1. The Analog World

The physical world is fluid, continuous, and unbroken. If you measure the temperature of a room, it does not jump instantly from 22°C to 23°C; it transitions through an infinite number of intermediate states. The pressure of sound waves hitting a microphone, the changing angle of a steering wheel potentiometer, the terminal voltage of a discharging lithium-ion battery, and the intensity of morning light are all continuous. They exist as physical quantities that can take any value within a range, changing smoothly across time.

Continuous physical waveforms vs discrete computer states

2. Why Computers Cannot Directly Understand Analog Signals

A digital processor, by contrast, is a machine of absolute division. Inside its silicon core, electricity is gated into binary states: a transistor is either fully conducting or cut off, representing a logical 1 or 0. A processor cannot store 'a slightly warm voltage' or 'a loud sound wave' in its registers. It understands only precise digital numbers. To bridge this gap between continuous physical reality and discrete computation, the system needs an interface that translates raw, infinitely variable analog signals into finite, structured digital words. This interface is the Analog-to-Digital Converter (ADC).

3. Sampling: Freezing Time

The first step of this translation is Sampling. Since a computer cannot observe the world constantly, it must observe it at discrete moments. Sampling is the process of reading the analog voltage of a signal at precise, regular intervals.

This is where we connect back to the Hardware Timer. A timer serves as the clock heartbeat of the system, asserting trigger events to dictate exactly when the ADC should observe. By sampling at a fast enough rate, the machine captures enough snapshots of the signal to represent its behavior across time.

Sampling timing trace showing periodic capture ticks

4. Quantization: Slicing the Continuous

Freezing time via sampling is only half the battle. At each sampling moment, the voltage level is still a continuous value with infinite decimal possibilities. A computer has a finite number of bits to represent this value, so it must map this infinite voltage to the nearest level on a pre-defined digital staircase. This mapping process is called Quantization.

Quantization discards the infinite sub-millivolt details, rounding the physical voltage to the closest discrete step. The tiny difference between the true analog voltage and the rounded digital level is called Quantization Error.

Continuous signal mapped onto discrete binary quantization levels

5. Resolution: The Fineness of the Mesh

How close can our digital staircase approximate reality? That depends on the ADC's Resolution. Resolution refers to the number of binary bits the converter uses to represent the signal, which dictates the number of steps on our measurement staircase.

- A 3-bit ADC has only $2^3 = 8$ steps. The staircase is coarse, jagged, and introduces substantial quantization noise. - An 8-bit ADC provides $2^8 = 256$ steps, which is sufficient for basic sensing but still relatively coarse. - A 12-bit ADC provides $2^{12} = 4096$ steps, offering a fine mesh that captures small changes with minimal noise. - A 16-bit ADC offers $2^{16} = 65,536$ steps, resolving microvolt fluctuations for high-fidelity audio or medical instrumentation.

By increasing resolution, we make the grid mesh finer, allowing the machine to capture a closer approximation of the physical wave.

Comparison of grid density for 3-bit vs 8-bit resolutions

6. Reference Voltage: The Calibration Ruler

To assign digital numbers to analog voltages, the ADC needs a ruler. This ruler is the Reference Voltage ($V_{REF}$). The reference voltage defines the maximum physical voltage the ADC can measure, which corresponds to the maximum digital count.

If we have a 12-bit ADC (0 to 4095) with a $V_{REF}$ of 3.3V: - An input of 0.0V resolves to 0. - An input of 3.3V resolves to 4095. - An input of 1.65V resolves to exactly 2048.

If the reference voltage fluctuates, our measurements fluctuate too. If $V_{REF}$ drops to 3.0V due to poor power supply regulation, an input of 1.5V will resolve to 2048 instead of 1861, causing a measurement error. Precise calibration of $V_{REF}$ is the cornerstone of accurate physical observation.

Reference voltage acting as a calibrated measuring scale

7. The Conversion Pipeline: Sample, Hold, Convert, Store

How does the physical conversion happen? Most microcontrollers use a Successive Approximation Register (SAR) ADC. The hardware processes each sample through a four-stage pipeline:

1. Sample: A physical switch closes briefly, connecting the external pin to an internal capacitor. 2. Hold: The switch opens. The capacitor 'holds' the captured charge steady so the voltage doesn't change during conversion. 3. Convert: A comparator compares the held voltage to a series of voltages generated by an internal DAC. Using a binary search (Successive Approximation), it tests the most significant bit first, deciding whether the signal is above or below half of $V_{REF}$, and repeats for each bit down to the LSB. 4. Store: Once all bits are decided, the binary result is copied to a data register, raising an interrupt or DMA request so the CPU can read it.

SAR ADC hardware blocks: Sample switch, Hold capacitor, Comparator, and SAR register

8. Accuracy vs Precision

In embedded sensing, engineers often confuse two critical terms: Accuracy and Precision. They are not the same.

- Accuracy is how close a measurement is to the true physical value. An accurate system has minimal calibration offset. - Precision is how consistent and repeatable the measurements are when the same input is read multiple times. A precise system has low noise.

A system can be highly precise but inaccurate (giving highly repeatable, low-noise readings that are calibrated incorrectly) or highly accurate but imprecise (averaging to the correct value but showing substantial noise on each individual sample). Below is a visual representation of these states:

9. Capturing Reality

Let's observe these conversion behaviors. In the interactive panel below, modify signal frequency, resolution, and sampling rate to watch how an analog signal is sliced and rounded into digital numbers.

10. The Cost of Observation

What happens when our observation parameters are configured incorrectly? In the simulator below, adjust the reference voltage, resolution, and sample rate to induce signal clipping, aliasing (undersampling), or high quantization stepping.

11. Real Embedded Applications

ADCs are the sensory organs of embedded silicon. They monitor the health of battery cells in electric vehicles, translate analog temperature sensors (like thermistors or RTDs) into precise degrees, decode phase currents for brushless DC motor control, digitize microphone voice data in smart assistants, and capture cardiac electrical potentials in medical ECG monitors. Without the ADC, a processor is deaf, blind, and isolated from physical reality.

The Sensory Handshake

Timers taught machines when to pay attention. ADCs taught machines how to translate reality into something computation could understand.

Yet observing reality is only half the story. Once a machine has observed its environment and decided on a course of action, it must eventually speak back—driving physical outputs and painting analog actions onto the physical world.