When people talk about the Internet, most think of wireless networks, satellites, and cell towers. But the reality is tougher: the foundation of the global web is fiber-optic cables laid underground and along the ocean floor. These thin glass strands transmit terabytes of data every second. And here is the important part: we are approaching the physical limit of how much information can be pumped through the existing infrastructure.
The problem isn't that someone slacked off. The problem is physics. Light in a fiber doesn't behave as simply as it seems. As you increase power, nonlinear effects appear – the fiber starts to distort the signal. Amplifiers generate noise. Dispersion smears the pulses. And all this happens over distances of thousands of kilometers. Systems working in the classic C-band (around 1530–1565 nm) have basically hit the theoretical Shannon limit. There is simply nowhere left to squeeze.
So, what do we do? Lay new cables? Expensive and slow. Switch to multi-mode or multi-core fibers? The tech is raw, still years away from commercial application. That leaves one practical path: expanding the utilized spectrum on the very same fiber that is already lying in the ground.
How Band Expansion Works
Imagine that optical fiber is a multi-lane highway. At first, we used only one lane – the C-band. Then we added the L-band (1565–1625 nm) and got two lanes. Now the proposal is to hook up the E-band (1360–1460 nm) and the S-band (1460–1530 nm). In total, this gives us the OESCL-band – from 1350 to 1629 nm. Almost 280 nanometers of spectrum instead of the usual 70–80.
Sounds logical: more spectrum means higher throughput. But the devil is in the details. Each band has its quirks. Fiber attenuates differently at different wavelengths. Amplifiers for new bands work with varying efficiency. The E-band, for instance, sits right next to an area where absorption due to hydroxyl impurities in the fiber is stronger. This means the amplifiers have to work harder to compensate for the losses.
And here is the main question that, for some reason, is rarely asked: how much energy does all this «guzzle»? Because it is one thing to pump through more data, and quite another to make it economically viable. If doubling the throughput requires ten times more electricity, such a solution is dead on arrival.
Energy as a Reality Check
The power consumption of fiber-optic systems isn't just an abstract number. It is real money in electricity bills, it is the need for cooling systems (especially for remote amplifiers), and it is the environmental footprint. Data centers long ago started calculating Energy Per Bit (EPB) as a key efficiency indicator. Backbone networks came to this later, but now it is just as critical a parameter.
Optical amplifiers are the main energy consumers in fiber-optic lines. They stand every 80–100 kilometers, compensating for signal attenuation. On a thousand-kilometer line, there are dozens of them. Each consumes energy not just for boosting the signal, but for its own operation – pump lasers, control systems, cooling. And if an amplifier for the C-band consumes 50 watts, while an amplifier for the E-band takes 80, this difference multiplies by the number of amplifiers and the operating time. Over a year, that racks up kilowatt-hours that need to be paid for.
So, we went ahead and measured the real power consumption of modern amplifiers. Not theoretical, not from datasheets, but practical – we hooked up wattmeters and watched how much the amplifiers «eat» in operating mode.
The Experiment: What the Measurements Showed
We tested two types of amplifiers. The first is the classic CL-EDFA, operating in the C and L bands (1528–1605 nm). This is a commercially mature technology, refined over years. The second is the experimental OESCL-EDFA, which covers the entire extended range from O to L (1350–1629 nm). This isn't a single amplifier, but essentially four separate modules (for S, E, C, and L bands) combined via an optical multiplexer.
Both amplifiers ran at full power – the worst-case scenario in terms of energy consumption. This is important because, in real systems, amplifiers often operate close to the max to ensure a sufficient signal level at the receiver after passing through all line sections.
What did we find out? Energy consumption depends nonlinearly on output power and varies greatly between bands. The E-band amplifier turned out to be the most «gluttonous» – to achieve the same output power as the C-band, it requires noticeably more energy. The reason lies in the physics of doping optical fiber with erbium and other rare-earth elements used in amplifiers. The E-band sits on the edge of the effective range for erbium amplifiers, and the efficiency there is lower.
The S-band is a different story. There are complexities there too, related to its location between the water absorption areas in the fiber, but technologically, amplifiers for the S-band are already more refined than for the E-band. This is visible in the power consumption.
Simulating a 1000 km System
Measuring the amplifiers is half the battle. We need to understand how they work as part of a real system. We simulated a backbone line 1000 kilometers long – a typical distance between major communication hubs. The line is split into 12 spans of 80 kilometers, with an amplifier at each junction. We used standard single-mode fiber (SSMF) – the exact kind already laid in most networks.
We factored in everything: fiber attenuation (it differs across bands), dispersion (pulse smearing over time), nonlinear effects (four-wave mixing, cross-phase modulation), and amplifier noise (ASE – Amplified Spontaneous Emission). We optimized the power in each span to find the balance between a sufficient signal level and an acceptable level of nonlinear distortion.
The results came out curious. The system on the CL-band (only C and L) provides a certain throughput at a certain EPB. That is our baseline. Now we take the OESCL-band – adding E and S to the existing C and L. Throughput grows by 2.98 times. Almost three times! On the same fiber, on the same line length.
But at what cost? Energy consumption per bit grows by 48%. That is, not three times, as it might seem, but less than one and a half. Why? Because a significant chunk of energy goes not to the data itself, but to maintaining the system's operation – powering electronics, cooling, pumping the amplifiers. These base costs get «smeared» over a larger volume of transmitted data.
What This Means in Practice
Let's calculate specifically. Suppose a system on the CL-band transmits 100 terabits per second and consumes, let's say, a hypothetical 10 kilowatts (real numbers depend on the specific implementation, but the proportions hold). Energy per bit is 0.1 microjoules.
