Author |
Message |
Hongtu Xie (tonali)
Member Username: tonali
Post Number: 11 Registered: 2-2010
| Posted on Sunday, April 11, 2010 - 12:53 pm: | |
thanks all, now I'm redoing the samples and test every parameters many thanks |
David Russell (russell)
Senior Member Username: russell
Post Number: 46 Registered: 2-2001
| Posted on Thursday, April 08, 2010 - 1:07 pm: | |
For the problem that you describe, the best that you can do is to redo the calibration using 4000-600 cn-1 at 8wn. The scans used to calibrate and the scans used to predict must match. Transfer between two MCT detectors shouldn't be a problem. Transfer between MCT and TGS is likely to result in additional prediction error. |
Cesar Guerrero (cesar)
Intermediate Member Username: cesar
Post Number: 16 Registered: 3-2006
| Posted on Thursday, April 08, 2010 - 12:49 pm: | |
Hi Hongtu, Many OPUS commands make spectra automatically compatible to each other. As I know, Quant2 does it. As rapid test, before other things, please verify the real frequency (limits) on your spectra. It is very probable that each spectrum doesn't start exactly at 4000 cm-1. For this, load one spectrum, then make a double click on History, and you will be able to see all the details of this spectrum. You can see which is the first (and last) frequency. Do the same for model's spectra and unknown spectra. It is very probable that they don't start at the same frequency. Have you check this? If this is the problem, cut the first frequencies or apply the "make compatible" command. Good luck! Best regards, C�sar |
Howard Mark (hlmark)
Senior Member Username: hlmark
Post Number: 315 Registered: 9-2001
| Posted on Thursday, April 08, 2010 - 12:18 pm: | |
Hongtu - Specifications for the wavelength range of a detector typically is the range for which the sensivity is some percentage of the maximum sensivity. This could be 50%, 10% or some other value that the manufacturer specifies. Outside that range, the sensivity is basically unknown, but is probably not zero. Only the manufacturer knows for sure. I recommend that you contact the manufacturer of your detector and ask them for the plot of sensitivity versus wavelength over the full range that you are concerned about. Some spectrometers allow you to plot a single-beam spectrum of the energy in one channel; this will generally be the (blackbody) energy of the source modified by the sensitivity of the detector. While not a "pure" measurement of the detector response, in practice this is actually more useful, since it represents the total instrument response under typical measurement conditions when you measure your samples. If the detector manufacturer can't or won't provide the sensitivity, and you can't measure a single-beam spectrum on your instrument, you can get some idea of how the energy is falling off outside the specified range by measuring a 100% line, i.e., the transmission spectrum you get with no sample in the beam for either the "sample" measurement or the "reference" measurement. Ideally such a spectrum will be a flat line at 100% transmission, but there will usually be noise superimposed on that line. The amount of noise at various wavelengths will be indicative of the amount of energy in the beam at those wavelengths, with more noise indicating decreased energy. A quantitative estimate of the relative energies can be calculated using the equations found in Appl. Spect., 56(5), p.633-639 (2002). If you solve those equations for the case of reference energy equals sample energy, and then plot the theoretical noise level as a function of energy, that will give you a pretty good idea of where you stand. Howard \o/ /_\ |
Hongtu Xie (tonali)
Junior Member Username: tonali
Post Number: 10 Registered: 2-2010
| Posted on Thursday, April 08, 2010 - 10:46 am: | |
Hi, Evgeny Thanks for your advice, I checked the MCT, the model is D316, the range is 12000-600 cm-1, i.e, the information between 600-400 cm-1 can't be used to make model? Regards Hongtu |
Evgeny (evgeny)
New member Username: evgeny
Post Number: 4 Registered: 3-2010
| Posted on Thursday, April 08, 2010 - 2:23 am: | |
Hi, Pay attention to the cut off point of the MCT detector. For some types of MCT detectors it is 850 cm-1. I suggest you to check this info in the specification documents. Cheers, Evgeny |
Hongtu Xie (tonali)
Junior Member Username: tonali
Post Number: 9 Registered: 2-2010
| Posted on Wednesday, April 07, 2010 - 8:58 am: | |
Hi, Cesar Thanks you! We used the soil to make model with 4000-400 cm^-1, 4 cm^-1 resolution, MCT detector want to predict: 1)4000-400, 8 cm-1 resolution, MCT detector; 2)4000-600, 4 cm-1 resolution, DLaTGS detector |
Cesar Guerrero (cesar)
Member Username: cesar
Post Number: 15 Registered: 3-2006
| Posted on Wednesday, April 07, 2010 - 4:24 am: | |
Hello, It seems that samples used for calibration/validation have a different spectral range than prediction samples. Which spectral range was used during PLS? (please indicate: in wavelengths, or in wavenumbers). Please, verify if this spectral range is the same in the �unknown� samples. Regards, C�sar |
Hongtu Xie (tonali)
Junior Member Username: tonali
Post Number: 8 Registered: 2-2010
| Posted on Tuesday, April 06, 2010 - 1:50 pm: | |
Dear all, I make a model of soil samples using OPUS6.5 Quant2 method, and I load the model to predict the unknow soil samples, there is an error "The X range is too small", What does this information indicate? Thanks |
|