uncertainty hides

A temperature dry block (also called a dry-well) heats or cools a metal insert with holes for inserting sensors — no liquid involved, unlike a temperature bath. It's portable, doesn't contaminate the sensors being calibrated, and is the standard field tool for RTD and thermocouple calibration. But most of its uncertainty comes from procedure, not from the block's own specifications:

Internal vs. external reference sensor. A dry block's built-in reference sensor sits in the surrounding metal block, not in the insert itself, so it reacts faster than the sensor being calibrated and can read a different temperature during a change. An external reference sensor placed in the same insert, at the same depth as the sensor under test, gives a far more accurate result.
Axial temperature homogeneity. Temperature isn't perfectly uniform along the depth of the insert — it's most consistent in a defined zone near the bottom (typically 40–60mm). A short sensor that doesn't reach that zone will read the wrong temperature unless the reference sensor is positioned at the same depth.
Difference between borings. Even in the same insert, one hole can run slightly warmer or cooler than another, especially opposite holes, due to uneven loading or heater tolerances.
Loading effects. Every sensor in the insert leaks a little heat to the environment (stem conductance). More sensors, or thicker ones, mean more leakage — and more disagreement between the insert's actual temperature and what any single reference reads.
Stability over time. Different sensors stabilize at different speeds. Read the result too early and you can be off by tens of degrees — not because anything is broken, but because the system simply hasn't reached equilibrium yet.

The short version: a dry block with excellent specifications can still produce a badly wrong calibration if the operator rushes it. Waiting for full stabilization and using a properly positioned external reference sensor eliminates most of the real-world error.

Calibration software: CMX vs. LOGiCAL

Beamex's two calibration management platforms serve different scales of operation:

LOGiCAL is a cloud-based, subscription platform for managing unlimited instruments and users without any IT infrastructure investment. It guides technicians through each calibration step, generates certificates automatically, and syncs data straight from Beamex documenting calibrators and the bMobile app. It suits maintenance teams, service companies, and organisations moving to a paperless process.
CMX is Beamex's more advanced tier, aimed at larger operations that need deeper analytics, business-system integration, and enterprise-scale calibration management.

Both can generate calibration curves and reports automatically from test results, instead of plotting them by hand.

Beamex bMobile

bMobile is Beamex's companion mobile app for field calibration work — it pairs with Beamex calibrators to pull up work orders, guide a technician through a calibration procedure on-site, and sync completed results straight back into LOGiCAL or CMX, cutting out manual data entry and paper records.