FAQ

What is NIR?

NIR is the abbreviation for Near InfraRed. NIR Spectroscopy, or NIRS, is a widely used method that can be applied to a wide range of research applications and industrial processes, helping to determine chemical and physical properties of raw materials without altering the sample. Within the electromagnetic spectrum, the NIR range is adjacent to the mid-infrared and extends up to visible region (from ~700 to 2500 nm). When NIR radiation is emitted to a given sample, this sample will absorb and reflect energy depending on its molecular composition in a very specific way, like a fingerprint. In chemometrics we call this fingerprint a spectrum. A spectrum is therefore a range of sequential wavelengths, showing a specific pattern of radiation absorbance, and therefore very rich in information about the composition of the measured sample. With specific sensors we can measure these spectra and, by applying chemometrics, we can extract the specific information we need from the sample.

What is single-point hyperspectral?

If we combine different spectral ranges in a measurement, we talk about hyperspectral measurements. A common example is to measure from 350 to 2500 nm, where radiation from the visible, NIR and shortwave infrared ranges would be acquired. The spectral range has been expanded, so the information we now have per sample has increased compared to NIR.

What shall I use, NIR or hyperspectral?

Whether to use NIR or hyperspectral depends on your specific needs, objectives, the nature of your application and therefore on the spectral range that might be optimal for you. Not sure what to use and whether you prefer single-point spectroscopy or imaging? It all depends on how you want to measure and work in the future. Contact us for further details.

What is chemometrics?

Chemometrics basically means interpreting any complex chemical data by applying mathematical and statistical methods. The main area where chemometrics is used is spectroscopy. As the information present in the spectra is very complex, not very selective and influenced by physical, chemical and structural variables, chemometrics is required to extract the maximum amount of relevant information from the spectra, applying appropriate statistical methods. The ultimate goal is then to obtain a spectral calibration that can be used in the future to predict a given property in a new set of samples.

What is a calibration?

A calibration in the context of spectral analysis is a critical step in ensuring the accuracy and reliability of spectral data and refers to the process of establishing a relationship between the measurements obtained from a spectrometer or analytical instrument and the known properties or concentrations of substances in a sample using a mathematical model.

Imagine that you have a set of samples, for example apples of different varieties. And you are interested in knowing the sugar content of these and other apples in the future. Now you measure your initial samples with NIRS. Each sample will have a specific energy absorption and reflectance fingerprint, or NIR spectrum. The differences between the spectra may be linked, for example, to the different sugar concentrations of the samples. These two data sets, spectra and laboratory data, will be correlated with each other, typically using PLSR. The result of applying PLSR will be an equation, the basis for the so-called calibration or predictive model. In this example, the calibration will later be used to predict the sugar content of new apple samples. Later, unlimited new samples can be measured with a spectrometer, and the sugar content will be predicted without the need for any additional laboratory analysis. This is why spectral analysis is considered a resource-saving technique.

Another important point is that to have a good calibration and, therefore, good predictions, the spectra need to be of high quality. This is because, not just the sensor used, but a carefully thought-out presentation of the sample, or in other words, the way a sample is measured, is essential.

How often do I need to recalibrate?

A recalibration is simply a calibration that instead of starting with new samples, starts from a previous calibration to which now, during the recalibration process, we will add new samples. The methods used and the procedure are the same as for a “from scratch” calibration. The goal is to improve a previous calibration by adding more variations to it. Whether and when a recalibration will be necessary depends on many factors, for example: what is my parameter of interest? What samples do I want to measure in the future? Does my calibration set already contain all possible variations that the prediction model will encounter in the future? Which error am I comfortable working with in the future? How many resources am I willing to invest to improve this error?

Especially when working with samples that can undergo any type of periodic or seasonal change, keeping the sample calibration pool updated by adding new samples when necessary is essential to maintain a good predictive capacity in calibrations over time.

Which industries can benefit from spectral analysis consulting?

NIR and hyperspectral analyses are spreading fast in medicine, agriculture, textile, chemical, pharmaceutical, recycling industries, etc. From basic research to practical applications, from public research centers to private companies, the possibilities are endless. By using spectroscopy, companies can authenticate products, detect frauds and non-conformities and control internal standards. Additionally, it is known that by using spectral analysis companies can save running costs by up to 90%. It is also knoen

Why shall we go for an external service instead of using internal know-how?

With an external service you can count on the assistance of experts, only when you need it. This can save your company fixed costs without losing data quality.

How many samples are needed to develop a model?

The calibration set must contain an adequate number of samples that uniformly cover a sufficiently wide range of analyte concentration. Inadequate calibration sets result in calibrations with poor predictive ability. The most representative samples that can be used to predict future samples should be included in the calibration set, and all expected sources of variability should be considered. There is no fixed number or rule of thumb for determining how many samples should be included in a calibration, as it can vary depending on the property of interest.