Read the original article here.

---

Understanding the Computer Bus: The Communication Highway

In the journey of building a computer from scratch, understanding how different components talk to each other is fundamental. Unlike simple circuits where connections might be point-to-point, complex systems with multiple devices require a standardized way for information to travel efficiently. This is where the concept of a bus becomes crucial. Think of it as the communication highway of your computer system.

Definition: Bus (Computing)

In computer architecture, a bus is a communication system that transfers data between components inside a computer or between computers. It encompasses both the physical hardware (like wires or traces on a circuit board) and the software or communication protocols that manage the data transfer. A bus acts as a shared pathway allowing multiple devices to communicate, utilizing protocols to prevent conflicts and ensure orderly data exchange.

Historically, buses were literally bundles of wires. The term itself comes from the electrical power "busbar," which is a conductor used to distribute electrical power to multiple connections within a switchboard or system. Just as a busbar distributes power, a computer bus distributes data and control signals.

Understanding bus architecture is vital because it dictates:

• How different parts (CPU, memory, peripherals) connect.
• How much data can be transferred at once (bandwidth).
• How fast data can travel.
• How devices are addressed and controlled.
• The complexity of adding new components.

The Three Pillars of a Traditional Bus

Most historical and many fundamental modern buses are conceptually divided into three main functional groups, often (but not always) corresponding to separate sets of wires or traces:

1. The Address Bus: Specifies the location (memory address or I/O device address) that the CPU or another device wants to access.
2. The Data Bus: Carries the actual data being transferred between components.
3. The Control Bus: Manages the timing and control signals necessary for the transfer process (e.g., read/write signals, timing clocks, interrupt requests, bus grant signals).

Let's look at each of these in more detail.

The Address Bus: Pinpointing Destinations

When the CPU needs to fetch an instruction from memory, read data from RAM, or send data to a peripheral device (like a graphics card or network chip), it first needs to tell the system where it wants to communicate. This is the job of the address bus.

Definition: Address Bus

An address bus is a bus used by the CPU or a DMA-enabled device to specify the physical address of a memory location or an I/O device register with which it intends to communicate (read from or write to).

Function: The sender (typically the CPU or a DMA controller) places the desired address onto the address bus lines. The memory controller or the addressed peripheral device "listens" to the address bus and responds if the address matches a location or register it controls.

Width and Addressable Space: The number of lines (or bits) in the address bus determines the total number of unique addresses the system can access. This directly impacts the maximum amount of memory a system can utilize or the number of I/O ports available.

• If an address bus has N lines, it can specify 2^N unique addresses.
• For example, a system with a 32-bit address bus can address 2³² = 4,294,967,296 unique locations.
• If each memory location stores one byte (the common architecture for PCs), this means the system can address approximately 4 gigabytes (GB) of memory (4,294,967,296 bytes). A 64-bit address bus expands this vastly to 2⁶⁴ bytes, which is approximately 16 exabytes – far more than typical systems utilize today, but necessary for future expansion and large datasets.

Implementation Note: Historically, a separate physical wire was used for each bit of the address. For example, a 16-bit address bus would have 16 wires. As address widths grew, this became unwieldy and expensive, leading to techniques like address multiplexing.

The Data Bus: Carrying the Information

Once the address bus has identified where to communicate, the actual data transfer occurs over the data bus.

Definition: Data Bus

A data bus is a bus used to carry data between the processor, memory, and peripheral devices. This data can be instructions being fetched by the CPU, data being read from or written to memory, or data being exchanged with an I/O device. Data buses are typically bidirectional.

Function: When reading from a location specified on the address bus, the addressed device places the data onto the data bus. When writing, the CPU or sender places the data onto the data bus, and the addressed device reads it.

Width and Throughput: The number of lines (or bits) in the data bus determines how many bits of data can be transferred simultaneously in a single operation.

• An 8-bit data bus transfers 8 bits at a time.
• A 64-bit data bus transfers 64 bits (8 bytes) at a time.