We switch to the OESCL-band. Throughput grows to 298 terabits per second. If power consumption scaled linearly, we would expect 29.8 kilowatts. But in reality, it comes out to about 14.8 kilowatts – a growth of 48%. The energy per bit is now 0.148 microjoules. Yes, it went up. But we got almost three times the throughput!
For a telecom operator, this is critical. Demand for bandwidth is growing exponentially – 4K and 8K video, cloud services, streaming, the Internet of Things. If a system is running at 80% load now, in a couple of years it will be overloaded. The options? Laying new fiber is millions of dollars and years of work. Or modernizing the existing line by installing new amplifiers and transceivers. The cost is incomparable.
And this is where the compromise – «plus 48% energy for plus 198% throughput» – looks reasonable. Yes, the electric bill will grow. But not threefold, only by one and a half times. Meanwhile, the network's capacity to serve clients will triple.
Technological Nuances
Of course, it isn't as simple as «plug in a new amp and watch it fly». The E and S bands are relatively new territory for commercial application. Amplifiers for them aren't as polished as for C and L. This is evident in efficiency – in our measurements, the E-band amplifiers showed lower efficiency than the C-band ones.
But this is a temporary situation. When we started mastering the L-band 20 years ago, there were difficulties there too. Then the technology matured, effective erbium amplifiers optimized for longer wavelengths appeared. The same will happen with E and S. It is a question of engineering refinement and scaling production.
Another point is components. The OESCL-band requires not just amplifiers, but multiplexers/demultiplexers capable of working across this entire spectrum, transceivers with broadband lasers and photodetectors, and dispersion management systems. But again, these are solvable engineering tasks. We don't need to invent new physics – we need to adapt existing technologies.
Alternatives and Their Problems
To be fair, spectrum expansion isn't the only path to increasing throughput. There are other approaches. Multi-core fibers – instead of one core, they make several in the fiber, essentially creating multiple parallel channels in one cable. Multi-mode fibers with spatial mode multiplexing – using different light propagation modes in the fiber as separate channels.
These technologies work in the lab and show impressive results. But they are far from commercial adoption. The problem is that they require replacing not just the amplifiers and transceivers, but the fiber itself. And that means re-laying cables – multi-billion dollar investments and decades of work. Even if the tech becomes commercially available tomorrow, the physical rollout will take a huge amount of time.
In contrast, extending the spectrum to the OESCL-band uses existing fiber. You only need to change the equipment at the ends of the spans – amplifier and transceivers. This can be done in stages, section by section, without stopping the network. From the standpoint of practical feasibility, the difference is colossal.
Why This Works in Siberia
I always test technologies against whether they can withstand Siberian conditions. Not because I am a local patriot, but because it is an honest reality check. If a system works at -40°C, with temperature swings of 60–70 degrees between summer and winter, with high humidity and dust – it will work anywhere.
Fiber-optic lines in Siberia are a harsh story. The fiber lies in the ground at the depth of frost penetration, amplifiers sit in containers at remote sites where the internal temperature can fluctuate by tens of degrees. Cooling systems in summer and heating in winter consume extra energy. And in these conditions, every extra watt of amplifier power consumption turns into an additional load on climate control.
Expanding the range to OESCL makes sense here precisely because the increase in power consumption is moderate. Plus 48% is not critical for the existing power supply infrastructure. But if power consumption had grown three times, proportional to throughput, we would have had to build new power lines to every amplifier station. Now that would have been a real problem.
Where to Go From Here
Our research shows that expansion to the OESCL-band is a working solution right now, with existing technologies. But there is room to grow. The efficiency of E and S band amplifiers can be improved. New doping elements, alternative fiber designs for amplifiers, and more efficient pump schemes are currently being actively researched.
If we manage to reduce the power consumption of E-band amplifiers by even 20–30%, the overall EPB of the OESCL system will improve noticeably. Perhaps the energy consumption growth can be cut not to 48%, but to 30–35% for that same tripling of throughput. That makes the technology even more attractive.
Another path is optimizing the system as a whole. Adaptive power control, where amplifiers run at full blast only when needed. Intelligent routing that uses less loaded bands to reduce nonlinear effects. All of this is a matter of software and control systems which can be improved independently of the hardware.
Practical Conclusion
Expanding the optical spectrum to the OESCL-band isn't a futuristic fantasy, but a real engineering opportunity available right now. We have shown with concrete measurements and calculations: a 1000 km system can transmit nearly three times more data with an increase in energy per bit of less than one and a half times.
It isn't a perfect solution – perfect would be tripling throughput with zero energy growth. But we live in a world of physical constraints, not a world of «spherical cows in a vacuum». In reality, you have to find compromises. And the compromise of «plus 198% throughput for plus 48% energy» is a damn good compromise.
For telecom operators, this means the ability to modernize existing lines without changing cables. For users – growth in available bandwidth and lower latency. For engineers – a clear technological roadmap for the coming years, without the need to wait for revolutionary breakthroughs in multi-core or multi-mode fibers.
The next few years will show how quickly the industry masters the E and S bands on a commercial scale. Technologically, everything is ready. What remains is scaling component production and lowering their cost. That is no longer a question of physics – it is a question of economics and engineering grit.
But for now, the main takeaway is simple: good old single-mode fiber, laid decades ago, still has life in it. Its potential is far from exhausted. We just need to learn to use more of the spectrum it is capable of transmitting. And we have shown that this is technically feasible and economically sound.
Verified at -40°C. Works.
