| 摘要 |
PURPOSE:To execute a filter arithmetic processing with a high accuracy at high speed by multiplying a coefficient group related to an input variable, or the coefficient group related to an output variable by 2S, and at the same time, executing the filter arithmetic processing by using the maximum number of digits, not producing an overflow even in the course of the operation of the multiplication of the input variable by 2NL. CONSTITUTION:An NL setting part 6, which decreases NL when it is decided by an overflow deciding part 5 that the overflow has been produced, and sets it at the value, which is 1 smaller than NL producing the overflow, when it is decided that the overflow is not produced, is provided. Then, the NL value, set by the NL setting part 6, is informed to an input signal setting part 3, and the filter arithmetic processing is executed by using the maximum effective number of digits, not producing the overflow, and the desired result of the filter arithmetic processing is obtained by multiplying the said generated result by 2<-(>S<+>NL<)>, and outputted. Thus, the highly accurate filter arithmetic processing, in which a digit omission, etc. is not produced even if a fixed decimal point operation by a digital signal processor, etc. is used, can be executed at high speed.
|
