This article explains which specifications are most often misunderstood in practice and which questions you should ask before choosing a sensor.
1. Accuracy: percentage of reading or full scale?
The most important distinction in the entire datasheet comes down to the reference value.
± 2% of reading means two percent of the current measured value.
± 2% of full scale means two percent of the upper limit of the measuring range.
For a sensor with a measuring range of 0 to 5,000 ppm CO₂, this makes the following difference: at an actual concentration of 500 ppm, the deviation is 10 ppm in the first case and 100 ppm in the second. The same percentage, but ten times the absolute error.
Many datasheets combine both, for example "± 2% of reading ± 50 ppm". This specification is more informative than it may first appear because it states the unavoidable baseline error.
What to check: In which part of the measuring range does your application operate? If you measure in the lowest tenth of the range, the reference value determines whether the sensor is suitable.
2. Response time: T90, T63 or system response time?
Response time describes how quickly a sensor responds to a change in concentration. It is commonly specified as T90: the time taken to reach 90 percent of the final value.
Some datasheets specify T63 instead. This value is consistently shorter and presents the same sensor more favourably. Comparing a T63 specification with a T90 specification is not meaningful without conversion.
There is also a second level to consider. The specified time applies to the sensing element, not to your measuring point. Tubing lengths, filters, dead volume and flow rates extend the actual response time, often considerably. In process measurement, the relevant parameter is the system response time, not the response time of the sensing element.
What to check: Which definition does the specification use, and under which flow conditions was it measured?
3. Drift: per month, per year or over the service life?
Drift describes how the measurement signal shifts over time without a change in gas concentration. A distinction is made between zero drift and sensitivity drift.
The time reference is critical. A value such as "< 2% per year" is fundamentally different from "< 2% over the service life". Likewise, "typical" does not provide the same assurance as "maximum".
What matters to you is the practical consequence: how often does the sensor need recalibration, and what does that cost over its planned operating life? A sensor with a slightly higher purchase price and twice the calibration interval can be the less expensive option over five years.
What to check: The time reference, the distinction between typical and maximum values, and the recommended calibration interval.
4. Measuring range: why wider is not better
A wide measuring range may look like an advantage at first. Technically, it is usually a compromise.
The optical path length, response curve and calibration of an NDIR sensor are designed for a particular concentration range. A sensor covering 0 to 100 vol.% cannot achieve the same resolution at ppm levels as a sensor designed specifically for that range.
Also check which part of the range the accuracy specification actually covers. Many specifications apply to a defined section, rather than the entire stated range.
What to check: Which concentration range must the application actually cover, including fault conditions and start-up?
5. Cross-sensitivity: a short list is not a good sign
The list of cross-sensitivities shows how strongly other gases influence the measurement signal. A short list does not automatically mean a more selective sensor. It may also mean that fewer gases were tested.
The test conditions are decisive: what concentration of the interfering gas was used, and at what target gas concentration? A specification without these conditions is difficult to assess.
It also matters whether selectivity is achieved optically, through absorption band selection and filter design, or compensated for mathematically afterwards. A mathematical correction assumes that the gas matrix is known and remains stable. This is often not the case in processes with a changing gas composition.
What to check: Tell the manufacturer your actual gas matrix, not just the target gas. Include moisture as well.
6. Temperature range: operating range is not compensation range
Many datasheets state an operating temperature range within which the sensor functions. This must be distinguished from the range for which temperature compensation is designed and the accuracy specification is valid.
The two ranges can differ considerably. A sensor may operate at minus 20 degrees Celsius, yet only achieve its specified accuracy between zero and forty degrees Celsius.
What to check: Does the accuracy specification apply across the entire operating temperature range, or only under reference conditions?
7. Reference conditions: the sentence everyone overlooks
Almost every datasheet contains a line such as "Specifications apply at 20 °C, 1013 hPa, dry gas". This line qualifies everything stated above it.
Your plant probably does not operate under these conditions. Changes in process pressure, moisture and temperature have a measurable effect. The question is whether the sensor compensates for these deviations, and how well.
What to check: Which compensation functions are built into the sensor, and which do you need to implement in the higher-level system?
The cross-check: five details you should provide
A reliable recommendation requires a description of the measurement task. In practice, five details are usually enough for an initial, informed discussion:
| Specification | Why it matters |
|---|---|
| Target gas and concentration range | Determines the measurement principle, optical design and calibration |
| Complete gas matrix, including moisture | Provides the basis for assessing cross-sensitivities |
| Temperature, pressure and their variation | Determines the compensation and gas conditioning required |
| Required response time at the measuring point | Determines gas flow design, sampling and sensor type |
| Installation conditions and interface | Clarifies integration and whether diffusion or extractive measurement is required |
The more precisely you describe these five points, the sooner you can find a suitable solution. It also reduces the chance of discovering after commissioning that a datasheet value applied under conditions that never occur in your process.
Frequently asked questions
What do "v. M." and "v. E." mean in a German datasheet?
"v. M." stands for "vom Messwert", meaning "of reading". "v. E." stands for "vom Endwert", meaning "of full scale", the upper limit of the measuring range. At low concentrations, the same percentage results in very different absolute deviations depending on the reference value.
What is the difference between T90 and T63?
Both describe response time. T90 is the time taken to reach 90 percent of the final value; T63 is the time taken to reach 63 percent. T63 values are consistently shorter and cannot be compared directly with T90 values.
How often does an NDIR gas sensor need calibration?
This depends on the required accuracy, drift and operating conditions. The manufacturer's recommended calibration interval and the requirements of the application or relevant standard determine the interval.
Is a wider measuring range an advantage?
Usually not. A range designed for the actual application generally provides better resolution and accuracy than an unnecessarily wide range.
Conclusion
A datasheet describes a sensor under defined conditions. Whether it is suitable for your measurement task depends on what each figure refers to: which value, which range, which temperature and which gas matrix.
Clarifying these references lets you compare sensors, rather than brochures.
Describe your measurement task and have the sensor selection checked → Send an enquiry