The notion of time reversal in discrete systems occasionally arises in the study of the discrete Fourier transform (DFT), the mathematical analysis of digital filters, and even in practice (straight time reversal is used in a digital filtering scheme described in Section 13.12). We give the topic of time reversal some deserved attention here because it illustrates one of the truly profound differences between the worlds of continuous and discrete systems. In addition, the spectral effects of reversing a time sequence are (in my opinion) not obvious and warrant investigation.

Actually, in discrete-time systems there are two forms of time reversal we need to think about. Consider the six-point x(n) time-domain sequence

Equation C-1. 

Due the periodicity properties of discrete sampled representations (discussed in Section 3.17), we can depict the x(n) time sequence as samples on a circle as shown in Figure C-1(a). There we arbitrarily assign positive time flow as counterclockwise rotation. (For our UK friends, counterclockwise means your anticlockwise.)

Time reversal, as defined here for sequences that are treated as periodic, means traveling clockwise around the circle (in the negative time direction) creating a new time sequence

Equation C-2. 

We call x_c(n) the circular time reversal of x(n), where the subscript c means circular reversal, and depict x_c(n) as in Figure C-1(b).

Figure C-1. Circular representations of periodic sequences: (a) original x(n) sequence; (b) circular time reversal of x(n).

The interesting issue here is that for real N-point time sequences, the DFT of x_c(n) is the complex conjugate of the DFT of x(n). That is,

Equation C-3. 

where the DFT index is 0 ≤ m ≤ N–1. Due to the conjugate symmetry of DFTs of real sequences, we should realize that X*(m) is a straight reversal of the X(m) samples.

Let's illustrate Eq. (C-3) with an example. With X(m) representing the DFT of x(n), we can write down X(m)'s m = 4 sample X(4) as

Equation C-4. 

Because e^–j2πk/6 has a period of 6, we can write Eq. (C-4) as

Equation C-5. 

Next, let's write down the (circular-reversed) X_c(m)'s m = 4 sample X_c(4) as

Equation C-6. 

or

Equation C-7. 

Replacing X_c(4)'s negative angles with their positive-angle equivalents yields

Equation C-8. 

which is the conjugate of Eq. (C-5) demonstrating that X(m) and X_c(m) are complex conjugates.

An alternate time reversal concept, that we'll call straight time reversal, is the simple reversal of Eq. (C-1)'s x(n) yielding an x_s(n) sequence

Equation C-9. 

where the subscript s means straight reversal. For real N-point time sequences, the DFT of x_s(n) is

Equation C-10. 

We can demonstrate Eq. (C-10) the same way we did Eq. (C-3), but consider Figure C-2. There we show the time-reversed version of the periodic x(n), showing both the 6-point x_s(n) and the 6-point x_c(n) sequences. Notice how x_s(n) is shifted backward in time by 5 samples from x_c(n).

Figure C-2. Periodic sequences x_s(n) and x_c(n).

Using the principle of the DFT's shifting theorem from Section 3.6, we know that X_s(m) is equal to X_c(m) times a linear phase shift of e^{–j2πm(5)/6} for our N = 6 example. So, in the general N-point sequence case

Equation C-11. 

which validates Eq. (C-10).