Rapid Scanning Fourier Transform Spectrometry

Rapid scanning serves to mitigate the problems of scintillation and dynamic range for Fourier spectrometry, as well as to simplify the measurement procedures. Rapid scanning means scanning at a rate such that the interferometric fringes act as their own chopper. It also means scanning at a rate such that the fringe frequencies exceed the scintillation frequencies, thereby avoiding confusion between the two. Of necessity the scanning becomes repetitive for the measurement of faint sources. During observation the repetitive interferogram signals are recorded on one channel of a stereo magnetic tape recorder, while timing information is recorded on the other channel. The interferograms are subsequently digitized and summed (in synchronism according to the timing channel) into a magnetic core memory. Fourier analysis of the resulting interferogram can be finally carried out by either analogue techniques or by digital computation. Astronomical spectra obtained by two instruments, one in the visible (S-20) range, the other in the lead sulphide range, are shown to demonstrate the efficacy of rapid scanning. Finally a curious systematic effect of erroneous digitizing to enhance weak spectral absorption lines was discovered and will be illustrated. The random noise which plagues spectral measurements divides into three categories ; proportional to the zeroth, half, and first power of the optical signal level respectively. As Fellgett pointed out in his thesis, Fourier spectrometry observes the various colors simultaneously. There is an increased observation time per wavelength resolution element compared to sequential observation. In winning this observation time the optical signal levels are proportionally increased. Thus for noise proportional to the zeroth power of the signal, as encountered with infrared detectors, the increased observation time is used to advantage, i. e., Fellgett's multiplex advantage. For noise proportional to the square root of the signal, such as photon statistics, increased observation time is just balanced on the average by increased noise and no advantage (or disadvantage) of the above sort occurs. For noise which is proportional to signal we lose. Scintillation noise due to atmospheric turbulence can easily relegate astronomical spectrometry to the last category. The crux of this problem is that scintillation fluctuations become spuriously interpreted as interferometer fringes. Limited dynamic range of the recording meter can also be thought of as falling in this last category. One digitizing bit will be a certain fraction of the maximum tolerable signal level. Although various gimmicks such as fringe chirping [l] may be used to alleviate dynamic range problems, it has always Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1967218