An electrochemical sensor listed for one target gas will, in practice, register a measurable response to several others, and the spread is wider than most datasheets imply: a representative cross-interference matrix shows H2S sensors responding at roughly 300% when exposed to HCl, 25% to SO2, and 25% to NH3, while CO sensors reading 5% on H2S and 20% on H2 [S3].
Cross-sensitivity is inherent to the sensing chemistry rather than a defect, and a single target-gas reading in a mixed atmosphere is therefore never a clean number without a supporting interference table [S2][S4]. The numbers that follow use the Industrial Scientific reference table collected at 20 °C, 50% RH, 1 atm, on new sensors, and they are the right starting point for sizing a toxic gas detector stack.
What Cross-Sensitivity Actually Means in an Electrochemical Cell
Cross-sensitivity is the sensor response to any gas other than the one it is specified for, expressed as a percentage of the reading the same concentration of target gas would produce [S2][S3]. A 100% reading on the cross-gas means identical sensitivity to the cross-gas and the target; a 300% reading means the cross-gas drives the cell three times harder per ppm than the target it is supposed to see [S3].
Three engineering facts drive the magnitude. First, the working electrode potential is not perfectly selective, so gases with similar oxidation/reduction potentials also react, which is why a phosphine (PH3) cell can show 80% response to CO, 415% to H2, and 425% to HCN depending on variant [S3]. Second, most cells include a chemical filter on the face to scrub the most common interferents, and that filter ages; once the scrubber is saturated, the cross-sensitivity numbers published for a "new sensor" no longer apply [S3]. Third, temperature, humidity, and pressure shift both the baseline and the cross-sensitivity coefficients, which is why Honeywell's TN-114 baseline of 68 °F (20 °C), 50% RH, 1 atm is the only condition under which the matrix is strictly valid [S2].
Reading a Cross-Interference Matrix: A Worked Example
Using the published matrix as a decision tool is faster than guessing [S3]. Take a 10 ppm H2S exposure: the same atmosphere will drive a CO cell to about 0.5 ppm equivalent (5%), a SO2 cell to about 0.5 ppm (5%), and an HCl cell to about 30 ppm equivalent (300%), which can be a far louder alarm than the H2S reading itself [S3].
The same matrix also shows why "H2 Low" CO sensors exist: a standard CO cell reads 20% on H2, but the H2-Low variant reads only about 3% on H2, at the cost of somewhat higher response to other interferents [S3]. For a refinery or chlor-alkali facility, that single column change is often the difference between weekly false alarms and a quiet detector, and it is a meaningful selection criterion when comparing multi-gas detector configurations.
Selection Criteria That Reduce, Not Eliminate, Cross-Sensitivity
No factory-published sensor has zero cross-sensitivity; the engineering job is to suppress the dominant interferents for the application [S4]. Four levers are commonly available.
First, pick the correct target-gas cell. PH3 cells show extreme cross-sensitivity to CO (80%), H2 (415%), and HCN (425%); a SO2 cell in the same atmosphere is far quieter, with a 0% CO response and roughly 40% NH3 response, which is why gas species selection is the single highest-leverage decision in any fixed gas detector design [S3]. Second, choose the H2-resistant variant where hydrogen is present; the CO (H2 Low) cell is a documented example, dropping H2 response from 22% to 3% [S3]. Third, use bias voltage and warm-up discipline. NO and NH3 sensors require a bias voltage and at least 6 hours of equilibration after installation before the baseline is stable enough to calibrate, while unbiased sensors need about 10 minutes [S2]. Skipping that step is a documented field source of nuisance alarms. Fourth, stack redundant sensors with overlapping coverage and apply a voting logic, an approach relevant to any portable gas detector carried into a confined space, and the same logic as the sensor-stack architecture used in chemical plant deployments.
Real Use Cases Where Cross-Sensitivity Drives the Design
In wastewater treatment, H2S and NH3 co-exist; an H2S-only toxic gas detector without an NH3-aware compensation table will over-read in digester rooms, so dual H2S + NH3 cells with a documented cross-interference review are standard [S5]. In semiconductor cleanrooms, HCl and HCN are used alongside solvents; an HCN cell shows 35% response to HCl and 30% to H2, which complicates leak detection during photoresist operations [S3]. In pharmaceutical and laboratory enclosures, alcohol-based cleaners drive false CO readings: a CO cell reads about 50% on ethanol once the scrubber is aged, which is why maintenance-window alarms are a known pattern and not a sensor fault [S3][S4]. In refrigeration service, refrigerant leaks and PH3 from rodenticides can co-exist, and PH3 cells cross-react strongly with both, requiring either a different target species or a multi-cell stack with voting.
Limits, Failure Modes, and What the Matrix Does Not Tell You
Published cross-sensitivity numbers are valid only for new sensors, at 20 °C, 50% RH, 1 atm, and they vary from cell to cell, so they should be treated as a guide rather than a tolerance [S2][S3]. Aged sensors can show dramatically different behavior: a "new" CO cell reads 0% on ethanol, but the same part aged or with a saturated filter reads around 50% [S3]. Filters also have finite capacity; ammonia sensors in particular are documented to fail when exposed above 200 to 300 ppm, which is well below the Lower Explosive Limit and within the range of process upsets [S2].
Non-chemical failure modes mimic cross-sensitivity. Condensation on the cell face, rapid ambient temperature swings, ventilation cycling, steam, dust, and the post-power-up stabilization period all produce transient signals indistinguishable from a real gas event [S4]. For flammable atmospheres, the same caution applies to combustible gas detector cells, which respond to any reducing gas, not just the calibration gas, and that is why field technicians are advised to walk the detector log against cleaning schedules, startup events, and HVAC cycling before assuming a process leak [S4].
Engineering Practice: Calibrate, Document, Re-Check Annually
For a defensible installation, three disciplines hold up. Calibrate with a traceable gas of certified concentration at the same 20 °C / 50% RH / 1 atm baseline the manufacturer used, and apply the published cross-sensitivity matrix as a software correction where the controller supports it [S2][S3]. Document the assumed interferent list for each sensor location, including cleaning solvents and known process by-products, so that future alarms can be triaged against that list [S4]. Re-evaluate the matrix annually, since the numbers are stated for new sensors and drift as the cell and its filter age, and a sensor that has been in service for 18 to 24 months is no longer described by the column heading marked "new" [S2][S3][S8].
Where cross-sensitivity cannot be reduced to an acceptable level, change the sensing principle: NDIR for CO2, PID for VOCs, and catalytic bead for combustible gases are all less prone to electrochemical cross-interference in their target ranges and are commonly paired with electrochemical cells in a four-gas instrument [S1][S7]. The realistic target is not zero cross-talk; it is a documented, audited, and reviewed matrix that lets a technician tell a real release from a cleaning event within the first ten minutes on site.
For related coverage, see Ethernet-APL vs 4-20 mA HART for temperature transmitters: a 2026 spec-level comparison.