【作者】 郜峰利；

【导师】 郭树旭；

【作者基本信息】 吉林大学 ， 电路与系统， 2008， 博士

【摘要】 本论文围绕半导体激光器的低频电噪声的检测技术展开研究。首先,作为论文的前期工作,在研究了半导体激光器驱动技术及光电特性的基础上,针对电导数测试过程中的强噪声的影响,提出把非线性扩散方程应用于系统中的电导数的降噪,取得了较好的降噪效果,在此基础上,分别基于虚拟仪器技术和嵌入式技术开发了两套大功率半导体激光器的光电特性检测系统,并把非线性扩散方程成功用于设备中的电导数降噪。研究了半导体激光器低频电噪声与器件可靠性的关系,通过所搭建的低频电噪声的直接检测系统,测量了大功率半导体激光器在低偏置电流下的低频电噪声,详细分析了低偏置电流下的电压1/f噪声与器件的可靠性的关系,提出了用低偏置电流下电压1/f噪声的幅度随注入电流的变化趋势来评价大功率半导体激光器的可靠性。针对低频电噪声的直接检测系统的缺点,为了提高1/f噪声的检测效率,开展了加性高斯白噪声背景下半导体激光器1/f噪声的估计研究,系统研究了1/f分形信号的计算机合成方法、基于小波变换的1/f分形信号的参数与波形估计等问题,在此基础上,提出了一种基于最小二乘的1/f分形信号的参数与波形估计方法,并分别通过计算机仿真和半导体激光器的实测数据实验验证了该方法的可行性,为利用数据采集卡开发半导体激光器低频电噪声的数字检测系统奠定了理论基础。更多还原

