MyTechDailyNews Communications: 24-Core Fiber Reaches 3.2 Pb/s Over 10.3 km
When a laboratory optical-transport trial reports 3.2 Pb/s, the number needs a decoder before it is mistaken for a customer-facing upgrade. In this result, data were carried across a 24-core fiber over 10.3 km, using space-division multiplexing and three wavelength bands—S, C, and L—while the system created thousands of parallel channels. For network engineers, the value is not that residential broadband changes next week. The value is that a multicore, multi-band system was measured in real time, with transmission components and modulation choices that are closer to commercial optical practice than earlier laboratory demonstrations.
What was actually demonstrated
The trial combined two scaling ideas: more spatial paths inside the fiber and more usable spectrum around the fiber. A single fiber strand held 24 cores, each treated as an independent transmission path. By operating across S, C, and L bands, the experiment used three wavelength bands as the spectral envelope. The total was reported as 3.2 Pb/s, equivalent to 3,200 Tb/s.
Several details matter more than the headline. The link used commercial coherent optical modules together with higher-order probabilistic shaping. The multicore fiber's core arrangement was optimized to reduce inter-core crosstalk, which is one of the central barriers to making multiple cores behave like separate fibers rather than coupled noise sources.
The team also completed real-time transmission validation using in-house optical modules. In those measured bands, S-band single-wavelength rate reached 400 Gb/s, while C-band and L-band single-wavelength rate reached 1.2 Tb/s. That gives operators a band-by-band view, rather than a single aggregate value that could be driven only by offline processing.
How to interpret 3.2 Pb/s
Aggregate optical capacity is easy to misunderstand when it is translated into user stories. A rate of 3.2 Pb/s, expressed as 3,200 Tb/s, is theoretically enough to move more than 20,000 4K movies in one second, or to carry about 3.2 million gigabit-broadband households simultaneously. These comparisons are arithmetic analogies. They help show scale, but they do not mean the experiment delivered gigabit service to 3.2 million homes.
The test distance was 10.3 km, and the fiber was a 24-core specialized multicore fiber. Because of that, the result cannot be read as evidence that existing home broadband has reached 3.2 Pb/s, or that ordinary single-core fiber in the field can now match the trial. The experiment proves feasibility under a carefully built testbed, not immediate reachability over installed access plant.
Baseline capacity versus multicore expansion
According to the researchers' comparison, traditional commercial single-core G.652 fiber has a practical C-band and L-band capacity ceiling of about 100 Tb/s. The reported multicore trial reached about 32 times that figure. This is the clearest way to communicate why the architecture attracts attention: one fiber can contain many parallel cores, so spatial dimensioning adds capacity headroom without simply pushing every wavelength channel to an ever higher symbol rate.
| Point | Trial architecture | Engineering significance |
|---|---|---|
| Fiber | 24-core multicore fiber | Multiple independent cores can multiply spatial paths in one cable strand. |
| Distance | 10.3 km | The capacity figure applies to the measured test link, not to every deployed route. |
| Bands | S, C, and L bands | Three wavelength bands expand the spectral space used by the trial. |
| Channeling | Thousands of parallel channels | Capacity is built from many simultaneous paths and wavelengths. |
| Aggregate rate | 3.2 Pb/s, equal to 3,200 Tb/s | The comparison with about 100 Tb/s for G.652 C/L capacity shows roughly a 32-times difference. |
| Band rates | S-band 400 Gb/s per wavelength; C/L 1.2 Tb/s per wavelength | Single-wavelength figures expose more than total capacity alone. |
Why this is useful for backbone, data centers, and future 6G transport
The trial is most relevant to scenarios where concentrated infrastructure needs large transport capacity. The research positioning highlights data center interconnect, backbone networks, and future 6G transport networks as potential beneficiaries. In those environments, a capacity ceiling is not just a physics debate; it shapes whether additional routes can be built, whether existing corridors are congested, and whether spectral and spatial resources can be expanded together.
