ADJOINT SPECTRUM .1. AN ALGORITHM TO EXTRACT TARGET SPECTRA UNDER SPECTRAL INTERFERENCES FOR USE IN ONLINE SPECTROMETRY

A spectrum measured with a sensitive device inevitably suffers from interference spectra which are both predictable and unpredictable. Amplitude of a target gas spectrum can be obtained as a result of the inner product of the measured spectrum and an ''adjoint spectrum'' generated by the algorithm proposed here. The result is independent of amplitudes of interference spectra as far as its profiles are known beforehand. The effectiveness is extended to unknown interferences if their profiles are gently sloping. Another important feature is that this algorithm can be executed in real-time on a microprocessor. Such features are favorable for field measurements of concentrations of atmospheric gases in real-time and in situ. If too many interference spectra are involved, the system becomes vulnerable to random noise included in the measured data. A compromise can be sought, but the solution depends on the intensity of the noise. A tunable diode laser absorption spectrometer (TDLAS) adopting this technique and designed to be used in field measurements of atmospheric methane, has shown an accuracy of 6 ppb in a laboratory test, which is more than one order better than the case without this algorithm.