Digital Signal Processing (DSP) is the processing of analog signals to digitize and the processing of digitized signals.
For example, as shown below, when the value of a sensor (such as a thermistor) is input to the control unit and some processing is performed,
Analog-to-digital conversion is performed on the sensor value divided by the pull-up resistor, etc. and input to the CPU.
At this time, the CPU recognizes the sensor signal, which is originally a continuous value, as a discrete value.
■Resolution, sampling period
Resolution and sampling period are important for digitizing analog signals with high accuracy.
＜Resolution＞
Resolution is the ability to express how many digits a value can be recognized in detail.It is determined by the performance of the AD converter and CPU, and the ability is expressed by the number of bits.
For example, suppose you have an AD converter and CPU with 8-bit performance, and the voltage range that the sensor can take is 0-5V.
How accurate can the sensor value be detected at this time? First of all, 8 bits in binary can be expressed in 2 to the power of 8 = 256 steps.
In other words, the range of 0-5V can be expressed in 256 steps, so the minimum possible value is 5/256=0.0195313V.
This minimum possible value is called LSB (Least Significant Bit). Also, the error that occurs when converting an analog signal to a digital signal is called a quantization error.
＜sampling period＞
The sampling period means how long the sensor value is input in time, and the shorter the sampling period, the better the accuracy.
The sampling period is represented by time or frequency (for example, if the sampling period is 0.1 second interval, the frequency is 10Hz).
The sampling period depends on the clock frequency of the CPU, but it is rare in actual use that the sampling time must be shorter than the clock frequency.
The limitation of the sampling period is the CPU processing load when reading many sensors at the same time or running many programs.
Slow down the sampling period when the processing load is over.
As for how long the sampling period can be extended, there is a theorem that the original signal can be restored if it is greater than twice the frequency of the signal.
This is called the sampling theorem.
TOP MENU
