348
A. L. Chakraborty and A. Roy
distorted and the error grows rapidly even for small changes to the LIA phase.
Recovery of the absorption line becomes increasingly more inaccurate and significant
errors arise in the measurement of the gas parameters. The phase adjustment is
done by inspecting the signals recovered on the two axes. Inaccurate phase settings
introduce large errors in the calculation of ψ 1 by the PD method and therefore cause
distortion of the recovered absorption line shape. The scenario worsens for low mole
fractions and when there are etalon fringes on the spectral wings (which is almost
invariably the case). The value of ψ 1 obtained from a separate characterization of
the laser cannot always be used to set the LIA phase because of additional phase
shifts that are introduced in the detection electronics. The other related issue in both
methods in the detection axis recovers only a projection of the full RAM signal.
Some recent work [54, 55], on the RAM method has however shown that the RAM
signals may turn out to be stronger than conventional 1f and 2f WMS signals under
the right operating conditions.
The main limitation with the RAM method is that the absorption-dependent signal
appears on a high, concentration-independent background RAM component due to
the linear IM. Amplification of the 1f X signal in Eq. 29 leads to saturation of the
detection electronics due to this large dc component and severely limits the detection
sensitivity. A method to fiber-optically eliminate the background RAM known as
RAM nulling, was demonstrated [57] to overcome this problem. Figure 14a shows
the setup. The modulated laser output is split into two paths using a 3dB coupler.
One part is directed through a gas cell and the other through a fiber-optic delay line
(single-mode fiber, SMF-28) of length L fiber = c/2n eff f m , where n eff is the effective
refractive index. The two parts are then recombined using a second 3dB coupler
and then coupled to the photo-diode detector. The fiber length ensures that a phase
difference of π is introduced between the IM components (not the electric field of
the light) in the two paths. Two polarization controllers (PC) ensure that the light
beams are polarized orthogonal to each other to minimize optical interference noise.
The amplitudes of the two anti-phase IM signals are carefully balanced using a
variable optical attenuator (VOA) in the delay line arm in such a way as to cause
the IM components to cancel at the combiner output in the absence of gas and in
the off-resonance regions. The resulting dc signal is rejected by the LIA and forces
the output to zero in the absence of the gas and in the off-resonance regions. This
eliminates the high background RAM signal. Introduction of a gas in the cell causes
this delicate balance to be disturbed and only the concentration-dependent RAM
component appears at the output. Note that this signal bears the signature of the line
shape and the relative transmission can be obtained from it [57].
Figure 16a shows the RAM signals with the high background RAM for 10, 1
and 0.1% methane. The concentration-dependent RAM is strong for 10% methane,
barely detectable for 1% methane and undetectable for 0.1% methane. The LIA
gain was set to the maximum possible value in each case. When background RAM
nulling is implemented, the three signals appear on a zero baseline. Figure 16c, d
show that line shape recovery is now possible for the lower mole fractions thereby
demonstrating the usefulness of the RAM nulling technique. A very small sloping
baseline does exist in the RAM nulled signals because the RAM nulling condition
A. L. Chakraborty and A. Roy
distorted and the error grows rapidly even for small changes to the LIA phase.
Recovery of the absorption line becomes increasingly more inaccurate and significant
errors arise in the measurement of the gas parameters. The phase adjustment is
done by inspecting the signals recovered on the two axes. Inaccurate phase settings
introduce large errors in the calculation of ψ 1 by the PD method and therefore cause
distortion of the recovered absorption line shape. The scenario worsens for low mole
fractions and when there are etalon fringes on the spectral wings (which is almost
invariably the case). The value of ψ 1 obtained from a separate characterization of
the laser cannot always be used to set the LIA phase because of additional phase
shifts that are introduced in the detection electronics. The other related issue in both
methods in the detection axis recovers only a projection of the full RAM signal.
Some recent work [54, 55], on the RAM method has however shown that the RAM
signals may turn out to be stronger than conventional 1f and 2f WMS signals under
the right operating conditions.
The main limitation with the RAM method is that the absorption-dependent signal
appears on a high, concentration-independent background RAM component due to
the linear IM. Amplification of the 1f X signal in Eq. 29 leads to saturation of the
detection electronics due to this large dc component and severely limits the detection
sensitivity. A method to fiber-optically eliminate the background RAM known as
RAM nulling, was demonstrated [57] to overcome this problem. Figure 14a shows
the setup. The modulated laser output is split into two paths using a 3dB coupler.
One part is directed through a gas cell and the other through a fiber-optic delay line
(single-mode fiber, SMF-28) of length L fiber = c/2n eff f m , where n eff is the effective
refractive index. The two parts are then recombined using a second 3dB coupler
and then coupled to the photo-diode detector. The fiber length ensures that a phase
difference of π is introduced between the IM components (not the electric field of
the light) in the two paths. Two polarization controllers (PC) ensure that the light
beams are polarized orthogonal to each other to minimize optical interference noise.
The amplitudes of the two anti-phase IM signals are carefully balanced using a
variable optical attenuator (VOA) in the delay line arm in such a way as to cause
the IM components to cancel at the combiner output in the absence of gas and in
the off-resonance regions. The resulting dc signal is rejected by the LIA and forces
the output to zero in the absence of the gas and in the off-resonance regions. This
eliminates the high background RAM signal. Introduction of a gas in the cell causes
this delicate balance to be disturbed and only the concentration-dependent RAM
component appears at the output. Note that this signal bears the signature of the line
shape and the relative transmission can be obtained from it [57].
Figure 16a shows the RAM signals with the high background RAM for 10, 1
and 0.1% methane. The concentration-dependent RAM is strong for 10% methane,
barely detectable for 1% methane and undetectable for 0.1% methane. The LIA
gain was set to the maximum possible value in each case. When background RAM
nulling is implemented, the three signals appear on a zero baseline. Figure 16c, d
show that line shape recovery is now possible for the lower mole fractions thereby
demonstrating the usefulness of the RAM nulling technique. A very small sloping
baseline does exist in the RAM nulled signals because the RAM nulling condition