| 摘要 |
Dibits (i.e., pairs of bits) in an input data stream are converted into a modulation signal (e.g., a pi/4-DQPSK modulation signal) using a technique that relies on pre-computed filter output values to reduce the processing required to generate each sample in the modulation signal. The pre-computed filter output values are stored in one or more look-up tables, which are accessed in real time during signal conversion to retrieve values as needed. In one implementation, the dibits, which are represented by complex values in a complex symbol space, are converted into two streams of state variables: one stream for the even symbols and one stream for the odd symbols. In addition, a third stream of state variables identifies whether each of the corresponding symbols is an even symbol or an odd symbol. The three streams of state variables are used to generate address pointers that are used to access the pre-computed filter output values from the look-up tables. Addition/subtraction operations are then applied to the retrieved values to generate the real and imaginary parts of the complex samples in the modulation signal. By greatly reducing the computations involved in generating each complex sample in the modulation signal, the present invention can be used to reduce the costs of implementing the modulation process and/or increase the rate at which the input data stream is over-sampled, thereby reducing distortion in the modulation signal.
|
