Introduction to Analog-to-Digital Processing Using a Microcontroller (Part 1)

Many modern embedded systems convert analog signals to digital, serving as a link between the analog world and the digital algorithm that evaluates or processes them. Sensors, measurement circuits, control systems, or device users generate analog signals that must be converted into digital form so that a microcontroller or a larger digital system can "understand," analyze, and process them. In a series of three articles, starting with this text, we will take a closer look at this topic.

An analog signal changes continuously over time. This means that its value can take any level within a given time interval and smoothly changes between levels. Such a signal is encountered when observing natural phenomena, such as sunlight, sound, or temperature. At the same time, it is susceptible to disturbances because any related external phenomena directly affect its value.

A digital signal represents information as discrete levels over time. Most digital systems have two levels, logical zero and one. A digital signal changes abruptly rather than smoothly, making it much more resistant to interference. In practice, this means the analog signal is more natural but harder to transmit and process stably, while the digital signal is simplified but thus more reliable, easy to memorize, transmit, and process by digital systems.

Processing "analog" to "digital"

The component providing the transition from the analog to the digital domain is the analog-to-digital converter (ADC, A/C). It plays a key role in modern electronic circuits by converting analog signals into digital data. Using the ADC, continuous analog signals in time and amplitude are converted into discrete time digital signals. The conversion process involves three main stages: sampling, quantization, and digital encoding.

Analog-to-Digital Converters

During sampling, the analog signal is measured (sampled) at fixed time intervals to obtain a sequence of instantaneous voltage amplitude values. The sampling period defines the sampling frequency, i.e., how often the signal is measured and processed. In most applications, the sampling frequency is constant, but—usually to save memory—in some applications, the sampling frequency can be variable. The Nyquist-Shannon sampling theorem states that to effectively process a signal and prevent aliasing, the sampling frequency must be at least twice the highest frequency component of the input signal. In practice, this minimum sampling period is shortened even 5- or 10-fold.

The next processing stage is quantization, during which the continuous amplitude of the sampled signal is divided into a series of discrete levels. The number of quantization levels is determined by the bit resolution of the ADC. An n-bit ADC has 2ⁿ quantization levels. For example, a 3-bit ADC has 8 (2³) different levels. During quantization, a discrepancy arises between the actual analog value and the approximate value resulting from the ADC bit resolution—this is called quantization error. This error is also described by the term "quantization noise."

The final stage of processing is digital encoding. In this stage, each quantization level is assigned a distinct binary code. As a result of encoding, the analog signal is converted into a numeric representation that can be processed by digital systems.

Important ADC parameters and phenomena during processing

Key ADC parameters determine how faithfully it converts an analog signal to digital form, with each describing a different aspect of measurement quality.

• Sampling frequency (period) defines how often the ADC takes signal samples and practically determines if the converter can capture all significant changes in the signal without aliasing. The relationship is intuitive – the faster the sampling, the better the representation of higher frequency signals.
• ENOB (Effective Number of Bits) describes the real resolution of the converter considering noise and nonlinearity. Often, ENOB is lower than the theoretical resolution given in the documentation, indicating the number of useful informational bits obtainable.
• INL and DNL parameters relate to converter nonlinearities: DNL shows how much the intervals between successive quantization levels differ from the ideal, whereas INL describes the deviation of the entire ADC characteristic from an ideal straight line. Large errors in these parameters cause distortions and uneven signal representation.
• The signal-to-noise ratio, or SNR, indicates how much of the signal is "clean" versus disturbed. The higher the SNR (usually given in dB), the more stable and repeatable the measurements.
• The THD parameter informs about the harmonic content additionally introduced by the ADC path into the signal, which is especially important in audio applications and measurements requiring high precision.
• The final important parameter is the sample acquisition time, the time needed to charge the sample-and-hold capacitor at the converter's input to the level corresponding to the current signal value. If this time is too short relative to source impedance, the sample will be distorted even if all other ADC parameters are ideal.

