Fourier transform spectroscopy works by measuring a spectral/light signal through a Michelson interferometer. In order to know the wavelength of the signal, one must use a stable reference, which is typically a metrology laser. In a standard Fourier transform spectrometer (FTS) system, the laser signal also runs through the interferometer and the laser beam is guided to a separate detector that is then used to trigger an analog-to-digital converter, which then captures the spectral signal.
One drawback of the standard method is one is restricted to measuring spectral content at half the frequency of the laser light or below due to the Nyquist sampling criteria. Also, velocity control must be very tight because one is sampling in the spatial as opposed to the time domain, which can lead to ghosting of the interferogram. This issue was first addressed by James Brault, who conceived digitizing the signal in the time domain with a high-resolution, 24-bit analog-to-digital converter, triggered by a crystal clock. This allowed him to apply filtering in the time domain, which eliminated the ghosting problem. He used information from an event counter that was triggered by a metrology laser to then resample the data linear in space. This removes the restriction of sampling at half the laser frequency sampling because the velocity variations are slow compared to the sample rate, and the sample rate can be as fast as it needs to be by adjusting the clock frequency independent of the metrology laser. The drawback to this method is it requires special hardware in the form of the high-speed event counter, and velocity correction points are restricted to metrology laser fringe crossings, which are used for the resampling of the data in the space domain. It also requires tuning to align the fringe crossing data with the sampled signal, which can be different from run to run.
In this newly improved method, the analog laser fringe signal is digitized in a separate channel along with the spectral data, which eliminates the event counter used in the earlier method. Then, the laser signal is demodulated using a heterodyning technique in the form of a software synthetic quadrature phase detector in combination with phase tracking to derive the slide position for each data point. Two synthetic signals are created, each being 90° out of phase with a frequency equal to the average frequency of the laser interference pattern. These two signals are then mixed with the laser signal. Fourier transform filters are then used to isolate the sidebands to derive the phase. Also, since both the laser data and spectral data are synchronous (unlike in the earlier improvement), no tuning is required. This makes for a very simple setup, and high-resolution data can be obtained at any wavelength, even while using 24-bit audio digitizers found on most modern computers.
The most unique feature of this innovation is to use a synthetic quadrature phase detector and phase tracker to determine an FTS slide position for each digitized point. The main benefit is that one is not restricted to a spectral range dictated by the frequency of the laser, and one does not need any additional hardware such as an event counter to linearize the data in space. It is also possible to use the same detector for both the metrology laser and the spectral signal since one can isolate each through filtering.
This work was done by Joel F. Campbell of Langley Research Center. LAR-17694-1
