The physical level is one of the most underestimated, yet fundamental levels. If something "breaks" here — nothing works above it anymore.

What the physical level is

The physical level is the level responsible for:

the transmission of raw bits (0 and 1) over the physical medium

It knows nothing about:

• HTTP
• TCP
• IP
• packets

It only knows:

"how to transmit a signal from point A to point B"

The main idea

A computer does not send "1" and "0" directly. It turns them into a physical signal:

• electrical (over a wire)
• light (fiber optics)
• radio waves (Wi-Fi, LTE)

Analogy

Imagine:

• Data link level → "the letter"
• Physical level → "the wires/air/light along which it travels"

What exactly the physical level solves

It solves tasks such as:

How to encode bits into a signal

• 0 and 1 → voltage / light / radio signal

How to transmit the signal

• continuously or in packets
• analog or digital

Transmission speed

• how many bits per second (bps)

The physical medium

• what the signal travels over

Wired transmission media

Twisted Pair

The most common (Ethernet):

• 2 wires twisted together → less interference
• cheap
• up to ~100 m

Used for:

• LAN
• Ethernet (Cat5e, Cat6)

Coaxial cable

• central conductor + shield
• better protection against interference

Used for:

• cable TV
• DOCSIS (internet over TV)

Fiber optics

• transmission by light
• very high speed
• enormous distances

Key properties:

• not afraid of electromagnetic interference
• speeds up to Tbit/s

This is the backbone of the internet.
In 2026, researchers from University College London set a new record for data transmission speed over already existing commercial fiber optic cables — 450 terabits per second, which is roughly ten times the figures of currently operating commercial networks. The record was made possible by equipment that used a broad spectrum of wavelengths — from 1264 to 1617.8 nanometers. An even more impressive result was achieved by Chinese researchers from China Information and Communication Technologies Group, who realized 2.5 Pbit/s (2500 Tbit/s) over a 24-core single-mode fiber, using more than 6000 parallel channels and the S-, C-, and L-bands. This is equivalent to transmitting more than 10,000 movies in 4K format in one second.

Power lines

• internet over the electrical network
• rarely used

Wireless transmission

The main idea

transmission through electromagnetic waves

Spectrum

The higher the frequency:

• the higher the speed
• the worse it penetrates through walls

Type	Example
Radio	Wi-Fi
Microwaves	4G / 5G
Infrared	remote controls
Visible light	Li-Fi

Main technologies

Radio communication

• Wi-Fi
• Bluetooth

Microwaves

• cellular communication
• satellites

Infrared

• short distances

The key problem: noise

Any signal:

• gets distorted
• gets lost
• mixes with noise

Therefore there is a speed limit.

Maximum channel speed

The idea

you cannot transmit infinitely fast — there are physical limitations

The Nyquist theorem (without noise)

C = 2B \ M

Where:

• C — speed
• B — channel width
• M — signal levels

The Shannon theorem (with noise)

C = B \ (1 + S/N)

Where:

• S/N — signal-to-noise ratio

The essence

even an ideal code will not be able to exceed this limit

Modern technologies are approaching the Shannon limit closely. For example, the programmable Nokia PSE-3 chipset with probabilistic constellation shaping (PCS) technology allows bringing the throughput of fiber optics closer to the theoretical limit, increasing capacity by up to 65% and reducing power consumption by 60% compared to existing networks. PCS provides practically infinite granularity of modulation control, allowing tuning the throughput of a wavelength from 100G to 600G.

Modulation

A computer does not simply "turn the signal on/off."
It modulates it:

• changes the amplitude (ASK)
• the frequency (FSK)
• the phase (PSK)

This allows:

• transmitting more bits at once
• using the channel more efficiently

Multiplexing

multiple data streams → one channel

Types

• FDM — by frequency
• TDM — by time
• WDM — by wavelength (optics)

Real systems

The telephone network

• old, but important
• the basis for ADSL

Cellular networks

• 1G → analog
• 2G → digital voice
• 3G → internet
• 4G → IP
• 5G → high speed + low latency

Satellites

• GEO (high, large latency)
• LEO (low, faster)

How this relates to development

A developer does not work directly with the physical level, but:

It affects everything

• latency
• packet loss
• bandwidth

Examples

Slow internet

→ a problem in the channel (noise, congestion)

WebSocket keeps dropping

→ an unstable physical channel

High ping

→ distance + transmission medium

Summary

The physical level is:

converting bits into signals and transmitting them through the real world

It is responsible for:

• the medium (cable / air)
• the signal
• the speed
• noise

How exactly bits turn into a signal

We already know:

0 and 1 → signal

Now the question: HOW exactly?