Together, these parameters create a full picture of the converter's capabilities and determine how accurately the embedded system can measure the real signal. It is also worth noting that while using an ADC embedded in a microcontroller is inexpensive and convenient, external ADC circuits usually have better parameters. The main reason is the lack of noise generated by the synchronously operating core and peripheral circuits directly near the converter structure.

Analog signal conditioning

Device designers often focus on digital processing, forgetting that processing quality largely depends on the whole analog input stage, which performs three functions.

First, the most important task of the analog path is to protect the device and its user from external disturbances or failure effects. Many good practices are used to secure device inputs mainly against discharges or user errors. This includes using protection diodes (including transient voltage suppressors), series resistors limiting input current, resistive dividers, amplifiers, buffer followers, and many others. The absolute principle is that the usually expensive microcontroller or microprocessor should be the last to be damaged after all protective measures fail.

Second, the analog path should adjust the measured signal level to the converter's capabilities, i.e., limit or amplify it as needed. Sometimes it is also necessary to introduce a DC offset because the signal oscillates symmetrically around the reference level (ground), but measurement is possible only in the range e.g., 0…3.3V. Energy matching is also very important—the source cannot be overloaded to prevent degradation of its parameters. A high input impedance and low output impedance amplifier stabilizes signal parameters and does not load the source.

Third, the analog path should limit the bandwidth of transmitted signals to eliminate components with frequencies above the safe Nyquist sampling frequency, thus preventing aliasing.

All these functions should be executed maintaining linearity (uniformity) of the signal transfer characteristic in the required frequency range, without a tendency to oscillate, preserving the proper phase and amplitude of the signal, appropriate impulse response, and available supply voltage. Proper design and implementation of the analog path and later correct ADC configuration largely determine conversion quality because these two elements form a single chain, where the weakest link immediately reduces overall quality. Before entering the converter, the signal must be clean, stable, and matched to its parameters. If the analog path introduces noise, distortions, or has too high output impedance, the ADC cannot properly charge the sample-and-hold capacitor, leading to distorted samples. On the other hand, improper converter configuration—such as too short sampling time, poorly chosen voltage reference, or incorrect clock frequency—causes even a perfectly prepared signal to be read with error.

In practice, the analog path ensures the signal is physically correct and stable, while the ADC must be configured to faithfully represent this signal digitally. If either element fails, problems such as measurement drift, high variance in readings, aliasing, nonlinearities, or disturbances arise, which are later difficult or even impossible to remove programmatically.

ADC in a microcontroller

Pipeline ADCs are used only in very fast, standalone ADC circuits, not in MCUs. Such converters are designed for very high sampling frequencies and require complex, power-consuming analog blocks, which do not fit the energy constraints of microcontrollers. In practice, manufacturers of modern microcontrollers embed only SAR or Sigma-Delta converters.

The SAR converter works on successive approximation, where successive bits of the result are determined by comparing the input voltage with values generated by an internal DAC. This provides fast conversion with moderate resolution, so SARs are commonly used in measurement systems, control, and low power applications.

The Sigma-Delta converter uses oversampling and pulse-density modulation followed by digital filtering to achieve very high resolution and excellent signal-to-noise ratio. It is slower than SAR but offers the highest measurement precision and thus dominates in precise audio systems, electronic scales, and measuring instruments.

Comparing these architectures, SAR offers a compromise between speed and accuracy, Sigma-Delta provides the highest precision at the expense of speed.

The internal ADC path in a microcontroller consists of several cooperating blocks that may differ functionally depending on the manufacturer, microcontroller family, ADC type, and controlling core. Usually, an analog multiplexer is placed at the input, allowing switching between different signal sources, such as analog pins, internal sensors, or voltage reference sources. Behind it works the sample-and-hold circuit, which "freezes" the voltage value briefly so the converter can process it stably. The key element is the ADC itself, responsible for analog-to-digital conversion. The whole operates with a precise reference voltage source that defines the measurement range and affects accuracy. The path also contains input amplifiers (often with programmable gain) and circuits compensating for interference and offset. Conversion results go to registers or DMA buffers, from where they can be read by software.

