Monthly Archives: November 2011

Crazy Idea: Apple should buy Clearwire, build iDevice LTE network

On a recent episode of Critical Path, it is noted that the fastest growing slice of the earnings pie for carriers around the world is data. Voice and SMS revenues are slumping, as users are turning to data networks for more and more of their communication. Phone apps like Apple’s iMessage and RIM’s BBM move text message traffic off SMS and to data networks. Phone calls will soon be replaced with Facetime calls when cellular networks are up to the task of carrying video traffic, with the exception of calling while driving.

If we look at Apple’s iPhone (and most cellular phones in general), the most disappointing facet of the device is often the carrier, specifically data traffic; followed closely by battery life (that’s another article entirely). So what is it that Apple can do to drive additional revenue as well as provide it a leg up on the competitions devices – tablets and phones, plus anything else they may think of in the future? It would need an end-run around the current cellular carriers. And this means owning and operating a cellular network.

This is initially difficult to do on a worldwide scale because of licensing issues. Each country has their own spectrum authority (FCC here in the USA), and the same slice of spectrum can be allocated for different uses around the world with the main exception of unlicensed ISM bands (2.4Ghz and 5Ghz for WiFi). Steve Jobsreportedly wanted to build their own network using these unlicensed ISM bands, but it was easy to see that it wouldn’t be technically possible.

Clear?

In the United States the obvious choice would be for them to acquire Clearwire’s spectrum and assets. Its market cap is incredibly low (less than $2B) and it doesn’t need a whole lot of cash to fix up ($900M in the next few yearsto build and operate a new LTE network), and is in desperate need of cash to pay its debt obligations, even choosing to skip a debt payment recently. Cheap considering how much spectrum they’re holding on to in major cities across the USA – 192MHz in many cities, 125MHz in NYC and as low as 75MHz in Detroit. The difficulty is that its majority owned by Sprint, however Sprint is in need of cash too and I expect it will have to be acquired by Verizon within the next five years if they don’t get their act together. Sprint seems less interested in Clearwire lately, especially since they announced they’re going on their own with LTE (using their own spectrum and Lightsquared spectrum instead of Clearwire spectrum). The downside to using Clearwire’s spectrum is that it is in the 2490-2690MHz band, which doesn’t have the best propagation characteristics (e.g. going through walls, into basements, etc). Apple would need to use their extensive antenna engineering knowledge to build a device that will still get fantastic reception even with poor signal strength.

The phone will still need (and should use) the voice networks from the old carriers. There is no need to build up that infrastructure again. Apple would roll out the TDD-LTE-Advanced (rel. 10) network on Clearwire’s 2.5-2.6Ghz spectrum in 2013 and provide tremendous speeds to end users – better than any of the current network carriers could offer. While LTE offers 10Mb/s down, the enormous spectrum holdings of Clearwire would allow speeds up to 50Mb/s on a regular basis, and peak speeds well above that. Putting their spectrum to use in a 50MHz TD-LTE-Adv configuration provides for over 250MB/s raw throughput (downlink, 2×2 MIMO) with user speeds around 20-50Mb/s and upload speeds around 10-15Mb/s.

How would the carriers react? A mixed bag – they’ve invested money in building up a network to handle tons of data, and while they might welcome Apple taking a load off their network (their CapEx would slow down dramatically, for a few quarters at least after rollout), they aren’t going to be happy with Apple taking revenue away – presumably because everyone could switch to no data plan or a minimum data plan for roaming outside of Apple’s initially incomplete network. But Apple recently just took a bite out of their revenue pie by introducing iMessage, reducing carrier revenue from text messages, though that is an order of magnitude smaller than the equivalent data revenues.

Killing Cable?

It also offers a hand in creating their own mini-cable system. With an abundance of spectrum, a separate 20MHz channel could be used just for broadcasting their own live TV on multicast – a 20Mhz channel (2×2 MIMO) with a 87:10 down/up ratio would have 120Mbs down, enough for 10 12Mbps 1080p feeds, the 8Mbps upstream channel would just be for device authentication and updates only. In true Apple/Pixar fashion, they’d only be showing a few choice channels with high quality content. During the low traffic periods of the day (would Apple sell informercials? I don’t think so…) they could turn off a few channels and stream prime content to the devices to be “unlocked” as prime-time TV shows. If they needed to increase throughput, they’d move to 4×4 MIMO and change the ratio to 90:7 for 255Mb/s down (21 channels 1080p channels) and a small control channel up.

Technical Difficulties

Apple would need to build dual-SIM devices, it would need a carrier SIM for voice and SMS, but an Apple SIM for data. However, Apple was rumored to be building a SIM replacement. This would allow for still one SIM card and Apple’s SIM would be based in software.

Building a network is no easy task, and considering that Clearwire is moving to a co-located configuration with Sprint (the same tower would have Sprint’s and Clearwire’s transmitters), any buy out might negate that cost-sharing benefit.

But overcoming one of the last poor aspects of the smartphone experience would be a huge deal, and give Apple a leg up on both other cellphone vendors and their carrier partners, at least here in the US.

Battery Magnitude

Recently, there was news on a new formulation of lithium-ion battery using a Si-Graphene anode that would provide for 10 times the charge in 10 times the power capacity that current lithium-ion batteries in the same size (and I’m assuming the same weight). Now the standard three to five years disclaimer applies – in that it has to be brought out of a lab, commercialized and they have to figure out how to mass-produce them and not lose any of their stand-out characteristics.

