Low Power Structure Design of GLFSR for BISR Based Application
Kakarla Hari Kishore, Fazal Noor Basha, Pavuluri Srinivas, Atluri Jhansi, Shaik Moulali · 2012
In this paper, structure design and optimization of a Built- In Self-Repair (BISR) design based on Generalized Linear Feedback Shift Registers (GLFSRs) are described. A new and effective pseudorandom test pattern generator, termed GLFSR, is introduced. These are Linear Feedback Shift Registers (LFSR's) over a Galois field GF. Unlike conventional LFSR's, which are over GF, these generators are not equivalent to cellular arrays and are shown to achieve significantly higher fault coverage. Experimental results are presented in this paper depicting that the proposed GLFSR can attain fault coverage equivalent to the LFSR, but with significantly fewer patterns. The proposed GLFSR structure will be implemented in SPARTAN-3E by using VHDL. The percentage improvement for fault coverage. This approach reduces the number of transitions in the scan chains and thus minimizing power consumption. By using encoding algorithm, the percentage improvement for power consumption. Linear binary Feedback Shift Registers (LFSR's) are extensively used for random pattern generation in the Built-In Self-Test (BIST) environment. This paper introduces a new pseudorandom pattern generator, called a Generalized Linear Feedback Shift Register (GLFSR), which is shown to outperform both LFSR and cellular arrays. These GLFSR's are defined over a higher order Galois field GF. Unlike the LFSR, these GLFSR's are not equivalent to cellular automata, therefore warranting investigation. These GLFSR's provide a uniform framework for study of LFSR, MISR, and multiple MISR. In addition, GLFSR's yield a new structure when the feedback polynomial is primitive and termed as MLFSR. Although this new structure's effectiveness has not yet been fully explored, its effectiveness in the context of test pattern generation is studied here. The results in this paper along with the previously reported work in demonstrate GLFSR as a new and powerful tool for BIST implementation. Importantly, what is shown in this paper is that GLFSR's are significantly more effective than the LFSR, commonly used as a test pattern generator and as an MISR for signature analysis. This is a significant finding when viewed in the context of earlier results which have established that the primitive GLFSR's outperform LFSR's as a signature analyzer. Here, we propose the application of the GLFSR to generate pseudorandom patterns for stuck-at and transition faults. A good pattern generator should generate patterns with high degree of randomness and should have efficient area implementation. A number of LFSR- based pattern generators proposed in the past use combinations of shift registers and XOR gates for particular applications. The GLFSR approach proposed here can be interpreted as a systematic alternative to the popular pattern generators used by IBM which comprises of the standard LFSR and extra XOR gates used at the output of the LFSR as shown in Fig. The outputs of the LFSR are transformed by these extra XOR gates to enhance randomness for the patterns. These XOR gates are added in a two step process in an adhoc manner. Our GLFSR design provides a one-step technique where the additional XOR gates are integral to the design and are the byproduct of the design methodology. This results in a better random distribution of the patterns and potentially lesser dependencies at the output. In the paper, we assume a BIST environment where most of the faults are to be detected using the proposed GLFSR's, and then the remaining few very hard-to-detect faults are to be detected by applying predetermined patterns, or by using random weighted patterns. For large circuits, where the testing time is a major concern, using GLFSR's to detect the large fraction of the faults can significantly reduce test time. Specifically, what is shown here is that the patterns generated by the proposed structure 1. can attain the same stuck-at fault coverage in combinational circuits with a small fraction of the number of tests required by LFSR, 2. can attain the same transition fault coverage as LFSR's with many fewer patterns, and 3. are more randomly distributed than LFSR's and cellular arrays. First, this paper describes the behavior of the GLFSR, suggesting a technique that exhaustively generates all the patterns using the GLFSR.