Real-time multi-channel simulation of sonar signals by spectral shaping of white noise using an array of digital signal processors

The signal throughput of a multi-channel sonar simulator is n*f where n is the number of channels and f is the sampling frequency. Under the simplifying assumption that higher order terms can be ignored, the processing load requirement for signal generation per sampling period is proportional to the number of signal sources - say, p - that we wish to simulate. Thus, the total instruction cycles required for the simulator may be taken as n*f*(a+p*q), where q is the average number of cycles needed to generate a signal value corresponding to a source and a is that for the ambient noise. For realistic scenario simulation, p cannot be too small. Similarly, q increases with increased sophistication of the effects to be incorporated. The total load therefore is in the range of GigaFlops. Parallel processing technology is most suited to handle such large throughput requirement. One of the attractive design approaches for the development of a Signal Noise Simulator is to use an array of standard DSP elements operating in parallel. Such a system can generate user specified scenarios with multiple moving targets, each with multiple tonals and varying degree of modulation. A number of facilities may be provided to tune the simulator output as required by the user. The modular design of such a system allows incremental building up of simulation complexity by the addition of parallel processing nodes in the form of standardised hardware modules. This paper describes the noise generation, inverse beam forming, filtering and related algorithms needed in the development of an SNS and the synchronisation issues of the parallel implementation.