Line coding (baseband)

This is when we directly encode bits into signal levels.

The simplest variant

• 0 → 0 V
• 1 → 5 V

But this is bad, because:

• there is no synchronization
• many errors occur

Real encoding schemes

NRZ (Non-Return to Zero)

• 1 = high level
• 0 = low level

Problem: if 0s or 1s go on for a long time → the receiver "loses the beat"

Manchester encoding

Every bit contains a signal transition

• 0 → low → high
• 1 → high → low

Plus:

• there is built-in synchronization

Minus:

• requires twice the frequency

PAM4 (a modern development)

For high-speed Ethernet channels (400G and above), PAM4 (Pulse Amplitude Modulation, 4 levels) is used — instead of two signal levels (0 and 1), four are used, which allows transmitting 2 bits per symbol. This has become a key technology for achieving speeds of 400G and 800G over copper and optical channels, although it requires more complex signal processing on the receiving side to compensate for distortions.

Conclusion

The physical level is not just a "signal," but encoding the signal in such a way that it can be reliably restored

Synchronization

Imagine:

• the sender sends bits
• the receiver must understand where a bit begins and ends

This is not obvious.

Solutions:

• Built-in synchronization (Manchester)
• A separate clock signal
• Clock recovery (restoration from the signal)

Important to understand

Without synchronization, even an ideal signal is useless

Analog vs digital signal

Digital:

• discrete levels
• resistant to noise

Analog:

• continuous
• sensitive to distortion

The key point
Even if you have a "digital network":

physically the signal is ALWAYS analog

Modulation

Let's break down why it is really needed

The task of modulation:

to transmit digital data through an analog channel

Main types

ASK (amplitude)

• we change the strength of the signal

FSK (frequency)

• we change the frequency

PSK (phase)

• we change the phase

The modern approach: QAM

Quadrature Amplitude Modulation

combines:

• amplitude
• phase

Why this is important

One signal can encode NOT 1 bit, but several:

Type	Bits per symbol
BPSK	1
QPSK	2
16-QAM	4
64-QAM	6
256-QAM	8

Conclusion

speed grows not only due to frequency, but also due to the "cleverness" of the encoding

Noise and distortion

Types of problems

Thermal noise

• physics, inevitable

Crosstalk

• neighboring wires interfere

Attenuation

• the signal weakens with distance

Interference

• other devices

Consequences

• errors
• data loss
• reduced speed

Signal-to-noise ratio (SNR)

The key metric:

SNR = Signal / Noise

• high SNR → good connection
• low SNR → errors

Example:

• Wi-Fi far from the router → low SNR
• LTE in the subway → low SNR

Bandwidth

It is important not to confuse:

Channel width (Hz)

• a physical characteristic

Speed (bps)

• how much data we transmit

The relationship

more bandwidth → higher possible speed

Multiplexing

The problem

one channel → many users

Solutions:

FDM (Frequency Division)

• everyone gets their own frequency

TDM (Time Division)

• everyone gets their own time

CDM (Code Division)

• everyone gets their own code

Example

4G / 5G use:

• OFDM (many frequencies + time)

Signal propagation

Speed of the signal

• copper: ~2/3 of the speed of light
• optics: ~2/3 of the speed of light
• vacuum: the speed of light

The delay formula

delay = distance / speed

Conclusion

latency is not only about servers, it is physics

Errors at the physical level

The physical level:

• does NOT correct errors
• does NOT check packets

It simply transmits

But:

It affects the probability of errors

The connection with the data link level

Here is the key idea:

physical level → "dirty channel"
data link level → "makes it reliable"

Overall summary

The physical level is:

Encoding

→ how to represent bits

Transmission

→ through which medium

Limitations

→ noise, distance, frequency

Optimization

→ modulation, multiplexing

Why the signal gets distorted

What exactly happens to the signal.

