| 摘要 |
In a string of modular fixed point data decimated calculation stages, operands of a later stage are formed by results of an immediately earlier stage. Reducing operations are executed to avoid a subsequent overflow condition. In particular, the results are matched to a uniform threshold value, excess of which could lead to said overflow condition in the next decimated calculation stage. If detecting actual excess, the results of the earlier stage are selectively downsized by a uniform factor to avoid a subsequent overflow. |
