Date of Submission


Date of Award


Institute Name (Publisher)

Indian Statistical Institute

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Subject Name

Computer Science


Advance Computing and Microelectronics Unit (ACMU-Kolkata)


Bhattacharaya, Bhargab Bikram (ACMU-Kolkata; ISI)

Abstract (Summary of the Work)

Today’s integrated circuits comprise billions of interconnected transistors assembled on a tiny silicon chip, and testing them to ensure functional and timing correctness continues to be a major challenge to designers and test engineers with further downscaling of transistors. Although substantial progress has been witnessed during the last five decades in the area of algorithmic test generation and fault diagnosis, applications of combinatorial and machinelearning (ML) techniques to solve these problems remain largely unexplored till date. In this thesis, we study three problems in the context of digital logic test and diagnosis. The first problem is that of fault diagnosis and can be stated as follows. Given the output syndromes for test responses in a circuit-under-test (CUT), localize the fault (root-cause) in the netlist; solution to this problem is required in order to fine-tune the process technology and to improve yield in subsequent production cycles. The second problem deals with the issue of unknown logic value (X) and answers the following question. Given a CUT and a test set T, how the fault-coverage of T is impacted when an unknown logic value (X) arrives at one of its inputs. Is there any computationally efficient mechanism to grade the CUT-inputs based on their X-sensitivity? A solution to this problem is needed in X-cancellation for optimizing test costs and to improve reliability. In the third problem, we investigate, given a CUT, how the structure of the underlying network can be represented in a compact and lossless form, so as to make them easily readable by ML-tools for test and diagnostic purposes. This is a classical problem of representing directed graphs that facilitates feature extraction. We present a combinatorial solution to the first problem, a learning-based technique to handle the second problem, and a novel solution to the third problem that relies on Pr¨ufer sequence.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


Included in

Mathematics Commons