A gas detector is the sensor-plus-transmitter assembly that turns a target gas concentration into a usable signal — analog 4–20 mA, HART, Foundation Fieldbus, or a wireless ISA100.11a / WirelessHART packet — so a control system can alarm, ventilate, or shut down [S2]. Selection is rarely a brand question first; it is a gas-and-hazard question that cascades into sensor type, certification class, and form factor.
Five gates decide the build in 2026: (1) target gas, range, and cross-sensitivity; (2) sensor technology matched to that gas and environment; (3) hazardous-area certification (ATEX 2014/34/EU, IECEx, or North American NEC Class/Division); (4) mounting form — fixed gas detector on a wall or duct, portable gas detector worn by a worker, or a multi-gas detector confined-space unit; (5) output protocol and integration with the plant DCS/PLC. Skip any gate and the spec almost always returns for rework within 18 months.
Gate 1: Target Gas, Range and Cross-Sensitivity
The target gas defines the entire instrument. Combustible hydrocarbons (methane, propane, pentane) call for a pellistor / catalytic-bead or a non-dispersive infrared (NDIR) sensor; oxygen uses an electrochemical or a galvanic cell; toxic gases such as H2S, CO, Cl2, NH3, SO2 are dominated by electrochemical cells; refrigerants and SF6 leaks are typically NDIR or heated-diode [S3]. Range selection should bracket 25–75% of the LEL for combustible work, 1× TLV-TWA for routine toxic monitoring, and STEL/IDLH ceilings for life-safety stands — a single instrument rarely covers all three.
Cross-sensitivity is the failure mode that wrecks budgets. An electrochemical CO cell will respond to H2 at a non-trivial fraction of its CO signal; an H2S cell can be poisoned by silicone vapours; a pellistor can be inhibited by halogenated solvents and silicones and lose sensitivity permanently [S3]. The data sheet's cross-sensitivity table — not the marketing line — is the gate. For mixed atmospheres, a multi-gas detector with four independent sensors is the standard response, not a single "universal" cell.
Gate 2: Sensor Technology Matched to the Application
Sensor choice is environment-bound. Catalytic-bead (pellistor) sensors are cheap and proven for combustible gas at 0–100% LEL, but they require oxygen to operate, can be poisoned by silicones, and burn out in lean atmospheres [S3]. NDIR sensors are oxygen-independent, immune to poisoning, and dominant for CO2, refrigerants, and SF6 leak detection, with typical response times in the tens of seconds and a service life measured in years [S1][S3]. Electrochemical cells are the workhorse for toxic gases at ppm level — compact, low-power, but limited 2–3 year life and temperature-sensitive.
Photoionisation detectors (PID) cover VOCs from ppb to thousands of ppm and are standard for confined-space entry and refinery turnarounds. Metal-oxide semiconductor (MOS) sensors are low-cost and used in combustible gas detector consumer-grade products, but they are non-selective and humidity-sensitive — fine for a parking garage, not for a hydrocarbon refinery. A practical gate: match the sensor's poisoning profile to the process stream, then verify with a 30-day field trial before buying at scale.
Gate 3: Certification — ATEX, IECEx, and North American NEC
Certification is non-negotiable in hazardous areas. In the EU, ATEX 2014/34/EU equipment categories 1 (zone 0/20), 2 (zone 1/21), and 3 (zone 2/22) gate the equipment allowed; outside the EU, IECEx equipment-protection-levels (EPL) Ga/Gb/Gc do the same work and are mutually recognised with ATEX under the IECEx scheme. In North America, NEC Class I/II Division 1 or 2 (explosion-proof / intrinsically safe) and Class I Zone 0/1/2 (UL 60079 family) are the parallel regimes. Specifying a detector that is "ATEX rated" without naming the zone is a common spec hole — the certificate must state the exact zone and gas group. [S1]
For a fixed gas detector head on a bulk-solvent tank farm, expect Ex db (flameproof) or Ex eb (increased safety) enclosure for the transmitter, with the sensor itself often Ex ia (intrinsically safe) on a zener barrier or galvanic isolator. For a worker-worn portable gas detector entering a sewer, the whole instrument is typically Ex ia IIC T4 Ga. Functional safety adds a layer: SIL 1 or SIL 2 certification to IEC 61508 is increasingly demanded for detectors that drive an automatic shutdown, and that SIL rating belongs on the same purchase order as the Ex rating — not as an upgrade quote six months later.
Gate 4: Form Factor — Fixed vs Portable vs Multi-Gas
Form factor follows the work to be done. Fixed detectors monitor continuously at known risk points — bulk storage, compressor skids, battery rooms, refrigerant plant rooms — and live on a 4-wire or loop-powered 4–20 mA bus, often with HART for diagnostics. A typical spec calls for one detector per 10–12 m of unobstructed perimeter in a battery room and one sensor per gas point in a process skid. Portable detectors are issued to workers for hot work, confined-space entry, and leak survey; they run on rechargeable Li-ion with 12–24 h endurance and dock to a bump-test station. [S2]
Multi-gas detector units combine four sensors (LEL, O2, CO, H2S is the confined-space standard) and are the right call for any non-routine entry. A useful decision rule: if the worker is the alarm recipient and the question is "is it safe to enter", a four-gas portable is the default; if the question is "is this asset still safe to run", a fixed point detector is the default. A toxic gas detector on a fixed loop is the right call for continuous ppm-level monitoring of H2S at a sour-gas wellhead, where the worker is too far away to act on a personal alarm.
