COMP3310/etc – #T2
Media: Copper
Dr Markus Buchhorn: markus.buchhorn@anu.edu.au
Why does it matter?
• Networks rely on analogue signals over/through physical media
• Many failure modes – and security challenges – are in the physical media
• Many weird behaviours in the physical media
– And they vary by physical media
– Dodgy CPUs, GPUs, RAM, HDD, etc. are fairly obvious – dodgy links are not.
• Time dependant, weather dependant, wildlife dependant, …
• Good network design starts from physical media options
– Costs (manufacture: materials and complexity/labour) – capital expenditure
– Maintenance, power – operating expenditure
– Benefits, risks
2
Putting those signals onto media
• Why wire, esp. copper?
– Ubiquity, great conductor, cheap to work with, flexible, lots of experience
– Can also be aluminium, steel, ceramics, wet string, …
• Used for electricity, so communications are easy
– Data over power lines (power-line communications [PLC] …)
• From ISP to homes, from devices to recorders
– Low data rates (<1Mb/s) but long distances (km)
– Developed in 1910’s
• Within a building
– High data rates (100Mb/s) but short distances (10’s m)
– Data over fence lines, data over train tracks, …
3
Characteristics of copper
• Light, malleable, reasonably robust to
oxidation – easy to connect
• Easy to make thin wires
– Cross-section 0.2mm2 solid, 0.02mm2 stranded
– = AWG 24 to AWG 33 – and lower
• Cheap? per kilo
• Tight bending radius
• Easy to add insulation and protection
– cloth to rubber to plastic
– And metal...
4
A shared medium
• One voltage over the whole wire(*eventually)
• Can (only) receive (RX) or transmit (TX) at a time?
– Time or Frequency Division Multiplexing
– TDM = Half-duplex = take turns
– FDM = Electronics at each end to separate them
• Need a “ common reference” or “return”
– Where’s zero?
– So cables tend to have a “pair” of wires:
RX/TX and Ground, or Reference,
or Common, or Earth, or Shield, or Return, or …
• Some early networks: single cable and ‘vampire taps’
– One tooth into each of the data & ground wires…
5
Electrical characteristics of copper
• Resistance
– Actually impedance (resistance, inductance, capacitance) – and others
– Things that hate electrical signals going through metal
• Some are frequency/change dependent, some are not
• Attenuation of a signal (aka “insertion loss”) as heat, from resistance
– And others
6
• Resistance varies with length and cross-section
• Hence stranded cables aren’t “as good” – effectively 5-10% thinner
• You don’t want your copper too long, or too thin – unless you do…
AWG Diameter (mm) Area (mm2) Resistance (Ω/km)
4 5 21 0.8
14 1.6 2 8.3
24 0.5 0.2 84
34 0.16 0.02 856
Resistance
7
#justTheVibeOfThis
Resistance(f)
• Resistance varies with frequency
Skin-effect – changes cross-section
(6MHz, skin ~ 25μm)
Leads to Frequency attenuation
8
Wire end-on
Wire end-on
at low frequency
Wire end-on
at high frequency
Noise
• A straight wire is a great antenna
– Can act as a receiver of other signals
– Can act as a transmitter of its own signals
– Electro-Magnetic Interference (EMI), Radio-Frequency Interference (RFI)
• Electrical “coupling” from adjacent wires
– Leads to “cross-talk” (=“XT”)
• Near end and Far end (NEXT, FEXT) mainly, but really all along the wire
• Impact goes up with distance
9
Jacket
Shielding
Dielectric
So how do we get around this?
• Start with Good Cable Design!
• 1st wire in middle, wrap it with the second
– “Coaxial” cables
– Upside: Well shielded, robust
– Downsides: single RX/TX – and pretty expensive
• Cables called ‘RG-nnn’ (RG-6, RG-58, …)
– (No numbering pattern in their characteristics)
– Varying attenuation, diameter, thickness, materials,
solid/stranded, …
10
Conductor
(wire)
Another approach
• Bring in Spatial Division Multiplexing
– Aka – more wires in the cable
• USB, ATA, HDMI, Ribbon cables, …
• Upside:
– Full duplex (RX, TX separately)
– Multiple paths to share =
Inverse multiplexing
(one-to-many, to-one)
• Downside:
– Long, straight unshielded wires =
antennas
– Lots of adjacent wires = crosstalk
11
That antenna problem
• What’s the problem?
– An antenna picks up/radiates every signal,
– Unless well shielded
– Shielding = materials = costs
• Can we avoid shielding? Yes, we can!
• First: Differential signalling
• Second: A noise source has a direction
– Impacts one wire more than the other
– Unless we add in Twisting
12
1
2
2
1
UTP Cables
• Unshielded Twisted Pairs
• Comes in multiple Categories (“Cat1”-Cat8)
– STANDARDISED: ISO/IEC 11801 (ed.3 2017) and TIA
– And some not, industry makes stuff up – sigh.
• Standardised Connectors
– A religious discussion with misused standards: RJ45
is common name (8P8C, GG45, …)
• “Structured Cabling”
– Overarching design concept for a building
– Run multiple functions over the same cables,
colour-coded
13
Structured cabling…
14What you want… What you get…
UTP Cables
• Standardised wiring within the plug
– STRICT rules about terminating cables
• Amount of insulation removed (XT)
• Amount of untwisting (Noise)
• Plenty of great youtube videos
• (Actually, if one standard is good, two (or more) are better!