ADC paths differ among microcontroller families primarily in architecture, accuracy, and sophistication of accompanying circuits. In ARM core microcontrollers, the analog path is usually the most developed. It includes fast high-resolution SAR converters, often with multiple parallel channels, advanced sample-and-hold circuits, programmable input amplifiers, and an extensive hardware trigger and DMA system. Thus, ARM microcontrollers perform well both in precise measurements and high-speed data acquisition.

The PIC family offers a wide variety, ranging from simple 8-bit devices to more advanced 16- and 32-bit ones. Basic PICs have a simple ADC architecture and moderate accuracy. More advanced PIC24 or dsPIC have much more complex and technologically advanced analog paths, with faster SAR converters, better input amplifiers, filtering, and the ability to work in automatic sampling and conversion modes. These higher PIC series are suitable for signal processing and motor control applications requiring fast and repeatable measurements. In AVR microcontrollers, the ADC path is simpler and geared towards energy efficiency and basic functionality. Their converters are generally slower and lower resolution, and the input circuit less advanced, with a simple multiplexer and basic sample-and-hold. AVR performs well where measurements need not be very fast and power consumption is a priority.

Reference voltage source

ADCs require a stable, precise reference voltage source. It can be built into the chip or supplied externally.

An internal reference voltage source is convenient to use since it does not occupy space on the board and requires no additional components. It also guarantees full integration with the microcontroller, simplifying design and reducing susceptibility to disturbances from external signals. However, its stability and accuracy are limited by the chip's technology process, so the reference voltage may have larger tolerance, poorer temperature coefficient, and higher noise than external solutions. As a result, measurements are sufficient for most typical applications but may be inadequate where high precision is required.

An external reference source (usually in the form of a specialized IC) allows much better stability, lower temperature drift, and lower noise levels, translating into greater repeatability and conversion accuracy. It also enables selection of the reference voltage to a specific measurement range, improving ADC resolution utilization. However, it requires an additional IC, proper trace design, filtering, and careful separation from disturbances, increasing design complexity and potentially raising costs. In practice, the choice between internal and external reference depends on whether simplicity and low power consumption or maximum measurement accuracy is prioritized.

Examples of microcontrollers with ADCs allowing external reference voltage include ATmega328P and ATmega32, produced by Microchip (formerly Atmel). In these microcontrollers, the reference voltage can be applied directly to the AREF pin or supplied from the internal source. Similar capabilities exist in STM32 microcontrollers from STMicroelectronics, e.g., STM32F103, which have separate VREF+ and VREF- pins for external ADC reference. Similarly, in the Microchip PIC family, e.g., PIC16F877A, VREF+ and VREF- pins allow using external reference voltage. The Analog Devices portfolio includes microcontrollers from families ADuC and ADuCM, such as ADuC7020 and ADuC7026 with built-in 12-bit ADCs and VREF+/VREF- pins, and ADuCM360 and ADuCM361 equipped with precise 24-bit Sigma-Delta converters and separate REF+/REF- pins, making them especially useful in measurement applications.

Read part 2 of the article

Read part 3 of the article

Transfer Multisort Elektronik (TME) is one of the world’s largest global distributors of electronic components, electrotechnical parts, workshop equipment, and industrial automation. The catalog includes over 1,500,000 products from 1,300 leading manufacturers. TME’s modern logistics centers in Łódź and Rzgów (Poland), with a combined area of over 40,000 m², ship nearly 6,000 packages daily to customers in more than 150 countries.

TME also invests in the development of knowledge and skills of young engineers and electronics enthusiasts through the TME Education project, and supports the tech community by organizing the TechMasterEvent series, promoting innovation and experience exchange.