How much 5GHz spectrum is enough?

Note: Post edited 2/15/13 with updated information from the NTIA report.

Recently, the chairman of the FCC proposed adding more spectrum to the 5GHz unlicensed ISM band. Currently it stretches from 5.15GHz to 5.85GHz, but with a 120Mhz gap (5.35-5.47GHz) and a 5MHz gap (5.73-5.735GHz), leaving about 575MHz of usable bandwidth (at various power levels subject to competing uses) in three parts – 5.15-5.35Ghz (200MHz), 5.47-5.73GHz (260MHz), 5.735-5.85GHz (100MHz). The FCC proposal was to fill in the 120MHz gap, as well as add 75MHz to the upper end of the band, extending it from 5.85Ghz to 5.925Ghz. However that upper 75MHz interferes with “smart car” connected vehicle program, and is opposed by AASHTO (the American Association of Highway & Transportation Officials – essentially the governing body for regulations covering roads and highways).

Filling in the first 120MHz is a good idea, and would leave us with more spectrum to add channels. It would allow for a 40% increase in the number of 80MHz channels available to broadcast on, and add a third available 160MHz channel (a 160MHz channel is made up of two 80MHz channels, its not cumulative). Currently, the specification and frequency allotment for the US only allows for five 80MHz channels and two 160MHz channels as shown below. Adding in the band between 5.35-5.47GHz would allow a contiguous span from 5.15-5.73Ghz (580MHz). The 580MHz usable span would allow for seven 80MHz channels. Those seven 80MHz channels could be paired into three 160MHz channels for even faster speed under an optional part of the 802.11ac specification.

From the NTIA report on the 5GHz reallocation

From the NTIA report on the 5GHz reallocation, via Steven Crowley’s post

The addition of the 75MHz to the upper end is more questionable, due to the conflicts AASHTO brought up (linked above). It would allow for an additional 80MHz channel (or one 160Mhz channel) in that band, but because of the drawbacks dealing with connected cars (I’d rather see connected cars drive me to work rather than faster wifi speeds at home) I’m on the fence as to whether or not this 75MHz offers as much benefit as the 120Mhz addition mentioned above.

What effects does the spectrum addition have on end users? Faster speeds, at least once the specifications are updated. Because more channels will be available for people to select, neighbors will have less collisions and more spectrum to share in a given area. WiFi access points with 2 or 4 antennas would be able to offer aggregate throughput speeds measured in gigabits to a number of devices throughout the home. Devices supporting the final 802.11ac spec should start to arrive in 2013, so its just a short wait; however the wait for this additional spectrum is likely to be a lot longer.

A different purpose for Clearwire requires a different set of ownership rules

When Clearwire’s goal was to sell wholesale access and provide fast Internet for everyone, the fact that it owned about 160MHz of 2.5GHz spectrum in major metro areas wasn’t a big deal because it was planning on wholesaling the spectrum to carriers it could come to financial terms with. That spectrum was essentially available to all carriers. But not anymore.

With Sprint (or Dish, maybe) acquiring them completely, the rules for ownership change – that spectrum is no longer available to everyone. Instead, it’s locked to Sprint and its MVNOs. And if you think from that viewpoint, the rules of ownership change. Verizon thinks that, and I think that way too.

I’ve already talked about my idea for rebanding the BRS/EBS range so that carriers can get access to the spectrum, while still giving Sprint a huge leg up (Sprint would still end up with 100MHz TDD spectrum, and 80MHz FDD would be either given to incumbent owners or auctioned off). The FCC might even have a stricter viewpoint than me, giving Sprint less than what I suggest. Other carriers (Verizon, T-Mobile, AT&T) could make good use of that spectrum instead of just letting Sprint sit on it.

Spectrum is a public good and should be treated as such. Giving Sprint nearly all the BRS/EBS band doesn’t serve the public interest if its going to take 10+ years to use it all (various technical reasons prevent them from using all of it now – most notably how much power would be needed to broadcast a signal from a device over a 100MHz range, the effective bandwidth currently is 20-40MHz; they could setup multiple 40MHz bands but you wouldn’t be able to aggregate them efficiently from a power standpoint).

