🔬 optics

Experimental Analysis of a Self-Coherent M-QAM Receiver by Means of Recurrent Optical Spectrum Slicing and Direct Detection

This paper experimentally demonstrates a self-coherent M-QAM receiver utilizing recurrent optical spectrum slicing and direct detection to achieve high spectral efficiency with minimal DSP and low driving voltages, resulting in a projected over 40% power consumption reduction for next-generation 1.6T pluggables compared to state-of-the-art coherent solutions.

Original authors: Kostas Sozos, Francesco Da Ros, Senior Member Optica, Metodi Yankov, Stavros Deligiannidis, George Sarantoglou, Charis Mesaritakis, Adonis Bogris, Fellow Optica

Published 2026-04-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Kostas Sozos, Francesco Da Ros, Senior Member Optica, Metodi Yankov, Stavros Deligiannidis, George Sarantoglou, Charis Mesaritakis, Adonis Bogris, Fellow Optica

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 send a complex message across a noisy, bumpy road. In the world of fiber-optic internet, this "road" is a strand of glass (fiber optic cable), and the "message" is a stream of data encoded in light.

For a long time, there have been two main ways to send this message:

The "Simple" Way (IM/DD): Like shouting a message. You turn the light on and off (or make it brighter and dimmer). It's cheap and uses little energy, but it can't carry very complex or high-speed messages over long distances without getting garbled.
The "High-Tech" Way (Coherent): Like whispering a secret while also changing the pitch and timing of your voice. This allows for massive amounts of data, but it requires a super-complex receiver (like a supercomputer) to decode the message. This consumes a lot of electricity and generates a lot of heat, which is a problem for the small, plug-in boxes (pluggables) used in data centers.

The Problem: We want the high speed of the "High-Tech" way but the low energy cost of the "Simple" way. Current attempts to mix them (like "Self-Coherent" methods) often still require too much digital processing power or expensive hardware.

The Solution: The "Recurrent Optical Spectrum Slicing" (ROSS) Receiver

This paper introduces a clever new trick to get the best of both worlds. Here is how it works, using some everyday analogies:

1. The "Squeezed" Message (The Transmitter)

Usually, to send complex data, you need to swing your light signal back and forth wildly (like a pendulum). This requires a lot of energy.

The Innovation: The researchers decided to "squeeze" the message. Instead of swinging the light all the way, they kept the signal in just one corner of the room (the first quadrant) and used a very gentle push (low voltage).
The Analogy: Imagine trying to send a message by waving a flag. Usually, you wave it wildly from left to right. Here, they just wiggle the flag slightly in the top-right corner. It's much easier to do (less energy), but because the wiggle is so small, the message gets blurry and hard to read at the other end.

2. The "Magic Prism" (The ROSS Receiver)

This is where the magic happens. Since the message is blurry and small, a normal camera (photodetector) can't read it. They need a special tool to "sharpen" the image before the camera sees it.

The Innovation: They use a device called a Recurrent Optical Spectrum Slicer (ROSS). Think of this as a smart, multi-layered prism or a musical equalizer for light.
How it works:
- The blurry light signal enters the prism.
- The prism splits the light into different "colors" (frequencies) and passes them through a loop (recurrent) that twists and turns them slightly.
- The Key Trick: This twisting process converts the phase (the timing/angle of the light wave, which is hard to see) into amplitude (brightness, which is easy to see).
- The Analogy: Imagine the message is written in invisible ink (phase). The ROSS prism is like a special UV lamp that doesn't just reveal the ink, but also rearranges the letters so they stand up tall and clear (amplitude) before the camera takes a picture.

3. The "Simple Camera" (Direct Detection)

Because the ROSS prism did all the heavy lifting to make the message clear and bright, the receiver doesn't need a super-computer. It just needs a simple, cheap camera (a standard photodetector) to take a picture of the now-clear message.

The Result: They successfully sent high-speed data (32 Gbaud) over 50km of fiber using this method. They even used a technique called "Geometric Constellation Shaping" (like rearranging the furniture in a room to fit more people comfortably) to make the signal even stronger against noise.

Why Does This Matter? (The Power Savings)

The biggest win here is energy.

The Old Way: To send this much data, a standard high-tech receiver acts like a heavy-duty server, eating up a lot of electricity and generating heat.
The New Way: Because the "Magic Prism" (ROSS) does the hard work optically (using light) instead of digitally (using electricity), the receiver is much simpler.
The Impact: The authors calculated that if they build this into a custom chip, it could save over 40% of the power compared to the best existing high-speed solutions.

The Bottom Line

This paper presents a new way to send internet data that is fast, simple, and energy-efficient.

Think of it like this: Instead of hiring a team of 100 accountants (digital signal processing) to decipher a messy receipt, they built a special scanner (the ROSS prism) that automatically cleans up the receipt so a single clerk (a simple camera) can read it instantly.

This technology could lead to faster internet connections that don't overheat data centers and don't drain our power grids, making next-generation 1.6 Terabit internet modules much more practical and affordable.

