| Author | Message | ||
     Tony Davies (td) Moderator Username: td Post Number: 252 Registered: 1-2001 |
Howard, This is unusual - I think we are close to agreeing! Jerry, How do you want to measure "performance"? In the 1970s people wanted better resolution than their dispersive instruments would provide. FT gave them better resolution so FT won the battle. In NIR we are more often interested in quantitative results and as Howard explained it is easier to obtain very low noise levels in a dispersive instrument so the FT has not won this battle, yet! Some time ago I proposed (at a Chambersburg meeting) a collaboration to develop a computer system which would generate synthetic data as though it had come from different spectrometers but no one was interested. If we had succeeded then we would have been able to transform spectra from any instrument to any other instrument. Best wishes, Tony | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 409 Registered: 9-2001 |
Tony - yes, it's starting to come back to me. Peter took on a much more precisely defined (and much harder) problem, not just to convert the wavenumbers to wavelengths correctly, but also to take into account the different instrument functions so that he could accurately reproduce the spectrum that would be measured on the alternate instrument. The problem was that the other instrument manufacturers believed what Myron Block said, when he said "All you need to do to change the wavelength range of the [FTS-14, which was the first instrument model Digilab produced] is to change the source, beamsplitter and detector." The major limitation of FT instruments is the dynamic range issue, which is often mentioned, but never really considered, especially in NIR usage. I don't really know what the situation is today, but certainly in the early days of FT-NIR, the instruments used the same 16-bit A/D converters that they used for mid-IR work. Few people seem to realize that that number of bits is the limitation on the allowable dynamic range, or alternatively, the instrument noise level. Coadding spectra to reduce the noise doesn't work unless the noise is larger than the A/D step size, and preferably 2-3 times the step size. In FT-NIR instruments that'a a major consideration. Grating-based NIR instruments don't have the large signal "spike" that interferometers have at zero retardation, so it's easier to deal with dynamic range issues. Even so, the (grating-based) InfraAlyzers needed 20-bit A/D converters to ensure that the noise contribution of the A/D was less than the other noise sources. Jerry - yes, the "real" spectrum of a material is unobsevable - in ANY spectral region. Quantum mechanics guarantees that. Nevertheless, good approximations can be obtained, and the conditions for doing that are known. The main one is that the instruiment's spectral resolution should be much narrower than the width of the narrowest band to be measured. If you can do that with both an FT instrument and a grating instrument, then you can essentially convert one spectrum to the other, without loss of information, so that you will get the "same" spectral profile, depending on how much you want to quibble. It doesn't negate the utility of NIR spectroscopy, or justify saying that it's not a "real" spectrum. \o/ /_\ | ||
| Jakob Schultz (schultz) New member Username: schultz Post Number: 2 Registered: 9-2005 |
Thanks for your responses Tony, I agree that all technologies has its pros and cons and the technology used to record a spectrum should always be stated. Both you, Howard and David mention history and a desire to differentiate from scanning monochromator instruments as one of the main reasons for the terminology used today. That was pretty much my own perception as well, but it is nice to get it confirmed. Cheers | ||
| Jerry Jin (jcg2000) Senior Member Username: jcg2000 Post Number: 40 Registered: 1-2009 |
Dear All: I am surprised to see "the �real� NIR spectrum is a theoretical idea and what we obtain depends on what system we are employing". If different NIR spectrometers give their own spectra for the same substance that cannot be comparable between dispersive and FT-NIR, that would cast serious doubt over NIR itself. I would expect to see difference in their speed, resolution, noise, etc. But I expect the same spectral profiles are produced regardless of the specific instrument. Also, can anyone elaborate "FT-NIR instruments work but their overall performance is on a par with dispersive NIR spectrometers"? I expect to see FT-NIR outperforms dispersive NIR, as I see in middle IR. Best, Jerry Jin | ||
| Tony Davies (td) Moderator Username: td Post Number: 251 Registered: 1-2001 |
Hello Howard, How was PittCon? You are quoting me out of context. I was discussing commercial systems. I am well aware of Peter�s work. In 2000 Tom and I published a �Tony Davies� column in Spectroscopy Europe which did it as an example of matrix algebra. http://www.spectroscopyeurope.com/images/stories/ColumnPDFs/TD_12_6.pdf The idea came from Fred�s 1981 paper on applying FT to data processing of NIR spectra he said � � in fact the Fourier transform can be searched to compute the magnitude of the computed spectrum at the required wavelength rather than re-computing the entire spectrum� We re-computed the 700 wavelengths that we required. We tested the program with spectra of a sample run on Heinz�s dispersive and FT spectrometers. The computed spectra demonstrates very clearly the better resolution of the FT spectrometer. Actually it is not that slow and with a 2010 computer it might be ten times faster! Back to the question. �Is there a difference?� History The first FT-NIR spectrometers where produced by companies producing mid-IR spectrometers and they assumed that there was little difference between the application of mid-IR and NIR data and so FT would be the superior technology and have the same success as it did with the dispersive mid-IR spectrometers of the 1970s. So they wanted to emphasise the FT. But they were wrong. FT-NIR instruments work but their overall performance is on a par with dispersive NIR spectrometers. Is there a difference? Yes, there are differences. Jakob hoped that the spectrum would be the same but at present the �real� NIR spectrum is a theoretical idea and what we obtain depends on what system we are employing. You need to know what type of instrument produced an NIR spectrum. Best wishes, Tony | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 408 Registered: 9-2001 |
