Designing a laboratory gas detection system: sensors, mounting height and what happens after the alarm

A detector on the wrong wall is worse than no detector, because it buys false confidence. How to choose sensing technology, place heads by gas density and airflow, set alarm levels, and keep the system honest with bump tests and calibration.

A detector mounted at eye level on the corridor wall, four metres from the cylinder store, will read zero through an entire hydrogen release. It will also pass its annual test, appear on the asset register and satisfy a casual inspection. That is the real hazard of a badly designed detection system: it fails while looking exactly like a working one.

Fixed and portable do different jobs

These are not alternatives and one does not reduce the number of the other you need. A fixed system watches a place continuously whether or not anyone is there. A portable instrument watches a person, and only while they are carrying it and it is switched on.

Question	Fixed system	Portable instrument
What is it protecting?	A room, a cylinder store, a pit, a plant room, the area around one process tool	The individual wearing it, wherever they walk
When does it work?	Continuously, including nights, holidays and during load-shedding if it is on the essential supply	Only during the shift, only if bump tested and switched on
Can it act?	Yes: sounders, beacons, extract boost, solenoid shut-off on the gas line, BMS and security desk	No, it warns one person
Where it is the wrong choice	Occasional maintenance work in a space with no permanent risk	Any unoccupied risk, and any space entered before the hazard is known
What it needs from you	Cause and effect matrix, calibration schedule, an owner for the response	Daily bump test and a docking or test station

The pattern we see most often in Bangladeshi laboratories is a good fixed system with no portables, so nobody can safely investigate the alarm it just raised. Confined space entry, cylinder store checks and roof plant visits all need a personal instrument that detects oxygen deficiency as well as the target gas.

Choosing the sensing technology

Sensor type is chosen by the gas and by what else is in the room. Each technology has a specific way of going wrong, and specifying around that failure mode is most of the engineering.

• Electrochemical cells cover toxic gases at parts per million and oxygen deficiency. They are the mainstay for carbon monoxide, hydrogen sulphide, chlorine, ammonia and oxygen. The cell is consumed as it works, so it has a finite life whether or not it ever sees gas, and heat and humidity shorten that life. In an unconditioned plant room in Dhaka, plan on replacement sooner than a European service interval suggests.
• Catalytic bead sensors measure flammables as a percentage of the lower explosive limit, and will see hydrogen. They need oxygen present to work, so they read low in an inert or oxygen-depleted atmosphere. They are poisoned by silicones, sulphur compounds and halogens, and silicone is everywhere in a laboratory: sealants, mould release, some hand creams. A poisoned bead reads low and gives no warning, which is why a bump test is not optional on this technology.
• Infrared sensors measure hydrocarbons and carbon dioxide without being poisoned, work in inert atmospheres, and fail safe when the optical path is blocked. They cannot see hydrogen at all, because a symmetrical diatomic molecule does not absorb infrared. Specifying infrared on a hydrogen risk is a mistake that survives design review more often than it should.
• Photoionisation detects volatile organic compounds down to low parts per million, which no other common technology does well. It is not selective, so it tells you that something organic is present rather than what it is, and the lamp needs cleaning. Humidity affects the reading, which is a live issue here for most of the year.
• Extractive sampling, as used by the Honeywell Midas and Midas S2, draws a sample down a tube to a detector mounted somewhere accessible: inside a tool enclosure, a duct, or a high ceiling void. It moves the calibration point to somewhere a technician can stand, which is why it is the standard choice on semiconductor and process tools. The trade-off is transport delay down the sample line and a pump to maintain, so confirm the maximum line length and the flow fault alarm at specification stage.
• Networked area monitors such as the Honeywell E3Point suit building-scale duties: car parks, plant rooms and utility spaces, where the value lies in many points reporting to one head end rather than in laboratory-grade sensitivity at one.

Placement: density sets the height, airflow decides the rest

Mounting height is the first thing to get right and the easiest to get wrong, because installers put detectors where the conduit run is convenient. Gas density relative to air decides where the cloud goes.

