Tag Archives: nbn

Should NBNCo be scared of gigabit LTE?

I saw a media article today about Telstra demonstrating one gigabit LTE and claiming this was somehow superior to the NBN. I can’t let a claim like that go unchallenged.

First, some history. In the early days of analogue mobile phones (AMPS) operators broke their radio spectrum into 21 blocks of voice channels and designed their networks to services “cells” from three“base stations” connected to antennas with overlapping coverage on three towers or buildings.

Every cell could see one to three towers and this allowed for a “handoff” between towers. There was no data service. Every call required a channel and that channel would have little interfere from another call on the same frequency in a distant cell.

I found a picture that illustrates this:


(Source: https://en.wikipedia.org/wiki/File:CellTowersAtCorners.gif )

The frequency reuse factor, k = 21 for this scheme. So you can use no more than 1/21 of the spectrum you bought in any one cell. By the end of AMPS carriers had dropped k to 7 at the expense of more interference but it improved the commercial yield on their spectrum investment by a factor of three. The CFO loved that. 

GSM, being digital, improved things such that nine or twelve sets of channels of digital data could form cells in a similar manner. Frequencies can be reused much more readily and the overlap doesn’t need to be so large. 114kbps data services can be delivered via this architecture from 200KHz data channels in 25MHz blocks. (You need 2 x 25MHz blocks, one for data UP and one for date DOWN.) You still need _relatively_ low interference so ironically this can work BETTER in a built up area where buildings block the interference. Combining two channels doubles the data rate and that is the service known as EDGE.


(Source: http://www.rfwireless-world.com/Tutorials/gsm-radio-frequency-planning.html)

The reuse factor, k = 12 for this scheme so you can use 1/12 of the spectrum you bought to service customers in any cell. In practice this meant you could service 124 channels / 12 = ~12 X 114kbps or a bit over a megabit from each sector of your base station. This is why mobile tower backhaul was only 2 megabits! 2 megabits for an entire tower with three cells but given the theoretical maximum data was 124 / 4 = 31 X 114 = 4.4Mbps it wasn’t a bad contention ratio. 

3G introduced “spreading codes” or code division multiple access aka CDMA which encode data on a radio carrier by multiplying it by a random digital number at a much higher bitrate than the data. To decode a specific data stream you multiply the received signal by the same random digital number and get the original data back. It’s like black math magic but it works!


(Source: https://goengineer.wordpress.com/2009/10/08/frequency-planning-gsm)

With 3G you can drive the frequency reused down to k=4. You take the 2 x 25MHz spectrum you bought for GSM and turn it into 2 x 4 x 5MHz. (2 x so you have UP and DOWN links.)

Data is transmitted on the entire 5MHz channel in frames that are 10ms long with 15 slots that contain data and control information directed towards users.

The 3G UMTS standard delivers 15 x 384Kbps data slots to the users in each cell. That’s a total of 5.8Mbps which was quite a thing from 5MHz of spectrum at the time.

HSDPA increases that to 7.2Mbps per user largely by squeezing more out of the 5MHz of spectrum  and joining a bunch of data slots together. Over the life of UMTS the modulation rate has been increased to get THREE times more bits out of the spectrum and to use spectrum from TWO towers so the pinnacle of 3G is DC-HSPA (Dual Channel / Dual Carrier HSPA) which is 42Mbps (six times the basic 7 Mbps service). That 7, 14, 21 or 42Mbps is shared by every user within a cell.

LTE is the logical evolution and convergence of these technologies.

Cell towers traditionally use three antennas per sector. Originally the strongest received signal from each user was selected. Then some beam steering was added to concentrate the signal to and from each user. This is called Multiple In and Multiple Out (MIMO). In LTE this beam steering is done for every user UP and DOWN continuously. 

The available spectrum is broken into numerous 180KHz channels (called resource blocks) with the maximum allocation being 100 blocks over 20MHz. These resource blocks are in turn modulated on subcarriers using orthogonal frequency division multiplexing which is rather like vectoring on VDSL2. It’s high-tech and it can carry 1 megabit over 180KHz which was theoretically impossible when I went to school but yields 100Mbps over 20MHz.

Together with MIMO,  LTE’s modulation technique you can set k = 1 i.e. you can use the same block of frequencies everywhere.