Re-rationalizing the 2.6GHz band

The BRS/EBS band went through one overhaul in the mid/late 2000s decade. But it seems like it needs another overhaul to deal with the current LTE realities.

Currently, the spectrum is divided into 5.5MHz, 6MHz and 4MHz blocks. The usable spectrum is from 2495MHz to 2690MHz. Not very clean and uniform. Its licensed on both BTAs (basic trading areas, of which there are 487 of them) and 35-mile radius geographic service areas (not something my database knows how to deal with, unfortunately).

I would license them more uniformly, using CMAs or cellular market areas (of which there are 734 of them), but with a uniform block size of 10MHz chunks. I would divide the spectrum into two parts – a middle band that is for TDD technologies (TD-LTE), and on either side FDD technologies (FD-LTE, what all carriers use today). The band plan would start at 2495MHz with four 10MHz uplink blocks, a 5MHz guard band, five 20MHz TDD blocks, another 5MHz guard band, and finally the other four 10MHz downlink blocks paired with the uplink blocks.

2.5ghz reband

Dealing with current licenses is what makes this tricky, if not impossible. Pending the Sprint/Clearwire merger, I’d venture to say that Sprint controlling nearly the entire 190MHz (they average 160MHz in major metro areas), I’d say that its time for them to divest some of those holdings (or terminate the leases or however they hold the spectrum). I really don’t see a use case for more than 100MHz for Sprint even in the next 10 years. There is a limit of 40-50MHz before the radios would consume too much power for use in a cell phone (thats not to say that in the aggregate, all the phones on the network couldn’t use it; they would just choose which 40MHz to broadcast in). In markets where they hold more than 100MHz of spectrum (most large ones), I’d simply give them the 100MHz TDD block. The remaining TDD and other FDD blocks would be auctioned or given to current non-Clearwire owners.

(what follows is a bit of a rant on spectrum ownership and usefulness for the average citizen)

Sprint owning most of the 2.6GHz band is a good strategic position for them, but the main goal for the FCC is (or should be) that the spectrum is being used in the best interests of the people. And I don’t think Sprint sitting on most of the spectrum, letting it go unused, is in the interest of the people of the United States. If they hold 190MHz of spectrum in a market, and they only plan on using 40MHz or even 100MHz over the next 5-10 years, then its time to give up enough of the spectrum so other companies can make use of it, and provide a more competitive marketplace for customers.

I look at the AWS-1 band – the only companies that really use it right now are 1) small/rural carriers and 2) T-Mobile (3G/HSPA+, LTE in 2013). The other major owners of AWS-1 spectrum – AT&T, Verizon and until recently the cable companies – don’t use it at all (some cable companies like Cox did launch a cellular company, but their 3G data service was provided by Sprint’s network, and their AWS-1 spectrum was unused since they never launched LTE before selling the spectrum). So from 2008 to today, large parts of the AWS-1 block have laid fallow. Verizon will likely build out their auxiliary LTE network (for metro areas) on their AWS-1 spectrum, but even that isn’t likely to begin until 2014 since it will likely be an LTE-Advanced network with 20MHz uplink/downlink channels where available (NYC, LA, etc). So the spectrum will likely lie fallow for a total of 7 years (2008-2015, when the first build-out requirements pass for Verizon) before its put to good use. A cautionary tale of letting companies buy and then sit on spectrum for competitive advantage. Exhibit 2: Verizon owning lower 700MHz “B” block licenses in major metro areas (LA, Chicago, Miami) and not doing anything with them. Likely, they’ll be sold to AT&T at the last minute and AT&T would expand their LTE from 5+5 to 10+10 (much to the pleasure of the residents of these cities). Again, sitting on spectrum for competitive advantage is a negative for the consumer. Build-out requirements are helpful in this, but four years is still a long time for carriers to sit on spectrum that could otherwise be used.

New Spectrum!

So this week was apparently a productive one for the FCC. Or maybe they wanted to clear off their docket for the new year.

