Monday, 2 September 2013

NBN: Fibre-on-Demand, at best $7,500/premises, more like $45,000/premises

Peter Martin tweeted that the ALP's claim of $5,000 to connect Fibre-on-Demand was "Mostly False".

I disagree and tweeted a short calculation. I find it surprising that the ALP has never made this argument:
$5000 is way low
NBN Co charge ~$110,000/km for "extensions".
800m = $88,000
At 26 prem/km, 20 per leg.
10 >400m
= $9000 ea

Here's how I came up with those figures.

Key Assumptions

The expectation is that the same rules as "Fibre Extensions" will be applied by NBN Co to "Fibre on Demand". Either a group of subscribers split the up-front cost between them, or one lead customer pays for the extension in full and then is given a rebate by NBN Co for every additional subscriber that connects using that cable. It's unclear to me just how costs are calculated for 2nd and subsequent customers, and if any rebates apply.

It's unclear whether for Fibre on Demand NBN Co would follow the Fibre Extension policy of full cost recovery from the first customer/group. This has major cost implications for NBN Co, and as a commercially viable business, they can't enter this to make a loss. They could assume a minimum connection percentage for all nodes and charge standard connection fees, or run a strict "cost-recovery" model.


ZDnet, May 2013, reported the first "Fibre Extension" had been completed:
NBN Co CEO Mike Quigley said at the time that it would be "very expensive" to deliver; one estimate that a business owner in South Australia received quoted that 1.3 kilometres of fibre would cost him AU$150,000.
$150,000 for 1.3 km is $115,384. There is too much uncertainty in both denominator and divisor to use this figure. A more reasonable figure is the rounded down: $110,000/km.

There's another figure needed, the number of premises per kilometre in Australia. Broadband Communities [Pg 22, "Advantage Fiber"] cite,  per km of paved road 26/km versus 24/km in the USA.

The 2012 NBN Co Corporate Plan says there are 148,000 route kilometres and 208,000 km of Fibre passing 12.2M premises, with 8.5M active. This gives raw numbers of 58 & 82 premises/km.

How to calculate the number of households that a single run of Fibre from a node, by NBN Co, could share? The figure lies between 26 and 58 premises/km, but probably 26-29 premises/km.

Reconciling the two numbers shows the routing and access density differences between wire and road networks. Roads need to observe geographical barriers and to provide lots of redundant interconnects in a 'mesh': you don't want to be forced into driving to/from a hub on every trip. There are also strong traffic management issues to avoid congestion at central hubs: road vehicles are different to Telecomms networks.

Copper distribution networks do follow road networks, at least in urban areas which serve the majority of subscribers. There are two problems:

  • For underground copper cables, separate runs are needed down each side of the street to minimise cross-street trenches and the chance of conduit collapsing under traffic.
    • Aerial cable, under power-lines, can easily cross the road with road-side poles, or
    • with backyard poles, as in Canberra, houses on two sides are easily served.
  • Distribution Areas are a hub-and-spoke layout/topology. "All roads lead to Rome", or all wires lead back to a central pillar.
    • Sets of 10-pair cables are run from the pillar at the centre of each DA. Multiple cables travel the same route, allowing for damage and additional demand.

For an access network, the wire-distance from premises to an exchange is always longer than direct "crows-fly" distances because of the many shared trunks, 'splits' and hub-and-spoke Distribution Areas. The total covered distance by the conduits is far less than the mesh road network, but the total distance covered by all pairs is far greater.

It's reasonable for one-off, on-demand installs to be run down each side of the street, reducing the 58 premises/km by half, to 29 prem/km. This is "within the ballpark" of the BC-mag figure of 26hh/km.

Turnbull has indicated somewhere, but not in any formal document, that his VDSL2/FTTN network will initially have an 800m cable-run limit. No statement has been made about how the 400m maximum effective range of VDSL2 Vectoring will be dealt with: either four times as many nodes are needed or the 75% of customers outside 400m won't have Vectoring.

