Wireless 101
M11 · L03
Module 11 · Lesson 3

Three Customers, One Radio

Every generation before it was designed for one service, so it could freeze its parameters — LTE fixed 15 kHz spacing for every band and every speed (M10-L2). NR was specified against three services whose demands contradict each other.

01 / 11
Wireless 101
M11 · L03
The three corners (ITU-R M.2410)

They Pull Opposite Ways

  • eMBB — 20 Gbit/s peak. Wants the widest channel and 256-QAM
  • URLLC — 1 ms and 99.999%. Wants the shortest transmission and spare margin
  • mMTC — 10⁶ devices/km² = one per square metre. Wants a narrow channel and sleep
  • 99.999% is 1 in 10⁵: at 1000 packets/s, one failure per 100 s
  • LTE’s HARQ round trip is 8 ms — eight budgets, for one retry (M6-L4)

No fixed parameter set is best at all three. So NR was made adjustable instead.

02 / 11
Wireless 101
M11 · L03
Flexible numerology

Double the Spacing, Halve the Slot

A slot is 14 symbols at every spacing
\Delta f = 15\cdot 2^{k}\ \text{kHz} \;\Rightarrow\; T_{slot} = \frac{14}{\Delta f} \approx \frac{1\ \text{ms}}{2^{k}}

15, 30, 60, 120 kHz → slots of 1, 0.5, 0.25, 0.125 ms. The resource block stays 12 subcarriers (M10-L2), so at 120 kHz it is 12 × 120 = 1.44 MHz wide.

03 / 11
Wireless 101
M11 · L03
The catch, priced

Latency Costs Prefix Reach

  • Overhead never moves: 4.69/71.36 = 6.6% at every spacing (M10-L2)
  • But the prefix duration halves: 4.69 → 0.59 µs
  • At 300 m per µs, reach falls 1407 m → 176 m
  • Doppler at 28 GHz, 100 km/h: 27.78/0.0107 = 2593 Hz
  • That is 17.3% of 15 kHz — but 2.2% of 120 kHz ✓

28/3.5 = 120/15 = 8, so scaling spacing with frequency holds the Doppler fraction constant. Short slots suit small cells and high bands, and nothing else.

04 / 11
Wireless 101
M11 · L03
Sub-6 versus millimetre wave

Bandwidth Beats SNR

Shannon, both cases (M6-L1)
\begin{aligned} 100\,\text{MHz} @ 20\,\text{dB} &= 10^{8}\log_2(101) = 666\ \text{Mbit/s} \\ 400\,\text{MHz} @ 10\,\text{dB} &= 4{\times}10^{8}\log_2(11) = 1.38\ \text{Gbit/s} \end{aligned}

1384/666 = 2.08× the capacity on 4× the bandwidth at 10 dB worse SNR. Matching 1.38 Gbit/s inside 100 MHz would need 41.7 dB — which no mobile link has.

05 / 11
Wireless 101
M11 · L03
Try it — no setting wins everything

Pick a Numerology

Spacing 30 kHz
Bandwidth 100 MHz
SNR 20 dB
slot 0.500 ms symbol 35.67 µs CP 2.34 µs = 6.6% RB 360 kHz C 666 Mbit/s CP reach 703 m balanced — wide-area eMBB
06 / 11
Wireless 101
M11 · L03
Why millimetre wave works at all

The Penalty Is the Cure

Fixed aperture, not fixed gain (M8-L1)
10\log_{10}64 = 20\log_{10}8 = 18.1\ \text{dB}
3.5 → 28 GHz
+18.1 dB
64 elements
−18.1 dB
Panel at 28 GHz
4.3 cm

λ = 10.7 mm, so 8 × 8 at λ/2 spans 7 × 5.36 = 37.5 mm. It does not cure blockage: a body still costs 20–40 dB (M8-L2).

07 / 11
Wireless 101
M11 · L03
gNodeB and the 5G core

The Radio Is Not the Budget

Fibre carries 200 m per microsecond
\frac{2\times 150\ \text{km}}{2\times10^{8}\ \text{m/s}} = 1.5\ \text{ms}

gNB splits into CU, DU and RU; the 5GC is services — AMF control, SMF sessions, UPF your packets. A gateway 150 km away burns 1.5 ms of fibre alone; an edge UPF 10 km away costs 100 µs.

08 / 11
Wireless 101
M11 · L03
Slicing, and the icon that lied

Standalone or Not

  • Slicing — separate logical networks, each with its own SLA, on one infrastructure
  • Honest limit: it is orchestration, not spectrum — one cell, one scheduler still
  • NSA anchors the control plane on an LTE base station and the LTE core
  • So early phones showed 5G while signalling over 4G — the throughput was real
  • SA is gNB plus 5GC end to end — and slicing and URLLC need it

Every non-radio feature in this lesson requires standalone. That migration matters more than any peak rate.

09 / 11
Wireless 101
Knowledge Check

Check whatstuck

Four questions on 5G NR — the three corners, flexible numerology, mmWave and the core.

Question 1 of 0
Score 0/0

10 / 11
Wireless 101
M11 · L03
Recap

What you learned

  • Three corners contradict, so Δf = 15·2k kHz and slots are 1 → 0.125 ms
  • CP overhead stays 6.6%; CP reach falls 1407 m → 176 m
  • 100 MHz at 20 dB = 666 Mbit/s; 400 MHz at 10 dB = 1.38 Gbit/s
  • +18.1 dB of path loss, −18.1 dB from 64 elements — the same 20 log₁₀8
  • 20 Gbit/s needs 800 MHz, 4 layers, 256-QAM and line of sight at once
Next in Module 11
11 / 11