Attenuation

The signal loses energy as it propagates.

Causes

• resistance of the conductor
• scattering in the medium
• absorption

Result:

the farther → the weaker the signal

Dispersion (distortion of shape)

Different frequencies of the signal travel at different speeds.
As a result:

• the signal "stretches"
• neighboring bits begin to overlap

This is called

intersymbol interference (ISI)

Example

You sent:

101010

And what arrived:

1??0?0

Why you cannot simply "speed up the signal"

Intuitively it seems:

"let's just increase the transmission speed"

But there are limitations:

Limitation #1 — bandwidth

The channel does not pass all frequencies.

• Sharp changes (bit 0 → 1) require high frequencies
• The channel "cuts" them

Limitation #2 — noise

The higher the speed:

• the smaller the "gap" between levels
• the easier it is to confuse the signal

Limitation #3 — interference

Signals begin to interfere with each other

Summary

speed is a balance between:

• frequency
• noise
• quality of the medium

Why fiber optics is better than copper

Copper

Problems:

• resistance → losses
• crosstalk (EMI)
• limited frequency

Fiber optics

Transmits light inside glass
Why this is effective:

1. No electromagnetic interference
2. Very high frequency (light)
3. Low attenuation

How it works

• light reflects inside the fiber
• almost nothing is lost

this is called:

total internal reflection

Summary

Parameter	Copper	Optical
Speed	medium	very high
Interference	present	none
Range	~100m	km

Why Wi-Fi is unstable

Causes

Interference

• other Wi-Fi networks
• Bluetooth
• microwave ovens

Multipath propagation

The signal:

• reflects off walls
• arrives by different paths

the signals superimpose

Result

• distortion
• packet loss
• speed jumps

Solution

Modern systems use:

• OFDM
• MIMO (multiple antennas)

MIMO (Multiple Input Multiple Output)

The idea

use reflections instead of fighting them

How it works

• several antennas transmit
• several receive
• the signal is "separated" mathematically

Result

• higher speed
• better stability

OFDM — the key to modern networks (Orthogonal Frequency Division Multiplexing)

The problem

one high frequency → sensitive to noise

The solution

split the channel into many small frequencies

What this looks like

instead of:

1 channel → 100 MHz

we do:

100 channels of 1 MHz

Advantages

• resistance to noise
• fighting multipath
• high efficiency

Used in

• Wi-Fi
• 4G
• 5G

Types of delays

Propagation delay

→ the time of signal propagation

t = distance / speed

Transmission delay

→ the time to "push" the data into the channel

t = data size / bandwidth

Processing delay

→ processing by devices

Queuing delay

→ waiting in a queue

Conclusion

"slow internet" is not always the channel, it can be a queue

Jitter

What it is

variation of delay

Why it is bad

• VoIP lags
• video stutters

Causes

• congestion
• an unstable channel

A real example: an HTTP request

We make:

GET /api

At the physical level

1. Bits are encoded
2. Turned into a signal
3. Travel over the medium
4. Get distorted
5. Are received
6. Are decoded

And all of this happens

billions of times per second

The key thought

a transmission channel is not an ideal pipe, but a physical system with limitations

Therefore the entire network is built like this:

1. Physical level → a noisy channel
2. Data link level → corrects errors
3. Transport → guarantees delivery

Units of measurement

It is important to distinguish

Bit/s (bps)

• data transmission speed

Baud

• the number of symbols per second

The difference

If one symbol encodes 3 bits:

