A client was having trouble getting correct readings with our SprintIR 60% CO2 sensor. They were attempting to use the sensor to read exhaled carbon dioxide in an application similar to capnography, but were getting unexpected results.
To test the sensor, the client was breathing directly on the sensor and graphing the CO2 levels in real time using our GasLab® software. The software reported that the sensor peak readings were flat (see red lines). After verifying the sensor was not damaged, the client asked us if we knew any reason why they might be getting bad data.
Fortunately, we were able to duplicate the result in our lab. In the first plot, we found that we could saturate the sensor by blowing hard on it about 1 inch from the sensor membrane surface. In the second plot, we blew on the sensor from a distance of about 4 inches and achieved more accurate results.
With some experimentation, we found that we cold saturate the sensor at will, and then desaturate the sensor by sucking very close to the surface. The hot breath was hazing the optics. Conversely, by sucking the air near the membrane surface we were changing the pressure so that the condensate evaporated and the sensor began operating correctly virtually instantaneously. The same effect can be found by exhaling close to a mirror.
Although we had solved the mystery, we recommended the client use a vacuum pump and hydrophobic filter setup similar to the one we've tested for capnography to achieve the most accurate results.