Loading learning content...
In 1973, Bob Metcalfe sketched a design for a local area network on a napkin at Xerox PARC. That sketch became Ethernet—now the most ubiquitous networking technology in human history. Billions of Ethernet ports exist worldwide, in everything from data centers to home routers, from industrial robots to medical devices.
At the heart of original Ethernet's physical layer was Manchester encoding. The choice of Manchester wasn't arbitrary—it solved specific problems in ways that enabled Ethernet to flourish. Understanding this application provides crucial insight into why encoding choices matter and how physical layer design influences entire network architectures.
This page examines how Manchester encoding enabled 10 Mbps Ethernet across three major physical layer standards: 10BASE5 (thick coax), 10BASE2 (thin coax), and 10BASE-T (twisted pair).
By the end of this page, you will understand Ethernet's physical layer requirements, how Manchester encoding enables CSMA/CD, the role of preamble and frame delimiters, specifications across all 10 Mbps variants, and why Manchester persisted from 1973 to present-day 10 Mbps implementations.
Before examining Manchester's role, we must understand the fundamental architecture that Ethernet's physical layer had to support.
CSMA/CD: The Ethernet Access Method
Ethernet uses Carrier Sense Multiple Access with Collision Detection (CSMA/CD) for shared medium access:
This access method places specific requirements on the physical layer that influenced encoding choice.
Why Manchester Was Chosen:
Manchester encoding satisfies all these requirements elegantly:
| Requirement | Manchester Solution |
|---|---|
| Carrier Sense | Continuous transitions provide clear "signal present" indication |
| Collision Detection | Collisions produce invalid transition patterns (multiple transitions per bit period) |
| DC Balance | Perfect balance enables transformer coupling |
| Self-Sync | Guaranteed transitions enable rapid clock recovery |
| Signal Detection | Transition-based detection works at any cable attenuation |
No other encoding available at the time met all requirements as completely. NRZ failed on DC balance and carrier sense. Return-to-zero codes wasted bandwidth. Manchester was the optimal choice for 1970s technology.
Original Ethernet used thick coaxial cable as a shared bus. All stations connected to the same cable, transmitted onto it, and received from it simultaneously. This true bus topology, combined with CSMA/CD, defined Ethernet's early architecture and drove the need for Manchester encoding's specific properties.
Understanding how Ethernet frames are transmitted reveals how Manchester encoding integrates with higher protocol layers.
Frame Structure at Physical Layer:
When the MAC layer hands a frame to the physical layer for transmission, the PHY prepends additional elements:
12345678910111213141516171819202122232425262728
Ethernet Frame at Physical Layer (10 Mbps) ┌─────────┬──────┬────────────────────────────────────────────┬─────────┐│Preamble │ SFD │ MAC Frame │ EFD ││ 7 bytes │1 byte│ (64 to 1518 bytes) │(carrier)│├─────────┼──────┼────────────────────────────────────────────┼─────────┤│ 56 bits │8 bits│ 512 to 12144 bits │ varies │└─────────┴──────┴────────────────────────────────────────────┴─────────┘ Preamble: 10101010 10101010 10101010 10101010 10101010 10101010 10101010 (seven octets of alternating 1-0 pattern) Purpose: Clock synchronization for receiver PLL lock SFD: 10101011 (Start Frame Delimiter) Purpose: Marks frame start; establishes bit position reference MAC Frame Contents:┌────────────┬────────────┬──────┬─────────────────────┬─────┐│Dest Addr │Source Addr │Type/ │ Data │ FCS ││ 6 bytes │ 6 bytes │Len │ 46-1500 bytes │4 byt││ │ │2 byte│ │ │└────────────┴────────────┴──────┴─────────────────────┴─────┘ EFD: Extended Frame Delimiter - carrier extension after FCS (varies by implementation; ensures collision detection) Total Transmitted Bits: 576 to 12208 bits (plus preamble/SFD)Transmission Time: 57.6 μs to 1220.8 μs at 10 MbpsThe Preamble's Critical Role:
The 7-byte preamble serves multiple essential functions:
1. Clock Acquisition: The alternating 10101010 pattern produces Manchester transitions at the maximum possible rate—once per half-bit period, or 20 million transitions per second. This gives the receiver's PLL abundant phase information to achieve lock rapidly.
