和积译码算法.pptx

4.3.1 Choice of Encoder Class The answer to the first question depends on the application. For example, many applications require that the encoder be systematic so that the data are readily observable in the code stream. As another example, parallel turbo codes (Chapter 7) require that the constituent convolutional codes be of the recursive type,and usually systematic. For many applications the encoder class matters very little,particularly applications in which the convolutional code is not a constituentcode of a turbo(-like) code. As seen in Example 4.3 and as we shall see in the discussion below of the third question, the choice of encoder class can affect the encoder memory μ. 4.3.2 Catastrophic Encoders 4.3.3 Minimal Encoders 4.3.4 Design of Convolutional Codes Regarding the fourth question listed in Section 4.3, for reasons affecting decoder complexity, typical values of n and k are very small: k =1, 2, 3 and n = 2, 3, 4. Typical code rates are 1/2, 1/3, 1/4, 2/3, and 3/4, with 1/2 by far the most common rate. 4.4 Alternative Convolutional Code Representations Under this section, we consider alternative representations of convolutional codes that are useful for many other aspects of convolutional codes, such as decoding, code design, and analysis. 4.4.1 Convolutional Codes as Semi-Infinite Linear Codes 4.4.2 Graphical Representations for Convolutional Code Encoders There exist several graphical representations for the encoders of convolutional codes, with each of these representations playing different roles. We start with the finite-state transition-diagram (FSTD), or state-diagram, graphical model for G(D). The encoder realization for G(D) was presented earlier in Figure 4.4(a) and the encoder state is defined to be the contents of the two memory elements in the encoder circuit (read from left to right). From tha