Fig. 1. Experimental setup. LP, linear polarizer; 50:50 FC, fiber coupler with 50:50 split ratio; LD, laser diode; ADC, analog-to-digital converter; DAC, digital-to-analog converter. The feedback control signal to the LD current controller for wavelength stabilization is generated digitally by the computer and converted to an analog signal through the DAC (see Section 4.C).

Fig. 2. The approximate wavelength step-input response of the diode laser at the linear section of the gas transmittance curve. A total wavelength change of ∼3.2  pm is produced. Approximate wavelength step-input response for (a) tens of ms scale when using an 8 Hz toggling rate and (b) tens of μs scale relevant for the 8 kHz toggling rate.

Fig. 3. ON/OFF toggling at 8 kHz. (a) Applied current pulse. The square current pulse peak-to-peak is approximately 43 mA, and the exponential part has an amplitude of 13 mA, with a time constant, τm=18.75  μs. (b) The considered HCN absorption line showing the ON/OFF toggling range. (c) Comparison of measured transmittance for a square pulse and the modified square pulse along with the theoretical curves. (d) Same as (c) but for a full period τp=128  μs.

Fig. 4. ON/OFF transmittance data at 8 kHz toggling rate sampled at 250 kSamples/s. (a) 32 data points during a single cycle. (b) Zoomed-in view of the OFF transmittance values. (c) Zoomed-in view of the ON transmittance values. The variation in transmittance values of the experimentally obtained raw 8 kHz data, 40 Hz averaged data and theory is in very good agreement.

Fig. 5. Investigation of the ON transmittance data for different mean or DC currents. (a) Experimentally obtained transmittance curves. (b) Corresponding theoretical curves. The yellow shaded section indicates the portion of the curves where the slopes (and sign of the slopes) are determined by the wavelength locking control loop.

Fig. 6. Slope of the ON transmittance versus mean or DC current. At Imean=225.0  mA, the slope is nominally zero. The feedback loop is extraordinarily sensitive to small thermal drift in the system, as shown by the error bars.

Fig. 7. Long-term gas sensing test with active wavelength locking feedback loop. (a) Experimentally measured concentration versus the recorded laboratory temperature. (b) Measured and residual drift corrected concentration versus time at 40 Hz update. (c) Allan deviation curve showing the VMR sensitivity as a function of measurement averaging time.