2. Signal Level Adjustment: Receiver automatic gain control (AGC) circuits use the preamble to set appropriate amplification before data arrives.
3. Bit Synchronization: The regular pattern establishes the precise timing relationship between transitions and bit positions.
4. Transient Settling: Initial signal transients (from cable reflections, transformer coupling, etc.) settle during the preamble, ensuring clean data reception.
| Parameter | Value | Significance |
|---|---|---|
| Preamble Length | 56 bits (7 bytes) | 5.6 μs at 10 Mbps |
| Transition Rate | 20 MHz (maximum) | One per 50 ns half-bit |
| PLL Lock Requirement | < 56 bits | Must lock before SFD |
| AGC Settling | Typically 20-30 bits | Signal level stabilization |
| Pattern | 10101010 repeated | Worst-case for cable loss, best for CDR |
Start Frame Delimiter (SFD):
The SFD (10101011) is carefully chosen:
In Manchester encoding, the two consecutive 1s produce:
Bit pattern: ...1 0 1 0 1 0 1 1
↑
Pattern break: no boundary transition
The receiver detects the missing boundary transition between the final two 1s, identifying the SFD and establishing the precise bit position for the following destination address.
The 7-byte preamble length accounts for worst-case clock acquisition time plus margin for preamble bits lost in initial detection. Repeaters and hubs may regenerate shorter preambles (as few as 55 bits), which is why the specification requires receivers to lock within 7 bytes but recommends designing for faster acquisition.
10BASE5 was the original Ethernet standard, using thick coaxial cable ("thicknet") as the transmission medium. Understanding this implementation reveals Manchester encoding's foundational role.
10BASE5 Physical Specifications:
| Parameter | Specification | Notes |
|---|---|---|
| Data Rate | 10 Mbps | 10 million bits per second |
| Encoding | Manchester (IEEE 802.3) | LOW-to-HIGH = 0, HIGH-to-LOW = 1 |
| Baud Rate | 20 MBd | Twice the bit rate |
| Cable Type | RG-8 coaxial | 10mm diameter, 50Ω impedance |
| Maximum Segment | 500 meters | Limited by signal attenuation |
| Maximum Stations | 100 per segment | Spacing: 2.5m minimum |
| Signal Levels | ±0.85V (typical) | DC-balanced on cable |
Medium Attachment Unit (MAU):
In 10BASE5, stations connected to the cable through a Medium Attachment Unit (MAU), also called a transceiver:
Manchester Signaling on Coax:
The thick coaxial cable carried Manchester-encoded signals as voltage variations:
Coaxial Cable Signal Levels:
Idle: 0V (no signal)
HIGH: +0.85V nominal (+0.70V to +1.03V range)
LOW: -0.85V nominal (-0.70V to -1.03V range)
DC Level (average): 0V (perfect balance)
This bipolar signaling ensured:
12345678910111213141516171819202122232425262728293031323334353637
Collision Detection in 10BASE5 Normal Transmission (Single Station):Signal amplitude: ±0.85VTransition rate: 20 MHz (Manchester) +0.85V ─┐ ┌───┐ ┌─── │ │ │ │ 0V ──┼───┼───┼───┼─── │ │ │ │ -0.85V ─┘ └───┘ └─── Collision (Two Stations Transmitting):Combined amplitude: up to ±1.7V (constructive) or near 0V (destructive) Case 1: Constructive interference (same phase) +1.7V ───────────── 0V ───────────── -1.7V ───────────── Case 2: Destructive interference (opposite phase) +0.85V ─┐ │ 0V ──┴─────────── (transitions cancel) -0.85V ─ Collision Detection Criteria:1. Signal amplitude exceeds normal range (> 1.5V peak)2. DC bias appears (transitions don't average to zero)3. Invalid transition patterns (multiple transitions in wrong intervals) MAU monitors for these conditions during transmission,signaling "collision" (COL) to the attached station.The 500-meter segment length in 10BASE5 was determined by Manchester encoding's 20 MHz spectral requirements combined with RG-8 cable's attenuation at those frequencies. Longer segments would attenuate the signal's high-frequency components, corrupting Manchester transitions and preventing reliable clock recovery.
