ELEC5508 Wireless Engineering
The Complete Exam Bible · Semester 2 2026ELEC5508 Overview
- University of Sydney
- Semester 2 2026
- Postgraduate unit
- Two-hour final exam
ELEC5508 Wireless Engineering is a postgraduate University of Sydney unit on the physical-layer ideas that make modern cellular links work.
- Largest assessment Six laboratory projects combine for 48%
- Final format Analytical calculations plus short explanations
- Core habit Classify the channel before selecting a method
How ELEC5508 is assessed
| Component | Weight | Format |
|---|---|---|
| Lab projects | 48% | Six assessable group labs, each worth 8%, with report submissions and oral verification |
| Matlab Wireless On-Ramp | 2% | Individual task requiring 100% completion by 20 August 2026 |
| Mid-Semester Quiz | 10% | Individual one-hour Canvas quiz, open for 24 hours in Week 9 on 9 October 2026 |
| Final Exam | 40% | Individual two-hour essay-type exam with analytical calculations and short explanations |
What ELEC5508 covers
The guide follows the physical-layer dependency chain: classify the propagation channel, protect symbols with diversity and OFDM, manage cellular interference, allocate rate and power, exploit users and antennas, then evaluate AI-assisted receivers.
5G Design Targets and Numerology
Service families, radio KPIs, spectrum bands and OFDM time-frequency choicesThis chapter translates the three 5G service families into measurable radio objectives. It connects eMBB with rate and spectral efficiency, mMTC with connection density and energy efficiency, and URLLC with latency and reliability.
- eMBB, mMTC and URLLC service pressures
- Peak rate, experienced rate and spectral efficiency
- Area capacity, latency and connection density
- Sub-6 GHz and higher-band coverage trade-offs
- Subcarrier spacing and useful symbol time
- Cyclic-prefix overhead and short-slot design
Wireless Propagation and Fading
Multipath, delay spread, Doppler, coherence scales and channel modelsThis chapter classifies a wireless channel along two independent axes. Delay spread and coherence bandwidth determine whether the signal sees flat or frequency-selective fading.
- Multipath addition and destructive fading
- Delay spread and coherence bandwidth
- Doppler shift, spread and coherence time
- Flat versus frequency-selective fading
- Slow versus fast fading
- Rayleigh and Rician amplitude models
Diversity and OFDM
Independent fading dimensions, cyclic prefixes, pilots and equalisationThis chapter separates reliability from equalisation. Diversity creates independently fading chances across time, frequency or antennas, while OFDM reorganises a frequency-selective channel into parallel subcarriers.
- Diversity order and independent branches
- Correlation and covariance rank
- Time, frequency and antenna diversity
- OFDM transforms and orthogonal subcarriers
- Cyclic-prefix sizing and overhead
- Pilot placement and one-tap equalisation
Cellular Reuse and Interference
Frequency reuse, SINR, OFDMA and interference-management choicesThis chapter treats frequency reuse as an explicit trade-off. Giving every cell more spectrum increases degrees of freedom, but closer co-channel transmitters reduce SINR. OFDMA can orthogonalise users within one cell while leaving out-of-cell interference.
- Cellular reuse patterns and co-channel distance
- Universal reuse and fractional reuse
- SINR numerator and denominator
- OFDMA in-cell orthogonality
- Interference averaging and avoidance
- Power, coordination and beamforming controls
Channel Capacity and Water-Filling
AWGN regimes, outage, channel knowledge and adaptive power allocationThis chapter reads Shannon capacity through bandwidth-limited, power-limited and fading regimes. It distinguishes ergodic averaging from slow-fading outage and makes channel knowledge part of every adaptation claim.
- AWGN capacity and logarithmic power gain
- Bandwidth-limited and power-limited regimes
- Ergodic capacity and slow-fading outage
- Receiver and transmitter channel knowledge
- Water-filling across parallel modes
- Channel inversion and fixed-rate trade-offs
Multiuser Diversity and Scheduling
Opportunistic selection, fairness, feedback and spatial opportunityThis chapter turns independent user fading into a scheduling resource. A max-rate scheduler follows the strongest instantaneous channel, while a normalised metric can reward a user whose current state is strong relative to its own average.
- Independent user peaks and selection gain
- Max-rate scheduling
- Normalised opportunity and proportional fairness
- Feedback timing and channel ageing
- Opportunistic beamforming
- Opportunistic nulling in cellular interference
MIMO Channels and SVD
Matrix rank, singular modes, array geometry and spatial multiplexingThis chapter treats the MIMO matrix as the resource, not the antenna count. Singular-value decomposition rotates antenna coordinates into orthogonal channel modes.