But it didn’t immediately seem apparent to most observers that it would be useful because of rapid capacity fade – the battery would only last 150 cycles before it had only 50% of it’s original capacity, although that is still five times current battery capacity. Traditional lithium-ion batteries last 300 cycles to 80%, lithium-polymer (the battery in your iPhone and Mac laptop) last about 1000 cycles to 80%, and lithium-titanate batteries can last 5000+ cycles to 80%.

The difficulty with the Si-Graphene battery is managing the user experience. If a user were to go through their entire battery in a day, every day, in 5 months they’ll only have half the capacity. So the device developers have to oversize the batteries but artificially clamp the energy storage to keep heavy users from destroying the battery in a short time frame.

Consumer Electronics? Sure…

Putting this battery into a smartphone to replace the current lithium-polymer battery would let average users go 10 days between charges. However every 250 days (25-30 charges) the user would notice they’ve lost a day of use before it was time to recharge, so from 10 days to 9 days. Will users be upset that they lost a day? Will they even notice? Or will they beat down the door of the store where they bought it demanding an exchange on a perfectly good product? How can we avoid this? By artificially limiting battery capacity.

If we were to limit the battery capacity artificially to a value that the 80th percentile user will have after 2 years of usage, we can save the trouble of users noticing their battery doesn’t last as long as it once did. I’ll assume this number is 75% of capacity, that is the 80th percentile user will go through enough battery capacity in two years that will cause a 30% capacity fade (this also factors in an increase in usage due to the bigger battery). So the phone will be setup so the average user can go 7 days before hitting the 25% warning and then recharging. Using 4Wh/day the user will go through 36.5 full cycles per year, which represents a 12% capacity fade per year. After two years, the capacity fade will be 24%. So after two and a half years, the average user will start to experience the battery holding less energy, and probably notice it around the end of year three – about which time the user will need to buy a new phone anyways as this one will be long out of warranty. An 80th percentile use will probably start to experience capacity fade earlier, around 18-24 months. A 95th percentile user is likely to do crazy stuff with their phone like stream audio all day and go through one cycle per day, and run into capacity fade in 6 months. This last case could actually be accounted for in software – if the phone notices its being used heavily it could ask the user to plug it in while engaging in the heavy activity, or just nerf capacity in software in the name of getting out of the warranty period without having to replace the battery.

Below is a chart of a traditional Li-Polymer batter (5.3) and a new Si-Graphene battery (53). You can see that the new battery has much larger capacity, it also fades much quicker. If you were to limit the Si-Graphene battery at 40Wh (40) capacity, the battery would get to two and a half years of average use before the user experienced any capacity fade.

The downside to this approach over traditional batteries is that users might increase their phone use and suck down more battery power per day knowing they have a lot more energy available to them, which is all the more reason to artificially limit capacity in the name of having the battery last long enough to have a useful device for 2-3 years.

The same results from the smartphone situation above would also apply to tablets. Laptop computers would probably see more agressive artificial capacity restrictions, as users usually run out of battery on the laptop before they are done with whatever they were doing (like doing internet research and writing a blog post about batteries ಠ_ಠ), so the issue of using more energy per day and higher annual cycle counts would apply.

Electric Vehicles? Not so fast…

If the approach of limiting battery capacity in the name of extending its life sounds familiar, it should, as it is how the battery in the Chevy Volt is managed. So what would happen if you applied this to the battery in the Volt? Not much difference, and probably an increase in cost.

If you recharge the Chevy Volt once a day, 365 days per year, it is equivalent to 237 full battery cycles per year (10.4kWh used for 35 miles, 16kWh capacity), and the battery type they use is expected to have a life of 1500 cycles (without any depth of discharge reduction bonuses). But if you were to drop-in a replacement battery with this new technology (assuming same size, weight, etc), you’d have a 160kWh battery. Now that doesn’t mean you drive 10 times as far, rather you just use an increasingly small portion of the battery, specifically an initial depth of discharge of 6.5%, and a rate of 25 full cycles per year. By the end of year 10, or 250 cycles, the battery would have degraded enough where it will start to run into problems storing and producing enough energy (assuming they can last that long from a calendar standpoint). This doesn’t appear to be a significant change from the current battery regimen, where the battery is warrantied for 8 years or 100,000 miles (12,500 miles per year). The only benefit to using the Si-Graphene batteries would be the increased power output – a Volt’s 9-second 0-60mph times could improve dramatically, along with faster recharging times.

What would help

The problem with this is that batteries are predominately priced in $/kWh, which would make the above scenarios prohibitive. The fundamental question is would it be more appropriate to charge by mass and volume? Does it cost 10 times the amount of making traditional batteries to these batteries? I don’t think it will. They might be able to charge more than the highest end batteries, but the $/kWh would need to be discounted compared to other types of batteries that have higher cycle lives. The best figure to use when it comes to battery prices is $/lifetime kWh, or the amount of energy a given battery will output until its no longer usable for the specific application (e.g. smartphone, EV, etc). A battery might cost $700/1500 kWh lifetime, and it might not matter that its 1kWh of storage for 1500 cycles or 10kWh for 150 cycles for certain applications – assuming other factors are held constant (volume, weight, safety). In fact, the latter configuration helps in applications where power demand is high (e.g. a car).

So the most basic thing to help these batteries would be an increase in cycle life. Even a relatively small increase in cycle life would dramatically impact the usefulness and increase the impact these batteries can have.