| Author | Message | ||
     Tony Davies (Td) |
The recent discussion about Beer's law reminded me of an idea that I have been intending to pose for some time. In 1994 I described a spreadsheet model that was developed to try to explain why the optimum absorption for whole grain analyzers is so high (log 1/T=2.0-4.0). It worked reasonably well. In 1996, Peter Hindle published a paper which used a model to compare different data treatments; with interesting results. In 1998 Fred McClure et al. published results of using a model to evaluate the effect of random noise and latter in 1998 Karl Norris described a model for evaluating interactions among instrument bandpass, instrument noise, sample-absorber bandwidth and calibration error. And I am sure there are others papers in other publications. I am convinced that models are very useful for determining the extent of knowledge and in what direction we should be aiming our concentration. All these models where programmed to answer specific questions; I would like to propose that we should construct a model (MATLAB prefered) that would emulate a spectrometer as exactly as we can make it. The model could then be used to explore any question, not just those we are already aware of. This would be quite an undertaking and so the question is: (I've got to it at last) Would any one be interested in collaborating in such a venture? It obviously needs a group 1) so that the model can be broken down into sub-models which would be developed by different members and 2) because only a group could hope to be able to master all the information it would require. Anyone interested? Best wishes, Tony | ||
| David W. Hopkins (Dhopkins) |
Tony, I think your proposal is a good one, and I am interested in participating. But, I would recommend that we tackle it in Excel. I think that would be useful to more people. Although I have considered learning Matlab, I still have not sprung all the money it takes, and so I am not interested in having the Matlab environment. | ||
| Michel Coene (Michel) |
Matlab does not come cheap, but if we want a useful model, it will be quite a big operation. Probably it can be split up in different parts that do not necessarily need to be made in the same environment/language. I am e.g. willing to �donate� some visual basic, ideal for things like file-format conversions, visualizations, and gluing other modules together. I believe it is possible to make Matlab �runtimes�, which do not require a license to operate. This would make the end result accessible for all. A good place to centralize all this is SourceForge (www.sf.net). | ||
| hlmark |
Guys - maybe it's just my natural pessimism, but what makes us think we can model the way real instruments behave? Having worked for an instrument manufacturer for a good number of years, I know two things: 1) There were several things about the instrument operation that we never really understood or sometimes even knew about 2) A lot of the things that we did know about and understood we never told anybody, because of the potential effect on the commercial aspects of the company's operation. I'm sure every other manufacturer has their own "secrets" too. So while the goal is laudable, I have some serious misgivings about our ability to achieve it, especially since it depends on a lot of volunteer effort. This group seems especially dedicated, but still, we all have other priorities which take precedence. Also, I appreciate Dave's problem with MATLAB. On the other hand, EXCEL is not really a programming language and is pretty limited in what can be done in that regard. VB is a programming language, but is missing a lot of nice features that MATLAB has that would make this feasible. I've written program in all three, and believe me, for a project like this, MATLAB is the way to go. On the other hand (that's three, now) besides some fingernail dirt, it's not common knowledge but MATLAB started out as a shareware project, and a primitive form of MATLAB exists as the shareware version. I don't know how useful it is, but I have a copy of it (somewhere) and could probably find it and give it to Dave; Dave could then evaluate whether it's worthwhie trying to use. Certianly, modern MATLAB has lots of features useful to us in such a project as we're contemplating (lots of matrix function, etc), but I don't know how much of that is in the shareware version. Re Ian's comment about making MATLAB "runtime": there is a compiler (at extra cost) that can turn a MATLAB program into a self-standing executable. I have that compiler and have used it, so I know a few things about it, too. Among them, you cannot (easily) make a MATLAB program that will be compatible with other programming languages. I'ts not impossible, but you not only need the MATLAB compiler and a 32-bit C compiler (such as Visual C), but there are also lots of special tricks you need to know to make it work. I've done it the other way 'round (a DLL written in C that can be called from MATLAB) and that's still tricky enough, although easier than making a MATLAB program that will work with VB, for example. Howard \o/ /_\ |