Sampling, artifacts uniform

Any sampling scheme, whether uniform or nonuniform, can be characterized by its effective bandwidth, dynamic range, resolution, sensitivity, and number of samples. Some of these metrics are closely related, and it is not possible to optimize all of them simultaneously. For example, minimizing the total number of samples (and thus the experiment time) invariably increases the magnitude of sampling artifacts. Furthermore, a sampling scheme that is optimal for one signal will not necessarily... 

It is important to stress that ATR absorbance is strongly affected by the sample/crystal contact. Quantitative results are thus difficult to obtain even if the contact is maintained during the sample rotation that is required to record all four polarized spectra. A reference band that does not show significant dichroism is thus most often used to normalize the polarized absorbances in order to obtain quantitative data. For instance, the 1,410 cm-1 band of PET has often been chosen for that purpose, not only for ATR studies but also for specular reflectance (see below) and even transmission studies when the sample thickness is not uniform. It was shown that an appropriate normalization is possible even if no such reference band is available, by using a combination of two bands with orthogonal dichroism [34]. When performing ATR experiments, one should also make certain that the applied pressure does not create artifacts by affecting the structure of the sample. 

For physically uniform samples, simple techniques such as offset or polynomial baseline correction may suffice to minimize the physical/optical artifacts from the spectral data, but there are also a number of more advanced techniques available, which are quite effective. For example, it is common to convert the spectral data to a Savitzky-Golay33... 

For arbitrary, non-uniform sampling it is no longer possible to obtain a spectrum that is equal to a spectrum of continuous signal, even if it is strictly band-limited. Spectral artifacts, depending on the sampling schedule, appear as part of Point Spread Function. 

Fig. 12 Peak amplitude, artifact level, and signal-to-artifact ratio for spectrum of non-decaying signal of frequency 10 Hz, sampled with (a) 512, (b) 256, and (c) 128 points. Uniform random sampling was used...

HIDY The filter media used in this study was a teflon-coated glass fiber substrate, manufactured for us under very strict specifications. It was used uniformly for all aerosol sampling in the SURE. The filter medium is not subject to a sulfur dioxide adsorption artifact, but it is subject to adsorption of nitric acid. It does not adsorb NO or NO2, however. 

---