>>Analog Signal Processing<<
By all practical means, this “old art of analog signal processing” can never be replaced by digital signal processing due to obvious cost and effort constraint. On the other hand, there are a lot of things you can do with digital signal processing that analog can never achieve. Any how, I am not aware of a “modern DSP” versus, I guess, “traditional DSP”. Perhaps, you can show us the difference.
>>Digital Signal Processing<<
I don't know what do you mean by more data points, more bits or higher frequency of sampling, but any how, it does not mean the signal is going to be cleaner. The trick is all in the algorithm itself that means the software. The majority of DSP algorithm is done by FFT (fast Fourier Transform) not these simple filters as you are thinking of, although you can make a good argument that FFT itself is a filter of some sort. DSPs are designed to do a multiply and addition very quickly. I suppose that these functions can be handled by the math coprocessor, but the question is speed. Could a processor be fast enough to take over the role of a DSP? Could a processor take over the role of a math coprocessor?
In some applications, real time processing is crucial such as motor controls. In such dedicated and specific applications, you are correct to point out that a dedicated microcontroller with an integrated DSP would be greatly advantageous. However, a P-II or a K-6 family of processors would certainly benefit from a powerful dedicated DSP integrated into the silicon, just like the math coprocessor done almost a decade before.
Now, what do you think this MMX is? Could it be some sort of a DSP? Obviously DSP's function does overlap with a math coprocessor, that is why the earlier generation of MMX instructions overlap with math coprocessor ones. Well, if MMX is a primitive DSP in action, don't you think there are still tremendous improvement over this field?
Time Traveler
|
