Algoritma memain peranan yang penting dalam Memori BIST dimana struktur algoritmanya berupaya mentakrif pengesan liputan kesalahan di dalam sesuatu sistem. Oleh yang demikian, jika penambahbaikan algoritma itu dibuat, ia membenarkan lebih banyak jenis kesalahan dapat dikesan walau hanya dengan satu ujian sahaja. Bagi algorithma MARCH C-, ia merupakan salah satu algoritma stabil yang mempunyai liputan jenis kesalahan yang tinggi kerana ia mampu mengesan banyak kesalahan seperti Address Decoder Fault, Stuck-At Fault, Transition Fault dan Coupling Fault namun begitu ia tidak berupaya untuk mengesan jenis kesalahan seperti Read Destructive Fault dan Data Retention Fault. Bagi meningkatkan prestasi algoritma MARCH C- untuk mengesan dua jenis kesalahan tersebut, beberapa pengubahsuaian perlu dilakukan. Melalui hasil analisa corak-corak kesalahan dan jenis algoritma lain, dua metodologi bagi penambahbaikan algorithma MARCH C- telah dikesan. Dengan menggunakan operasi pembacaan yang berganda dalam elemen tunggal MARCH yang terdapat didalam algoritma MARCH-NU boleh membantu untuk mengesan Read Destructive Fault manakala penggunaan HOLD time dalam algoritma MARCH 9N mampu mengesan Data Retention Fault. Melalui gabungan kedua-dua cara tersebut dalam algoritma MARCH C- membolehkannya untuk mengesan kesalahan Data Retention Fault dan Read Destructive Fault secara 100% tanpa menjejaskan prestasi asal MARCH C-. Selain daripada itu, penambahbaikan algoritma MARCH C- juga boleh mengesan jenis-jenis kesalahan secara struktur dan mengasingkan kesalahan diantara kesalahan tingkahku (Data Retention Faults) dan kesalahan umum memori serta dianatara kesalahan Stuck-At Fault dan Transition Fault. Fault Injection test telah dijalankan bagi memastikan Enhanced MARCH C- boleh mengenal pasti corak-corak kesalahan. Enhanced MARCH C- mempunyai kadar lulus 100% dalam mengenal pasti Stuck-At Fault, Transition Fault, Data Retention Fault, Inversion Read Fault, Incorrect Read Fault dan Read Destructive Fault. Ia mempunyai kadar 95.16% dalam mengenal pasti State Coupling Fault dan Idempotent Coupling Fault. Sebagai konklusi, Enhanced MARCH C- algorithm telah berjaya meningkatkan liputan jenis kesalahan dalam Data Retention Fault dan Read Destructive Fault manakala ia juga mempunyai abiliti untuk mengenal pasti jenis kesalahan secara struktur.
Algorithms plays an important role in the Memory Built-In Self-Test as its structure will define the fault coverage of the system. Thus, the improvement of the algorithms will allow more fault types being identified in single test. For MARCH C- algorithm, it was a quite balance algorithm as it has the ability to cover Address Decoder Fault, Stuck-At Fault, Transition Fault and Coupling Fault but it was not able to identify Read Destructive Fault and Data Retention Fault. To increase the fault coverage of the MARCH C- algorithm on that two fault types, some enhancements were needed to make on the algorithm. Through the analysis of the fault patterns and other algorithm, it was found that multiple read operation within a single MARCH element from MARCH-NU algorithm can aid in the identification of the Read Destructive Fault whereas having HOLD time in the MARCH 9N algorithm would expose the Data Retention Fault. By integrating both new fault identification methodologies into the MARCH C- algorithm and enhance it so that it would able to increase the identification of the Data Retention Fault and Read Destructive Fault by 100% and without causing any performance to interfere with the original fault identification ability. Besides that, the enhanced MARCH C- algorithm can structurally identify the fault types and differentiate among the behavioural faults (Data Retention Faults) and common memory faults as well as among Stuck-At Fault and Transition Fault. Fault injection test were carried out to ensure the coverage of the enhanced MARCH C- algorithm. It was having high passing rate which are 100%
identification of the Stuck-At Fault, Transition Fault, Data Retention Fault, Inversion Read Fault, Incorrect Read Fault and Read Destructive Fault. It also obtained 95.16% for the State Coupling Fault and Idempotent Coupling Fault. As a conclusion, the enhanced MARCH C- algorithm had increased its fault type identification in Data Retention Fault and Read Destructive Fault while having the ability to structurally identify the fault types.
Fsm-based enhanced march c- algorithm for memory built-in self-test _Ch’ng Ming Zong_E3_2017_MFAR