For example, if an analog signal is sampled once per millisecond (ms), then the sampling frequency is 1 kHz and the sampling rate is 1000 samples per second. If the sampled signal has a frequency of 1 kHz, then the signal is sampled once per period. It cannot be reproduced. If, on the other hand, the signal frequency is 100 Hz, then the signal is sampled ten times at the same sampling frequency. The signal is therefore easily reproducible. The sampling rate must therefore be in a certain relation to the signal frequency. This relation is given by the sampling theorem. According to this theorem, a signal reproduction requires a sampling rate that is at least twice as high as the signal frequency. This is true for sinusoidal signals for their 1st harmonic, but not for pulse or square wave signals.
In the case of voice transmission via ISDN with a maximum frequency range of 4 kHz, the sampling rate is 8 kHz, which corresponds to a sample interval of 125 µs. In the case of audio with a maximum frequency range of 20 kHz, the sampling rate is 44.1 kHz (22.67 µs) and 48 kHz (20.83 µs). For high-quality multi- channel audio, the sampling rate can be as high as 192 kHz. Much higher values result for video and HDTV. For example, digital video with a bandwidth of 6.5 MHz for the luminance signal results in a sampling rate of over 13 MHz and a sample interval of 74 ns. The sampling rate is even higher for HDTV with 74 MHz and a sample rate of 13.5 ns.
For pulse-shaped signals, the sampling rate must be many times higher than their fundamental frequency, otherwise important pulse characteristics cannot be determined. If the sampling rate is many times higher than the theoretically required sampling frequency, this is called oversampling.