[信息与通信]DSP与嵌入式处理器基础及ADI DSP介绍.ppt

接下来，用xn-5乘以c0，xn-6乘以c1，xn-7乘以c2。然后将三个得数相加得到下一个输出。 * * Once again, the very same thing here, and at a high level, this is pretty much how a filter, an FIR in this case, works. 按上述方法类推，一直递增到高阶数，这就是滤波器、也就是本例中FIR的工作原理。 * * So, alternatively to FIRs, we also have IIRs, as I mentioned earlier during the presentation. IIR stands for infinite impulse response filter, so IIRs have an impulse response function that is a known zero over an infinite length of time. So in principle, it’s just like an FIR but with infinitely many taps, which in return, will require infinitely many multiplies and infinitely many samples. But obviously this isn’t realistic in today’s DSP world. And therefore, we have to cheat. IIRs use feedback enabling the system to remember infinitely many previous inputs. The way it works is that instead of working on an infinite number of inputs, which is simply impossible to do in real life, it feeds the output back into the input path, achieving in this way an infinite impulse response. An FIR filter has a number of useful properties which sometimes makes it preferable over an IIR filter. FIRs, as we have seen, require no feedback, and this means that any rounding errors in an IIR filter are carried on because of the feedback, while in FIR filters, this doesn’t happen, which also makes FIR filters easier to implement. The main disadvantage of FIR filters is that they require considerably more computational power compared to an IIR filter with similar characteristics. That said, many DSPs today provide specialized hardware features to make FIR filters just as efficient as IIRs for many different applications. By the way, not to forget that all the usual analog issues apply here as well—stability, phase distortion, ripples, etc. We’ll actually talk more about digital versus analog filters in a second. 除了FIR之外，还有之前提到过的IIR。IIR的全称是无限脉冲响应滤波器，IIR的脉冲响应函数在无限长时间内收敛为零。因此，基本上它和FIR很相似，只是具有无限多个抽头，这就需要无限多次乘法运算和无限多个样本。但是很明显，这在当今的DSP领域并不现实。因此，我们需要采用特别的方法。IIR采用反馈的方式使系统记住无限