Image Sensors Explained: CCD vs CMOS | Computer Vision Basics

Semiconductors are materials that are neither fully conductors nor fully insulators, possessing unique controllable conductivity properties. Materials are classified by resistivity: conductors (~10^-8 ohm-meters), semiconductors (~1 ohm-meter), and insulators (~10^19 ohm-meters). Semiconductor materials include elemental types (silicon, germanium with four valence electrons), compound semiconductors (inorganic like gallium arsenide, organic like anthracene), and organic polymers (discovered post-1990). According to Bohr's atomic theory, electrons occupy specific energy levels around the nucleus. When atoms form crystals, interatomic forces cause their individual energy levels to split and broaden into continuous energy bands. For n atoms with four valence electrons each, there are 4n valence electrons total distributed across 8n possible energy levels (4n in the lower band and 4n in the upper band). The energy gap between these bands determines the material's electrical properties and classifies it as conductor, semiconductor, or insulator.

Semiconductor physics focuses on three central questions: (1) determining the number of charge carriers (electrons and holes) in a semiconductor, (2) understanding their spatial distribution and movement mechanisms (drift and diffusion), and (3) learning how to manipulate these properties to create useful electronic devices like MOSFETs, diodes, and BJTs; the study employs quantum mechanics, statistical mechanics, density of states calculations, energy band diagrams, effective mass concepts, and Fermi statistics to analyze carrier behavior and derive fundamental equations such as Ohm's law, the continuity equation, and ambipolar transport equation.

This lecture introduces fundamental concepts of semiconductor physics, covering thermal voltage (VT = KT/Q, approximately 26mV at room temperature), ionization potential (energy to remove valence electrons), and crystalline structure of materials. The instructor explains the classification of materials based on energy band diagrams: insulators have large energy gaps (>5eV) with completely filled valence bands and empty conduction bands, preventing current flow; semiconductors have small energy gaps (~1eV) where thermal energy breaks bonds, creating free electrons and holes (bipolar conduction), with silicon and germanium being primary examples; metals have overlapping valence and conduction bands with abundant free electrons (unipolar conduction). The energy gap in semiconductors decreases linearly with temperature, making them temperature-dependent conductors.

Under low-level injection assumption, the injected minority carrier current in an extrinsic semiconductor is essentially the diffusion current, as minority carriers move from regions of higher concentration to lower concentration due to illumination or external excitation.

Electrical conduction requires charge carriers: positive ions and electrons in gases, dissociated ions in electrolytes. Solids classify as insulators (zero conductivity), metals (excellent conductors), or semiconductors (intermediate conductivity increasing with temperature). Metals conduct through free electron gas where electrons move randomly among vibrating positive ion cores. At absolute zero, electrons move freely with zero resistance. As temperature increases, ion vibrations cause inelastic collisions, reducing electron mean free path. The drift velocity u = -μE relates to mobility μ. Current density J = nqμE combines carrier concentration, charge, mobility, and electric field. Conductivity σ = nqμ, leading to Ohm's law J = σE and resistance R = ρ(L/A). Silicon and germanium (group IV elements) form stable tetrahedral crystal lattices through covalent bonding, with energy gaps of 1.1 eV and 0.67 eV respectively. At absolute zero, semiconductors behave as insulators with no charge carriers.

Analog-to-digital conversion (ADC) is the process of converting continuous analog signals into discrete digital values. Digital signals have only two states: 0 (representing 0 volts) and 1 (representing 5 volts). Analog signals, however, can take on many different values within a range. The ADC on Arduino Uno has 8 bits of resolution, meaning it can represent 2^8 = 256 different values, ranging from 0 to 255. This conversion is essential for reading sensors that output continuous voltage values.

ADC converts continuous analog signals to discrete digital quantities through filtering, sampling, holding, and quantization. The Nyquist Sampling Theorem requires sampling at least twice the highest signal frequency. Quantization divides the signal range into discrete levels—more bits provide finer resolution. Flash ADC uses multiple voltage comparators connected to a priority encoder with a resistor divider network. All comparisons occur simultaneously, making it the fastest method but requiring many comparators for high resolution. This makes flash ADC ideal for applications needing immediate signal conversion without delay, though it is hardware-intensive.

