⚡ electrical engineering

A Comparative study on THz Communication Systems: Photonics versus Electronics Approaches

This paper provides a comprehensive comparative analysis of electronics- and photonics-based Terahertz communication systems for 6G networks by reviewing experimental demonstrations, modeling hardware impairments, and deriving analytical expressions to evaluate and contrast their signal-to-noise ratio and bit error rate performance.

Original authors: Talha Rahman, Murat Uysal

Published 2026-04-30

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Talha Rahman, Murat Uysal

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a super-fast highway for data, one that can carry the massive amount of information needed for things like holographic video calls and immersive virtual reality. This highway exists in a part of the spectrum called the Terahertz (THz) band. It's a "super-highway" because it has a massive amount of space (bandwidth) to carry data, but it's also a very difficult terrain to navigate.

This paper compares two different ways to build the cars and engines that will drive on this highway: the Electronics Approach and the Photonics Approach.

The Two Approaches: Two Different Engines

Think of the THz signal as a very high-pitched whistle. To create this whistle, you need a source.

The Electronics Approach (The Gearbox):
- How it works: Imagine you have a steady, low-pitched engine (a standard radio oscillator). To get the high-pitched THz whistle, you put this engine through a series of gears (frequency multipliers) that speed it up.
- The Problem: Just like a bicycle chain gets noisy and wobbly when you shift gears too fast, this "gearbox" amplifies the engine's natural imperfections. The paper finds that the noise floor (the background static of the engine) gets magnified significantly. If the base engine isn't perfect, the final high-speed signal is full of static.
- The Analogy: It's like trying to whisper a secret through a long, bumpy tunnel. The louder you try to speak (increase signal power), the more the tunnel walls shake and distort your voice, eventually making it impossible to understand.
The Photonics Approach (The Laser Duo):
- How it works: Instead of gears, this method uses light. Imagine two very precise laser beams (like two different colored flashlights). When you mix them together in a special crystal (a photomixer), they "beat" against each other, creating a new, high-pitched sound (the THz signal) based on the difference between the two laser colors.
- The Problem: Lasers aren't perfectly steady either. They have "jitter" (phase noise) and "flicker" (intensity noise). Also, the amplifiers used to boost the light add their own "hiss" (Amplified Spontaneous Emission).
- The Analogy: This is like trying to create a perfect rhythm by clapping two slightly shaky hands together. If your hands shake too much (laser noise), the rhythm gets messy.

The Big Discovery: "Signal-Dependent Noise"

The most important finding in the paper is a concept called Signal-Dependent Noise.

In normal radio (like your Wi-Fi): The background noise is like a constant rain. Whether you are whispering or shouting, the rain falls at the same rate. If you shout louder, you drown out the rain.
In THz systems (both types): The noise is like a shaking table. The harder you push on the table (the stronger your signal), the more violently it shakes.
- Electronics: The shaking comes from the amplified noise of the oscillator gears.
- Photonics: The shaking comes from the interaction of the lasers and the mixing process.

Why this matters: In these systems, simply turning up the volume (increasing power) doesn't help forever. Eventually, the shaking gets so bad that it drowns out your message, no matter how loud you get. This creates a "ceiling" or a performance floor where the error rate stops getting better, no matter how much power you add.

The Battle of the Modulations (QAM)

The paper tested different ways of packing data into these signals, called QAM (Quadrature Amplitude Modulation). Think of QAM as different patterns of dots on a map.

Simple patterns (QPSK, 16QAM): These are like large, widely spaced dots. They are very tough. Even if the table shakes a bit, you can still tell which dot you are on. Both Electronics and Photonics handle these well.
Complex patterns (64QAM, 256QAM): These are like tiny, tightly packed dots. They are very efficient but very fragile.
- The Result: When the "shaking table" (noise) gets too strong, the tiny dots blur together. The paper shows that for these complex patterns, the Electronics approach suffers more from the shaking than the Photonics approach. The noise in electronics scales up more aggressively with the signal strength.

The Role of the "Digital Brain" (DSP)

Since the hardware (lasers and oscillators) is imperfect, the paper emphasizes the need for a powerful Digital Signal Processor (DSP)—the "brain" of the receiver.