  • The use of WCS licenses for LTE was approved by the FCC, along with the spectrum transfers from various companies to AT&T. This will give AT&T 20MHz of LTE (10 up, 10 down) in most parts of the country. Future deals with the remaining spectrum holders in the WCS band could give them a nationwide footprint for WCS. Those remaining companies include Sprint. AT&T expects to deploy this in three years. This spectrum is located at 2.3GHz, and will be considered a “high band” with respect to Qualcomm’s WTR1605L (the other high band is the 2.6GHz spectrum that is mostly owned by Clearwire/Sprint). 
  • The approval of the AWS-4 band currently owned by Dish Network. This will add 40MHz of LTE-Advanced within four years. Dish fought hard but failed to keep the FCC from placing limits on the lower 5MHz of their 20MHz upstream channel. Dish will likely only be able to use 15MHz of upstream and 20MHz of downstream. Dish is currently mulling whether or not they will start their own cellular network, or sell the spectrum (and see a huge windfall, since the spectrum was bought out of bankruptcy for around 2B but could fetch a price as high as 6B). This spectrum is at 2Ghz and 2.2GHz, and is considered a “mid band” like PCS and AWS.
  • The approval of the PCS-H block (10MHz, 5+5), likely to be sold to Sprint in an auction by the end of 2013. This 5MHz block will allow Sprint to increase their LTE downstream channel from 5MHz to 10MHz. However, power restrictions on the upper 3MHz of their upstream channel may force them to keep their uplink at 5MHz (again, not a big deal, bandwidth is in demand from the internet to the user, not the user to the internet). This is an extension of the currently used PCS band.

It was a great week for wireless warriors – 70MHz more for LTE will be available to various companies to increase throughput and user speeds. The downside is that it’ll take 3-4 years for all of it to get here for the major metro areas (40% population), and 6-7 years for it to reach 70% of the population.

New Feature + Minor Data Update

I added a new feature – the ability to deep link directly into band/block data. Simply add #band/band-code/block-code to the end of the URL to have it automatically load that data up. So for example, to load up AWS-1 block C, add #band/aw/c to the end of the URL. FWIW, I’m using “AZ” for the two letter band code for AWS-4 (“AW” is AWS-1, so I figure “AX” for AWS-2, “AY” for AWS-3 and “AZ” for AWS-4, makes sense to me).

I also updated the database to move all the WCS licenses from the various companies that sold them to AT&T. Not all the WCS licenses moved, and Sprint still owns several large blocks in the southern US (notably, both of the ones covering Dallas, AT&T’s home turf. I didn’t move over the few AWS licenses transferred from Nextwave, nor did I include any recent 700MHz transfers made recently.

December Data Update

Data updated for December, doesn’t look like there were too many changes. Still having duplicate data issues for the Leap/T-Mo/Verizon license swap. I updated the carrier maps as well as added one for T-Mobile + MetroPCS.

Battle Royale: PCS-H vs AWS-4

I’m not entirely sure if this is technically possible (I think it might but I’m not sure), but here is my idea so far on the whole Dish/Sprint PCS-H vs AWS-4 battle royale with the FCC.

  1. Move Dish uplink up 5MHz (2005-2025MHz).
  2. In a concession to Dish for dealing with the inconvenience, also give them PCS-J block (really AWS-4 “C” block, now 2000-2005/2175-2180MHz), subject to transmit power and other restrictions to keep it from interfering with PCS-H.
  3. Auction PCS-H (presumably to Sprint, for around $1B) with rules about filtering off at the high end (2000MHz) necessary for adjacent small cell compatibility.

The idea is that the new AWS-4 “C” block could be used for small cells sometime in the future, adjoining Dish’s existing spectrum, and buffering the uplink from the PCS downlink. Transmit power would be limited on the 2000-2005MHz range, but that should be tolerable for small cells (and maybe just one 3MHz uplink channel). Since smaller cells are a part of LTE-Advanced anyways (HetNets and DASes and all), this spectrum will likely be useful to someone, and it might as well go with Dish now rather than the FCC sit on it for 5 years and we have another squabbling match in 2017. As long as Sprint and Dish can play nicely together about avoiding interference  then they can both get something from this deal.