Still, the same fact that makes the result important also limits its commercial meaning: the work is a validated experiment. Scaling from a 10.3 km testbed to operational networks requires engineering that converts a measured multicore link into repeatable field behavior. That is the boundary between a conference result and deployable infrastructure.
How the paper stands among earlier records
The result was selected for the 2026 International Optical Fiber Communications Conference (OFC) and received a high score. The team was also invited to submit a paper to the Journal of Lightwave Technology. These milestones indicate technical significance within the research and conference program.
Earlier laboratory studies on similar multicore fiber have reported higher transmission rates, but they relied on experimental devices and signal-processing schemes. The present work is positioned differently: its focus is real-time transmission validation in a system closer to commercial optical architecture. This does not automatically make it a product, but it makes the result more directly useful for engineers assessing what may be transitioned from testbed to network.
Checklist: what to check before calling it deployment-ready
- Confirm the measured link. The reported trial used 10.3 km of 24-core fiber. Any extrapolation to longer routes or installed cable needs separate validation.
- Separate aggregate capacity from user throughput. 3.2 Pb/s equals 3,200 Tb/s at the system level, but it is not a customer speed.
- Check the wavelength envelope. The trial used S, C, and L bands, which adds spectral width beyond the conventional C/L focus.
- Look for spatial-path control. Core arrangement optimized to reduce inter-core crosstalk is central to treating 24 cores as usable parallel paths.
- Assess modulation and optics. Commercial coherent modules and higher-order probabilistic shaping are relevant; in-house optical module real-time validation adds architectural credibility.
- Keep peer-review status distinct from field readiness. OFC selection, a high score, and an invitation to submit to the Journal of Lightwave Technology are scientific milestones, not deployment announcements.
Common misconceptions
- Misconception: "3.2 Pb/s will appear in broadband packages."
Correction: The figure is laboratory capacity over a specific 10.3 km multicore testbed, not an access-network service rate. - Misconception: "Multicore fiber means single-core fiber is immediately replaced."
Correction: The result suggests a path for capacity headroom in data center interconnect, backbone, and future 6G transport, but existing G.652 plant and ordinary single-core fiber are not automatically superseded. - Misconception: "Any faster lab number is equally meaningful."
Correction: Earlier higher-rate experiments used experimental devices and signal processing; this work is notable for real-time validation closer to commercial architecture. - Misconception: "Journal and conference approval means the product is ready."
Correction: The trial remains at the validation stage, with further engineering required before scale deployment.
FAQ
What did the 3.2 Pb/s trial demonstrate?
It demonstrated aggregate transmission capacity of 3.2 Pb/s, or 3,200 Tb/s, over a 10.3 km 24-core fiber using space-division multiplexing, three wavelength bands—S, C, and L—and thousands of parallel channels.
Is 3.2 Pb/s the speed users will see?
No. The user-facing translations are analogies: 3.2 Pb/s could theoretically move more than 20,000 4K movies in one second, or support about 3.2 million gigabit-broadband households at the same time. They do not describe service offered by the test.
How does it compare with G.652 fiber?
The researchers stated that traditional commercial single-core G.652 fiber has a C/L-band capacity ceiling of about 100 Tb/s. The multicore trial's total capacity was about 32 times that value.
Why is single-wavelength rate included?
It clarifies how the aggregate was built. With in-house optical modules in real-time validation, S-band single-wavelength transmission reached 400 Gb/s, and C-band and L-band single-wavelength transmission reached 1.2 Tb/s.
Where could such capacity matter?
The trial points to potential expansion paths for data center interconnect, backbone networks, and future 6G transport networks, where additional spatial and spectral dimensions can relieve capacity pressure.
Is multicore fiber now commercially deployable?
Not on the evidence presented. The work is a real-time experimental validation. Scaling to deployment requires further engineering, so the result should be treated as a milestone in the capacity roadmap rather than a shipping product.