top of page


This post explains how we calibrate high end sensors and we will also look at the pros and cons of using high end calibration on low cost sensors

In our previous blog post, we explained major challenges faced by low cost AQ sensors. We have used the words sensor/monitor/instrument interchangeably.

Calibrating high end sensors

Normally, for high-end (expensive) sensors, the calibration lasts for little over a 1 year. Assuming they are not suffering from any mechanical damage e.g. due to vibration or tilt. Periodical calibration is recommended if you depend upon high accuracy and precision.

Example: NO2 sensor calibration for high end sensors

Here are two methods we use to calibrate an NO2 sensor.

Method 1: Zero and Span Check

In this method, we calibrate the instruments with the primary standard (certified gas cylinders). Main components needed are:

  • Dilution system with certified Mass Flow Controls (MFC).

  • NO2 Gas cylinder (1000 ppb) with the national and/or international standards such as Bundesanstalt für Materialforschung und -prüfung/BAM, National Institute of Standards and Technology/NIST, Nederlands Meetinstituut/NMi, NPL.

  • “Zero air” Gas cylinder with the above mentioned standard.

  • Sampling manifold.

A dilution system, fed by the gas cylinders generates different NO2 mixing ratios. This is then fed to the instrument together with Zero Air. This is typically called a "zero and span check". A calibration curve (x-y plot) is fitted to get a linear curve (y=ax+b). The parameters of this curve can then be used to scale the latter instrument output.

Method 2: Intercomparision with certified instruments

This typically requires the following steps.

  • Use a professional (typically very expensive) calibrated instrument, with certified standards.

  • Put your instrument(s) inlet side-by-side with the calibrated instrument for a while to measure the ambient NO2.

  • Plot and fit the linear curve with both outputs. Be careful about the conclusions you make under these conditions. e.g. You may want to use your sensor to measure 100ppb NO2, but your environment may never reach that level. It is like trying to measure 100m with a 5m long ruler.

Calibrating low cost sensors

For most of us, a low-cost device is more accessible compared to the high-end instruments. Unfortunately, today there are no standards for such sensors, only vague guidelines. This has led to a flood of low quality monitors and confusion for the consumer.

Additionally, even if some manufacturers claim they are tested and calibrated, not all manufacturers offer genuinely good quality. The calibration may only last for a weeks/ months due to different calibration methods and different instrumentation quality.

Can we calibrate low cost sensors in the same way as the expensive sensors?

It is possible to apply the 2 calibration methods used for high end sensors, however there are pros and cons that must be considered.

Method 1. Zero and span check

We already know that the low-cost sensors are highly affected by environmental parameters. The gas generated from gas cylinders is normally 25 degree Celsius, 0% RH, which is ideal for calibration but not for a real world application. So you have to manually change the environmental parameters for a wider range calibration.


We can easily control the parameters such as Temperature, relative humidity, gas mixing ratio.


The low-cost sensors have a feature called “cross-interference”. e.g., the NO2 sensor may response to NO gas and O3 gas. This can make things unpredictable.

2. Intercomparision with certified instruments


It is possible to test and calibrate the sensor under real world conditions e.g. high traffic.


We cannot simulate every ambient condition e.g. extreme weather. So tuning the sensor can take some time while we can get enough samples from the required conditions

We often use different calibration contexts e.g. field calibration using high and low cost instruments. Here are some pictures from field calibration...

Whats next ...

We have been continuously improving the methods and algorithms required for Calibration. Moreover an automatic calibration method is also undergoing field trials. Stay tuned.

Sheng Ye, Research Scientist, Signify GmbH München

Rise of the low-cost sensor

Air pollution is a hot topic in the recent years. A growing number of people care about the surrounding air quality. Normally, reliable air quality data can only be obtained from the Environmental Protection Department (EPD), which measure air quality using traditional air monitoring instruments.

It is a challenge for government(s) to service the rising public demand for air quality information. Traditional air monitoring stations generally use optical instruments that offer precise and accurate air quality data, but at a high cost. Due to this limitation, dozens of devices/instruments for air quality measurement made of low-cost sensors have popped up in the market.

Low-cost sensors react to the target pollutants/gases via chemical reaction (electrochemical), optical absorption (NDIR), light scattering or absorption (optical particle counter) and so on. They are portable, low-cost & highly customized.

But they face some key challenges.


Ambient interference & Calibration

Firstly, they are highly affected by the ambient environmental factors (e.g. Most PM sensor may fluctuate with the relative humidity, and may output “crazy” values while raining). Also, some of the low-cost sensors will not only react to one pollutant, in the other word, they are facing cross interference issue (e.g. nitrogen dioxide sensor is normally showing the reaction with ozone and/or other nitrogen oxides). So:

• If you directly purchase the sensors and DIY your own device: The manufacturer may offer the initial converting factors, which is most likely obtained under a stable laboratory environment. That means you will probably get completely wrong measurement values (even negative) if only using the manufacturer parameters.

• If you spend more money, and purchase the ready-to-go device: The seller may (or may not) solve the calibration issue we mentioned above with several traditional expensive reference air quality monitoring instruments. You are getting a “black-box” device and have to determine whether to trust the measurement values you get.

A bunch of researches/projects could be found online which are utilizing these sensors without mentioning the data quality.

Aging & Drift

Secondly, as we know, almost all electronic devices are encountering the aging issue. The low-cost sensors are certainly among this group. People should realize that these sensors are “low-cost” as a kind of consumables. For example, a high quality (with a higher price) electrochemical sensor cell could last for two years with 80 percent decayed outputs. Hence, a drift of the measurement values will raise at a time point.

Regulation & Trust

Finally, a key problem is that there is no regulation for the low-cost sensors/devices, which leads to the chaotic market. Undeniably, the low-cost sensors is a “gold mine” if they could be used correctly. Currently, researchers and manufacturers are putting a lots of effort into regulation.

What we do

We collaborate with Tracegas group, LMU, who are experts in air quality monitoring and assessment to tackle these challenges. Our algorithms are undergoing field trials in Munich with positive results. We want to enable a new era of air quality measurement.

Stay tuned.

- Sheng Ye, Research Scientist, Signify GmbH.

Understanding log, machine or sensor data is not trivial. Each team has different perspectives on data.

Engineers that can make sense of it are often far away from the product that generates the data. System managers are interested in knowing what happened. Operators and testers need to store data securely and catch failures.

Our toolchain helps different teams get better data insight in 3 steps:

  1. Operators and test teams can Collect and Inspect data with Warehouse and Sarge in a secure and scalable way.

  2. System managers can understand what happened with Dashboards. Vectors, Significants and Rangoli lets them create charts easily.

  3. Engineers can understand why it happened with Parade. They can slice and dice data. Even invite other engineers to investigate a problem together.

Our toolchain helps teams reduce effort and collaborate better. We helped a client reduce time from 2 hours down to a few seconds for a recurring process.

bottom of page