November Data Update

I’ve updated the bands and metro area (but not carriers, I swear I’ll get it done next month – work+school+social life means I don’t have much free time). Also you may need to clear your browser cache if the new bands.json file wont download.

It appears that the Verizon/T-Mobile AWS swap is included in this months data, which makes the maps incredibly messy (also I think the FCC hasn’t removed some of the old data – so it may come up as shared when its really not). Also included is some of AT&T’s 700MHz B/C block transactions with smaller carriers – their goal with this is to acquire 700MHz spectrum in rural areas where they have wireline service.

Updated Data for October

I updated the band & metro area (but not carrier) data for October. I wrote some code that downloads and processes the data, so future data updates should be a bit easier. The data includes the transfer of SpectrumCo and Cox spectrum to Verizon, and Verizon’s exchange with Leap.

Spectrum Crunch?

Tim Farrar pointed out that bandwidth demands aren’t increasing on a per-device basis (though as more people get smartphones, overall traffic continues to increase). To me it seems obvious this is a result of data caps, and given the performance I’ve had with my iPhone 5 so far it seems like the spectrum crunch is easing (I’ll reserve final judgment until after Christmas and the holidays to see how the LTE networks hold up). The telecoms are pushing hard on things like LTE-Adv, hetnets, microcells, and 40MHz LTE, but what good is all that coverage and 25Mb/s data speed all if you still have a 2GB or 3GB monthly cap for $30/mo? Or 10GB of data to share between 4 people for $120/mo?

Given that both AT&T and Verizon in most areas are only using a 10+10MHz LTE network (20MHz total, or about 20% of their total spectrum capacity), and speeds are still fairly fast for me while out on the go (10Mb/s down, 10Mb/s up on AT&T for me), I don’t see what the big deal is over spectrum anymore. Yeah, we’ll need a little more spectrum, I’d venture to say about another 50MHz per major carrier before 2020, but as long as the prices for data transfer caps stay where they are, people are going to adjust their habits to keep their monthly bills in check. And as the unlimited data plans fade away (due to devices failing and upgrades requiring capped plans), those users are going to have to check their data usage (no more spending all day watching or listening to movies on Netflix over the cellular network).

With the plans currently in place, all four major carriers probably have enough spectrum for the next 4 years. By 2016 the picture will look quite different — AT&T has 20MHz of LTE now, and will have another 20 with WCS, and then whatever it can refarm on PCS (10MHz per market), so an average of 50MHz of LTE per market. Verizon will have 40-60MHz of LTE in metro areas – 20 in 700MHz and 20-40 in AWS. Sprint will likely end up with 20MHz nationwide (PCS G+H blocks) plus what it holds in the SMR spectrum (17MHz avg nationwide), plus Clearwire’s TD-LTE. T-Mo will have 40MHz LTE in the AWS band in most major areas, plus I expect them to pick up more spectrum in the future AWS-3 (extended AWS-1) and the 600MHz incentive auctions. Beyond the big four, we’ll also see Dish Network try to get in the game with their 40MHz of spectrum, building a pure LTE-Adv network, plus whatever they may pick up in auctions.

So why not divert some of the spectrum away from cellular carriers, and towards metro-area wireless broadband? I’d love to see the 70MHz in the 1.3GHz spectrum identified in the recent PCAST report not used for cellular but rather for metro WiFi. Lower frequency, better propagation, set stuff up on light poles and let people hook up to it, use an updated protocol that can handle 100+ devices, etc. Some of the spectrum identified might be better suited for backhauls to sites as well (3.6GHz), rather than tower to handset.

There is still a lot of work ahead for the transition to LTE, a process that will take a long time (3G networks wont be shut off until the early 2020s). And there are spectrum needs for the next few years to manage that and other future transitions. But with the growth in mobile traffic trailing off due to carriers successfully altering consumer behavior, the worries about a future wireless apocalypse doesn’t seem realistic anymore.