Spectrum Deficit Disorder
The discussion about the wireless spectrum crunch can be hard for the average citizen to follow. While it’s easy to understand the straightforward fact that every network has a capacity limit, it’s not so easy to understand how trends in wireless device utilization stress network capacity, let alone how the limits of wireless networks compare to those of wired networks. When this topic is raised a number of rat-hole arguments immediately surface regarding the forecasts for wireless demand and the various means of expanding the supply of bandwidth on wireless networks.
Finally there’s a report that addresses the fundamental issues and only the fundamentals that will help the lay person put these issues in perspective. The report is written by Michael Kleeman of the Global Information Industry Center at UC San Diego, and it’s titled “Point of View: Wireless Point of Disconnect.” The report contrasts how easy it is to add capacity to an optical network to the rather cumbersome process that’s involved in adding capacity to wireless systems.
In the optical world, it’s simple:
Fiber optics is an amazing technology. It can send 40 billion bits per second 100 miles down a highly engineered, beautiful glass fiber by simply using pulses of laser generated light. Need to double the capacity? No need to add another fiber; just get another color laser and send multiple colors (or frequencies) of light down the fiber. If you need more distance just add a booster to amplify the laser light and you can go thousands of miles via a medium that is protected from weather and other interference.
In the wireless world, you have two choices: Add some spectrum to an existing system, or add some towers and backhaul:
Increasing capacity, given the limited radio spectrum, costs and constraints of smaller cell diameters and increased interference, is important in order to meet public demand. A common strategy for increasing capacity is to divide cells into three sectors, which provides more capacity but only temporary relief as the number of users continues to grow. The Federal Communications Commission recently made more spectrum available through auction. Adding new spectrum is the least expensive way to grow capacity because it can utilize much of the same infrastructure, e.g., no new towers, cell sites, generators, etc.
The key insight here (which could have been made more clear) is that the cost per megabit/second of radio-based bandwidth is several times more than it is for wired capacity. In fact, the two costs are additive, as every increase in wireless capacity also requires a corresponding increase in wired capacity on the backhaul side, both new towers and new backhaul are bricks-and-mortar projects; You can’t turn on a new color if there’s no fiber on which to do it.
Granted, micro-cell architectures are part of the picture, but their utility is sharply limited as they primarily serve stationary and “nomadic” uses rather than truly mobile ones.
What’s driving the demand for more Internet bandwidth? Mainly it’s video streaming, as we can see on the wired side. Netflix is up to a third of Internet use in the USA during prime time, up from the fifth that was previously reported. In the mobile space, we’re also witnessing the emergence of media-rich social and augmented reality apps.
So the bottom line is clear: The demand for mobile bandwidth is on the rise, but the supply is not growing nearly as fast as it needs to grow in order for the user experience to continue its arc of improvement. Rather than squabble over the details, it’s wise to increase the supply of spectrum, because more spectrum is the cheapest way to keep the supply of mobile bandwidth growing.
It’s also worth noting – although the report doesn’t do this – that the Internet grew from a “lab toy” into the indispensable engine of innovation and social transformation that it is – against the backdrop of cheap bandwidth enabled by rapid innovation in fiber optics. We’ve become accustomed to cheap bandwidth because that’s what we’ve had for the wired Internet, and we’ve developed expectations on both the technical side and the policy side that cheap bandwidth was going to solve all our problems. That worked out to be roughly the case.
We cannot expect that technology alone will produce the same effects for the wireless Internet that we’re used to. In fact, we can very well expect that technology will not produce these effects or anything close to them. Wireless technology advances, but not at the same speed as optical has over the past 15 years. The principal driver of optical innovation is the almost perfectly noise-free channel inside the optical fiber, a channel that is largely immune to environmental noise because noise is electro-magnetic while the optical signal is, well, optical, operating at an entirely different range of frequencies than the common noise produced by the earth’s magnetic field and common transmitters. Optical fiber is a wonderful shield against noise, but air is not. And air just happens to be the medium through which radio waves have to pass.
So we face a conundrum when it comes to increasing the capacity of wireless systems, which are in the same sort of place that wired systems were before the invention of optical fiber. Every increase in wireless capacity requires the equivalent of a new wire, which comes about in the form of a tower, a backhaul connection, or a frequency. Of these options, more spectrum is by far the cheapest.
“In fact, we can very well expect that technology will not produce these effects or anything close to them.”
Absolutely. What we’re seeing in increased wireless efficiency (bits/hz) is largely realized at shorter ranges and higher signal to noise levels. HSPA+ and LTE pretty much share the same radio modulation methods i.e., QPSK/QAM-16/QAM-64. Much of the gains in LTE come from increased spectrum and improved backhaul wiring.