I posit that NO customers closer than 400m will consider Fibre-on-Demand. Those who win the "VDSL2 Node Jackpot" won't be motivated to pay a considerable sum, both upfront and we can presume in additional rental charges as per BT Openreach, for a marginal increase in access rate and NO increase in Busy Hour sustained throughput.

VDSL2, without Vectoring, will provide reasonable accesses rates up to 400m. With the limits of the Node electronics and uplinks, it's very unlikely that much more than 100Mbps will be offered with Fibre-on-Demand. Offering 1Gbps, when the uplink is that speed, is an invitation to Retailers and NBN Co to be savaged, rightly, by the ACCC for "false or deceptive advertising". Subscribers expect sustained throughput to be near the advertised access rate. 1Gbps Fibre-from-the-Node will fail miserably to give reasonable sustained throughput at peak hours due to Contention & Congestion within the Node and it's uplink.

We know from the 19-Apr NBN Co data that the trial Fibre installs cost $4,000 on top of the $1,100 for the lead-in. This is consistent with the $110,000/km figure, giving 27 premises/km.

We also know from the 19-Apr NBN Co data that 31% of consumers select 100Mbps services when it is offered at a higher rate.

  • NBN Co will charge roughly pro-rata for distance for each 800m run, plus a connection/upgrade fee for the Node.
  • Prudent Financial management says NBN Co will only install short fibre runs after a firm order is placed. A minimum number of premises, no fewer than 4, might be set, otherwise "full cost to first connection".
  • Good engineering practice dictates running a single cable along the entire run of copper from the Node to the end of the leg. This minimises later costs, pipe & pit overfilling and follows the Statement of Expectations on "upgrading to FTTP".
  • Cables will conform to the NBN Co standard: bundles of 12 ribbons of 12-fibres (min 144 fibres in a bundle).
The cost of 800m leg: $110,000 * 0.8 = $88,000
Add node connection/upgrade costs = $95,000/leg + per-premises lead-in of $1100.

Per 800m leg, the maximum number of subscribers is 21 at 27 premises/km.
Only those outside the "magic circle" of 400m will ever seek a "Fibre on Demand" upgrade: 10 subscribers maximum per leg.

The current NBN Co "take-up rate" expected for Fibre is 70% of premises passed will become active subscribers.

The maximum take-up,  on this basis, of Fibre-on-Demand would be 7.5 subscribers per leg.

A minimum shared connection cost, for all 7 signing up is:
= $95,000 ÷ 7
= $13,571 + $1,100 lead
= ~$14,500/subscriber
The very best, and rare, scenario will be a two-sides of the street being served.
Still $6,500 + $1,100 = $7,600, minimum.

What is more likely is that only 25%-30% of active subscribers, as shown by the real-world data on take-up of 100Mbps services, will take-up Fibre-on-Demand.

That's at best  2 (1.9-2.25) subscribers per 800m leg, or $45,000/subscriber, 1100 hours of take-home average wages plus $3,600/year in interest at 8%.

For a service that cannot guarantee 1Gbps, why would any subscriber be that desperate for 100-250Mbps?


  1. Very interesting analysis.

    You can come at figures for FOD as exceeding "up to $5000" in several ways - your's is just one of the more detailed ones.

    I've been trying to argue the absurdity of the $5000 figure for yonks, just by applying common-sense economy of scale logic. If it takes $3,600 per premises to supply fibre in a mass deployment(Turnbull's Whyalla-factor bullshit figure), then how much would it be to supply fibre to a single premises as a one-off?

    Well, obviously, you just roll in the construction crew and fibre you up for a grand extra, no questions asked. Malcolm said so.

    I have a ready response for all the whingers who'll be complaining to the "guvmint" over the next few years: you voted for them.

    1. Dedalus,

      Thanks for the comment.

      As you say "you voted for them", great response.



Note: only a member of this blog may post a comment.