| 摘要 |
In order to reduce the computational complexity of maximum a posteriori processing in a turbo decoder the operation is often switched to the log domain and known as log-MAP decoding. However this still involves the calculation of a log of a sum of exponents which is computationally intensive. The invention uses Pad’ approximants wherein power series derived from received data and coefficients are divided by one another. In one embodiment the Pad’ approximant is convened into the closely related continued fraction form. In this form the approximation can be derived using a series of cascaded computation units (500a-c) each comprising an adder whose output feeds a divider which then feeds the adder of the next unit. This unit can be easily scaled up to provide more accuracy by adding further units. |
