Electrochemical (EC) ammonia sensors measure NH3 via a redox reaction on a working electrode, producing a current proportional to concentration; commercial units reach detection limits of 0.1 ppm and accuracy of ±2% reading in the 0-100 ppm range [S1][S4]. Semiconductor (MOS) sensors instead use a heated metal-oxide film whose resistance changes when NH3 adsorbs onto the surface, and are specified for long-life, low-cost monitoring rather than low-ppm safety [S4][S7].
Both technologies compete in the same end-uses: industrial refrigeration, cold storage, fertilizer plants, livestock farms and sewage treatment [S4][S5]. The decision between them is driven almost entirely by the alarm set points and selectivity requirements of the installation, not by sensor cost alone.
Working Principle and Detection Limits
Electrochemical cells contain an electrolyte (aqueous or, in newer designs, an ionic liquid) that consumes NH3 at a sensing electrode and generates a current read by a potentiostat [S7]. This mechanism gives EC sensors a detection floor near 0.1 ppm and a usable range up to roughly 5,000 ppm in Danfoss' NH3 modules, with stated accuracy of ±2% [S1][S4]. MOS sensors detect gas through resistance change of a tin- or tungsten-oxide layer held at 300-450 °C; they register a percent-level change in conductance rather than a true ppm-resolved signal, so their effective resolution is coarser and their zero drift larger [S4][S7].
For LoRaWAN-class wireless nodes, most commercial ammonia products in 2026 ship with an EC front-end; some vendors offer a semiconductor variant in the same housing for cost-sensitive sites [S4]. NDIR (non-dispersive infrared) modules sit above both in accuracy and exceed 5 years of service life, but their higher cost keeps them out of the mainstream EC vs MOS decision [S4].
Accuracy, Range and Resolution Compared
Resolution options are 0.01 ppm and 0.1 ppm; 0.01 ppm is mandatory for early warning in enclosed livestock rooms, while 0.1 ppm is sufficient for trend tracking and general alarms [S4]. MOS datasheets rarely publish a meaningful ppm resolution; their output is treated as a relative concentration band rather than a calibrated value [S4][S7].
Range scaling also differs: 0-50 ppm modules are sized for indoor farm environments with low safety thresholds, 0-100 ppm for general industrial, and 0-1,000 ppm or 0-5,000 ppm for compressor rooms and cold storage where over-range damage from a sudden relief event is a real failure mode [S1][S4]. EC cells can be specified across all three bands; MOS sensors are most commonly offered in the wider 0-100 to 0-1,000 ppm bands because their signal-to-noise ratio collapses at sub-50 ppm concentrations.
Service Life, Calibration and Total Cost
EC ammonia cells degrade because the electrolyte is consumed by every exposure, including long-term background NH3; typical service life is 2-3 years in clean environments, dropping toward 12-18 months in a continuously exposed compressor room [S2][S4]. Calibration intervals of 6-12 months are the industry norm, and field replacement of the cell only (not the transmitter) keeps spares cost low [S1][S4]. Danfoss cites an expected 2-year life for its dedicated NH3 EC module in the 0-5,000 ppm range [S1].
MOS sensors are rated for 5-10 years because the gas-sensing reaction is largely reversible and there is no electrolyte to be depleted, but they require more frequent field verification because selectivity drift is harder to bound [S2]. The lower unit price of MOS modules is offset by the false-alarm cost they impose in safety loops, and the higher unit price of EC cells is offset by their selectivity and the regulatory acceptance of their output for IIAR 2 set-point actuation [S2].
Selectivity, Cross-Sensitivity and False Alarms
EC ammonia cells are described as highly selective to NH3 and rarely triggered by other gases or solvents in a refrigeration plant [S2]. MOS sensors, by contrast, show broad cross-sensitivity to carbon monoxide, hydrogen (from battery charging), alcohols, paints, cleaning solvents and engine exhaust, all of which are common in machinery rooms and cause nuisance trips [S2]. The selectivity penalty is severe enough that engine-room guidance written for ammonia machinery rooms calls out MOS cross-sensitivity as a primary cause of false alarms [S2].