An Analog-to-Digital Converter (ADC) is an electronic device that converts continuous analog signals (such as voltage from sensors measuring temperature, light, or pressure) into discrete digital values that can be processed by microcontrollers or computers; the conversion process involves sampling the analog signal at regular intervals and quantizing it into a finite number of digital levels based on the converter's resolution (measured in bits), where higher bit counts provide finer granularity and better representation of the original analog signal.

An Analog-to-Digital Converter (ADC) is the device responsible for creating the stream of digital samples from an analog audio signal. It takes the voltage coming from microphones and converts it into a stream of discrete numerical values. Inside devices like the Tascam DR-05, the ADC continuously measures the microphone voltage and writes down precise numerical readings at regular intervals determined by the sample rate.

Analog-to-digital converters (ADCs) convert continuous analog signals into discrete digital values. The simplest ADC is a comparator that outputs logic 1 when input equals reference voltage, otherwise 0. Key ADC specifications include resolution (6-18 bits depending on speed), accuracy, conversion speed, linearity, and monotonicity. Selection depends on required precision and application needs.

Digital images contain three-dimensional information despite appearing two-dimensional. Geometric resolution is determined by total pixel count (columns × rows), while radiometric resolution represents pixel intensity values (typically 8-bit with 256 scales). Color images use three RGB channels producing 16.7 million colors. Casual photography introduces geometric distortion (mencong/skew), which can be prevented using flatbed scanning. Geometric corrections using casual tools like Photoshop may introduce new errors, as demonstrated by the UGM logo distortion example.

Images are composed of pixels, which are square points that make up pictures. When you zoom in on an image, you eventually reach a point where no additional detail can be revealed because the image contains a finite amount of information (bits). This is why simply enlarging an image in movies doesn't create new information. To create color images, computers need more than just one bit per pixel. Since a single bit can only represent two values (0 or 1), multiple bits are required to represent different colors. JPEG files have a specific signature where the first three bytes are always 255, 216, and 255 (in decimal), which is a sequence of 24 bits that computers use to recognize JPEG files. Computers use binary (0s and 1s) to represent all data, but hexadecimal (base-16) provides a more compact representation. Hexadecimal uses digits 0-9 and letters A-F (representing 10-15), where each digit represents four binary bits. In C programming, structures (structs) allow programmers to create custom data types that combine multiple variables into a single unit. Macros in C are defined using the #define directive and represent constants or code snippets that are replaced by the preprocessor before compilation. C provides functions for reading from and writing to files, enabling programs to persist information beyond execution. The fopen() function opens files with different modes: 'r' for reading, 'w' for writing (which overwrites existing files), and 'a' for appending (adding to existing files). The fprintf() function formats and writes data to files. The fclose() function is used to properly close files after operations. The malloc() function allocates memory dynamically at runtime, requiring length + 1 bytes for strings to account for the null terminator. String comparison in C is case-sensitive, meaning 'mom' and 'Mom' are considered different strings. The strcmp() function compares two strings character by character and returns 0 when strings are identical, and a non-zero value when they differ.

Digital computers represent continuous analog images through two discretization processes: (1) Spatial discretization - sampling the image on a regular grid to convert continuous spatial coordinates into discrete pixel locations; (2) Intensity discretization (quantization) - converting continuous intensity values into discrete numerical representations. Without discretization, the infinite points and continuous intensity values of analog images cannot be stored or processed digitally.

This segment covers the foundational concepts of digital imaging. A pixel is the smallest unit of any digital image, and more pixels mean more information and clarity. Higher megapixels generally indicate better quality, though this isn't always true. Digital zoom works by extracting megapixels from a specific area, but it cannot improve image quality because the original pixel count remains unchanged. This explains why digital zoom cannot make images clearer - it simply enlarges existing pixels without adding new information.

Digital images are composed of pixels—small individual locations with specific color or grayscale intensity values. Images are stored in memory as contiguous blocks of data, typically starting with a header containing format information, dimensions, and metadata. The pixel data follows as a long list, allowing direct access for processing. Pixel indexing uses a stride value (row width) with the formula: pixel_position = y * stride + x. RGB images use three color channels (Red, Green, Blue) corresponding to human color perception, while grayscale images use a single channel for memory efficiency. Understanding these fundamentals enables image manipulation, filtering, and processing operations.