Frequency Offset: The lasers or oscillators might drift slightly, like a runner starting a race a few steps ahead of the line. The brain has to calculate this drift and correct it.
Phase Noise: The signal might wobble in its timing. The brain uses algorithms (like a Phase-Locked Loop) to track this wobble and smooth it out.
The Catch: The brain can fix the slow, predictable wobbling (correlated noise), but it cannot fix the random, sudden static (white noise floor). If the hardware noise is too high, even the smartest brain can't save the signal.

Summary of the Comparison

Feature	Electronics Approach (The Gearbox)	Photonics Approach (The Laser Duo)
Main Weakness	The "gearbox" amplifies the base oscillator's noise floor.	Lasers have "jitter" (phase noise) and amplifiers add "hiss" (ASE).
Noise Behavior	Noise gets worse as signal power increases (Signal-Dependent).	Noise gets worse as signal power increases (Signal-Dependent).
Who Wins?	Good for simple signals, but struggles with complex, high-speed data due to oscillator noise.	Generally handles complex, high-speed data better, provided the lasers are high-quality.
The Limit	You hit a wall where more power just creates more noise.	You hit a wall where laser imperfections create a noise floor.

The Bottom Line

The paper concludes that while both approaches are necessary for the future of 6G, they have different "Achilles' heels."

If you use Electronics, you need incredibly quiet, high-quality base oscillators, or the noise will ruin your high-speed data.
If you use Photonics, you need very stable, low-noise lasers and careful management of optical amplifiers.

The paper doesn't say one is strictly "better" for all situations, but it provides a mathematical map showing exactly where and why each system fails. This helps engineers design better "brains" (DSP algorithms) and choose the right "engines" (lasers or oscillators) to build the 6G super-highway.

1. Problem Statement

Terahertz (THz) communication (300 GHz – 3 THz) is a critical enabler for 6G networks, promising terabit-per-second data rates for applications like holographic telepresence and immersive XR. However, the "THz gap" historically lacked efficient devices. While recent advances have enabled two primary generation methods—electronics-based (using frequency multipliers) and photonics-based (using photomixing)—both suffer from severe hardware impairments that degrade performance.

Existing literature often treats these impairments in isolation or assumes a simple Additive White Gaussian Noise (AWGN) model, which fails to capture the complex, signal-dependent noise characteristics unique to THz systems. There is a lack of a unified framework that:

Compares the two architectures side-by-side.
Models specific hardware impairments (e.g., laser Relative Intensity Noise vs. oscillator phase noise floors) and their propagation through the transceiver chain.
Derives analytical Signal-to-Noise Ratio (SNR) expressions that account for these non-linear and signal-dependent noise sources.

2. Methodology

The authors developed a comprehensive analytical framework involving signal modeling, noise characterization, and performance evaluation:

Signal Modeling:
- Photonics-based: Modeled the generation of THz signals via photomixing of two optical lasers. The model accounts for optical beat noise, laser Relative Intensity Noise (RIN), Amplified Spontaneous Emission (ASE) from optical amplifiers, and shot noise.
- Electronics-based: Modeled the generation via frequency multiplication of a base oscillator. The model accounts for the amplification of the oscillator's white noise floor and phase noise by the multiplication factor ( $N$ ), as well as thermal noise in mixers.
Noise Characterization:
- The authors distinguished between Signal-Independent Noise (SIN) (e.g., thermal noise) and Signal-Dependent Noise (SDN).
- They demonstrated that in both architectures, the aggregate noise is heteroscedastic (variance depends on the signal amplitude), violating the standard AWGN assumption.
- Photonics: SDN arises from the beating of optical noise with the signal (RIN/ASE) and shot noise.
- Electronics: SDN arises from the multiplication of the oscillator's white noise floor with the signal.
Analytical Derivation:
- Derived closed-form analytical expressions for SNR at both the transmitter and receiver for both architectures.
- Explicitly tracked how noise sources (RIN, ASE, shot noise, oscillator noise floor) propagate and scale through the system.
Simulation & Validation:
- Conducted Monte Carlo simulations for various square QAM formats (QPSK to 256QAM).
- Evaluated Bit Error Rate (BER) performance under realistic parameters (e.g., laser linewidths, oscillator noise floors, I/Q imbalance).
- Validated the analytical models against experimental data from existing literature (e.g., [26], [36]).