【Abstract】 With the development of optoelectronic technology, simconductor laser diodes (LDs), as a kind of important light source devices, are widely applied in the domains of high-technology such as industry, medicine treatment, military science and information technology etc., and their applied domains are also expanded. It is always expected that LDs will work for a long time reliably, whether they are used in any domain, so, their reliability is of great concern in practical applications. At present, the usual method to evaluate and screen the reliability of LDs is electrical aging for them, but in the electrical aging, the devices having good reliability are usually destructed, and moreover, electrical aging is time-consuming. So, electrical aging method can not only affect the quality and using-life of LDs, but also reduce the efficiency to evaluate and screen their reliability.The factors affecting the reliability are one of the primary sources producing low-frequency electrical noise, which intensity can reflect the quality and reliability of devices. Therefore, low-frequency electrical noise, as a diagnosis tool of quality and reliability of devices, has more and more demonstrated its study advantage because of its fast, non-destructive, sensitive merits. Furthermore, the research on the correlation between low-frequency electrical noise and reliability of devices, in evaluating and screening reliability of devices, can also improved the manufacture technics of low-noised devices, which has been an important investigated subject. It is said that, it has very high value and science significance that characteristics of low-frequency electrical noise are used in the representation of reliability of LDs. So, it is the precondition of the research on correlation between low-frequency electrical noise in LDs and their reliability how to detect this kind of very low stochastic signal accurately and fast.The work in this thesis is supported by the National Natural Science Foundation of China (Grant No.60471009)“Research on detection method of noise in high-power semiconductor laser diodes and their reliability correlation”and the high technology key project of Jilin province (Grant No. 20040301-4)“The quality detection and screening equipment exploitation of high-power semiconductor laser diodes”. This work is carried out mostly in detection technology of low-frequency electrical noise in LDs.Firstly, the production and development of LDs is summarized, and their application in industry, medicine treatment, military science and information technology and so on is introduced. The representation of low-frequency electrical noise used in reliability of LDs is described, and consequently, the significance of characteristics and detection of low-frequency electrical noise in LDs is presented. The several kinds of common methods to dectect 1/f noise in LDs are also summarized.Secondly, to better understand the essential characteristics of LDs and carry out the following the research on detection of low-frequency electrical noise of LDs, from the driving technology of LDs, the characteristics of V-I and optic-power in LDs are investigated. According to the principle about the circuit of voltage-controlled constant-current source, a programmable driver for LDs is accomplished, which flexible controlling of the driving current is achieved, and some protective methods to LDs, such as current limit, avoiding electrical pulse and delay start-up etc. problems, are completely finished by the software, insteading of mostly considering the solved means on hardware circuits. By performance testing, the linearity and stability of the driver are competent for driving needs of LDs. The photoelectric sampling circuit is designed by using a photoelectric cell as a sensor and connecting it to a multilevel amplifier, which achieves the linearity conversion from the optic power to the voltage. Based on the above designs, respectively by using PCI-6014 DAQ and LPC2136 MCU which has an ARM7 kernel, we developed two sets of equipments to detect the electrooptics characteristics of LDs.The nonlinear diffusion equation is applied to denoising of electrical derivative data according to denoising theory of nonlinear diffusion equation in digital image processing. The experiments indicat that this denoising method can not only effectively reduce the noise in electrical derivative data, but also preserve the dip hight to conquer the disadvantage of median filter. The nonlinear diffusion equation has been successfully used in the above two electrooptics detecting systems.Thirdly, we mostly study the correlation between 1/f noise and reliability of LDs at low bias currents, for the particularity of high-power LDs needing heat-emission because of high operating currents.The low-frequency electrical noise basis of semiconductor devices is introduced; two importrant models of 1/f noise, surface carrier number fluctuation model and mobility fluctuation model are described; the correlation between 1/f noise and reliability of LDs and impact of every parts of devices over 1/f noise are investigated, and then the relation of voltage 1/f noise vs bias current at low bias currents is deduced, and in principium the detection results of 1/f noise is analysed compendiously; on this basis, a set of system to directly detect the low-frequency electrical in LDs is developed, in which the key technologies such as grounding and shielding the devices, parameters getting are introduced detailedly.Using this direct detection system, the voltage low-frequency noise power spectral density in high-power InGaAsP/GaAs quanta well semiconductor laser diodes (QW-LDs) was measured below one-fiftieth threshold current. The experiment results show that, the low-frequency electrical noise in LDs is 1/f noise ,which magnitude is decreasing along with the increasing of bias currents on the whole, and g-r noise is not in existence. According to the value of dynamic resistance of LDs at low bias current, the 1/f noise model at low bias currents is deduced. Based on the model, it is analyzed that 1/f noise at low bias current mainly arises from the active region and the creepage resistance, and moreover, it is proposed that the magnitude and its variety trend of 1/f noise magnitude vs the bias current are nearly correlative with the reliability of LDs.Finally, we carried out the studies of estimating the parmenters and wave of 1/f-type fractal signal submerged in Gauss-white noise, for there are the disadvantages of the above derect detection system such as complicated detection process, longer detection time, part backward detection technique and so on. The fundamental theories of 1/f process and wavelet transformation are introduced; the correlation and variance characteristics of binary wavelet coefficient are deduced; the methods to synthesize 1/f fractal signal are investigated, especially, the computer simulation experiments of synthesized 1/f fractal signal based on wavelet reverse transfromaion are carried out in detail, and other methods are also introduced; concentrating on the problems of the parameters and wave estimation for 1/f fractal signal in additive white noise background, we carry out experiment research, and the maximal likehood estimation is introduced.A novel method to estimate the parameters of 1/f-type fractal signal is proposed according to characteristics of the wavelet coefficient variance versus scale of 1/f-type fractal signal in white noise. Then the parameter estimation of 1/f-type fractal signal is according with parameter estimation conditions of least square method. The simulation experiment results demonstrate that, this method can effectively estimate the parameters of 1/f-type fractal signal to separate it from additive white noise. The experiment results of the actual measured data indicate that, this method can effectively extract the 1/f noise submerged in white noise of LDs, and moreover, the estimated 1/f signal is better accordant with the measured results of the parallel apparatus to further validate the feasibility of least square method.更多还原