We will now derive the convolution representation for LTI filters in its full generality. The first step is to express an arbitrary signal as a linear combination of shifted impulses, i.e.,
If the above equation is not obvious, here is how it is built up intuitively. Imagine as a 1 in the midst of an infinite string of 0s. Now think of as the same pattern shifted over to the right by samples. Next multiply by , which plucks out the sample and surrounds it on both sides by 0's. An example collection of waveforms for the case is shown in Fig.5.4a. Now, sum over all , bringing together the samples of , to obtain . Figure 5.4b shows the result of this addition for the sequences in Fig.5.4a. Thus, any signal may be expressed as a weighted sum of shifted impulses.
Equation (5.4) expresses a signal as a linear combination (or weighted sum) of impulses. That is, each sample may be viewed as an impulse at some amplitude and time. As we have already seen, each impulse (sample) arriving at the filter's input will cause the filter to produce an impulse response. If another impulse arrives at the filter's input before the first impulse response has died away, then the impulse response for both impulses will superimpose (add together sample by sample). More generally, since the input is a linear combination of impulses, the output is the same linear combination of impulse responses. This is a direct consequence of the superposition principle which holds for any LTI filter.

We repeat this in more precise terms. First linearity is used and then timeinvariance is invoked. Using the form of the general linear filter in Eq.(4.2), and the definition of linearity, Eq.(4.3) and Eq.(4.5), we can express the output of any linear (and possibly timevarying) filter by
where we have written to denote the filter response at time to an impulse which occurred at time . If we are to be completely rigorous mathematically, certain ``smoothness'' restrictions must be placed on the linear operator in order that it may be distributed inside the infinite summation [37]. However, practically useful filters of the form of Eq.(5.1) satisfy these restrictions. If in addition to being linear, the filter is timeinvariant, then , which allows us to write
The infinite sum in Eq.(5.5) can be replaced by more typical practical limits. By choosing time 0 as the beginning of the signal, we may define to be 0 for so that the lower summation limit of can be replaced by 0. Also, if the filter is causal, we have for , so the upper summation limit can be written as instead of . Thus, the convolution representation of a linear, timeinvariant, causal digital filter is given by
for causal input signals (i.e., for ).
Since the above equation is a convolution, and since convolution is commutative (i.e., [84]), we can rewrite it as
or
This latter form looks more like the general difference equation presented in Eq.(5.1). In this form one can see that may be identified with the coefficients in Eq.(5.1). It is also evident that the filter operates by summing weighted echoes of the input signal together. At time , the weight of the echo from samples ago [ ] is .