A/D converters are essential when dealing with analog signals. When using the A/D converter to quantize the analog signal, that is, when digitizing, if the signal contains frequency components higher than 1/2 of the frequency used, then as shown in Figure 1.3, completely different frequency components will be generated. Thus, a quantization error occurs. This phenomenon is called the confounding effect
(aliasing effect), the device to prevent this phenomenon has an anti-aliasing filter (anti-aliasing effect), which is composed of LPF .
Figure 1.3(a) is the case when a 95kHz sine wave is sampled (A/D converted) at 100kHz intervals. Connect each sampling point, there will be a 5kHz sine wave with a half cycle of 100us. This 5kHz waveform is based on the frequency component generated by the aliasing error.
Figure 1.3(b) shows this situation as a spectrum. If a 95kHz signal wave is sampled at 100kHz, a spectrum will be generated at both ends of the sampling frequency as the center and the deviation distance as the signal frequency. In addition, the spectrum also appears at both ends centered on an integer multiple of the sampling frequency, but this is not shown in Figure 1.3.
In order to prevent the generation of aliasing errors, LPF can be inserted as shown in Figure 1.3(d) so that the input signal component is lower than the system resolution in the frequency range above 1/2 of the adopted frequency. This LPF is to put the aliasing filter.
For example, when sampling at 100kHz with a 12-bit A/D converter, you should use an LPF that ensures an attenuation of 1/4096=-72dB at 50kHz.
Of course, if the known signal does not contain frequency components above 50kHz, there is no need to use LPF.
Recently, with the advancement of semiconductor technology, A/D converters have been able to achieve high-speed sampling frequencies relatively easily. If the frequency bandwidth of the signal is the same, the higher the sampling frequency, the easier it is to improve the slope of the filter and reduce the burden on the filter. This is called over sampling.