Gate 5: Output Protocol and Plant Integration
Output is the gate that locks a detector into a control system — or strands it as a stand-alone annunciator. The dominant plant protocol is still 4–20 mA analog, frequently with HART superimposed for remote calibration, range changes, and diagnostic trending. A HART detector can be queried from a handheld or a DCS asset-management screen for tag, range, last calibration date, and end-of-life signal — information that would otherwise need a physical visit. Foundation Fieldbus and PROFIBUS PA detectors integrate fully into a DCS for control and diagnostics on a single digital bus, at the cost of more complex commissioning. [S3]
Wireless options — WirelessHART (IEC 62591) and ISA100.11a — are now routine for remote or temporary points where cable tray is uneconomic: tank-farm perimeter, turnaround coverage, construction-phase monitoring. Wireless detectors typically run on a lithium battery with a 3–5 year life at a 30 s update rate. The spec gate is the plant standard: a greenfield refinery in 2026 will usually run 4–20 mA + HART as the workhorse, with wireless for <20% of points on tanks, water-treating areas, and remote well pads. Confirming the protocol against the DCS I/O card list before issuing a PO is the cheapest hour spent on a detector project.
Selection Comparison: Common Detector Types Side by Side
Four common instrument archetypes, lined up against the criteria that drive the purchase: [S1]
Catalytic-bead combustible detector (fixed): low unit cost, 0–100% LEL, requires oxygen, poisoned by silicones and halogens, 2–3 year sensor life, ATEX/IECEx Ex db or Ex ia variants available, 4–20 mA + HART typical. NDIR combustible/CO2/SF6 detector: oxygen-independent, poisoning-immune, 5+ year life, higher unit cost, T90 in tens of seconds, fixed or portable form factors [S1][S3]. Electrochemical toxic gas detector: ppm-level resolution, 2–3 year cell life, temperature- and humidity-sensitive, intrinsically safe variants dominate, 4–20 mA + HART. Four-gas confined-space portable: combines LEL/O2/CO/H2S, 12–24 h battery, audible/visual/vibrating alarms, mandatory for hot work and entry permits.
For a refinery, the spec usually lands as catalytic-bead LEL at process points, NDIR at storage, electrochemical H2S/CO at the wellhead, and a fleet of four-gas portables for entry. A gas detector inventory built to this template scales linearly with the number of risk points and rarely needs re-specification at the sensor-technology level — only at the protocol and SIL layer as the DCS evolves.
Limits, Failure Modes, and What Not to Do
The recurring failures in 2026 are not exotic — they are spec gaps. Detector specified for a gas it cannot see (NDIR for H2, electrochemical for methane, MOS for ppm toluene). Detector specified for the wrong zone (Ex eb in a zone 0 location, Ex d in a zone 2 cost-down). Detector specified without a calibration interval written into the contract — most electrochemical cells drift 2–5% per month and a 12-month calibration cycle is the floor, not the ceiling. Detector with a 4–20 mA output that the DCS I/O card cannot accept; the issue surfaces only at commissioning, and a card swap plus re-termination is a four-figure line item per point. [S2]
A second cluster of failures is environmental. Pellistor in a silicone-rich paint-booth atmosphere. Electrochemical CO cell in a -30 °C Canadian winter without cold-weather conditioning. NDIR with condensate on the optics in a humid coastal plant. Each of these is a gas detector that passes the FAT and fails the SAT. The procurement gate that catches them all is a 30-day in-service trial at the actual installation point, with bump-gas verification at the end. Bump test takes seconds; full calibration takes minutes; the wrong sensor in service takes years to discover.
Standards, Sourcing, and Audit Trail
For a procurement-grade spec, the document set is short and well-defined. ATEX 2014/34/EU for the EU, IECEx scheme for international, UL/CSA certifications for North America, IEC 60079 family for the underlying Ex protection concepts, and IEC 61508 / IEC 61511 for the SIL argument. The detector data sheet should state the exact certificate numbers, the gas group (IIA / IIB / IIC), the temperature class (T1–T6), and the EPL — vague "ATEX approved" language on a quote is a reject. [S3]
For sourcing, the 2026 supply side is concentrated: a handful of Western incumbents (Dräger, MSA Safety, Honeywell, Industrial Scientific, RKI) carry the bulk of the fixed gas detector and portable gas detector installed base in oil & gas and chemicals, while a deeper Chinese tier supplies regional combustible gas detector and toxic gas detector product at lower price points with comparable ATEX/IECEx certification [S1]. The decision gate is not origin — it is whether the certificate is on the nameplate and whether the sensor has a verifiable calibration certificate traceable to a national standard.
The downstream signal that the spec is being executed is the bump-test log: every detector, every 30–90 days, gas applied, response recorded, sensor replaced at end of life. A 2026 audit-grade site runs this log digitally, ties it to the maintenance CMMS, and shows the trend to the inspector in one screen. A site that cannot produce it has a detector, not a gas detection system — and that distinction is what separates a passing PSM audit from a finding.
For process engineers mapping a greenfield spec, a useful parallel is the spec-gates discipline used in adjacent equipment categories — the same five-gate structure (function, type, certification, form factor, integration) governs control valve selection criteria and VFD buying guide decisions, so a unified gate template is worth carrying into the detector purchase. The next trackable signal is the IECEx CoPC (Certificate of Personnel Competence) for the calibration crew, which has been tightening across major EPC owners since 2024 and is the audit item most likely to surface in a 2026 PSM review.