Make your paper interactiveCreate

Text from the interactive explanation

Turbo Trellis-Coded Hierarchical Modulation Assisted Decode-and-Forward Cooperation

Hua Sun, Soon Xin Ng, Lajos Hanzo

Original paper

A radio paper, made playable

The source packs a coarse and a fine message into every symbol: the far receiver catches the coarse one, while a nearby relay catches the fine one and passes it on.

Drag the relay. Turn the source down. Watch what each listener can still make out.

Cluster tightness R = 3.0, the paper’s best setting with a real code.

Chapter 1

Every symbol is a dot

A radio does not send bits one by one. It sends short pulses called symbols , and each symbol is a point on a flat map called a constellation . With four points, each dot names one of four combinations — so every symbol carries 2 bits . The receiver looks at where a dot lands and picks the nearest point.

On the way, noise nudges every dot at random. The signal-to-noise ratio (SNR) compares signal power with noise power, measured in decibels : +3 dB is about twice the power, +10 dB is ten times.

Radio waves also bounce off buildings, so each symbol arrives randomly weaker or stronger: fading . The receiver knows the fading and undoes it — but undoing a deep fade magnifies the noise riding on that symbol too, so those dots fly far off. We draw them faint.

Uncoded decisions on symbols simulated in your browser (rolling window).

Pack 16 dots into the same space — will they need more or less signal to read?

Chapter 2

Two messages in one dot

Hierarchical modulation arranges 16 dots as 4 tight clusters of 4. Which cluster (quadrant) the dot is in carries the coarse layer L1 : 2 bits, robust. Which dot inside the cluster carries the fine layer L2 : 2 more bits, fragile.

The HM ratio R is the gap between clusters divided by the gap inside a cluster. A larger R squeezes each cluster tighter: the coarse layer gets easier to read, the fine layer harder.

Watch the two error meters as you change R and the SNR. Colours show which fine-layer message (b1b0) each dot carried.

Bit labels, laid out like the constellation (b3b2 | b1b0):

Uncoded decisions on symbols simulated in your browser (rolling window).

How loud must each layer arrive?

Raw dots are only half the story. Each layer gets its own rate-1/2 Turbo Trellis-Coded Modulation (TTCM) code: of the 2 bits per layer in every symbol, 1 is information and 1 is a check bit. The decoder iterates 4 times, and errors fall off a cliff once the signal passes a threshold. “Reliable” in the paper means a bit error ratio below 10−6 — one wrong bit in a million.

R = d1/d0, the ratio of the gap between clusters to the gap inside one.

δ1 = 1 / (√2 · (1 + R)), and α = 1 / √(1 + 2δ12) keeps the average symbol energy at 1.

The 16 points are α · [ (±1 ± j)/√2 + δ1 (±1 ± j) ]: the first term picks the quadrant (coarse bits b3b2), the second the dot inside it (fine bits b1b0, the same set-partition layout in every quadrant).

The ideal-code coarse threshold used here is 1.81 dB − 10·log10(α2): the 1.81 dB a capacity-achieving rate-1/2 code needs on a 4-point link, plus the energy the inner offsets take away from the quadrant decision.

Chapter 3

Distance is the expensive part

Radio power spreads out like light from a bulb: double the distance and only a quarter of the power arrives. With this inverse-square law, a relay at fraction p of the way hears the source (1/p)2 times louder than the destination does — and its own hop to the destination is (1/(1−p))2 times stronger than the full path.

Mini challenge: place the relay so it hears the source about 11.7 dB louder than the destination does.

Chapter 4

Two turns to talk

The scheme runs in two time slots and uses decode-and-forward : the relay fully decodes what it hears before passing it on. Follow the bits.

Original redraw of the system in the paper’s Fig. 2. Amber = coarse layer, cyan = fine layer, small grey beads = check bits.

Chapter 5 · the laboratory

Find the valley

Two knobs set the bill: cluster tightness R and the relay position . The source must reach the destination with the coarse layer and the relay with the fine layer; the relay must then reach the destination. Average the two slots’ powers and you get a landscape. For each R, the best spot is the farthest point from the source where the relay can still just hear the fine layer — the red valley line. Can you find its lowest point?

