Recent research results show that the low-frequency noise of optoelectronic coupled devices has become an important
sensitive parameter affecting its operational status and reliability. Screening optoelectronic coupled devices is an
effective method by measuring the noise power spectrum. However, this method is based on the traditional Fourier
analysis and spectrum analysis, its measuring speed is quite slow, and the method used to establish screening threshold is
more complicated. In this paper, a set of measurement and analysis system based on virtual instrument is set up, which is
composed of dual-channel low-noise pre-amplifiers, dynamic signal analyzer Agilent35670A and PC. According to the
wavelet analysis method, the different kinds of noise can be identified. Through the GPIB control, separating the 1/f
noise, the g-r noise and the burst noise is performed and the noise analysis process is finished by the LabVIEW
procedure. Experimental results demonstrate that this system can not only improve reliability of screening device to
satisfy higher reliability and quality requirement, but also the testing and analyzing process is finished faster and more
accurately than the traditional method.
In this paper the theoretical analysis of low frequency noise sources in Optoelectronic Coupled Devices (OCDs) is given and the relation between typical defects and low frequency noise that consists of 1/f, generation-recombination (G-R) and burst noise is described. A novel measurement system for low frequency noise is introduced here, and nV-level measurement precision can be achieved with the typical low-frequency noise measuring system, which is based on virtual instrumentation. According to statistical and experimental results, a threshold to screen potential devices with excess noise is derived, which has been proved theoretically that the screening criterion is reasonable. Contrasting the former screening criterion, some familiar noise parameters of devices are adopted to establish a novel screening criterion in this paper. At last, the Levenberg-Marquardt regression algorithm is used for the low-frequency noise parameters fitting. The noise parameters are very useful for analyzing the defects.