Is a 10-bit ADC Sufficient for Microcontrollers?

1. Introduction

Currently, most microcontrollers come with built-in AD converters that are typically between 10 to 12 bits. In general applications, how much does the bit depth of AD conversion affect measurement errors? Is a 12-bit ADC sufficient?

We will discuss the impact of AD conversion bit depth on measurement error through practical measurements.

2. Circuit Design

Design a test circuit using the F030 microcontroller for testing. The input analog signal is sent to the ADC’s channel 0 after passing through an RC filter. Two resistors are used to separate the analog ground and power supply. A 3.3V voltage regulator chip is also set up on the circuit board. The PCB is designed to be compatible with a one-minute board-making method.

AD\Test\2024\April\TestADCBit.PcbDoc

After one minute, the test circuit board is manufactured perfectly.

3. Measurement Results

Next, we will solder the circuit board, clean it, and proceed with measurements.

First, use a probe clip to download the test program. The test program is downloaded, and the measurement data is uploaded to the computer through the serial port in the ST-LINK. Using Python programming, we utilize the DH1766 to supply voltages from 0.1 to 3.2V, and the actual voltage values are measured using the DM3068. The microcontroller averages the 512 collected data points. Before averaging, we take the first 1 to 12 bits separately, allowing us to obtain ADC sampling values at different resolutions.

Within the measured voltage range, a total of 2000 data points were collected, and the relationship between voltage and ADC was plotted. It can be seen that when the ADC sampling bit depth is relatively low, the errors caused by quantization are more significant. Although this is the average of 512 data points, the errors primarily stem from the quantization of the ADC. However, as the bit depth increases, fluctuations caused by noise also become evident in the data. At this point, the errors caused by noise exceed those brought about by the conversion bit depth.

By performing linear fitting between the microcontroller’s AD data and the voltage, we calculate the variance of the error between the fitted voltage and the actual voltage. It can be observed that as the AD conversion bit depth varies from 1 to 12, the variance decreases sharply. However, the rate of decrease slows down significantly, and when the bit depth exceeds 10 bits, the error remains relatively constant. This is the data line plotted after taking the logarithm of the variance. It becomes clearer that once the AD conversion bit depth exceeds 10, the errors in the obtained data primarily stem from circuit noise. Therefore, for data conversion, higher ADC bit depths become ineffective.

4. Conclusion