Analog to Digital Conversion of Time? Learn about Sampling!

When digitizing analog data that changes continuously over time, a technique called "sampling" is used, in which the data is divided into regular intervals (sampling periods) and converted into digital data. (The vertical voltage axis will be explained in the next column on quantization.)

An example where the sampling period is 1/8 of the analog data period

Although the data is slightly off, the shape of the original waveform is clearly visible.
The shorter the sampling period, the more accurately the original analog waveform is reproduced, resulting in improved accuracy.
However, the shorter the sampling period, the greater the amount of data per unit time, which means that the conversion speed of the AD converter itself and the communication speed of the digitized signal must also be increased, leading to increased costs.
Now let's see how far we can lower the sampling frequency.

Example where the sampling period is 3/8 of the analog data period

The error has become quite large, but we can see that it is fluctuating at the same frequency as the original data.
Let's try extending the sampling period further.

An example where the sampling period is 1/2 of the analog data period

This means that it is no longer possible to know how the original analog data was changed.
Let's try extending the sampling period further.

An example where the sampling period is 3/4 of the analog data period

It looks like a clean waveform, but it appears to be changing at a cycle that is 1/3 of the original cycle.
This phenomenon of waveforms being observed with incorrect periods is called aliasing.

If the sampling period is longer than half the period of the measurement target, the original waveform cannot be reproduced and erroneous data may be obtained.
The parameter for the sampling speed of an AD converter is usually expressed as the frequency, which is the reciprocal of the period, as follows:

Minimum required sampling frequency of AD converter > measurement target frequency x 2

This principle is called the Nyquist theorem, and the minimum required sampling frequency is called the Nyquist frequency.

The sampling frequency of an AD converter is usually given in Ksps or Msps on datasheets.
The initial K or M is the unit of size, the second letter s is the number of samples, p is per, so it stands for /, and the fourth letter s is second, so it stands for seconds.
5Msps is 5 mega samples per second, or 5 million samples per second.
20Ksps means 20 kilo*samples per second, which means 20,000 samples per second.