The width of the data bus, combined with the bus speed (frequency) and the number of transfers per clock cycle, determines the bandwidth – the amount of data transferred per unit of time.

• Add Context: Early microprocessors often had 8-bit data buses (like the Intel 8080 or MOS 6502). The original IBM PC XT used an 8-bit data bus for its expansion slots (ISA) but had a 16-bit internal data path on its CPU (Intel 8088). The IBM AT and later systems moved to a 16-bit data bus externally (16-bit ISA), followed by 32-bit (EISA, VLB, PCI), and now commonly 64-bit internally within the CPU and memory interface.

The Control Bus: Orchestrating the Exchange

The control bus is the conductor of the orchestra. It carries various signals that manage the flow of information and synchronize the operations on the address and data buses.

Definition: Control Bus

A control bus is a bus used to carry control and timing signals from the command source (usually the CPU) to other components, and status signals from components back to the command source. These signals manage the read and write operations, indicate the status of devices, handle interrupts, and ensure orderly communication.

Function: Examples of signals on the control bus include:

• Read (RD): Indicates that the CPU wants to read data from the addressed location/device.
• Write (WR): Indicates that the CPU wants to write data to the addressed location/device.
• Clock (CLK): Provides the timing pulse for synchronizing operations.
• Reset (RST): Initializes the system.
• Interrupt Request (IRQ): A peripheral device signals the CPU that it needs attention.
• Bus Grant (BG): Indicates that the bus is free for a device (like a DMA controller) to use.
• Memory Request (MREQ): Indicates that the current operation involves memory.
• I/O Request (IORQ): Indicates that the current operation involves an I/O device.

Necessity: Without the control bus, the address and data buses would be chaotic. The control signals tell devices when to look at the address bus, when to put data on or take data off the data bus, and what kind of operation is being performed (read or write).

Bus Types by Role: Internal vs. External

Buses can also be categorized based on where they operate within the computer system:

1. System Buses (Internal Buses): These are the primary buses connecting the core components of the computer – typically the CPU, memory, and sometimes high-speed peripherals directly integrated onto the motherboard.

   Definition: System Bus

   A system bus (also known as an internal bus or memory bus) is a bus that connects the major components of a computer, such as the central processing unit (CPU), memory, and input/output controllers, allowing them to exchange data and control signals.

   • Add Context: Early system buses often combined address, data, and control lines into a single, integrated design. Modern high-performance systems may have separate dedicated buses for CPU-to-memory communication (the memory bus) and CPU-to-chipset communication, which then branches out to other buses. The memory bus, for example, is highly optimized for speed and defined by standards bodies like JEDEC for specific DRAM types (DDR4, DDR5, etc.).

2. Expansion Buses (Peripheral Buses): These buses are designed to connect external devices or expansion cards (like graphics cards, network cards, sound cards) to the system bus. They provide a standardized interface for adding functionality to the computer.

   Definition: Expansion Bus

   An expansion bus (also known as a peripheral bus) is a bus that provides connections for expansion cards or external peripheral devices to interface with the computer system. They extend the capabilities of the system bus and often include standardized slots or ports.

   • Add Context: In the "from scratch" context, understanding the expansion bus means understanding how to design or interface with standard slots (like historical ISA or modern PCIe) to add custom hardware. Examples include PCI, AGP, PCI Express (PCIe), and external buses like USB, FireWire, SATA, etc.

Implementation Details: Parallel vs. Serial

Buses differ significantly in how they transmit data:

1. Parallel Buses: Data words are transferred simultaneously across multiple wires. A bus with a 32-bit width uses 32 wires for data transmission, plus additional wires for addressing, control, and power.

   Definition: Parallel Bus

   A parallel bus is a bus that transmits multiple bits of data simultaneously over separate parallel wires or traces. Its width is defined by the number of data lines used for simultaneous transfer.