Tony - when you said: "It is possible (but tedious) to compute a point by point transfer between scales but I have never seen any claim that anyone does it." I concluded that you must have missed Peter Griffiths' paper on transfer of spectra from the rare-earth NIR standard from an FTIR to a grating instrument-equivalent spectrum in JNIRS; 11(4); p.229-240 (2003). He, Tomas Isaksson, Richard Jackson and his then-student Husheng Yang did a superlative job of having the computer do THE WHOLE THING! I can't imagine how long it took, though! They also used published spectra of water vapor where the vibration-rotation bands are known to 0.001 cm-1 to ensure the accuracy of the bands of the NIR standard. A monumental work! They then published the rest of a trio of papers on the topic; besides that one, there were papers in JNIRS; 11(4); p.241-255 (2003) and in Appl. Spect.; 57(2); p.176-185 (2003) That said, I agree with Dave Cameron that the reasons for the nomenclature are mainly historical. What we now know as FTIR can trace its roots back to the early 1900s. After Michaelson invented the interferometer, it was used by the astronomical community to measure the spectra of stars. Back then, they had to do the FT without computers, which weren't invented for quite a while. The Fast Fourier Transform didn't exist yet then, either. So it was used, because it was the only thing available that would work, but it wasn't a very popular technique! Chemists didn't latch on to the term "FTIR" until Block Engineering, and then Digilab, developed an instrument specifically for use in chemical analysis, in the 1960's. Those instruments were specifically for the mid-IR region. When modern NIR analysis became commercially available in the late 1970's, roughy 20 years later, all those initial instruments were based on either interference filters or diffraction gratings for wavelength selection. When FT instrumentation started to be applied to NIR analysis, the term "FT-NIR" came along with it, both to distinguish the application of the interferometric measurement to the NIR region from its historical application to mid-IR, and also to distinguish it from the generations of grating-based instruments that were previously exclusively used for NIR analysis. \o/ /_\ | ||
| Tony Davies (td) Moderator Username: td Post Number: 250 Registered: 1-2001 |
Hello Jakob, Welcome to the forum! I think you need to know what technology was used to measure your spectrum. First of all there was an (unwritten) agreement when FT was introduced to IR spectrometers that the spectrum would be plotted backwards (high wavenumbers to low wavenumbers) so that the spectra would look similar from the different spectrometers. This is not always apreciated by people who came with FT instruments directly for the NIR region, Their FT instruments plotted spectra the "natural way" from low to high wavenumbers. However, even when applying the convention, the spectra will not look exactly the same because dispersive instruments will measure at equal wavelength (nm) intervals while the FT spectrum will be plotted at equal wavenumber intervals. It depends which area of the spectrum is most important to you. FT spectra compared to wavelength spectra will have more measurements at low wavelengths and fewer at high wavelengths. Probably the more important technical point is that the wavenumber scale in an FT instrument is measure by laser while the spectrum is being recorded and this inherently more precise than the wavelength calibration in dispersive instruments. Conversely the noise level in dispersive instruments will often be lower than an FT instrument so the photometric measurements may be superior to the FT measured absorptions. Many FT instrument manufactures provide the option of plotting the spectrum on a wavelength scale but in my experience these are interpolations and are only approximate. It is possible (but tedious) to compute a point by point transfer between scales but I have never seen any claim that anyone does it. The point is that you cannot make a direct comparison between spectra produced by the alternative systems. The other point to note is that wavenumber is proportional to�frequency�and to�photon�energy. So if we are considering combination bands, wavelengths need to be converted to wavenumbers. (there is much discussion about this in another thread last post 19Aug09). Best wishes, Tony | ||
| David Cameron (david_cameron) Junior Member Username: david_cameron Post Number: 7 Registered: 3-2011 |
The short answer - no. Any spectrum is the convolution of the spectrum as we would measure it with a perfect spectrometer with the instrument lineshape function. In my mind, more historical. Dispersive instruments were the first technology in IR, NIR, visible, and UV spectroscopy. FT instruments are easier operate at longer wavelengths - the Infrared, and the first commercial lab Ft Machines were IR. Around the early 1990s some labs decided to try FT in the NIR, for reasons initially relating to calibration transfer. These machines were called FT-NIR machines, they worked, but the name has stuck. One feature which is different, the natural X scale of an FT machine is cm-1 (wavenumbers). You measure signal as a function of distance, when you do an Ft you get the inverse of distance, cm1. | ||
| Jakob Schultz (schultz) New member Username: schultz Post Number: 1 Registered: 9-2005 |
Dear all A thing has puzzled me for some time. A NIR spectrum can be recorded in a number of ways using different types of spectrometer technologies (DDA, scanning monochromator, interferometer etc). For me a NIR spectrum is a NIR spectrum, irrespective of the technology used to record that spectrum. But it seems that when an interferometer is used an the resulting interferogram is fourier transformed, then it is very important to emphasize that it is FT-NIR and not just NIR. Is there a good (or bad for the sake) reason why it is emphasised that it is FT-NIR and not just NIR? Does FT-NIR spectra have properties which makes it necessary to distinguish them from other types of NIR spectra? Regards |