That’s “LTE”, a hundred megabits more or less shared by all the users receiving their service from a given sector (or cell) on a given tower.

What if you had lots of money and could buy lots of spectrum? Well you would find 5 lots of 20MHz blocks of spectrum and send the data in parallel over 500 resource blocks at once. That’s known as “LTE Advanced Cat15” and it delivers a gigabit shared by all the users receiving their service from a given sector (or cell) on a given tower.

Is this enough to scare NBNCo?

No! Because even using fibre to the node NBNCo can deliver a separate 100 megabits to each and every house served by a sector on a tower, potentially hundreds of houses, which would be tens of gigabits.

LTE off mobile towers and roof tops isn’t going to put the NBNCo out of business just yet.

What if you could make your cells quite small? What if they were mounted on power poles or on top of NBN Nodes, if you happened to own them?

Well that’s something that should scare NBNCo and I think it’s one of the NBN defeating or eating endgame options Telstra have in their kit bag.

BUT, all those basestations serving all those cells need backhaul and that needs gobs of fibre. Physically small LTE basestations will emerge as a fibre to the node or distribution point technology but at the same time fibre will always be able to drop infinite bandwidth to individual end users. This suggests that fibre to the premises will inevitably return to favour as bandwidth demands keep spiralling up. 

In the mean time mobility is the killer app and the mobile operators can charge a premium for it. LTE is going to satisfy low volume users but the price will be prohibitive for hundreds of gigabytes per month. That said, they are the very users NBNCo is counting on to cross subsidise the roll out. 

We live in interesting times.

Broadband Speed Claims

On July 27th Ry Crozier published an article about the ACCC consultation paper on Broadband speed claims. It contains a number of quotes from this short essay I wrote in response to some questions he put to me. I have edited that essay a bit and publish it here. It may form the basis for a response to the consultation paper. 

Over the more than twenty years I have been designing, building, operating and managing ISPs the constant demand from users has been “make it faster!”

The pinnacle of this fetishising speed was the National Broadband Network, which would deliver fibre, the thing everyone saw as synonymous with speed, to everyone’s home.

Sadly selection for one characteristic often comes at cost to another. While everyone obsessed with how many megabits or even gigabits of bandwidth would be delivered to their houses very few people listened to those of us who questioned what other parts of the network would cost to use and how they would be monitored and managed.

The NBN’s 121 points of interconnect added great cost and complexity to retail service provider networks as well as the duplication of cost during the migration from legacy networks to the NBN.

I believe the key issues for RSPs since NBNCo started designing and building it are:

  • The high cost of Concentrating Virtual Circuits per megabit creates a huge incentive to provision them scarcely.
  • The numerous points of interconnect scattered all over the country impose large standing costs with very few customers, certainly at first, again creating a huge incentive to provision them scarcely.
  • NBNCo refused visibility to the utilisation of their Passive Optic Network or their “transit network” but still insisted that RSPs didn’t need to see that information because they would manage it to suitable SLAs.
  • The worst case scenario would be that a change in NBN technology or project goals would see an RSP with customers spread across their own DSLAMs, wholesale DSLAMs, NBN FTTP and any other access technology in the same region, each with its own overheads.

We are seeing the effects of scarcity of NBN RSP POI backhaul and CVC bandwidth in many parts of Australia. For every example of an end user with a high speed port seeing slow downloads there is an example somewhere else of a user with brilliant performance. We are seeing the effects of the worst case scenario where few customers are downstream of a POI and the initial “free” 150 megabits of CVC hasn’t been exhausted. Those customers are the lucky ones! They have abundant bandwidth in their access network so their Internet experience is governed largely by the quality of their ISP’s network.

In my time as a regulatory manager for an ISP I had numerous arguments with the enforcement branch of the ACCC about “speed” claims. They hated “up to” and could not understand that ISPs had no visibility to the Telstra copper that would be used to provide the Telstra wholesale DSLAM service.

Readers may remember iiNet and Internode publishing line sync heat maps of Sydney and both organisations published network traffic graphs for much of their lives showing how hard the network was being used. Ironically those heat maps became one of the reasons the NBN came into existence and with the NBN there is no transparency to access network performance at all.