Cable DOCSIS on the other hand has the benefit of shielded coaxial cables which start at QAM-64 and handle up to QAM-256.
“What we’re seeing in increased wireless efficiency (bits/hz) is largely realized at shorter ranges and higher signal to noise levels.”
True. Those shorter ranges and higher signal to noise levels obtain when using femtocell and Wi-Fi offloading. It is hopeless to obtain such performance (e.g., 10x capacity increase within a macrocell) by using more spectrum — the, say, doubling of which will only result in a 2x capacity increase. Since the Global Information Industry Center is sponsored by AT&T, they probably can’t say that.
[…] the top of our tech reading list, Richard Bennett has a great post on the “Spectrum Deficit Disorder.” It’s worth a full read but in a nutshell, Bennett cites an FCC chart showing that […]
“It is hopeless to obtain such performance (e.g., 10x capacity increase within a macrocell) by using more spectrum”
That’s like saying because it is hopeless to achieve fiber optic performance on mobile broadband, then don’t bother with more spectrum and just lay fiber on every square foot of the country.
Of course we’re not going to see typical 802.11n raw 300 Mbps performance – though shorter range 20 MHz LTE can – but double the spectrum will double the mobile bandwidth which is what counts. Wi-Fi performance is meaningless when you’re outside of Wi-Fi range. Fiber performance is meaningless when you don’t have a cable.
The fact is that all three of these scenarios (wired, wireless LAN, mobile wireless) are important.
We’d like to deploy new spectrum to increase the performance of systems such as LTE and LTE-Advanced that are capable of using additional spectrum effectively.
Adding more towers to the existing networks without adding new spectrum to them does not do that at all. This is the major problem that I have with the argument that we can build our way out of Spectrum Deficit: It makes LTE a useless exercise.
New tower infrastructure is already being built at a cost of tens of billions per year per major carrier. It’s still not enough.
My point is that new towers don’t address the LTE problem at all.
I wasn’t disagreeing with your point, in fact I agree with it and there is no way around more spectrum for LTE.
What I was saying was that even if new towers did address the LTE spectrum deficit, what can be done on adding infrastructure is already being done without putting the carriers in the red financially. So the people saying that tower investment are the answer is wrong in two ways.
[…] noteworthy: Spectrum Deficit Disorder (HIGHTECH forum.org) […]
First, I think it is fine for spectrum rights to be more property-like and for additional spectrum to be there on the table when an operator is considering options for improving their network.
That said, I’m more concerned about the user experience than with which technology is used. The best way to improve the use experience is to use topology, not technology, and get the user closer to the base station. More spectrum doesn’t do this.
In a model developed by Qualcomm for a representative HSPA+ scenario on a loaded network, a home user would get 180 kbps from a macrocell. Putting a femtocell in a user’s home would allow that user to get 14.5 Mbps, a gain of 80x. Relying on spectrum alone, if I give the mobile broadband industry the entire 300-3000 band, the gain will only be 7x. As noted above, this is not mobile. But, we’re at home or office most of the time, and that percentage of time is expected to increase. Indoor broadband connections increase in number and improve. Etc.
LTE doesn’t add much to this scenario. LTE-Advanced will due to improved interference mitigation techniques. Yes, LTE-Advanced can support up to 100 MHz bandwidth; that’s nice, but what is the use case and business model?
LTE can also support odd, unpaired chunks of narrowband spectrum in TDD mode; TDD makes more sense sense given upload/download asymmetry.
That’s femtocells. As we know, there is a lot happening in Wi-Fi, such as Hotspot 2.0. In addition, IEEE 802.11ai is looking to reduce initial authenticated link setup time from a up to a few seconds today down to 50 ms. Not mobile, but on foot from AP to AP, not bad. Wi-Fi is becoming more cellular-like.
Spectrum vs. towers is apples and oranges, and femtocells don’t do anything that Wi-Fi can’t do already. We need wider channels for LTE and LTE Advanced, regardless of how many towers we have.
While LTE and LTE Advanced can bond frequencies, it can’t make the signals transmitted over all patches of spectrum behave the same way, so there’s really no substitute for fat channels.
Transmitting over all patches of spectrum has an advantage. It provides frequency diversity. When propagation on one is poor, it can be good on another since frequencies propagate differently. Narrowband channels can substitute for fat channels, and TDD can substitute for FDD.
If we really want wider channels for LTE and LTE-Advanced to add capacity, we should be looking more at the frequencies around 2 GHz. They allow for smaller antennas at the base station and user device. Also, interference doesn’t propagate as far. This is one reason 1800 MHz is the most popular frequency for LTE deployment worldwide.
In a channel-bonding scenario, using different frequencies for different parts of the packet is not a win because the packet isn’t useful until it’s complete. A uniform virtual channel is the goal.