- Narrowband matrix model y equals Hx plus n
- Singular-value decomposition
- Rank and spatial degrees of freedom
- Beamforming versus multiplexing
- Channel conditioning and correlation
- Angular resolution and array aperture
MIMO Receivers and Beamforming
Matched filtering, zero forcing, MMSE, SIC and spatial nullsThis chapter compares MIMO detection strategies through the interference-noise trade-off. Matched filtering preserves desired energy but leaves cross-stream interference. Zero forcing cancels interference but amplifies noise along weak singular directions.
- Matched filtering
- Zero-forcing channel inversion
- Linear MMSE filtering
- Successive interference cancellation
- V-BLAST and layer ordering
- Transmit and receive beamforming
AI Signal Processing for Wireless
Learning objectives, neural receivers, training mechanics and robustnessThis chapter begins with the learning objective rather than the network name.
- Supervised, unsupervised and reinforcement learning
- Regression, classification and loss choice
- Forward pass, backpropagation and updates
- Convolution across OFDM time and frequency
- Conventional and learned receiver pipelines
- Model-driven learning and mismatch testing
The 2026 materials begin with 5G service goals and radio performance measures, then develop multipath channels, delay and Doppler spread, diversity, OFDM, cellular interference, Shannon capacity, water-filling, multiuser scheduling and MIMO. The final technical block connects classical channel estimation and symbol detection to neural and model-driven receivers.
This is a mathematics-intensive unit: useful answers join an explicit channel model, a calculation with units and a short explanation of the design consequence. The laboratory projects make the same demand in practical form.
Their marking guides reward correct implementation, organised spectra or BER evidence, and a clear explanation of why the result follows from coherence, cyclic prefix, diversity, equalisation or antenna geometry. The official 2026 assessment totals 100%: six group lab projects combine for 48%, Matlab Wireless On-Ramp is 2%, the Mid-Semester Quiz is 10%, and the two-hour Final Exam is 40%.
The published final description calls it essay-type and says analytical questions require calculations and/or short explanations. It does not publish a question count, section structure or permitted materials, so those operational details should be checked in the live Canvas instructions.
This guide uses original numbers and scenarios, while its methods follow the lecture, tutorial and lab question types in the captured University materials.
Worked example · free
Choose a cyclic prefix from bandwidth and delay spread
- +1Subcarrier spacing is 2.4 MHz divided by 192, which equals 12.5 kHz.
- +1Useful symbol time is the reciprocal of 12.5 kHz, which equals 80 microseconds.
- +1Sample time is 80 microseconds divided by 192, or 0.4167 microseconds.
- +1The prefix needs ceil(18/0.4167) = 44 samples, giving 18.33 microseconds.
- +1Overhead is 18.33 divided by 98.33, or 18.64%; rounding down would fail the delay requirement.
Key terms
- Coherence bandwidth
- The frequency separation over which channel gains remain strongly related; smaller values indicate stronger frequency selectivity.
- Coherence time
- The time scale over which a channel remains strongly related; larger Doppler spread makes it shorter.
- Delay spread
- The span of significant multipath arrival delays that drives frequency selectivity and guard-interval needs.
- Doppler spread
- The range of path frequency shifts caused by motion and arrival geometry, which drives time variation.
- Diversity
- Use of independently fading observations so all branches must be poor together before information is lost.
- Cyclic prefix
- A copied tail placed before an OFDM symbol to protect the useful FFT window from channel memory.
- Frequency reuse
- Repeated use of spectrum in separated cells, trading bandwidth per cell against co-channel interference.
- Water-filling
- Capacity-maximising power allocation that gives more power to better parallel modes and can disable poor ones.
- Multiuser diversity
- Scheduling gain created by selecting among independently varying user channels at favourable times.
- Singular mode
- One orthogonal scalar channel exposed by singular-value decomposition of a MIMO channel matrix.
- Zero forcing
- A receiver or precoder that nulls interference by channel inversion, potentially amplifying noise.
- MMSE
- A filter balancing residual interference and noise to minimise mean squared estimation error.
- Successive cancellation
- Detection that decodes one stream, subtracts it and repeats on the reduced channel.
- Neural receiver
- A learned mapping from received samples or an OFDM grid to channel, symbol or bit estimates.
ELEC5508 FAQ
How is this unit assessed?
The published 2026 structure is six group lab projects worth 48% together, Matlab Wireless On-Ramp worth 2%, a Mid-Semester Quiz worth 10%, and a Final Exam worth 40%. The numeric rows total 100%. Lab marking guides reward implementation, evidence and clear interpretation, while the exam combines calculations with short explanations.
What is the final exam format?
The official information sheet states a two-hour, essay-type Final Exam worth 40%. It describes analytical questions with answers requiring calculations and/or short explanations. The available material does not state a question count, sections or permitted aids, so confirm those operational details in the current Canvas exam instructions.
Which OFDM calculations should I practise?
Practise deriving subcarrier spacing, useful symbol time, sample time, cyclic-prefix samples and overhead from bandwidth, FFT size and delay spread. Then compare the symbol or frame duration with coherence time and the subcarrier spacing with Doppler spread. The explanation of those inequalities is as important as the arithmetic.