   • Advantage: Potentially high bandwidth at lower frequencies, conceptually simpler for early designs (direct connection from CPU pins).
   • Disadvantage: Problems arise as speeds increase:
     • Timing Skew: Signals traveling on different wires arrive at slightly different times due to variations in wire length, impedance, or driver characteristics. This makes synchronization difficult at high frequencies.
     • Crosstalk: Signals on one wire interfere with signals on adjacent wires, corrupting data.
     • More physical wires, leading to larger connectors and more complex circuit board routing.

2. Serial Buses: Data is transmitted bit by bit sequentially over a single wire or a small number of wires (often a differential pair for noise immunity).

   Definition: Serial Bus

   A serial bus is a bus that transmits data sequentially, one bit at a time, over a single communication channel (typically one wire or a pair of wires for differential signaling).

   • Advantage: Simpler wiring (fewer wires), inherently avoids timing skew issues between data bits transmitted serially on the same line, better signal integrity at high speeds.
   • Disadvantage: Requires complex encoding and decoding logic (SerDes) to convert parallel data from internal circuits into serial data for transmission and back again. Potentially lower bandwidth per wire, but the ability to run at much higher frequencies often results in higher overall bandwidth than comparable parallel buses.
   • Add Context: The transition from parallel (like standard PCI) to serial (like PCI Express) was a major shift driven by the speed limitations of parallel signaling. Modern high-speed interfaces like USB, SATA, FireWire, Thunderbolt, and PCIe are all serial. This was made possible by advancements in integrated circuits (Moore's Law) allowing for complex SerDes chips to be integrated easily.

Bus Speed, Frequency, and Bandwidth

The performance of a bus is often described by its speed or frequency and its bandwidth.

• Frequency: Measured in Hertz (Hz, often MHz or GHz), this indicates how many cycles of the bus clock occur per second. Operations on the bus are synchronized to this clock.
• Data Transfer Rate (or Throughput/Bandwidth): The amount of data moved per unit of time, typically measured in bytes per second (B/s) or bits per second (b/s).

Bandwidth is calculated based on:

• Frequency: How fast the clock ticks.
• Bus Width: How many bits are transferred per transfer cycle.
• Transfers per Cycle: How many data transfers occur during each clock cycle (Single Data Rate vs. Double Data Rate).

Definitions: Speed and Bandwidth

• Frequency (Bus Speed): The rate at which the bus clock cycles, measured in Hertz (Hz).
• Bandwidth: The maximum rate at which data can be transferred over the bus, typically measured in bits per second (b/s) or bytes per second (B/s). It is a function of frequency, bus width, and transfers per clock cycle.
• Single Data Rate (SDR): A signaling method where data is transferred only once per clock cycle, typically on the rising or falling edge of the clock signal.
• Double Data Rate (DDR): A signaling method where data is transferred twice per clock cycle, typically on both the rising and falling edges of the clock signal. This effectively doubles the data rate for a given frequency.

Calculation Example: A parallel bus with a 32-bit width operating at 100 MHz using SDR: Bandwidth = 100,000,000 cycles/second * 32 bits/cycle * 1 transfer/cycle = 3,200,000,000 bits/second To convert to bytes/second (1 byte = 8 bits): 3,200,000,000 b/s / 8 bits/byte = 400,000,000 B/s = 400 MB/s.

A serial bus operating at a much higher frequency might achieve higher bandwidth despite having a smaller "width" (often considered 1 bit per lane, but multiple lanes can be grouped in modern serial buses like PCIe). Modern serial buses also use sophisticated encoding schemes (like 8b/10b or 128b/130b encoding) and complex signaling (like PAM4 in some high-speed interfaces) which further increase the effective data rate, though the encoding adds overhead.

Bus Multiplexing: Sharing Wires

To reduce the number of physical pins on chips and traces on circuit boards, a technique called multiplexing is often used. This involves using the same physical wires for different purposes at different times.

