In Tuesday’s Wireless Review Column I made the mistake of attributing all of the iPhone’s 1 million sales to AT&T customers in the U.S. I then extrapolated the effects of 1 million new customers on AT&T’s 3G network, resulting in some nonsensical numbers. Needless to say, I’ve corrected the numbers below using the assumption that roughly half were sold in the U.S. Hopefully the analysis will now make more sense. I apologize for my bone-headed mistake.

Over the weekend hundreds of thousands new 3G devices suddenly appeared on the AT&T network. I’m referring to the new iPhone 3G, which sold out in many stores in a matter of hours last Friday (check out associate editor Sarah Reedy’s attempts to buy one).

Let’s put that in perspective. AT&T UMTS phones accounted for 16% of its overall base in Q1. AT&T had 71.4 million subscribers at the end of March, meaning 11.4 million were using a 3G phone if not a 3G service. Let’s assume AT&T accounted for roughly half of global iPhone sales. Discounting the AT&T subscribers that may have upgraded to the iPhone from a different AT&T UMTS device, AT&T just saw its 3G base go up about 8%. And that’s just actual phones with 3G radios. A lot of people are walking around with 3G feature phones they use primarily for text messaging and voice calls. iPhone users consume data by the boatloads thanks to the niftiness of the Safari browser and other apps.

Finally we should take into account that Apple just sold shipped 10 million applications over the weekend to new and old iPhone owners. Not all of those apps were sold to AT&T iPhone users and some of them were downloaded to newly upgraded 2G iPhones, but it’s safe to assume many of those applications were downloaded directly over the air, adding significant traffic to the 3G network. That’s just the initial install. Though some of those apps are stand-alone programs like games, a lot of them like Facebook and AIM connect back to the network, fueling even more 3G traffic.

The point I'm trying to make is the demands on AT&T's 3G network just shot up dramatically in literally three nights. AT&T, however, says it's well-prepared to meet this demand. Spokesman Mark Siegel said that general 3G traffic from data cards, smartphones and feature phones — not just the iPhone — is on the rise, leading AT&T to upgrade its network in key markets. The end result is that nearly half of AT&T 3G markets — about 150 cities and towns — will have double the current 3G capacity today.

What Siegel would not reveal is how exactly AT&T plans to double that capacity. It's rather odd that an operator would make a point of saying they're upgrading their network but then not provide any details on what upgrade they're using. Siegel said AT&T can't show its cards due to competitive reasons — there is a very big operator whose name starts with a "V" that would like to know how AT&T is tinkering with its network. Assuming AT&T isn't bluffing, then how's it pulling this little miracle off?

The easiest way to double your capacity is to double your network resources; i.e., add another 10 MHz UMTS carrier. I highly doubt that's the case, though. Cost aside, we should keep in mind spectrum is a scarce commodity. Even CDMA operators with their tiny 1.25 MHz 1X channels are reluctant to beef up their EV-DO networks with an additional channel because that's one less channel they have for voice. Here, though, UMTS has the advantage. It not only supports voice, but it does so much more efficiently than GSM. So why doesn't AT&T just switch everything to UMTS? The answer is handsets. With only 84% of all phones transmitting GSM, any additional 10 MHz allocation to 3G would drain a lot of capacity away from a lot of 2G customers.

So what about using new spectrum? AT&T has all of that Advanced Wireless Services and 700 MHz spectrum lying around collecting dust, right? Again there's the cost issue — new network, new base stations — but the biggest problem is devices. The only AWS UMTS devices out there are just the handful T-Mobile has on its new network, and Apple just got around to supporting 3G on PCS and global 3G frequencies — it ain't messing with AWS anytime soon.

Cell-splitting? Adding sectors? That's some serious network planning we're talking about: not just new base stations but new cell sites. Frequency reuse becomes an issue, too, because AT&T isn't already running UMTS on multiple carriers in its markets. And of course, cost again becomes a factor.

So what's left? I suspect that AT&T is mucking about with downlink software in its base stations while it makes the upgrade to high-speed uplink packet access (HSUPA). High-speed downlink packet access (HSDPA) has software iterations that add increased downstream capacity to the network through a series of upgrades. I believe that AT&T is making the upgrade from the 7.2 Mb/s iteration to the final 14.4 Mb/s, maximizing the full potential of its current 3G architecture. It's cheap, it's easy, and it doubles overall network capacity, but it's hard to describe in a press release.

Contact me at kfitchard@telephonyonline.com.