| Author | Message | ||
     Theo van Kempen |
My lab is interested in using FTIR as a quality control tool for feed samples but we are struggling with what accessory to use that does not require sample preparation (e.g., mixing with KBr). We have tried an ATR, which worked reasonably well, but the limits of this technology were apparent. Any other methods out there that are worthwhile considering? | ||
| Fernando Morgado |
Dear Mr. Kempen : For feed analisys the more common is work in the NIR range. Probably you can use filter, dispersive or FT-NIR instruments. I work a lot of time developing calibrations using a FT-NIR instruments using accesory for feed samples. With NIR technology you don´t need prepare the samples, sometime is necesary grind the sample. I recomend to you ask for some FT-NIR instruments, because dispersive have problems when you need transfer the calibrations to other similar equipment. FT-NIR don´t have drifting problems. For choose the better instrument is necesary first you define the aplication for choose the correct accesory. I work with ABB BOMEM, for measuring protein, fat, moisture, ash, and others parameters in food, and the result are very good. Best Regards Fernando Morgado Gnecco Biochemistry. | ||
| George Mahalinobis |
Theo, Just to add to Mr Morgado's comments, NIR is indeed an excellent approach for your needs. I am not sure what he means regarding "problems" in dispersive instruments. Foss's digital dispersive NIR systems have been transfering calibarations successfully for over 10 years. FT-NIR has just recently solved that problem. | ||
| Theo van Kempen |
Although NIRS does okay, our research indicates that FTIR with an ATR does even better, even though ATR is not an ideal sample accessory. So I would like to find out how well the FTIR can do compared to the NIRS with a good sample accessory. E.g., we get prediction errors for methionine in animal meals that are almost twice as good with FTIR than with NIRS. | ||
| Subodh |
Theo, I do not agree with your comments dated 14 Nov. A good calibration on FT-NIR will produce very good results those are comparable with FT-IR and HPLC. For FT-NIR applications you may see the following web site of Buchi-Switzerland www.ft-nir.com Please let me have your comments again. Subodh | ||
| Peter Tillmann (Tillmann) |
Dear Theo, als mentioned before most people use NIR for feed analysis: forages and mixed feeds. If you are looking for a running application and you are not in science and if you can live with NIR, get hooked up to a network. It will allow you to get running in days instead of months with the whole power of the instrument/calibration. Forages: Wende analysis, van Soest fibers, sugar, starch and energy (caluculated from above and other in vitro constituents) Mixed feeds: as above but no van Soest of course Sample preparation: 1 mm grinding that's it Some people do it without grinding, but than you will most of the time loose some of the named infromation. In addition there some around telling on amino acid analysis with NIR or even better digestible cystein, but I have seen none proving that it's not solely based on a correlation to crude protein, what would be obvious. Your Peter Take for example a look at www.vdlufa.de/nirs if you do read German (the English pages at the same site are less prepared). | ||
| Tony Davies (Td) |
Peter has made a good point which has not yet (to my knowledge) been properly address by the NIR community. When you have a number of analytes -such as amino acids - which have very strong correlations with another analyte - protein in this case - it is very difficult to prove that you are really measuring amino acids and not just relying on the in-built correlation. The validation graphs may look good but it is a mistake to have confidence in the results. One day this will be important. NIR analysis will fail to detect an abnormally low value for an essential amino acid, which will be discovered after it has caused trouble. Tony | ||
| hlmark |
Tony makes a fairly good point, but I agree with it only partially. Being able to "predict" the minor component from a measurement of a major component based on their intercorrelation can be fairly reliable IF you can also rely on the correlation being maintained in future samples. That is going to depend on how much you know about the chemistry (and possibly other properties) of your samples. Even in that case, however, there's a situation in which you will not be able to use that sort of calibration: if you've developing calibrations for use in regulated industries: the pharmaceutical industry being one of the main areas here. Tony, you'll recall the meeting that you and I and Tom F. were all at, at the beginning of the year where some of these issues were disucssed. Most of the discussion group is probably not too aware of the considerations involved, but we three, at least, know that the FDA is very strong on analytical methods meeting certain criteria, including one they call "specificity", which essentially means that you have to show that you are in fact measuring the species you are reporting. Measuring by correlation is not permitted there, even if you can show that it works. So I agree with you that that sort of problem is going to become much more important in the future, I think. Howard | ||
| Peter Tillmann (Tillmann) |
Howard, I think you are wrong with your statement that analysing minor components by analysing a major and using a correlation will fulfill its job. Of course this approach will help in many situations (at least as long as the correlation between the minor and major component is given), but this is no analysis of the minor component. It is the interpretation of the analysis of the major component. In agriculture we have many situations were due to missing information or money constrains we use correlations or tabulated values. This is o.k as long as it is named correctly. But no one should state he analysed f.e. the energy content and got x MJ without doing an animal test. All other energy data are calculated from formulars using lab analyses or tabulated values. But the later is outside the lab and no analysis anymore. I think the NIR community can't get respect from other analytical disciplines when think are mixed up this way. The FDA is correct when it asks for a specificity of the analytical method. An interpretaion of results can always be done on top of analysis. But don't claim you are doing things you aren't doing. Yours Peter | ||
| hlmark |
Peter - I can't really quarrel with you, since I agree with what you say. But in my defense, if you read my message carefully, you'll note that I never used the work "analyze". I only used the word "predict" and even then, only in quotes, and then gave a whole bunch of caveats. | ||
| Michel Coene (Michel) |
If you allow me to go back to the beginning of the discussion, question is also what parameters you are interested in. If you just care about basics (H2O, Fat, Protein), you could probably measure it on-line, say on a conveyor belt, or while it is falling down from a silo. In that case I can imagine ATR is maybe not your favourite solution... NIR will give you the sampling liberty you need, but a bigger calibration effort will be the price to pay. | ||
| Fernando Morgado |
For food and feed constituent NIR predictions work very good. If the idea is install some instrument in line for control the production process is necesary first considerate if the plant have the capacity to regulate automatic the process using the data obtained with the NIR instruments. If not possible make this, install NIR instrument in line don´t have sence. Better is install some Lab NIR instruments near to the process and analize samples obtained manualy. The time for obtain the result is almost in line analisys, and you can inform the result for make the adjust necesary in the production. For traditional parameters in feed industry as Fat, Protein, Moisture ( not only H2O), Ash, NIR give good result. Models created for this parameters can be validated using ASTM E 1655 procedure, this mean the result obtained with NIR analize are similar to obtained with traditional method. For give some answer to the first question make for Mr.Kempen for predict constituent in row material , process product , and finish product I think the better solution is move to NIR technology. Is necesary decide now wich is better for your plant, in line or Laboratory instrument. A little coment about aminoacid, I´m accord with Tony and Howard. The NIR feed spectra principal show bands related to N-H, C-H, O-H. Is not possible identify special bands for aa in NIR spectra. Maybe PLS can look it, but the correlation obtained for aa probably is related to the total protein concentration, and not to one specific aa. Raman technology can identify specific band for aa, but in pure proteins samples, and this is not the case because feed don´t have only proteins. |