Definitions: Bus Multiplexing

• Bus Multiplexing: A technique where the same physical wires are used to carry different types of information (e.g., address and data) at different points in time during a communication cycle. This reduces the number of required connections but adds complexity to the timing and control logic.
• Address Multiplexing: A specific form of bus multiplexing commonly used for DRAM, where the memory address is split into two parts (row address and column address) and sent sequentially over the same set of address lines. Control signals (like RAS and CAS) indicate which part of the address is currently being sent.

Example: Early processors like the Intel 8086 and buses like conventional PCI multiplexed address and data lines. First, the address would be placed on the shared lines, along with a control signal indicating "address is valid." Then, the control signal would change, and the same lines would be used for the data transfer. This reduced the number of pins and traces compared to having dedicated address and data buses but required more complex timing logic. Serial buses take this to the extreme, multiplexing address, data, and control information into a single bitstream.

Advanced Bus Concepts

As computers evolved, several concepts were developed to improve bus efficiency and performance:

• Direct Memory Access (DMA): Allows peripheral devices to read from or write to memory directly, without involving the CPU in each data transfer step.

  Definition: Direct Memory Access (DMA)

  Direct Memory Access (DMA) is a feature of computer systems that allows certain hardware subsystems (like disk controllers, network cards, or graphics cards) to access main system memory (RAM) independently of the central processing unit (CPU). This offloads the CPU from data transfer tasks, significantly improving system performance, especially for high-speed I/O.

  • Add Context: A DMA controller is a specialized chip or function within a device that takes control of the bus from the CPU for a short period to perform the transfer. This requires bus mastering capabilities.

• Interrupts: A mechanism where a peripheral device can signal the CPU (via the control bus) that it needs attention, rather than the CPU having to constantly poll the device. This significantly improves CPU efficiency. Interrupts need a system for prioritization, historically done via daisy-chaining or dedicated interrupt controllers.

• Bus Mastering: The ability of a device (like a DMA controller or a sophisticated peripheral) to take control of the bus from the CPU to perform operations (like DMA transfers) independently. This is a feature of many modern bus architectures (like PCI and PCIe).

A Brief History of Bus Evolution

Understanding the historical progression helps illustrate the challenges engineers faced and the solutions they devised.

First Generation (Early Computers & Microcomputers):

• Separate Buses: Initially, memory and I/O often had separate buses with distinct instructions and timings.
• The Need for Efficiency: Introduction of interrupts to avoid CPU polling loops. Channel controllers (like those on IBM mainframes) as dedicated I/O processors to reduce CPU load.
• Unification (PDP-11 Unibus, S-100): The idea of a single, unified system bus emerged. Devices were mapped into the memory address space (memory-mapped I/O), allowing the CPU to use the same instructions for memory and I/O access.
• Passive Backplanes: Early microcomputers like the Altair 8800 (with the S-100 bus) used a passive backplane where expansion slots were essentially wired directly to the CPU pins.

  Definition: Passive Backplane

  A passive backplane is a printed circuit board containing connectors into which other boards (like a CPU card, memory cards, and I/O cards) are plugged. It provides the physical bus connections (address, data, control, power lines) but does not contain active circuitry for bus arbitration or control itself. The CPU board typically acts as the bus master.

• Limitations: All devices on the bus had to operate at the same speed. Increasing CPU speed was limited by the slowest device on the bus. Configuration was complex (jumpers for addresses, interrupts).

Second Generation (ISA, EISA, VLB, PCI, AGP):

• Bus Controller/Bridge: Systems introduced dedicated bus controllers (often part of the chipset) that acted as bridges between the faster CPU/memory side and the slower peripheral bus side. This isolated the CPU, allowing it to run faster.
• Improved Performance: Wider data paths (16-bit, 32-bit).
• Modularity and Plug-n-Play: Buses like PCI introduced more sophisticated arbitration and configuration mechanisms, moving towards automatic setup (Plug-n-Play) instead of manual jumper settings.
• Bottlenecks: Despite improvements, the shared bus still became a bottleneck as CPU and graphics card speeds outpaced the general expansion bus. This led to dedicated high-speed buses like AGP specifically for graphics.
• Rise of Dedicated Peripheral Interfaces: High-speed storage and other devices moved off the main expansion bus onto dedicated interfaces like SCSI and IDE, reducing the load on the main bus.