– TIA/EIA-568 A and B – and others
– So not totally standardised. )
• And … bring back some shielding??
15
UTP -> SF/FTP ??
• (U/)UTP:
• Add Foil: U/FTP and F/UTP
• Add Braided Shielding: S – only on outside
• Up to SF/FTP…
– Nearly every combination is sold somewhere
• Take home messages:
– More F and S is great for performance
– More F and S is a pain for making cables.
• Makes them really expensive and stiff, too
16
UTP cables by category
Category Cable Bandwidth Data rates
3 UTP 16MHz 10-100Mb/s
4 UTP 20MHz
5 UTP 100MHz 100-1000Mb/s
5e UTP 100MHz
6 UTP 250MHz 10Gb/s
6A UTP, F/UTP, U/FTP 500MHz
7 S/FTP, F/FTP 600MHz 10Gb/s + stuff
7A S/FTP, F/FTP 1000MHz
8/8.1 F/UTP, U/FTP 1600-2000MHz 40Gb/s
8.2 S/FTP, F/FTP 1600-2000MHz
17
Do not memorise, just get the taste of it
Time
– Speed of electricity in copper is not infinite = ~ 2/3 c = 0.3 μsec/100m
• Propagation delay – takes time for everyone to get the signal
• Which leads to collisions
– @1Gb/s => 100bits
• And leads to skew across multiple pairs
– UTP: 2+ pairs have to be the same length, within tolerance
– Reflections
• “Impedance mismatch”
• Send a signal, it comes back to you, later
– Used on landline phones for positive feedback!
– Otherwise it’s a real nuisance.
18
Building: Capital and Operating costs
• Everything costs (materials, equipment, labour)
• Capital and operational costs
– Buying the stuff
– Putting it in
– Taking care of it
• What happens?
– Trenching ($-$$$$), then ducting/pulling, or direct burial
• But metal rusts, junctions leak, ducts collapse, insulation perishes
• Insects/rodent get hungry, excavators get sloppy
– Aerial cables?
• Antennas, lightning, birds/mammals, flying excavators, cable thieves
19
The great outdoors…
20
• Easy to make a shared backbone network – you only need one (or two).
• From the backbone (exchange), get to every house…
• Trade-off costs vs performance:
– How much does an exchange cost
– How long is that final cable run
– What’s the cost to make and deploy
– What can you get through it
• Quick scaling calculation:
– 10M houses/offices in Australia,
– Up to ~4km from their exchange… (can’t go much further!)
The Last Mile
21
Exchange Exchange
What’s it like?
22
Putting Bits on Copper Cables: DSL
• Digital Subscriber Line
– Based on PSTN cables = 1 pair of wires to your home, but “digital”
• Telco standards = International Telecommunications Union
– ITU-T (with names like G.992, H.264, …)
– Many flavours, mostly asymmetric
23
DSL ITU Down Up
ADSL G.992.1 10Mb/s 1Mb/s
ADSL 2 G.992.3 12Mb/s 3.5Mb/s
ADSL 2+ G.992.5 24Mb/s 3.5Mb/s
VDSL G.993.1 52Mb/s 16Mb/s
VDSL 2 G.993.2 200Mb/s down+up
S(H)DSL G.991.2 5.7Mb/s per pair, up to four
When living
next to the
exchange!
Asymmetric?
• Commercially:
– Consumption is >>> production
– Except for telecommuters, home backup, healthcare, aged care, lockdowns, …
– Can be tuned…
• Technically:
– Wire to your house: largely alone
– Wire into the exchange: lots of adjacent wires
• Crosstalk
– Exchange can send strong signal down
• Homes cannot = weaker signal + noisy neighbours
24
DSL+
• DSL based on
– 65536 QAM with FDM
– With increasing number of overlaid features
– G.inp = reduce impact of external noise, error-correction (profile 17a)
– G.vector = reduce impact of crosstalk = 100Mb/s at 500m
– G.fast = 1Gb/s to 100Mb/s (down+up) at 500m
– XG.fast = 10Gb/s at 130m
• Need a DSL modem at home, maybe a filter for the phone
• Need a DSL Access Multiplexer (DSLAM) at the exchange/node
25
The magic inside DSL
26
• Multiple (4kHz) frequency carriers – some up, some down
• Each carrier can take 2-16bits/symbol
• Each version of DSL adds more channels = more bandwidth
The NBN today – VDSL2
27
Frequency Attenuation
28
What’s going on here?
29
Downstream
Data rate?
30
41.4Mb/s 36.2 Mb/s 19.0 Mb/s
• Each channel carries N bits per symbol per Hz
• Each channel is ~4.3kHz
• So datarate = #channels * N * 4.3kHz
Sum = 96.6Mb/s
Actual = 94.3Mb/s
Moving on: Some (copper) last-mile alternatives
• Power-line carrier:
– Note earlier performance limitations in wide area.
– Shared medium; transformers block it, not every circuit in a house is joined up,
maybe seen by your neighbour, can be a transmitter, can be an antenna
• Fences, Train tracks, wet-string, …
• Mixed-Technology approaches
– NBN Fibre-to-the-…
– Hybrid Fibre Coax [HFC, cable networks]
• Coax cable, shared medium, …
– After we’ve discussed optical fibre…