Case study: Benefits of NDIR over TC CO₂ sensors inside high humidity CO₂ incubators

Inside cell culture laboratories, CO₂ incubators are one of the most important tools to grow living cells. Those cells vary from simple cell lines to human embryos grown for IVF purposes. The incubator mimics the human body as closely as possible so the cells feel comfortable and multiply successfully. Alongside a stable temperature, pH level is critical. CO₂ sensors, in combination with CO₂ gas, bring the pH value to the correct level, which makes an accurate CO₂ reading inside the incubator essential.

XiltriX International has been installing temperature and CO₂ monitoring systems inside incubators for more than 30 years. Over that time a lot has changed, especially the sensor technology.

Laboratory staff who witnessed this evolution can still remember the difficulty of measuring CO₂ manually with a chemical Fyrite meter. Used correctly, an accuracy of roughly ± 0.5% could be achieved. One of the first measuring principles to supersede the Fyrite meter was the widely used TC (thermal conductivity) CO₂ sensor. It used the electrical resistance through the air to measure how much CO₂ was present inside the incubator. When it detected too little CO₂, it let in more gas. With all other parameters correct, the accuracy of that sensor was acceptable, but it had one fatal flaw: it was terribly susceptible to changes in humidity.

The next evolution was the NDIR (non-dispersive infrared) CO₂ sensor. At first it had one downside compared with the TC sensor: it was a lot more expensive, and it suffered from drift. After many years of development, current generations of NDIR sensors are very stable and run for prolonged periods without frequent manual calibration.

Now back on topic: why an incubator is considered a black box. Ask a lab technician what the CO₂ percentage in their incubator is and they will look at the display and read out the value they see. They usually have no idea which sensor technology produced that value, and therefore cannot judge whether it is correct. Most incubators on the market today use an NDIR sensor, but not all.

XiltriX relies solely on NDIR sensors for monitoring inside CO₂ incubators. Depending on the brand and type of incubator, multiple form factors are used: some sensors withstand high-temperature decontamination, others fit inside a smaller access port, but all are very accurate.

One of our customers contacted us with a question. The XiltriX sensor was consistently measuring 3% where their incubator displayed 5.5%. The incubator had not been opened for several days, so as not to disturb the culture, and the lab did not own a handheld CO₂ meter to check which value was correct.

Not having an accurate handheld CO₂ analyser in this kind of situation is tricky. Without the ability to measure CO₂ or pH, it is nearly impossible to gauge which device is right. The customer had no reason to suspect the incubator: it had recently been serviced and calibrated by the manufacturer, it sounded no alarms and showed no external signs of fault.

After remote consultation, the XiltriX technical team asked the customer to check whether the sensor inside the incubator was positioned correctly and whether any condensation was present. Condensation can influence an NDIR reading if condensate drips inside the sensor tip and blocks the light source or the detector. After that check the CO₂ values slowly started to recover. Nothing appeared to be wrong, aside from a fleeting comment that the customer had refilled the water tray of the incubator.

That comment triggered alarm bells. Our engineers inquired about the CO₂ sensor and found it was of the TC type. This is not always clearly marked on the outside of the incubator, and one incubator model is sometimes supplied with different sensor types. The customer was not aware of the type of sensor used. Without that knowledge the graph made no sense; knowing a TC sensor was involved made everything clear.

The incubator had slowly run out of water. The water had evaporated and had not been refilled in time. Because the cells inside were not very susceptible to excess evaporation of culture media, the lab technicians thought nothing of it. What they failed to recognise is that the lack of humidity also changed the thermal conductivity of the air, making the incubator "think" it was running at 5.5% CO₂ while it was actually operating at 2.5%.

Why the water tray had run dry is not clear. It may have been the amount filled in the first place, human error in forgetting to refill it, or simply more frequent door openings than usual for this device. The graph shows it takes multiple hours for the CO₂ values to return to their setpoint, because evaporation is slow. Without an NDIR CO₂ sensor as an independent monitoring device, this failure would probably have gone unnoticed, and running the incubator at such an incorrect CO₂ value has a profound effect on the pH of the growth media, and on the cells growing in it.

In 20 years in laboratory monitoring we have seen many devices and technologies fail. Some through unfortunate events, others through human error. One thing stands out every time: one equals none. Relying on a single sensor means running into problems at some point. In 2012 the Harvard Brain Bank lost a -80 °C freezer with priceless donated brain samples, throwing autism research back by years. Harvard had invested in monitoring technology; the available information describes separately failed circuits, but does not say whether those circuits were driven by a single temperature sensor and controller or by independent ones.

When only the onboard controller and its sensors are monitored, in other words the small computer that operates the freezer or the incubator, you are led to believe monitoring is in place. If the controller or the circuitry itself fails, no alarm sounds. There is no independent referee in place, nor in control. The device assumes all parameters are running at optimum settings. The empty water pan never caused an alarm, and the TC CO₂ sensor could not detect the difference: it was properly calibrated for use at high humidity, as our example shows.

Running a lab without monitoring is an unmitigated risk. Running a lab with a monitoring system built on dependent sensors is equally risky. A monitoring system that relies on dependent sensors reduces risk on average, but serious damage remains very likely. A critical incident leads to reputation damage and a very real risk of a liability case, not counting the direct losses in labour, investment and science. On top of that, precious time is lost because experiments have to be redone: their results are no longer reproducible.

Customers regularly ask about connecting devices digitally through their available communication ports. That is not a bad idea in itself, because it gives the monitoring system more metadata and more powerful insight into what happened. It should be seen as an extra option, though, and never as a replacement for independent monitoring sensors.

XiltriX will continue to develop and evolve partnerships with our customers to provide the most reliable and accurate monitoring available.

Do you want help or advice with the monitoring of incubators or other devices? Contact XiltriX at sales@xiltrix.com or visit xiltrix.com.