Third Generation (PCI Express, HyperTransport, InfiniBand, USB, SATA, Thunderbolt, CXL):

• Shift to Serial, Point-to-Point (often): Modern high-performance buses moved towards serial communication over dedicated links or "lanes" between devices, or sophisticated switched fabrics, rather than a single shared parallel bus.
• High Speed and Scalability: Serial links can run at much higher frequencies, and multiple links ("lanes") can be grouped together for increased bandwidth (e.g., PCIe x1, x4, x16).
• Network-like Protocols: Higher protocol overhead is required for error correction, flow control, and managing complex connections, making them conceptually closer to networks in some ways.
• Flexibility: Many third-generation buses can be used for both internal communication (PCIe) and external connections (Thunderbolt, which uses PCIe signaling).
• Specialized Interconnects: Emergence of standards like CXL (Compute Express Link) designed for high-speed CPU-to-device and CPU-to-memory communication in data centers.

Examples of Buses

The world of computing uses a vast array of buses, depending on the era and purpose. Here are a few mentioned in the source, categorized for clarity:

Internal Computer Buses (primarily within the machine):

• Parallel: ISA (historical PC expansion), EISA (historical PC expansion), VLB (historical PC expansion), PCI (historical/legacy expansion), AGP (historical graphics), VMEbus (industrial), S-100 (early microcomputer), STEbus (industrial/embedded).
• Serial: PCI Express (PCIe - modern standard for expansion cards like graphics, SSDs, etc.), SATA (storage), SLDRAM (memory interface, largely superseded), RDRAM (memory interface, largely superseded), HyperTransport (processor interconnect, legacy), Compute Express Link (CXL - high-speed processor/memory interconnect).

External Computer Buses (primarily for connecting devices outside the main chassis):

• Parallel: Centronics (historical printer port, although the source argues against calling it a "bus" due to power handling), IEEE 1284 (update to Centronics).
• Serial: USB (Universal Serial Bus - ubiquitous for peripherals), FireWire (IEEE 1394 - historical for video/storage), eSATA (external SATA), ExpressCard (laptop expansion, legacy), Thunderbolt (high-speed interface combining PCIe and DisplayPort signals), RS-232 (legacy serial port). Field buses (CAN bus, Modbus, ARINC 429, MIL-STD-1553, IEEE 1355) are specialized serial buses used in industrial, automotive, and avionics systems.

Internal/External Buses (can bridge both domains, often via connectors):

• Thunderbolt (as mentioned, bridges external devices using internal PCIe signaling)

The Bus in "Building From Scratch"

When undertaking the task of building a computer from scratch, understanding buses means:

1. Choosing Components: You need to select components (CPU, memory, peripherals) that use compatible bus interfaces. You can't directly connect an old ISA card to a modern PCIe slot, for example.
2. Designing the Interconnect: If you're building a simple system, you'll need to design the physical connections (traces on a PCB) for your chosen bus architecture (e.g., connecting CPU pins to memory and I/O pins).
3. Implementing Control Logic: You need to understand the timing and control signals on the control bus to correctly sequence read and write operations. If using multiplexed buses, you need the logic to de-multiplex addresses and data.
4. Handling Addressing: You must map memory locations and I/O device registers within the system's address space, respecting the width of the address bus.
5. Considering Performance: The choice and implementation of the bus architecture directly impact the system's overall speed and bandwidth.

In essence, the bus is the nervous system of your computer. Without a solid understanding of how it works – from the basic address/data/control lines to complex serial protocols and historical evolution – building or even truly comprehending a computer system at a fundamental level remains elusive. It is indeed a core piece of the "lost art."

---