Following the question and answers between Richard Duerst and Mike Ford and Andy Turner featured in last issue’s Dear Readers [Edition1], there’s been more toing and froing of emails. Richard’s response to Mike and Andy’s answer came in at the end of February…
An extension to the first question, i.e. “What are the trade-offs in going from an 8- to a 16- to a 20-bit A/D convertor?” is: Can the same signal-to-noise ratio be achieved with a spectrometer with an 8-bit A/D as with a spectrometer with a 16-bit A/D if signal averaging is employed, with random ‘noise’ at the level of one or two least significant bits? I realize that there will be a significant difference in accumulation time.
Mike then came back with this answer…
I have discussed this question with Dr Andy Turner and we agree the answer. In principle, it would be possible to achieve the 8 bit improvement in S/N by summing 64,000 scans (the improvement in S/N goes as the square root of the number of scans added, so you need 2 to the 16 scans). However, almost invariably, other factors come into play which limit the maximum improvement to the S:N ratio to much less than 8 bits. Such factors include the noise not being truly random and non-linearity of the A/D convertor which, ideally, should be as good as the most significant 8 bits of a 16 bit convertor. Of course, the memory into which the signal is summed would need to be 24 bits (or the equivalent) in order to avoid overflow.
Then Richard had another question…
But what are the non-random noise sources relevant to IR?
Mike’s response this time was…
I have had another chat with Andy and here is the best we can do as a quick answer – we do not guarantee that it is complete! We have assumed that ‘relevant to IR’ means FTIR – some of the things we mention are specific to FT.
Possible non-random ‘noise’ sources include: anything which causes a repetitive disturbance of the initial data collection interval – such as unequal mark/space ratio in the reference laser fringes or a repetitive scan velocity variation; any non-linearity in the total process of conversion of the optical (IR) signal to its digital form including non-linearity in any of the components, i.e. the detector, amplifier(s), A/D convertor; any spurious signal picked up which is synchronous with the scan or data collection interval. Some of the above might not be truly classified as ‘noise’, but they will still cause spectral distortion, as will numerous optical effects.
Back came another question from Richard….
Would the cooling of a detector – it rarely reaches a steady state – and room
temperature variations also be included as non-random noise?
And Mike’s answer was…..
The best answer I can give is as follows. Room temperature fluctuations will, almost certainly, be slow compared with single scan times and their effects (probably minor changes in the spectral profile) will tend to be averaged out by summing multiple scans. They are best considered as slow, random noise which will not tend to limit the ability to improve S:N by multiple scanning. On the other hand, detector temperature fluctuations can be much faster, due to such things as bubbling of the coolant, and could be of a similar time scale to scanning. However, it is very unlikely that any such variations would be even partly synchronised with the scanning, and any associated
noise would therefore be random and would average out.
All seems now quiet from Richard at the moment! Actually it’s great that IJVS can get people conversing with each other (albeit via computer) from all over the world. We’re happy to help.
Quite of few of you have emailed asking about the impact factor of IJVS and quite right too as we are a bone fide journal. We have been trying to get an answer out of the American Chemical Society and in fact wrote an official letter to them at the end of November 2001, but they haven’t replied. So we’ve tried by email and still no answer! If you read Patrick’s Editorial, you’ll hear his thoughts on the matter! Now on another matter, we had a plea from Professor Salvatore Romano from Florence, Italy….
I am a professor of medical physics at the University of Florence. I was assigned the job to develop within one year a laboratory for practical lessons and group operation beginning with simple to more sophisticated instrument. Say: validation of the Poiseuille Law to IR-UV-Vis, fluorescence and X and gamma spectrometry. Some of these experiments are to be confined to a 3-year course for medical-biotechnology students. My personal research topics are mainly on FTIR and autofluorescence micro-spectrometry and imaging of biological samples. I ask you for some, even obsolete but still functioning, instrumentation such as a fluorescence microscope, an FTIR (or grating) spectrometer, a densitometer (for radiography digital conversion) a beta-counter and so on.
Salvatore Romano, Prof. of Medical Physics, Dept of Clinica Physiopathology
University of Florence, viale G. Pieraccini, 6 50139 Florence – Italy
Phone: +39 (0)55 4271 465 (room), +39 (0)55 4271 /218; /219; /217 (Lab)
Fax: +39(0)55 4271 413
Can anyone help Salvatore? Recently we have advertised the availability of IR kit in the States. I’m sure some must be lying around in Europe.Please help and keep us informed.
Another request for information came in , this time from Jorge Villafuerte from Ecuador..
I am interested in information about the analysis Diesel fuel using infrared – the analysis to be quantitative and qualitative.We want to determine elements such as sulphur, gum, etc. Can you help me with these problems?????
Again is there anyone who can help Jorg, his email is firstname.lastname@example.org. If you can, then please don’t forget to
copy us here at the Office.
|Please keep sending your comments and questions, sometimes it may take a while for us to answer, but we will.
All your questions and our replies are published here in Dear Readers.