仪器仪表学报
儀器儀錶學報
의기의표학보
CHINESE JOURNAL OF SCIENTIFIC INSTRUMENT
2009年
11期
2372-2378
,共7页
肖继学%谢永乐%陈光(礻禹)
肖繼學%謝永樂%陳光(礻禹)
초계학%사영악%진광(시우)
DSP%累加器%可测性设计%测试%数据通路
DSP%纍加器%可測性設計%測試%數據通路
DSP%루가기%가측성설계%측시%수거통로
DSP%accumulator%design-for-testability%test%data path
在综述VLSI结构可测性设计方法的基础上,提出了DSP数据通路基于累加器测试的结构可测性设计方案:利用选择器或三态门实现电路测试、工作模式的切换;在测试模式时,电路中的寄存器复用为扫描链以完成测试矢量的传送从而提高电路的可测试性能.基于本方案的FFT处理器、IIR滤波器、DF-FPDLMS自适应滤波器的数据通路的可测性设计,若忽略数据线延迟,其关键路径仅比原来的分别增加了1、2、0倍的选择器或三态门门延迟.实验表明,若字宽、阶数均为8,它们所需额外硬件开销分别为原来的5.416%、4.969%、4.783%,关键路径分别增加了1.839%、2.382%、0.036%.结果表明,该方案通用性好,扩展性强,额外硬件开销小,几乎不会影响原电路的性能.
在綜述VLSI結構可測性設計方法的基礎上,提齣瞭DSP數據通路基于纍加器測試的結構可測性設計方案:利用選擇器或三態門實現電路測試、工作模式的切換;在測試模式時,電路中的寄存器複用為掃描鏈以完成測試矢量的傳送從而提高電路的可測試性能.基于本方案的FFT處理器、IIR濾波器、DF-FPDLMS自適應濾波器的數據通路的可測性設計,若忽略數據線延遲,其關鍵路徑僅比原來的分彆增加瞭1、2、0倍的選擇器或三態門門延遲.實驗錶明,若字寬、階數均為8,它們所需額外硬件開銷分彆為原來的5.416%、4.969%、4.783%,關鍵路徑分彆增加瞭1.839%、2.382%、0.036%.結果錶明,該方案通用性好,擴展性彊,額外硬件開銷小,幾乎不會影響原電路的性能.
재종술VLSI결구가측성설계방법적기출상,제출료DSP수거통로기우루가기측시적결구가측성설계방안:이용선택기혹삼태문실현전로측시、공작모식적절환;재측시모식시,전로중적기존기복용위소묘련이완성측시시량적전송종이제고전로적가측시성능.기우본방안적FFT처리기、IIR려파기、DF-FPDLMS자괄응려파기적수거통로적가측성설계,약홀략수거선연지,기관건로경부비원래적분별증가료1、2、0배적선택기혹삼태문문연지.실험표명,약자관、계수균위8,타문소수액외경건개소분별위원래적5.416%、4.969%、4.783%,관건로경분별증가료1.839%、2.382%、0.036%.결과표명,해방안통용성호,확전성강,액외경건개소소,궤호불회영향원전로적성능.
Approaches of structural design-for-testability are overviewed. Then a scheme of structural design-for- testability for DSP data path to be tested based on accumulation is presented. In the scheme, multiplexers or tristate gates are utilized to switch the modes of the data path between test and work, and some registers in the circuit are reused as scan chains during testing to deliver test patterns to the circuit, and then the circuit testability is improved. Based on the scheme, the data paths of FFT processors, IIR filters and DF-FPDLMS adaptive filters were designed. The critical paths become a little longer and the increased computation times are one, two and zero multiples of the delay of a multiplexer or tristate gate, respectively if the data bus delay is ignored. The experiments of 8-order DSP data paths with 8-bit word width show that the additional hardware overheads are 5.416%, 4.969% and 4.783% of the original circuit hardware overheads, and the critical paths are increased by 1.839%, 2.382% and 0.036% respectively. Results show that the performance of the scheme is good in generality and extension with a little additional hardware overhead and degradation of the original circuit performance.