Gas	Behaviour relative to air	Where the head belongs
Hydrogen	Much lighter, rises quickly and collects in ceiling pockets	High, at the highest point of the ceiling or void above the source
Methane and natural gas	Lighter, rises	High, allowing for beams and coffers that trap gas
Ammonia	Lighter than air, but a cold liquid release behaves heavier at first	High for the vapour risk, plus a low head near a liquid release point
Carbon monoxide	Close to air density, mixes with room air	Breathing zone height, in the occupied area
Chlorine, sulphur dioxide, hydrogen sulphide	Heavier, sinks and pools	Low, near floor level and near drains, pits and stairwells
LPG, propane, butane	Heavier, flows along the floor	Low, and in any pit or trench the gas can reach
Most solvent vapours	Heavier at room temperature	Low, close to the bench or store where the release starts
Oxygen depletion from nitrogen, argon or liquid nitrogen	Displaces air, fills low spaces first before mixing	Breathing zone, plus a low head in basements, pits and cold rooms

Density gets you to the right height. Airflow gets you to the right wall. Put the head between the likely release and the people, in the path the air actually takes, and confirm that path with smoke before you drill. Do not mount a head inside the extract duct where the sample is already diluted, in the direct throw of a supply diffuser, or where the wall gets hosed down during cleaning. And mount it where a technician can reach it with a calibration gas cylinder without a ladder and a work permit, because a head that is awkward to calibrate is a head that will be skipped.

Alarm levels and what happens next

An alarm with no defined action is noise, and noise gets silenced. Set two levels, write down what each one means, and name the person who does it.

1. First level, alert. For toxic gases this is normally set against the occupational exposure limit for the substance, and for flammables against a low fraction of the lower explosive limit. It means investigate: local sounder, panel indication, a message to the responsible officer. Nobody evacuates.
2. Second level, act. A higher setting that means leave and do not re-enter. Beacon and horn, and this is where the automatic responses belong: close the solenoid valve on the gas supply at the cylinder, stop the process tool, boost the extract if boosting will help and inhibit it if it will draw the cloud through occupied space.
3. Fault. A separate, distinguishable signal for sensor failure, flow fault on a sampling detector, or loss of supply. A detection system that cannot tell you it has stopped working is the failure mode described at the top of this article.
4. Escalation out of hours. Most laboratories here are unoccupied from evening to morning. Route the panel to the security desk and to a duty engineer's phone through the building management system, and rehearse who calls whom.
5. Post-alarm procedure. Write down who may re-enter, with what instrument, and what reading permits normal work to resume. It is the step that is almost always missing, and the one people need at two in the morning.

Two Bangladesh-specific points. The gas panel needs its own battery backup and the essential supply, or a load-shedding event silences detection at exactly the moment the extract fans stop and gas can accumulate. And the sounder must be audible over the fume hoods and the air handling plant, measured with everything running rather than in a quiet building on a Friday.

Bump testing and calibration are not the same thing

A bump test applies gas and confirms the sensor responds and the alarm activates. It answers one question: is this thing alive? A calibration applies a certified concentration and adjusts the reading so the number means something. You need both, on different schedules.

• Portables should be bump tested before each period of use. A docking station makes this a ten-second task and keeps the record automatically, which is the only way it will actually happen every day.
• Fixed heads follow the manufacturer's stated interval, and additionally after any alarm event, after a sensor change, and whenever poisoning is suspected on a catalytic bead.
• Shorten the interval for the environment. Heat, humidity, dust and corrosive atmosphere age sensors faster than the datasheet assumes. Use the as-found readings from previous calibrations to justify the interval you set.
• Calibration gas is a lead-time item. Certified cylinders carry an expiry date, they are hazardous goods, and they are imported. Order the replacement while the current cylinder is still in date, and keep the certificate with the calibration record.
• Pre-calibrated sensor cartridges, as used on the Midas S2, move the skilled work to the factory. The cartridge arrives calibrated, carries its own part and serial number and date, and a technician swaps it without opening a live enclosure. For a laboratory without a dedicated instrument technician, that is often the difference between a system that is maintained and one that is not.
• Keep a register per detection point: location, gas, range, setpoints, sensor serial number, last bump, last calibration, next due, and who signed. Auditors ask for this by name and they check the dates.

Commissioning: the tests to insist on

1. Apply gas at every head, not a sample of them. The point is to find the head wired to the wrong channel, and that head is never in the sample.
2. Sign off a cause and effect matrix with the laboratory before commissioning, listing every input and every output. Test each row against the matrix and mark it off.
3. Verify each point at the building management system and at the security desk, with the correct location text. A panel that says nothing more than 'Zone 3' at three in the morning has not been commissioned.
4. Prove audibility and visibility in the occupied space with the ventilation running, and from behind a closed fume hood sash.
5. Confirm the automatic actions physically happen: watch the solenoid close and the tool stop, and time it.
6. Take handover of as-built drawings showing every head with its height above floor level, the sample line routes and the setpoints as commissioned. Without heights on the drawing, the next engineer cannot tell whether a head is in the right place.

Back to all Insights