10BASE2 (also called "Cheapernet" or "Thinnet") was developed to reduce cost while maintaining compatibility. It uses thinner, more flexible RG-58 coaxial cable with similar Manchester signaling.
10BASE2 vs. 10BASE5 Comparison:
| Parameter | 10BASE5 | 10BASE2 |
|---|---|---|
| Data Rate | 10 Mbps | 10 Mbps |
| Encoding | Manchester | Manchester (identical) |
| Cable Type | RG-8 (thick, 10mm) | RG-58 (thin, 5mm) |
| Impedance | 50Ω | 50Ω |
| Max Segment Length | 500m | 185m |
| Max Stations/Segment | 100 | 30 |
| Connection Method | Vampire tap + drop cable | BNC T-connector |
| Cost | High | Lower (hence 'Cheapernet') |
Why Shorter Segments:
RG-58 cable has higher attenuation than RG-8:
| Cable Type | Attenuation at 10 MHz |
|---|---|
| RG-8 | ~1.9 dB/100m |
| RG-58 | ~4.6 dB/100m |
With more than double the attenuation, RG-58 required proportionally shorter segments to maintain adequate signal quality for Manchester decoding. The 185m limit (often rounded to 200m in discussion) ensures the worst-case received signal still has sufficient amplitude and transition integrity.
BNC Connectors:
10BASE2's use of BNC T-connectors meant:
Missing terminators caused reflections that corrupted Manchester transitions, making the entire segment unreliable. This was a common troubleshooting issue in 10BASE2 networks.
Manchester Encoding Unchanged:
The encoding itself was identical to 10BASE5:
This physical layer compatibility meant that 10BASE5 and 10BASE2 segments could be connected through repeaters, forming a unified network despite different cable types.
In coaxial Ethernet, improper termination was the leading cause of network failures. Without 50Ω terminators, signals reflected from cable ends, creating echoes that interfered with Manchester transitions. The resulting errors appeared as collisions to CSMA/CD, often bringing the network to a standstill.
10BASE-T revolutionized Ethernet deployment by using unshielded twisted pair (UTP) cable—the same wiring used for telephone systems. This enabled Ethernet to leverage existing building infrastructure and adopt a star topology.
10BASE-T Physical Specifications:
| Parameter | Specification | Notes |
|---|---|---|
| Data Rate | 10 Mbps | Same as coaxial variants |
| Encoding | Manchester | Same encoding scheme |
| Cable Type | Cat 3 (or better) UTP | Two pairs used |
| Connector | RJ-45 | 8-pin modular jack |
| Max Segment Length | 100 meters | Station to hub |
| Topology | Star (via hub) | Point-to-point links |
| Signal Levels | ±2.5V (differential) | Higher than coax for noise immunity |
Topology Transformation:
10BASE-T fundamentally changed Ethernet's topology while maintaining CSMA/CD semantics:
Coaxial Ethernet:
[Station]─────┬─────┬─────┬─────┬─────[Station]
│ │ │ │
[Station][Station][Station][Station]
Physical bus: all stations share one cable
10BASE-T:
[Station] [Station] [Station]
│ │ │
└─────┬─────┴─────┬─────┘
│ │
┌────┴───────────┴────┐
│ HUB │
└────┬───────────┬────┘
│ │
┌─────┴─────┐ └─────┐
│ │ │
[Station] [Station] [Station]
Physical star: point-to-point links to hub
Logical bus: hub repeats signals to all ports
The hub received Manchester-encoded signals from any transmitting station and repeated them to all other ports, maintaining the logical bus behavior required for CSMA/CD.