• 1 baud = 3 bps

Why this is important

In reality:

speed ≠ transmission frequency

This is directly related to QAM and modulation

The channel is a filter

The physical channel:

does not transmit the signal as is, but distorts it

What it does

• removes high frequencies
• smooths the signal
• adds noise

Consequence

rectangular signals (ideal bits) do not exist in reality

This leads to

• ISI (overlapping of bits)
• errors
• speed limitation

Analogy

The channel = a smeared pipe

What is sent:

|_| |_| |_|

What you receive:

~~~~~~~ ~~~

the task of the system is to restore the bits

Why information theory matters

The meaning:

a channel has a maximum information capacity

This means

• you cannot "cheat physics"
• any protocols are limited by this bound

Practical consequence

• Wi-Fi will not get faster infinitely
• 5G has a limit
• cables too

Equalization

This is real engineering magic that everyone often forgets about.

The problem

the channel distorts the signal

The solution

the receiver "restores" the signal

How

• analyzes the distortions
• compensates for them

Where it is used

• Ethernet
• Wi-Fi
• LTE

Result

we do not only transmit the signal — we "heal" it

The spectrum is a limited resource

Frequencies

• licensed (cellular networks)
• unlicensed (Wi-Fi)

Consequence

• Wi-Fi is overloaded
• LTE is more stable

Practice

Why Wi-Fi slows down:

• many devices
• one band

Satellites

Types

GEO (geostationary)

• high (~36,000 km)
• large latency

LEO (low orbit)

• faster
• a more complex system

The key point

latency is physics, not "bad internet"

The last mile

The problem

how to bring the internet to the user

Options

• ADSL (over the telephone)
• cable (DOCSIS)
• fiber optics
• 4G / 5G

Sharing the channel in real networks

In life

• Wi-Fi → is shared among everyone
• LTE → dynamically distributes resources

Consequence

speed depends not only on the channel, but also on the neighbors

Errors are inevitable

The physical level:

does not guarantee delivery
does not correct errors

But:

it determines the probability of an error

This is called:

BER (Bit Error Rate)

Why it matters:

• affects TCP
• affects retransmissions
• affects latency

Theoretical foundations of transmission

We have already become acquainted with the Shannon and Nyquist theories, but there are also a couple more important things.

Representing a signal as a sum of harmonics

Any signal can be decomposed into sines (Fourier).
Consequence:

• the channel "cuts" part of these harmonics
• the signal gets distorted

The key thought

the shape of the signal depends on the available spectrum

This directly explains:

• ISI
• the speed limitation

The signal spectrum and bandwidth

A rectangular signal (ideal bits)

→ requires infinite bandwidth

Consequence:

ideal digital transmission does not exist

Digital transmission through an analog channel (modems)

The problem

the telephone network is analog

The solution

a modem (modulator-demodulator)

How it works

• computer → digital signal
• modem → converts it to analog
• channel → transmits it
• modem → back to digital

Example

• dial-up internet (old, but conceptually important)

Multiplexing - an addition

WDM (fiber optics)

• different wavelengths (colors of light)
• one cable → dozens of channels

this is the basis of backbone internet

Static vs dynamic

• TDM → fixed slots
• modern networks → dynamic distribution

The telephone network (PSTN)

Architecture

• subscriber line (local loop)
• switches
• trunk lines

PCM (Pulse Code Modulation)

• analog → digital signal
• 8 bits × 8000 times/s = 64 Kbit/s

the standard of the telephone channel

Why it is important to know this

• the historical base of networks
• affects ADSL

ADSL and DSL technologies

The idea

use the existing telephone line

How

• voice → low frequencies
• internet → high frequencies

spectrum division

Consequence

• the internet and the telephone work simultaneously

Cellular networks

We became acquainted earlier with the generations; let's examine the mechanisms.

Cells

• the territory is divided into zones
• each is served by a base station

Handover (transfer of service)

• the user moves
• the connection is "handed over" between towers

Paging

• the network searches for the device

The key idea

the network "watches" you in order to deliver the signal

Cable networks (DOCSIS)

The idea: internet over the TV cable.

Feature:

• a shared channel for all users

Consequence:

in the evening the speed drops (load)

Satellites — differences of orbits

Comparison

Type	Altitude	Latency
GEO	high	large
MEO	medium	medium
LEO	low	low

Practice

• Starlink → LEO
• TV → GEO

Regulation

Frequencies are distributed by states

you cannot simply "take a frequency."

Examples

• Wi-Fi → free bands
• LTE → licensed ones

Consequence

Wi-Fi overload is not a bug, but a consequence of spectrum policy

The main conclusion

The physical level is:

not about networks, but about the limitations of reality

And the rest of the network is an attempt to work around these limitations.