Three different activities get called calibration, and only one of them is calibration.
A bump test exposes the sensor to gas above the alarm point to confirm it responds and that the alarm sounds. It proves the instrument is alive. It does not prove it is accurate.
A zero sets the instrument’s baseline. If the instrument is told that the current atmosphere is 0 ppm, whatever is in that atmosphere becomes the new zero.
A span adjusts the sensitivity so that a known concentration reads correctly. This is the step that needs a certified gas.
Confusing the three is how a plant ends up with an instrument that passes every check and reads wrong.
Clean air is an assumption, not a reference
Zeroing in ambient air assumes two things: that the air contains 20.9% oxygen and none of the target gas. Both assumptions fail in the places detectors are installed.
A detector in a pump station wet well is installed where hydrogen sulphide is expected. Zeroing it in place, in the morning, after the well has been standing overnight, sets the baseline at whatever the well was producing. The instrument now reads 0 ppm at a concentration above zero, and every subsequent reading is low by that amount. If the well was at 4 ppm when it was zeroed, the first-stage alarm at 10 ppm will not sound until the atmosphere reaches 14 ppm.
Oxygen has the same problem from the other direction. A cell zeroed in nitrogen reads 20.9% in air as something other than 20.9%. An oxygen channel zeroed in a confined space that has been purged will be wrong by the difference for the rest of the calibration interval.
There is a legitimate use for ambient zero: an instrument with a fresh air point that is genuinely fresh, outdoors, upwind of the plant, with a known oxygen concentration. Everything else should use a zero gas cylinder.
What the cylinder has to be
A standard gas cylinder is a gas mixture whose concentration has been analysed and certified, with a tolerance and an expiry date. Three properties matter.
Concentration. It should be in the range the instrument works in, ideally 30% to 70% of full scale. A cylinder at the alarm point is also a useful choice, because it tests the setpoint directly.
Balance gas. This is the one that gets overlooked. For a catalytic bead %LEL sensor, the balance must be air, because the sensor measures the heat of combustion and needs oxygen to burn. A 50% LEL methane cylinder balanced in nitrogen will produce a low or zero reading on a perfectly good sensor. For most electrochemical cells the balance can be nitrogen, but an oxygen channel obviously needs an oxygen mixture, and some cells behave differently in the absence of oxygen.
Certification and date. A cylinder has a shelf life and a certification validity, and both expire. A cylinder that has been in a cupboard since the last audit is not a reference. Reactive components also decay: hydrogen sulphide mixtures in aluminium cylinders are notorious for losing concentration over months as the gas reacts with the walls.
Four ways a correct calibration still goes wrong
Flow rate. A diffusion sensor calibrated through a cap needs a modest flow, typically in the region of 0.3 to 0.5 litres per minute. Too little and the reading drifts up slowly as gas accumulates. Too much and the sensor sees a slight over-pressure that pushes gas in faster than diffusion would, which produces a reading that is high. Either way, the instrument leaves calibrated to a condition it will never see in service.
Tubing. Hydrogen sulphide adsorbs onto many polymers. A long run of the wrong tubing will deliver a concentration at the cap well below the cylinder’s label, and the instrument will be span-adjusted upwards to compensate. Short runs, and PTFE or the tubing the manufacturer specifies.
Time. A span adjustment must be held until the reading is stable, not until it looks stable. Electrochemical cells have a T90 in the tens of seconds, and some gases take several minutes to reach equilibrium. A technician under time pressure who reads the display at 60 seconds and presses accept has calibrated to a transient.
Humidity and temperature. A cell calibrated in dry cylinder gas and then installed in a humid wet well will read differently, because water vapour affects both the electrolyte and the diffusion rate. Let the gas reach ambient temperature before use, and do not calibrate a cold instrument that has just been brought in from a vehicle.
Bump test and calibration are on different schedules
| Activity | Interval | What it catches |
|---|---|---|
| Bump test | Before each use, or daily in continuous service | A dead cell, a blocked inlet, a failed alarm |
| Zero | At each calibration | Baseline drift |
| Span calibration | Typically every 3 to 6 months, site conditions permitting | Sensitivity loss |
| Full verification | Annually, against a second gas or an independent instrument | Systematic error in the calibration set-up |
The bump test interval is short because the failure it catches is sudden. The calibration interval is longer because the failure it catches is gradual. Using one to replace the other removes half the protection.
A plant that span-calibrates twice a year and never bump-tests is protected against drift and unprotected against the cell that dies on a Tuesday. A plant that bump-tests daily and never calibrates knows the instrument responds and does not know to what.
The record is part of the calibration
An instrument is calibrated when the certificate exists. The record should carry the cylinder number, the certified concentration, the cylinder expiry, the reading before adjustment, the reading after, the gas temperature and the name of the person who did it.
The “before” value is the most valuable field on the form. A channel whose before-value is drifting steadily across three intervals is telling you the cell is going, and that is a replacement scheduled in advance rather than a failure discovered on a Sunday.
Calibration intervals, flow rates, balance gases and tubing materials vary by sensor type and by manufacturer. Confirm the procedure against the datasheet and the calibration instructions for the specific module, and against the site’s own permit requirements.