In this step, the higher frequency of the signal is filtered out to make the conversion of the analog signal easier. When people speak, the energy in their voices is normally between the range of 200 to 300 hertz for the lower limits and 2,700 to 2,800 hertz for upper limits (Tomasi 275).
In the case of standard voice or speech communication, this reaches up to 3000 hertz bandwidth. The purpose of this bandwidth limiting is to eliminate aliasing or anti aliasing. Aliasing refers to the frequency of fold over distortion and in order to avoid aliasing, the following equation should be satisfied: fs 2fa , where fs=minimum Nyquist sample rate (hertz) and fa = maximum analog input frequency (hertz). Note that aliasing and antialiasing occurs due to under sampling of the input analog signals (voice). The Nyquist criterion fs 2fa happens because frequency of sample signals is less than the maximum frequency of the analog input signal (Cisco). Thus, an overlapping of the sampling frequency spectrum and the maximum frequency of the analog input signal. Overlapping happens due to inaccuracy of the output low-pass filter to detect the occurrence of overlapping between the frequency spectrum of samples and the analog input signals. The output low-pass filter is used in the reconstruction of the original input signal. As a result, a new signal (false signal) is created from the original source which is then termed as aliasing (Ciscosystems).
Sampling. After the process of filtering in which the higher components of signal frequency are filtered out, the next step is the sampling. In this second stage of converting the analog input signal to digital output voice signal, a sample is taken from the filtered input signal at a condition of constant sampling frequency. The sampling of filtered input is done by using the pulse amplitude modulation (PAM) process. In sampling, the original analog signal is used in modulating the train of pulse amplitude having a constant frequency and amplitude (see figure 1.).
Figure 1. Sampling of Analog Input
The train of pulses moves at the same frequency which is termed as the frequency of sampling. The sampling of voice analog signal can be done at several million times per second. The sampling of the frequency was first determined by a scientist named as Harry Nyquist. He discovered that reconstruction of signal is possible by using the output low-pass filter. Additionally, the signal reconstruction can be done if the frequency of the sample is twice the highest or the maximum frequency of initial input voice signal. The following is the Nyquist criterion (Rumsy 210):
Fs > 2(BW)
Fs = Sampling frequency
BW = Bandwidth of original analog voice signal
Voice digitizing. The next step after the filtering and sampling (using the pulse amplitude modulation, PAM) of the analog input voice signal, is the digitization of samples to be transmitted over a telephone network. This process of digitization of analog input voice signals is termed as the pulse code modulation or PCM. Note that the single difference between the pulse amplitude modulation, PAM and the pulse code modulation, PCM is the fact that the PCM takes one step further in the process (Cisco systems).