Made available in DSpace on 2014-06-12T15:52:41Z (GMT). No. of bitstreams: 1 license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2009 / A busca por novas funcionalidades no que diz respeito a melhoria da confiabilidade dos sistemas eletrônicos e também a necessidade de gerir o tempo gasto durante o teste faz do mecanismo Built-in-Self-Test (BIST) um característica promissora a ser integrada no fluxo atual de desenvolvimento de Circuitos Integrados (IC). Existem vários tipos de BIST: Memories BIST, Logical BIST (LBIST) e também alguns mecanismos usados para teste as partes analógicas do circuito. O LBIST tradicional usa um hardware on-chip para gerar todos os padrões de teste com um gerador pseudo aleatório (PRPG) e analisa a assinatura de saída gerada por um registrador de assinatura de múltipla entradas (MISR). Essa abordagem requer a inserção de pontos de teste extras or armazenagem de informação fora do chip que tornará possível alcançar uma cobertura de teste > 98%. Também a geração de todos os estímulos de teste implica no sacrifício no tempo aplicação do teste, o qual pode ser aceitável para pequenos sistemas executarem auto-teste durante a inicialização do sistema mas pode tornasse um aspecto negativo quando testando System-on-chip (SOC) ICs. O fluxo corrente de desenvolvimento de um IC insere scan chains e gera automaticamente padrões de teste de scan para alcançar uma alta cobertura para o teste de manufatura. Técnicas de compressão de dados provaram ser muito úteis para reduzir o custo de teste enquanto reduzem o volume de dados e o tempo de aplicação dos testes. Esse trabalho propõe o reuso de padrões de teste comprimidos usados durante o teste de manufatura para implementar um LBIST com objetivo de testar o circuito quando ele já está em campo. O mecanismo LBIST proposto objetiva descobrir defeitos que podem ocorrer devido ao desgasto do circuito. Uma arquitetura e um fluxo de desenvolvimento semi-automático do mecanísmo LBIST baseado em padrões de teste de scan são propostos e validados usando um SoC real como caso de teste

Test and Diagnosis are critical areas in semiconductor manufacturing. Every chip manufactured using a new or premature technology or process needs to be tested for manufacturing defects to ensure defective chips are not sold to the customer. Conventionally, test is done by mounting the chip on an Automated Test Equipment (ATE) and applying test patterns to test for different faults. With shrinking feature sizes, the complexity of the circuits on chip is increasing, which in turn increases the number of test patterns needed to test the chip comprehensively. This increases the test application time which further increases the cost of test, ultimately leading to increase in the cost per device. Furthermore, chips that fail during test need to be diagnosed to determine the cause of the failure so that the manufacturing process can be improved to increase the yield. With increase in the size and complexity of the circuits, diagnosis is becoming an even more challenging and time consuming process. Fast diagnosis of failing chips can help in reducing the ramp-up to the high volume manufacturing stage and thus reduce the time to market. To reduce the time needed for diagnosis, efficient diagnostic patterns have to be generated that can distinguish between several faults. However, in order to reduce the test application time, the total number of patterns should be minimized. We propose a technique for generating diagnostic patterns that are inherently compact. Experimental results show up to 73% reduction in the number of diagnostic patterns needed to distinguish all faults. Logic Built-in Self-Test (LBIST) is an alternative methodology for testing, wherein all components needed to test the chip are on the chip itself. This eliminates the need of expensive ATEs and allows for at-speed testing of chips. However, there is hardware overhead incurred in storing deterministic test patterns on chip and failing chips are hard to diagnose in this LBIST architecture due to limited observability. We propose a technique to reduce the number of patterns needed to be stored on chip and thus reduce the hardware overhead. We also propose a new LBIST architecture which increases the diagnosability in LBIST with a minimal hardware overhead. These two techniques overcome the disadvantages of LBIST and can make LBIST more popular solution for testing of chips. Modern designs may contain a large number of small embedded memories. Memory Built-in Self-Test (MBIST) is the conventional technique of testing memories, but it incurs hardware overhead. Using MBIST for small embedded memories is impractical as the hardware overhead would be significantly high. Test generation for such circuits is difficult because the fault effect needs to be propagated through the memory. We propose a new technique for testing of circuits with embedded memories. By using SMT solver, we model memory at a high level of abstraction using theory of array, while keeping the surrounding logic at gate level. This effectively converts the test generation problem into a combinational test generation problem and make test generation easier than the conventional techniques. / Ph. D.

With the increase of device complexity and test-data volume required to guarantee adequate defect coverage, external testing is becoming increasingly difficult and expensive. Logic Built-in Self Test (LBIST) is a viable alternative test strategy as it helps reduce dependence on an elaborate external test equipment, enables the application of a large number of random tests, and allows for at-speed testing. The main problem with LBIST is suboptimal fault coverage achievable with random vectors. LFSR reseeding is used to increase the coverage. However, to achieve satisfactory coverage, one often needs a large number of seeds. Computing a small number of seeds for LBIST reseeding still remains a tremendous challenge, since the vectors needed to detect all faults may be spread across the huge LFSR vector space. In this work, we propose new methods to enable the computation of a small number of LFSR seeds to cover all stuck-at faults as a first-order satisfiability problem involving extended theories. We present a technique based on SMT (Satisfiability Modulo Theories) with the theory of bit-vectors to combine the tasks of test-generation and seed computation. We describe a seed reduction flow which is based on the `chaining' of faults instead of pre-computed vectors. We experimentally demonstrate that our method can produce very small sets of seeds for complete stuck-at fault coverage. Additionally, we present methods for inserting test-points to enhance the testability of a circuit in such a way as to allow even further reduction in the number of seeds. / Master of Science