Readers may also want to ponder the limited availability of prawns and oysters at the all-you-can-eat salad bar.

While I have great sympathy for the ACCC’s position that ISPs should be able to inform customers of what “speed” their Internet service will work at even with an abundance of bandwidth in a private access network the entire global Internet is not under the ISP’s control and the performance of individual services on the Internet will vary massively. While this is not a reason for ISPs to not disclose their traffic management practices and utilisation it is certainly something that makes it a “wicked problem”.

Traffic shaping is all about picking losers. All that giving a packet “priority” means is it experiences less delay within the network. The least worst case for a packet is to be pushed back in time until a hole in the data stream can be found. This is like a spatula applying icing to a cake. The worst case for a packet is to be shaved off like a plane flattens a piece of wood by removing all the bits that aren’t flat. The effect of this on your Internet is to slow your page loads and file downloads, cause your video player to “buffer”,  your videoconference stream to break up and your voice over IP to stutter. It might even cause some applications to stop entirely or be unusable.

In our with-us or against-us world it is very hard to have a rational discussion about these issues. I’m not laying “blame” or suggesting any player is operating maliciously in the ecosystem.

Retail ISPs can buy global Internet access in capital cities for a few dollars per megabit. They have no incentive to create artificial scarcity in this layer. This isn’t the place to look for a problem with NBN performance.

Operators of legacy DSLAMs have some mixed incentives. Their entire network has been declared obsolete by the very existence of the NBN. There is little incentive to make capital expenditure on improving the backhaul capacity of those networks for the (hopefully) months of operation left before FTTN migration starts and relieves the demand.

NBNCo will doubtless claim that all the performance problems end-users have with RSPs would go away if only RSPs would purchase enough $15 to $17.50 per megabit CVC. Or should I say “up to $17.50”?

In politics you never have an inquiry unless you know the likely outcome. This review is going to make it plain that RSPs need to purchase more CVC. This is only affordable for them if retail ISP prices rise or if the NBN operating company reduce their charges.

Given Telstra warned earlier this year that NBN costs would reduce their Net Profit by around $2b pa once the NBN is rolled out this enquiry is likely the first phase of a plan that will see a significant rise in the price of NBN delivered ISP services in Australia.

It’s time for the NBN company to reduce the CVC charge significantly and to be more transparent about the link utilisation within its network. That will make it clear to consumers which RSPs have sufficient capacity.

Everything on demand?

Grahame Lynch, publisher of CommsDay wrote a piece yesterday that got me thinking. I wrote the following as a reply to his post to Facebook.

The future of lean back entertainment is everything on demand. Broadcasters today can still earn good to great money for advertising on live sports events but it’s a race to the bottom for the filler ads on the reruns of Gilligan’s Island on the digital dividend junk channels. I’m not currently buying ad spots but I believe the News shows and early evening still command a reasonable premium because they are watched live. Other entertainment shows are being recorded on PVRs and the ads skipped. The PVR industry, under the influence of broadcasters under the Freeview brand have tried to limit the amount of storage in PVRs to 1TB because they know every show stored in the box is another lost opportunity to display ads during live shows. When you add VOD services operated by the PVR vendor there is an added incentive to reduce recording because it impacts VOD revenue. Along comes Netflix with the equivalent of a infinite back-catalogue of TV shows and old movies. Binge watching ensues.

The ISPs deluded themselves that they would get to clip the ticket on video services but when the opportunity to do it via their monthly access charges came along they stuffed it up. They _are_ dumb pipes and every time they try to be anything else it’s either by being highwaymen holding customer performance to ransom or by building the sorts of bundled-with-crap services only telcos can dream up. To win, build a set of products that customers need, price it to make a profit, provision, rate and bill it correctly first time every time and provide awesome customer support services. But the commercial model of the NBN makes it very hard to make a profit unless you have vast numbers of highly profitable low consumption customers. Many of those are quietly shifting to mobiles as the PSTN is turned off in NBN service areas.

Fixed line ISP retailers are going to have to raise their retail prices or face sharply falling profits.

Years of begging the ACCC for increased margins by forcing the reduction of Telstra’s wholesale prices in a market with a dominate ex-incumbent with three parallel private access networks have left them with no way out.