Sporametri, processed datasets


This sensor data has been processed by applying the following steps:
* Data including artefacts deriving from sensor power cycles has been removed.
* Sensor offset level (0 level) drift caused by temperature and humidity effects has been compensated.
* Sensor noise level has been filtered slightly to remove signal alteration which derives from effects not connected to sensor response towards gas concentration change.
 
The remaining dataset still contains noise, which can be distinguished as a somewhat symmetrical (both positive and negative readings) shape around 0 ppm level. The data with this kind of noise consists always of false positive and false negative readings with an amplitude smaller than the limit of detection (LOD) of the sensor in each moment. The LOD is the minimum concentration that can be detected as significantly different from the background. The background is the sensor signal in the absence of target compounds. IUPAC (International Union of Pure and Applied Chemistry) [AS1]defines the limit of detection as three times the standard deviation from the background and this is one option to be used to determine LOD for these sensors. In this dataset the amplitude of noise is not constant but dynamic in nature due to complex radiofrequency (RF)[AS2] interference during the data collection. The sensitivity towards this kind of interference is due to early prototype design of used sensor electronics.
 
Signals can be distinguished from noise by searching for positive readings which are above the LOD of the sensor at the time of the observation. Please note that the LOD and background are not constant values and should be calculated for each potential signal separately. Also, a signal is always longer than a single data point and increasing and decreasing edges of signal can be distinguished. Also, in many signals (but not all!) several sensors have reacted at the same time indicating a mixture of target compounds.
 
Please note that further processing of data is recommended and used algorithms and tools should be selected depending on the objective at hand.