Sampling, artifacts random

Case III - Analyte Detection (A - random). One of the most reliable modes of simple (single) analyte detection obtains when y, B and A are each measured (observed) for every sample processed. Such is often the case, for example, in radiocarbon dating where the age of each unknown artifact is estimated from sequential measurements of the sample, the background, and the radiocarbon dating standard. Thus,... 

FT IR instruments can have very high resolution (<0.001 cm-1). Moreover since the data undergo ana-log-to-digital conversion, IR results are easily manipulated Results of several scans are combined to average out random absorption artifacts, and excellent spectra from very small samples can be obtained. An FT IR unit can therefore be used in conjunction with HPLC or GC. As with any computer-aided spectrometer, spectra of pure samples or solvents (stored in the computer) can be subtracted from mixtures. Flexibility in spectral printout is also available for example, spectra linear in either wavenumber or wavelength can be obtained from the same data set. 

As mentioned previously, partial glycerides are artifacts of the extraction process, especially the stages prior to sterilization. Oil obtained from unbruised sterilized fruits shows trace levels of partial glycerides. Random analyses of samples of refined palm oil, pahn olein, and palm stearin have shown the presence of about 2% of 1,2-diglycerides and about 4% of 1,3-diglycerides with trace amounts of monoglycerides. These partial glycerides are important as they are known to affect the crystallization behavior of the oil. 

It is knownis that small metallic particle effects arising from random depolarization factors in highly anisotropic metallic grains in ceramic samples can lead to artifacts in the infrared reflection (and absorption) spectrum. Since the samples used in the photoinduced absorption experiments are insulating micron sized crystallites embedded in an insulating medium, such small particle effects will not affect either the infrared spectrum or the photoinduced IR absorption spectrum where the measured AT/T is of order i0-3. 

As mentioned above, the nuFT does not find the optimal solution that fits to the experimental data. Spectra obtained by nuFT suffer from additional artifacts, which, in the case of random sampling, take a noise-hke form. Luckily, they also reveal similar properties like thermal noise, i.e., the artifact level is proportional to Jn (see Fig. 12) and does not depend on a dimensionality of a signal, maximum evolution times nor spectral widths [45]. This fact may be proved in various ways (two of them were presented in [45]), and below we will present a new, simpler proof, based on known properties of Monte Carlo integration [66]. 

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...

Kazimierczuk and co-workers apphed their semi-automatic CLEAN procedure to suppress artifacts in a randomly sampled N-labeled NOESY-HSQC spectrum of ubiquitin [60]. It was demonstrated that the process does not systematically influence relative peak amplitudes, and is therefore applicable to NOESY spectra. Similar conclusions were later drawn by Stanek and Kozmihski [85], and by Werner-AUen and co-workers [84], who compared their reconstructions with conventionally sampled three-dimensional spectra of the same spectral resolution. The algorithm proposed by Kazimierczuk and co-workers was later also applied to higher-dimensional experiments [80]. 

---