Why is it Good to Combine Networks?
[Note: This post covers the technology side of the proposed AT&T - T-Mobile merger, and only the technology side. To see what I think about the policy side, check out the Innovation Policy Blog on Monday, March 28th.]
Many of the questions coming up around the AT&T – T-Mobile merger have to do with spectrum. AT&T clearly wants T-Mobile’s spectrum and spectrum-related assets such as towers so much that they’re willing to outbid the other suitor. But why is it better to have one big network to serve 140 million customers than to have two smaller networks serving 100 million and 40 million customers respectively?
The easiest way to explain this is with a personal anecdote. I’ve been a customer of both AT&T and T-Mobile recently, and therefore had the opportunity to do some impromptu testing of both networks. I found that talking on the phone while driving from Sunnyvale, CA to the East Bay community where I live would typically result in a call dropping four days out of five. T-Mobile calls would drop on the south side of the Sunol Grade on I-680 more often than not, but nowhere else. AT&T calls did fine at the Sunol Grade, but would drop on the north side of Pigeon Pass on CA-84 more often than not. When you consistently get call drops like these, you’re driving through a dead zone where your RF coverage isn’t good due to the combination of topography and tower placement; it’s no accident that they happen most often on the wrong side of a hill, because hills cast RF shadows just like trees do with sunlight.
If the AT&T and T-Mobile networks were combined, it’s not hard to imagine that both of these call drops would be eliminated since the dead zones are both very small relative to the footprints of the relevant towers – neither is in a populated area. The combined network wouldn’t have any dead spots on the commute from Silicon Valley and Pleasanton/Livermore.
There are other reasons for call drops, such as over-booked towers and overloaded spectrum, and these would be alleviated in many circumstances as well: If AT&T’s tower is overloaded in one area, but T-Mobile’s isn’t, then load balancing would eliminate the overload. Similarly, if the overload is on the T-Mobile side but not on the AT&T side, joy would also ensue because the same traffic volume would be distributed across a greater pool of spectrum with lower average load. So that’s good as well.
There are also scenarios – more common with the rise of 4G – where spectrum is OK, but backhaul from the tower to the Internet is not. It’s common on 3G to provide backhaul with bonded T1s, multiple copper pairs with 1.44 Mb/s each. With 4G, backhaul is more likely to be provided by DS3 (45 Mb/s) or fiber in the multiple hundreds of Mb/s range. However, this high-speed backhaul is not universally deployed by any means. Where AT&T has fiber backhaul but TM doesn’t, then TM can be switched over to AT&T backhaul and vice versa; this enables the allocation of spectrum to be unlocked and more or less immediately deployed. TM has invested a lot in backhaul, so this is a real and immediate benefit.
Overall, 4G is poor in both spectrum and backhaul, depending on region, so averaging across a larger pool is helpful by reducing the gaps that exist today.
It’s also helpful to combine spectrum into larger pools than to draw from smaller ones. It’s like the difference between having one credit card with a high limit and two cards with smaller ones; you can divide your purchases between two cards, but two consumers can’t divide their spectrum use that way. So just as it’s easier to use the one big card that has a high limit, it’s easier to manage a network with a large spectrum pool. Moreover, if you have a large block of contiguous spectrum, you can use it more efficiently than you can if you have to leave dead spaces (or “guard bands”) between your allocation and your neighbor’s: You can use it all.
That’s why Wi-Fi is so fast relative to 3G. In a Wi-Fi system, you have access to 80 MHz of spectrum every 300 feet of propagation, but in 3G you have access to 20 MHz for every 1500-3000 feet. Considering that you typically share a Wi-Fi channel with two or three people and you share 3G with a thousand or so, this is obviously a big boon.
Obviously, there are some enormous questions about combining the number two and four networks in a country into a single network, mainly around consumer choice, billing, customer service, handset choice, and a host of other things that are far from straightforward as far as the AT&T/T-Mobile deal is concerned, and I don’t want to sell those questions short. But from a pure technology standpoint, the deal is a winner. And did I mention that AT&T and T-Mobile use compatible technology, GSM/GPRS/HSPA+? Well, they do, and that makes all the difference in the world for the two of them.
As Commissioner McDowell said, this is “the mother of all roaming agreements.” So right. The FCC is undertaking a data roaming inquiry at the moment, and one of the key road blocks is the existence of dual standards for every mobile radio system short of LTE. No matter what the rules are, real-time roaming from a CDMA network with EV-DO to a GSM network with HSPA+ isn’t going to work, even for dual-mode handsets. So there’s that to consider as well. I’m sure Sprint worked that all out because of their experience with engorging the Nextel network with its non-CDMA technology, and they’ve got the scars to prove it.
Nevertheless, technology is only one side of the question. The merger of the nation’s second and fourth largest mobile networks also raises questions about competition, regulation, and spectrum, which I will deal with in a post in a policy venue.