1. Problem Statement

The transition from Intensity Modulation and Direct Detection (IM/DD) to coherent technology for short-reach applications (e.g., data center interconnects) has increased spectral efficiency but introduced significant challenges:

Power Consumption: Coherent transceivers require heavy Digital Signal Processing (DSP), high-speed Analog-to-Digital Converters (ADCs) with oversampling, and complex optical hybrids, leading to power consumption that exceeds the thermal limits of standard pluggable form factors (OSFP, CFP8).
Complexity: Existing self-coherent or phase-retrieval solutions (e.g., Kramers-Kronig receivers, Stokes vector receivers) often suffer from high DSP complexity, large overhead, or require full-duplex fiber pairs, hindering commercialization.
Driver Requirements: Conventional coherent transmitters require full-swing driving voltages ( $2V_\pi$ ), necessitating high-power RF drivers.

The authors aim to bridge the gap by demonstrating a self-coherent receiver that combines the spectral efficiency of coherent formats with the simplicity and low power consumption of direct detection, utilizing minimal DSP and low driving voltages.

2. Methodology

The proposed solution utilizes Recurrent Optical Spectrum Slicing (ROSS) as a photonic accelerator to perform optical equalization and phase-to-intensity conversion before detection.

Transmitter Configuration:
- Uses a standard IQ modulator but biases it at the quadrature point (instead of the null point used in coherent systems).
- Operates with low driving swings (below $V_\pi/3$ , specifically $<1V$ ), confining the signal constellation to the first quadrant. This preserves a carrier component for direct detection without needing a separate local oscillator or carrier insertion.
- Employs Geometric Constellation Shaping (GCS) using DBSCAN clustering to pre-compensate for the non-linear phase-to-amplitude conversion introduced by the low-swing modulation and the ROSS filter.
Receiver Architecture (ROSS):
- Optical Processing: The signal passes through a reconfigurable photonic processor (iPronics Smartlight) implementing recurrent MZI (Mach-Zehnder Delay Interferometer) nodes.
- Mechanism: These nodes act as spectral slicers with central frequencies slightly detuned (approx. 5 GHz) from the carrier. The recurrent nature of the filter creates a frequency-dependent transfer function that:
  1. Linearizes the channel (mitigating dispersion and colored noise).
  2. Non-linearly converts phase information into amplitude information, allowing recovery of IQ data via simple photodetection.
- Detection: Uses single photodetectors (PD) followed by low-complexity DSP (Feed-Forward Equalization - FFE) or light machine learning algorithms. No optical hybrid or balanced detection is required.
Experimental Setup:
- Signal: 32 Gbaud QAM-4 and QAM-16.
- Transmission: C-band over 25 km, 50 km, and 75 km of Single-Mode Fiber (SMF).
- Hardware: Implemented using the iPronics Smartlight photonic processor (programmable unit cells) to create the ROSS nodes.

3. Key Contributions

First Experimental Demonstration: Validated the concept of using ROSS neuromorphic receivers for M-QAM demodulation with direct detection, achieving QAM-16 reception up to 50 km.
Low-Voltage Operation: Demonstrated successful reception with driving swings as low as $0.3V_\pi$ (approx. 1V), eliminating the need for high-power RF drivers required by conventional coherent systems.
DSP Simplification: Achieved complex signal recovery using simple photodetectors and linear FFE, avoiding the heavy DSP and carrier phase estimation algorithms typical of coherent receivers.
Geometric Constellation Shaping (GCS): Introduced a GCS technique using DBSCAN to optimize the constellation for the specific non-linearities of the low-swing ROSS system, improving BER performance.
Power Consumption Analysis: Provided a detailed comparative analysis for next-generation 1.6T pluggables, projecting significant power savings.

4. Key Results

Performance:
- QAM-4: Achieved Bit Error Rate (BER) $< 10^{-5}$ in back-to-back and maintained $< 10^{-4}$ at 75 km.
- QAM-16: Achieved BER $\approx 10^{-3}$ in back-to-back and $\approx 10^{-2}$ (within FEC threshold of $2 \times 10^{-2}$ ) up to 50 km.
- Distance: Successfully demonstrated 32 Gbaud transmission over 25 km, 50 km, and 75 km.
Robustness:
- Laser Drift: The system showed tolerance to laser frequency drifts up to 1 GHz with minimal BER degradation, and up to 3 GHz with <50% performance loss, suggesting compatibility with semi-cooled lasers.
- Bandwidth: The system maintained performance with receiver analog bandwidths as low as 0.35x the baudrate (11 GHz for 32 Gbaud), significantly improving Electrical Spectral Efficiency (ESE) to 11.4 bits/s/Hz.
Power Consumption (1.6T Module Projection):
- The proposed ROSS architecture is projected to consume ~6.22 W (5.22 W without TEC) for a 1.6T module.
- This represents a >40% reduction compared to lightweight coherent solutions (10.9 W) and a **1.5 W reduction** compared to IM/DD solutions (~7.7 W).
- Key savings come from driver-less operation, reduced ADC sampling rates (baudrate vs. oversampling), and lighter DSP requirements.

5. Significance

This work presents a viable pathway for energy-efficient high-speed optical interconnects. By moving the equalization and phase recovery tasks from the digital domain (DSP) to the optical domain (ROSS), the authors demonstrate a receiver that:

Reduces Power: Critical for meeting the thermal budgets of next-generation 1.6T/3.2T pluggables.
Simplifies Hardware: Removes the need for optical hybrids, balanced detectors, and high-power RF drivers.
Maintains Performance: Offers spectral efficiency comparable to coherent systems while utilizing the cost-effective components of IM/DD systems (low-bandwidth PDs, standard lasers).

The study suggests that ROSS-based self-coherent receivers could be the enabling technology for future short-reach coherent-like links, offering a "best of both worlds" solution for data center and metro applications.