This is the single most important technical reason EC dominates safety-instrumented loops: an EC output can be trusted to actuate emergency ventilation or equipment shutdown, while an MOS output should be treated as a trend indicator only [S2]. A related engineering reference details how to read and suppress EC false alarms at the algorithm level, which is the right follow-up read for spec engineers [S2].
Alarm Set Points Under ANSI/IIAR 2
ANSI/IIAR 2 mandates three ammonia alarm set points in a refrigeration machinery room: Level 1 at 25 ppm (ACGIH TLV-TWA, monitored alarm, signal to a constantly attended location), Level 2 at 150 ppm (automatic emergency mechanical ventilation, latched until manually reset inside the room), and Level 3 / ESD at 40,000 ppm or the detector's upper limit, which corresponds to 25% of ammonia's 150,000 ppm Lower Flammable Limit [S2]. Older installations often used the detector's upper range (commonly 10,000 ppm) as the ESD trigger; the current IIAR 2-2021 specification raises this to 40,000 ppm or the detector's upper limit [S2].
These set points drive sensor range directly. A detector that is expected to actuate Level 2 must resolve 150 ppm with adequate margin, which is well within EC capability; a detector that must also actuate Level 3 needs 0-50,000 ppm or 0-100,000 ppm coverage, and many EC cells are not offered at that span, so the spec is typically a second higher-range EC or an IR cell [S2]. MOS devices are not specified for any of the three IIAR 2 levels in safety-of-life applications [S2].
Environmental Limits and Installation
EC cells are sensitive to temperature and humidity extremes and to condensing conditions; refrigerated spaces and outdoor compressor skids need IP65 or higher enclosures, while dry indoor control rooms can use IP30 modules [S4]. MOS sensors tolerate harsher temperature and humidity, which is why they are still preferred for unconditioned sheds and agricultural buildings where the air is dirty and the cost of a false alarm is low [S2][S4].
Detector placement per the Danfoss refrigeration guide calls for sensors near the compressor, at the low point of the room where NH3 is heavier than air, and at the breathing zone of any operator workstation, with the number of detectors sized to room volume and air-flow pattern [S1]. For wireless LoRaWAN nodes, IP65 housings with 4.48 µA sleep current and 15-20 km open-air range have become the 2026 default for remote farm and tank-farm sites where running 4-20 mA loop cable is impractical [S4].
Selection Matrix: EC vs MOS vs NDIR
On four decision criteria drawn from the research: (1) Selectivity, EC wins outright, MOS is poor, NDIR is excellent; (2) Service life, MOS wins (5-10 years), EC is 2-3 years, NDIR exceeds 5 years; (3) Sub-1 ppm resolution, EC reaches 0.1 ppm, NDIR does as well, MOS does not; (4) Unit cost, MOS is lowest, EC is mid, NDIR is highest [S1][S2][S4]. The implication is a clean three-way split: EC for IIAR 2 safety loops and 0-100 ppm toxic-gas monitoring, MOS for residential or non-safety trend monitoring, NDIR for high-accuracy industrial and laboratory use where the budget supports it [S4].
A practical rule of thumb for spec engineers in 2026: specify EC for any ammonia loop that must actuate a regulatory alarm or interlock under EN 378:2016, ISO 5149:2014, ASHRAE 15-2016 or ANSI/IIAR 2; specify MOS only for household, agricultural trend or simple linkage monitoring where false alarms are tolerable; specify NDIR when ppm-level accuracy must be held over multi-year calibration intervals in dirty or wet environments [S1][S2][S4]. Two trackable signals to watch over the next quarter: vendors introducing EC cells with ionic-liquid electrolytes to extend service life past 3 years, and LoRaWAN ammonia nodes shipping with on-device selectivity diagnostics to flag EC drift before it crosses the 25 ppm Level 1 set point [S4][S7].