Differential Signaling:
10BASE-T uses differential Manchester signaling over twisted pairs:
Signal levels are higher than coaxial Ethernet to overcome twisted pair limitations:
Differential Signal Levels:
Transmit: ±2.5V (5.0V peak-to-peak differential)
Receive: ±585mV minimum (after cable loss)
The twisted pair's coupling provides several advantages:
1234567891011121314151617181920212223242526
10BASE-T RJ-45 Pinout RJ-45 Connector (Front View) ┌─────────────────────────┐ │ 1 2 3 4 5 6 7 8 │ │ │ │ │ │ │ │ │ │ │ └──┴──┴──┴──┴──┴──┴──┴──┴─┘ Pin Signal Pair ───────────────────────────── 1 TX+ (Transmit+) Pair 2 ─┐ 2 TX- (Transmit-) Pair 2 ─┘ Orange pair 3 RX+ (Receive+) Pair 3 ─┐ 4 (unused) │ Blue pair (unused) 5 (unused) │ 6 RX- (Receive-) Pair 3 ─┘ Green pair 7 (unused) Brown pair (unused) 8 (unused) Transmit Path: Station → Hub via pins 1,2Receive Path: Hub → Station via pins 3,6 Full Duplex Note (later enhancement):Simultaneous TX and RX on separate pairs enablescollision-free, true full-duplex operation whenswitches replace hubs (no CSMA/CD needed).When 10BASE-T stations are idle (not transmitting data), they send periodic "link integrity" pulses—100ns signals every 16±8ms. These pulses verify the link is active and allow hubs to detect disconnected ports. Without data traffic, no Manchester transitions occur, so these pulses maintain the link state indication.
CSMA/CD collision detection must work regardless of the physical medium. Manchester encoding enables this, but the detection mechanism differs between coaxial and twisted pair implementations.
Coaxial Collision Detection (10BASE5/10BASE2):
On shared coaxial cable, collision detection relies on signal amplitude and DC shift:
Amplitude Monitoring: During transmission, the MAU monitors the cable for signal levels exceeding normal ranges. Two overlapping Manchester signals can constructively interfere, producing up to 2× normal amplitude.
DC Shift Detection: When Manchester signals collide, their DC balance may be disrupted. The resulting DC offset indicates a collision.
Transition Pattern Analysis: Valid Manchester produces one transition per half-bit period. Collisions create additional transitions or cause expected transitions to cancel.
10BASE-T Collision Detection:
10BASE-T uses separate transmit and receive pairs, fundamentally changing collision detection:
123456789101112131415161718192021222324252627282930313233
10BASE-T Collision Detection Station-to-Hub Communication (Half-Duplex Mode): ┌─────────────┐ ┌─────────────┐ │ Station │ TX+ TX- │ Hub │ │ │ ───────────────►│ │ │ (TX) │ RX+ RX- │ (Port) │ │ (RX) │ ◄───────────────│ │ └─────────────┘ └─────────────┘ Collision Scenario: Time T0: Station A starts transmitting (TX pair active)Time T0: Station B also transmits to hub (different port)Time T1: Hub detects signals on multiple portsTime T1: Hub transmits JAM signal on ALL ports RX pairsTime T2: Station A detects signal on RX pair while transmittingTime T2: COLLISION DETECTED (TX and RX active simultaneously)Time T3: Station A ceases transmission, schedules retransmit Collision Detection Logic:┌──────────────────────────────────────────────────────────────┐│ IF (Transmitting on TX pair) AND (Receiving on RX pair) THEN ││ COLLISION = TRUE ││ Assert COL (collision) signal to MAC ││ Begin JAM transmission (32-bit pattern) ││ Abort frame, schedule binary exponential backoff ││ END IF │└──────────────────────────────────────────────────────────────┘ JAM Signal: 32 bits of any pattern that violates CRC Ensures all stations recognize collision stateThe SQE (Signal Quality Error) Test:
In original Ethernet, the SQE test (also called "heartbeat") was a mechanism for verifying collision detection circuitry:
Minimum Frame Size and Collision Domain:
For CSMA/CD to work, colliding stations must still be transmitting when collision evidence reaches them. This constrains network design:
Minimum Frame Time = 2 × Maximum Propagation Delay + Safety Margin
At 10 Mbps:
Minimum frame = 64 bytes = 512 bits = 51.2 μs transmission time
Maximum round-trip delay = ~51.2 μs = 2500m of cable
This is why 10 Mbps Ethernet limits:
- Maximum 5 segments with 4 repeaters
- Maximum 2500m end-to-end
- Maximum 2.5km cable path between any two stations
Manchester encoding's self-clocking property ensures that collision evidence (the combined signal) propagates at the same rate as data, maintaining this timing relationship.
