Frequency definitions In signal processing we use several types of frequencies. This may seem confusing at first, but it is really not that difficult.

Analog frequency The frequency of an analog signal is the easiest to understand. A trigonometric function with argument Ω t 2 F t generates a periodic function with a single frequency F. period T the relation T 1 F Frequency is then interpreted as how many periods there are per time unit. If we choose seconds as our time unit, frequency will be measured in Hertz, which is most common.

Digital frequency The digital frequency is defined as f F Fs , where Fs is the sampling frequency. The sampling interval is the inverse of the sampling frequency, Ts 1 Fs . A discrete time signal with digital frequency f therefore has a frequency given by F f Fs if the samples are spaced at Ts 1 Fs .

Consequences In design of digital sinusoids we do not have to settle for a physical frequency. We can associate any physical frequency F with the digital frequency f, by choosing the appropriate sampling frequency Fs. (Using the relation f F Fs ) According to the relation Ts 1 Fs choosing an appropriate sampling frequency is equivivalent to choosing a sampling interval, which implies that digital sinusoids can be designed by specifying the sampling interval.

Angular frequencies The angular frequencies are obtained by multiplying the frequencies by the factor 2: Angular frequency Ω 2 F Digital angular frequency ω 2 f

Signal periodicity Any analog sine or cosine function is periodic. So it may seem surprising that discrete trigonometric signals not necessarily are periodic. Let us define periodicity mathematically. If for all k ℤ we have Analog signals x t x n k T0 , then xt is periodic with period T0. Discrete time signals x n x n k N , then xn is periodic with period N. Consider the signal x t 2 F t which obviously is periodic. You can check by using the periodicity definition and some trigonometric identitites. Consider the signal x n 2 f n . Q:Is this signal periodic? A: To check we will use the periodicity definition and some trigonometric identities. Periodicity is obtained if we can find an N which leads to x n x n k N for all k ℤ . Let us expand 2 f n k N . 2 f n k N 2 f n 2 f k N 2 f n 2 f k N To make the right hand side of equal to 2 f n , we need to impose a restriction on the digital frequency f. According to only f N m will yield periodicity, m ℤ . Consider the following signals x t 2 1 8 t and x n 2 1 8 n , as shown in .

a) 2 1 8 t b) 2 1 8 n Are the signals periodic, and if so, what are the periods? Both the physical and digital frequency is 1/8 so both signals are periodic with period 8. Consider the following signals x t 2 2 3 t and x n 2 2 3 n , as shown in .

a) 2 2 3 t b) 2 2 3 n Are the signals periodic, and if so, what are the periods? The frequencies are 2/3 in both cases. The analog signal then has period 3/2. The discrete signal has to have a period that is an integer, so the smallest possible period is then 3. Consider the following signals x t 2 t and x n 2 n , as shown in .

a) 2 t b) 2 n Are the signals periodic, and if so, what are the periods? The frequencies are 1/π in both cases. The analog signal then has period π. The discrete signal is not periodic because the digital frequency is not a rational number.

Conclusion For a time discrete trigonometric signal to be periodic its digital frequency has to be a rational number, i.e. given by the ratio of two integers. Contrast this to analog trigonometric signals.

Matlab file periodicity.m

Take a look at Introduction Discrete time signals Analog signals Discrete vs Analog signals Energy & Power Exercises ?