3. Key Contributions

Unified Signal Models: Developed comprehensive signal and noise models for both photonics and electronics approaches that explicitly incorporate implementation-specific hardware impairments, moving beyond the simplistic AWGN assumption.
Analytical SNR Expressions: Derived exact analytical expressions for transmitter and receiver SNR, categorizing noise into signal-independent and signal-dependent components to explain performance limits.
Comparative Performance Analysis: Provided a direct comparison showing that while both systems suffer from signal-dependent noise, the electronics-based approach is significantly more sensitive to oscillator noise floors, leading to a steeper degradation in high-order modulation performance.
Impact of Impairments: Quantified the impact of:
- Phase Noise: Modeled as a Wiener process for lasers and a multi-slope process for oscillators.
- I/Q Imbalance: Analyzed its effect on error floors for high-order QAM.
- Noise Scaling: Demonstrated that noise variance scales with symbol power, causing error floors in high-SNR regimes.

4. Key Results

Noise Characteristics:
- In Photonics, the dominant impairments at the transmitter are RIN and ASE. At the receiver, the noise is a mix of thermal, shot, and optical noise. The noise variance scales with the symbol power, but the effect is moderate compared to electronics.
- In Electronics, the oscillator noise floor is the critical bottleneck. Frequency multiplication amplifies this noise by $N^2$ , making it the dominant source of Signal-Dependent Noise (SDN).
BER Performance:
- Low-Order Modulation (QPSK/16QAM): Both systems perform close to the theoretical AWGN limit, as thermal noise dominates.
- High-Order Modulation (64QAM/256QAM): Both systems exhibit error floors at high received power due to signal-dependent noise.
- Comparison: The electronics-based system shows a much larger penalty for high-order QAM. For 256QAM, the electronics approach suffers a ~5 dB penalty compared to AWGN, whereas the photonics approach suffers ~2 dB.
Phase Noise Sensitivity:
- Photonics: Requires low-linewidth lasers (e.g., <100 kHz) or optical frequency combs to mitigate phase noise accumulation.
- Electronics: While PLLs can track near-carrier phase noise, the white noise floor cannot be compensated. To match the performance of a 100 kHz linewidth laser system, an electronics-based system requires a base oscillator noise floor of $\le -166$ dBc/Hz (after multiplication), which is extremely challenging to achieve.
I/Q Imbalance: Higher-order modulations (64QAM, 256QAM) are highly sensitive to I/Q imbalance in both systems, creating error floors that cannot be overcome by increasing transmit power.

5. Significance and Implications

Design Guidelines: The paper provides a systematic basis for designing robust 6G THz transceivers. It highlights that simply increasing transmit power yields diminishing returns due to signal-dependent noise; instead, improving component quality (low-RIN lasers, ultra-low-noise oscillators) is paramount.
Architecture Selection:
- Photonics offers superior spectral purity and tunability but faces challenges in power efficiency and component cost. It is better suited for scenarios requiring very high spectral efficiency with high-order QAM, provided low-noise lasers are used.
- Electronics offers better integration potential (CMOS/InP) and higher output power but is severely limited by oscillator noise floors. It may be more suitable for lower-order modulations or shorter-range links where thermal noise dominates.
DSP Requirements: The study underscores the necessity of advanced Digital Signal Processing (DSP) for carrier recovery, phase noise mitigation, and I/Q imbalance compensation. However, it also notes the trade-off between DSP complexity and power consumption.
Future Research: The models can be extended to hybrid architectures (photonic TX + electronic RX) and multi-carrier waveforms (OFDM), though the latter faces additional challenges with phase noise and Peak-to-Average Power Ratio (PAPR).

In conclusion, this work bridges the gap between theoretical modeling and practical hardware constraints in THz communications, proving that hardware impairments are the fundamental limiters of 6G THz performance and that the choice between photonics and electronics depends heavily on the required modulation order and the availability of low-noise components.