Power budget (bar length = actual power)

A slice at your R: power vs relay position

average per slot source relay best spot (the kink)

Ideal code: thresholds recomputed from the paper’s capacity equations (match Fig. 6); landscape recomputed with the paper’s power formula (compare its Fig. 5).

Chapter 6

What the simulations showed

All results come from calculation or computer simulation (Rayleigh fading, receivers that know the channel, relay on the straight line). Power is the transmit SNR per time slot, with the direct source–destination path loss set to 0 dB.

Power per time slot for reliable transmission

Coarse layer first. The destination decodes L1 right after slot 1; in the authors’ earlier design it had to wait for slot 2 (that design carried 1.5 bit/symbol).

2.85 dB from the bound. The real code’s 3.62 dB per slot is 2.85 dB above the ideal-code lower bound of 0.77 dB.

Code choice matters. Swapping in a rate-1/2 LDPC code with 20 decoder iterations gave slightly worse error rates than TTCM with only 4.

Throughput is the price. Uncoded 16QAM carries 4 bit/symbol; the cooperative scheme 1 bit/symbol per slot — but at about 55 dB less power.

Chapter 7

A link that changes gear

The real-code thresholds also tell the system when to cooperate at all. Slide the signal quality the destination receives from the source and watch the job of the relay change.

Regime boundaries from the paper’s simulations with the real code (4.23 dB and 12.85 dB, Fig. 9(d)).

Back to the whisper

So — can a radio whisper and still be understood?

Yes, if it whispers in layers. The source speaks just loudly enough for the far receiver to get the important coarse message, and a cheap helper standing closer catches the fine detail and passes it on. With a real code, the paper’s simulations put the whole system at 3.62 dB per time slot — 0.61 dB below sending directly with the same code and about 11.3 dB below the authors’ earlier relay design. The important layer arrives after the first slot, and the symbols stay compatible with ordinary 4-point receivers.

What next? The authors suggest that more relays could bring the system closer to the 0.77 dB bound, though at the cost of throughput and extra complexity. They also point to spatial modulation , where a transmitter with several antennas switches on just one at a time, and the choice of antenna itself carries extra bits (log2 Nt for Nt antennas).

About the paper

Turbo Trellis-Coded Hierarchical Modulation Assisted Decode-and-Forward Cooperation

Hua Sun, Soon Xin Ng and Lajos Hanzo · University of Southampton

This page is based on the authors’ manuscript (PDF, southampton.ac.uk). The work was published as: IEEE Transactions on Vehicular Technology, vol. 64, no. 9, pp. 3971–3981, 2015, doi: 10.1109/TVT.2014.2366689.

How this page was made

The live dot clouds are simulated in your browser with the paper’s channel model (Rayleigh fading, receiver knows the channel).

Ideal-code thresholds are recomputed from the paper’s equations and match its Fig. 6.

Real-code thresholds and all real-code results are read from the paper’s simulations (Figs. 9–10).

The 3-D landscapes are recomputed from those thresholds with the paper’s power formula (eqs. 33–40).

All diagrams are original redraws ; no figures from the paper are reproduced.

Topics in this iPaper

Engineering
we propose a single-relay aided hierarchical modulation based cooperative communication system

Abstract, page 1

Computer science
the average mutual information of receiving the output Y due to the input X may be derived

Section IV (DCMC based system analysis), page 5

hierarchical modulation
Our twin-layer model of the HM-16QAM constellation seen in Fig. 3

Section III (Twin layer HM modulation), page 3

cooperative communication
The general model of our TTCHM aided DAF RN based cooperative communication system is depicted in Fig. 1.

Index Terms and Section II, pages 1-2

decode and forward relaying
while the relay invokes the Decode-and-Forward protocol

Abstract, page 1

turbo trellis coded modulation
The source employs a pair of Turbo Trellis-Coded Modulation schemes relying on specially designed hierarchical modulation

Abstract, page 1

channel capacity
The capacity lower bound of our DAF cooperative system is derived based on the Discrete-input Continuous-output Memoryless Channel (DCMC) capacity analysis

Section I contributions, page 2

power allocation
a power allocation plan is provided and it is demonstrated that the power dissipation of the entire system may be readily optimized

Section I contributions, page 2

Turbo Trellis-Coded Hierarchical Modulation Assisted Decode-and-Forward Cooperation — Interactive explanation | iPaper.ai