Why do students find MIMO difficult?
MIMO requires several ideas at once: matrix rank, singular values, channel knowledge, antenna correlation and receiver design. The reliable approach is to decompose the question. First inspect useful singular modes, then decide whether the objective is beamforming, diversity or multiplexing, and finally choose a receiver whose noise-interference trade-off fits the SNR.
What earns marks in laboratory reports?
The captured marking guides consistently reward a correct implementation, clear presentation, measured or simulated evidence, and an explanation of the underlying mechanism. For OFDM that includes cyclic prefix, synchronisation, frequency correction and equalisation. For MIMO it includes channel classification, BER curves, diversity gains and antenna comparisons.
How should I study the AI receiver topic?
Begin with the conventional pilot, channel-estimation, detection and demapping chain. Then define the learned receiver's input, output target and loss. Compare fully data-driven and model-driven designs, and test on held-out channel conditions using BER or block error rate. A low training loss without communication evidence is not enough.
Questions students ask in each ELEC5508 chapter
- Why does 5G use more than one subcarrier spacing?
- Different services and carrier bands impose different timing, phase-noise and delay-spread pressures. Wider spacing shortens symbols and slots but makes a fixed cyclic prefix consume a larger fraction. Narrow spacing improves that efficiency but lengthens the symbol, so the best choice depends on propagation and latency. Chapter 01 · 5G Design Targets and Numerology
- Can a channel be slow and frequency selective?
- Yes. Slow or fast describes time variation relative to a signalling interval, while flat or selective describes variation across frequency. A wideband link can see different gains across subcarriers yet remain nearly constant for many OFDM symbols. Chapter 02 · Wireless Propagation and Fading
- Does every extra antenna add diversity?
- No. Diversity depends on independent channel dimensions, not hardware count alone. Strong correlation can reduce covariance rank, so several antennas may behave like fewer independent branches. Antenna placement and propagation geometry determine the effective gain. Chapter 03 · Diversity and OFDM
- Why can universal reuse improve capacity?
- It gives each cell the full available band, increasing degrees of freedom. The benefit survives only if interference management keeps SINR from falling too far. Cell-edge users may lose even when the cell-average throughput rises. Chapter 04 · Cellular Reuse and Interference
- Why does doubling power not double capacity?
- The AWGN capacity expression is logarithmic in SNR when bandwidth is fixed. Additional power therefore has diminishing rate return at high SNR. At low SNR or in the power-limited wideband regime, the relationship is closer to linear. Chapter 05 · Channel Capacity and Water-Filling
- Why does multiuser diversity improve with more users?
- With independent channel variation, more eligible users increase the chance that at least one has a strong instantaneous state. The gain depends on timely feedback and distinct fading; identical or stale reports erase the selection opportunity. Chapter 06 · Multiuser Diversity and Scheduling
- Does a four-antenna link support four streams?
- Only if the channel matrix has four useful independent modes and the receiver has enough dimensions. Correlation or limited angular spread can reduce rank, leaving fewer streams even though the hardware count remains four. Chapter 07 · MIMO Channels and SVD
- Why can zero forcing perform poorly?
- When H is poorly conditioned, inversion divides by a weak singular value and greatly amplifies noise. Zero forcing removes inter-stream interference, but that benefit can be outweighed at low or moderate SNR by the amplified noise. Chapter 08 · MIMO Receivers and Beamforming
- Why are CNNs suitable for OFDM receivers?
- An OFDM grid has local correlation across neighbouring subcarriers and symbols because of coherence bandwidth and coherence time. Two-dimensional convolution exploits that structure with shared filters instead of flattening every observation into an unrelated feature. Chapter 09 · AI Signal Processing for Wireless
How to study for the exam
Build revision around a single dependency map rather than isolated formulas. Start every problem by classifying the channel in frequency and time. Put delay spread, coherence bandwidth, Doppler spread and coherence time on one page, followed by the OFDM inequalities and cyclic-prefix calculation. Add a capacity panel that states the channel-information assumption before water-filling or beamforming.
For MIMO, sketch H = UΣV* and annotate rank, singular-mode strength, diversity and spatial multiplexing as distinct ideas. Rework tutorial calculation types with fresh numbers, then explain each answer aloud because the lab oral verification rewards causal reasoning. For laboratory preparation, predict the spectrum, constellation or BER change before running the model and compare observation with that prediction.
Finish with the AI topic only after the conventional receiver is clear: input grid, pilots, channel estimate, detector, bit likelihood, target and loss. In the last revision pass, check units, decibel conversions, active water-filling modes, cyclic-prefix coverage and whether any adaptation claim silently assumes transmitter channel knowledge.
Your AI Electrical Engineering tutor for ELEC5508
Stuck on a hard ELEC5508 question? Sia is AskSia’s AI Electrical Engineering tutor — ask any ELEC5508 Wireless Engineering question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.