Degree Name

Doctor of Phylosophy


School of Electrical, Computer and Telecommunications Engineering


This thesis explores two measurement applications of optical feedback self-mixing system with semiconductor lasers. One is the displacement measurement of a remote target; the other is the measurement of the linewidth enhancement factor which is a characteristic specification of a semiconductor laser.

The above mentioned two types of measurements are performed based on the self-mixing signals acquired with an optical feedback system set-up. Apparently better signal quality will facilitate the proposed algorithms and yield improved accuracies. Hence, the issue of signal pre-processing is addressed before the measurements are conducted by means of digital signal processing techniques. The experimental signals are firstly processed with median filter to achieve basic level of data smoothing and the elimination of the impulsive noises. A high level of data smoothing is achieved by the employment of an artificial neural network. A good accordance is found between the pre-processed noisy self-mixing signal and its clean counterpart in computer simulations.

In order to investigate the displacement measurement with the laser self-mixing system, the analytical solution for the displacement of an external moving target is firstly examined with the Lang-Kobayashi theory. One difficulty that is associated with direct utilization of this solution is the wrapped phase values between 0 and π as a result of the inverse cosine function that is involved in recovering the phases of the reflected light. The basic idea of the phase unwrapping algorithm is to locate the points where the phase is wrapped and thus recover it to its real value by adding or subtract multiple numbers of 2π for the target movement away and towards the laser respectively. The accuracy of this phase unwrapping algorithm under difference levels of laser feedback was then investigated by computer simulations. It was shown that the target displacement can be reconstructed with the accuracy λ/25 under weak feedback regime λ/20 under moderate feedback regime.

The measurement of the linewidth enhancement factor is based on the data-to-model fitting paradigm. That is, the experimental self-mixing data is applied to the theoretical model of the feedback interferometry system and a few model parameters are identified accordingly so that a pre-defined cost function is minimized. In order for the method to be applicable under a universal situation, a genetic searching approach is proposed to locate the global minimum so as to yield the estimation of a set of parameters in the theoretical model including the linewidth enhancement factor. The thorough investigation on the error surface revealed multiple local minima in the cost function and uneven sensitivities of the cost function with respect to different parameters. Thus a global searching procedure is proposed by performing multiple rounds of genetic algorithm with the later round concentrating the searching within the areas that is obtained from the previous round of running the algorithm. It is shown with computer simulations and experiments that the proposed algorithm achieves the accuracy of 3.8% for the measurement of the linewidth enhancement factor.