I/O Devices and Interfaces

An input device converts a user action or physical quantity into digital data. An output device converts digital results into a form usable by a person or another physical system. An I/O interface connects such a peripheral to the processor and memory system.

Classes of I/O Device

Device	Input supplied
Keyboard	Text, numeric data and commands
Mouse or touchpad	Pointing position, motion and button actions
Scanner	Printed text or images converted to digital form
Microphone	Sampled audio signal
Camera or webcam	Still-image and video data
Barcode/QR reader	Encoded product, ticket or identity data
Sensor	A measured quantity such as temperature, pressure or light

Input-device examples.

Device	Output produced
Monitor	Visual text, graphics and video
Printer	Permanent hard copy
Speaker	Audible sound
Plotter	Large engineering drawings, plans or maps
Actuator	Physical motion or control in an automation system

Output-device examples.

Devices may also be classified as storage devices (SSD or disk), communication devices (network interface or modem), and control/sensing devices. These classes can overlap: a storage device both accepts output writes and returns input reads.

Functions of an I/O Interface

Peripherals differ greatly from a CPU in speed, word format, voltage and protocol. The interface therefore provides:

• Address decoding: selects the intended device and one of its data, status or control registers.

• Buffering and speed matching: holds data while a fast bus and a slow peripheral operate at different rates.

• Format conversion: performs serial/parallel, word-width, coding or electrical-level conversion where required.

• Control and status: starts an operation and reports ready, busy, completion or error conditions.

• Synchronization: uses clocking or handshakes so both ends know when data are valid.

The processor commonly sees a data register, a status register and a control register. In memory-mapped I/O these registers occupy normal memory addresses; in isolated or port-mapped I/O they occupy a separate I/O address space and use special instructions or control signals.

Programmed, Interrupt-Driven and DMA I/O

Method	Operation	Main advantage	Main disadvantage
Programmed I/O	CPU issues a command and repeatedly polls the status register until ready, then transfers each item	Very simple hardware and software flow	Busy waiting wastes CPU time
Interrupt-driven I/O	CPU starts the device and does other work; the device interrupts when ready, and an ISR transfers data or records completion	Better CPU utilization for intermittent events	Context-switch and per-interrupt overhead
DMA	A DMA controller becomes bus master and transfers a programmed block directly between I/O and memory	High block throughput with little CPU work per item	Additional hardware, arbitration and cache-consistency concerns

Principal I/O transfer methods.

For programmed I/O, the CPU executes a loop such as “read status; test ready; repeat; read data.” It is suitable for a simple device or a very short operation when waiting costs less than interrupt setup.

For interrupt-driven I/O, the CPU enables the device interrupt and continues another task. On readiness, the interrupt-service routine saves context, identifies/services the device, transfers data or acknowledges the event, clears the request and returns. Excessively frequent interrupts can still consume substantial CPU time.

For direct memory access (DMA), the CPU initializes source or destination address, transfer count, direction and device channel. The DMA controller requests the bus, moves data between the interface and RAM, updates address/count, and interrupts the CPU once the block completes or an error occurs. Cycle stealing takes occasional bus cycles; burst mode holds the bus for several transfers and gives higher throughput but may delay the CPU.

Parallel and Serial Interfaces

A parallel interface carries several bits at once on separate data conductors. It can transfer a word in one interval on a short link, but uses many pins and suffers inter-wire timing skew at high speed or long distance. A serial interface sends ordered bits over one data path (or a small number of lanes). It uses fewer pins and is easier to route over distance, but one lane needs several bit intervals for a word. High symbol rates and multiple lanes allow modern serial links such as PCIe, USB and SATA to exceed older parallel buses.

Parallel and serial timing for the word $10110100$, followed by straight-through RS-232 data, handshake and ground connections.

In the figure, the raw eight-bit word occupies one parallel transfer interval but eight serial bit intervals:

$$T_{word,serial}=8T_{bit}=\frac{8}{R_b},$$

where $R_b$ is bit rate. Real asynchronous serial characters also include framing bits, so their wire time is longer than this raw-data comparison.

Note

Example — Serial framing time

At $9600\,\text{bit/s}$, eight raw data bits take

$$\frac{8}{9600}=0.833\,\text{ms}.$$

With common 8N1 framing (one start, eight data, no parity, one stop), each character uses 10 bits and takes

$$\frac{10}{9600}=\boxed{1.042\,\text{ms}},$$

giving a maximum of about 960 characters/s before higher-level overhead.

RS-232 Interface

RS-232 is a single-ended serial electrical and signaling standard originally defined between data terminal equipment (DTE), such as a terminal or PC, and data communication equipment (DCE), such as a modem. It commonly carries asynchronous characters consisting of idle, start, data, optional parity and stop intervals.

Signal	Pin	Direction at DTE	Function
TXD		Output	Transmitted serial data from DTE to DCE
RXD		Input	Received serial data from DCE to DTE
RTS		Output	Request to send; hardware flow-control request
CTS		Input	Clear to send; DCE permission to transmit
GND		Reference	Common signal reference

Selected DB-9 RS-232 signals, named from the DTE viewpoint.

A straight-through DTE–DCE cable connects corresponding DB-9 pins ($2$–$2$, $3$–$3$, $7$–$7$, $8$–$8$, $5$–$5$). A DTE–DTE null-modem connection instead crosses data and, when used, handshake signals so that each transmitter reaches the other receiver.

RS-232 uses inverted bipolar levels rather than TTL logic: a mark or logic 1 is negative (typically $-3$ to $-15\,\text{V}$), while a space or logic 0 is positive (typically $+3$ to $+15\,\text{V}$). The region between $-3\,\text{V}$ and $+3\,\text{V}$ is undefined at the receiver. A level translator is therefore required between an RS-232 connector and ordinary low-voltage digital logic. The standard defines the interface, not the meaning of the application data carried within each character.

Quick Review

• Programmed I/O: CPU polls and transfers; simplest but wastes cycles while waiting.

• Interrupt I/O: device signals readiness, improving CPU utilization at the cost of ISR overhead.

• DMA: controller transfers blocks directly between I/O and memory and interrupts mainly on completion.

• Parallel versus serial: parallel spends conductors to move bits together; serial spends bit intervals and uses fewer wires.

• RS-232: bipolar, inverted, single-ended DTE–DCE serial interface with TXD, RXD and optional hardware handshaking.

Tip

Exam Focus — I/O answer

Define the interface before comparing transfer methods. For DMA, mention CPU initialization, bus ownership, direct memory transfer and completion interrupt. For RS-232, identify DTE/DCE, directions and pins, straight versus null-modem wiring, and the inverted bipolar voltage convention.