A 'late collision' occurs if a collision is detected after the first 64 bytes (512 bits) have been transmitted. Late collisions indicate network design violations—the collision domain is too large. Manchester encoding cannot help here; network topology must be corrected.
Manchester encoding served 10 Mbps Ethernet excellently, but higher speeds demanded different approaches. Understanding this transition illuminates both Manchester's limitations and the properties that made it work so well.
Why Manchester Couldn't Scale:
At 100 Mbps, Manchester encoding would require:
The solution was adopting more bandwidth-efficient encodings:
| Standard | Data Rate | Encoding | Baud Rate | Bandwidth |
|---|---|---|---|---|
| 10BASE-T | 10 Mbps | Manchester | 20 MBd | ~20 MHz |
| 100BASE-TX | 100 Mbps | 4B/5B + MLT-3 | 125 MBd | 31.25 MHz |
| 1000BASE-T | 1000 Mbps | PAM-5 + 4D | 125 MBd × 4 pairs | ~62.5 MHz |
| 10GBASE-T | 10 Gbps | PAM-16 (128-DSQ) | 800 MBd × 4 pairs | ~400 MHz |
100BASE-TX: 4B/5B + MLT-3
Fast Ethernet (100 Mbps) replaced Manchester with a two-stage encoding:
4B/5B Block Coding: Every 4 data bits are mapped to 5 code bits. The code guarantees no more than 3 consecutive identical bits, ensuring sufficient transitions for clock recovery.
MLT-3 Line Coding: The 5-bit codes are transmitted using Multi-Level Transmit 3, which cycles through three voltage levels (+V, 0, -V). Each bit of "1" causes a level change; bits of "0" maintain the current level.
Result: Clock recovery is maintained with only 25% overhead (instead of Manchester's 100%), and the fundamental frequency is reduced to 31.25 MHz—compatible with Category 5 cabling.
1000BASE-T: PAM-5 + 4D
Gigabit Ethernet uses all four pairs of Category 5 cable simultaneously:
Manchester's Enduring Role:
Despite these advances, Manchester encoding persists where simplicity matters:
Manchester's key innovations—self-clocking, DC balance, guaranteed transitions—live on in every modern encoding scheme. 4B/5B, 8B/10B, 64B/66B, and PAM encodings all incorporate these properties in more bandwidth-efficient ways. Manchester was the prototype; modern encodings are optimized implementations of the same essential ideas.
We've explored Manchester encoding's foundational role in Ethernet from original coaxial implementations to modern 10BASE-T. Let's consolidate the essential knowledge:
What's Next:
With our understanding of Manchester encoding in Ethernet complete, we'll now examine efficiency trade-offs—quantifying Manchester's bandwidth penalty and comparing it systematically with alternative encoding schemes to understand the design decisions that shaped modern networking.
You now have comprehensive knowledge of how Manchester encoding enabled Ethernet—from the original Xerox PARC design through coaxial standards to ubiquitous 10BASE-T. This understanding connects encoding theory to real-world networking and explains the evolution of physical layer technologies.