Research On Built-in Self-test Of Embedded Memory Based On Test Time Optimization

With the application and innovation of System on Chip(So C),the proportion of embedded memories in the So C is increasing.Memory built-in self-test is a testing technology,which is commonly used in embedded memories.It is necessary to improve test efficiency and reduce test time for a large number of embedded memories of So C.Therefore,this research studies the embedded memory built-in self-test based on test time optimization.When static and dynamic faults coexist in a memory,the diagnostic data has a problem of data redundancy.When faced with fault rows or fault columns,the redundancy rate increases sharply,which causes a long test time.Aiming at the problem,combining the lossless compression of diagnostic data and fault patterns recognition,a memory built-in self-test design that can identify the fault modes and compress their diagnostic data losslessly is proposed.The problem of redundancy of diagnostic data is solved by identifying the fault patterns as row faults,column faults or unit faults,and performing lossless compression on their diagnostic data.The experimental results show that the design compresses the diagnosis data losslessly and reduces the test time of a memory.In the 8k×16 memory model,the output time of the design is 35.16% lower than that of the MEB design,and the area overhead ratio is only increased by 0.45%.In addition,multi-memory built-in self-test of So C seeks a test scheme with the shortest test time,which is constrained by the area overhead,test power consumption and test time of So C.Aiming at the problem,the multi-memory built-in self-test is modeled as a multi-objective optimization problem,and a multi-objective clustering genetic annealing algorithm is proposed.On the basis of genetic algorithm,memory compatible groups are obtained by memory clustering,and high quality initial solutions are obtained by heuristic method.An objective function with different weights under multiple constraints is proposed.Simulated annealing algorithm is used to avoid the risk of local optimal solutions for better individuals.The experimental results show that the proposed algorithm gets better results than genetic algorithm,and obtains a set of memory solutions,whose test time is 48.7% lower than the existing method,saving the test time of multi-memory.