Tuesday 6 November 2012

NBN: Costing the Coalition FTTN. It's not Good News.

Mr Turnbull, in his CommsDay address, seemed to be proposing the Paul Fletcher, "Wired Brown Land" FTTN:
  • 25Mbps guaranteed, VDSL2
  • 800m max wiring (approx. 500m road distance)
  • we can infer 40-60 subs per node, consistent with the 160 subs/node of the first Telstra FTTN.
  • The "golden screwdriver" is "Multi-Service Nodes": field upgradable from copper to Fibre.
So, what's it going to cost to roll this out to 12M subs, the 93% of premises addressed by NBN Co?

Will Mr Turnbull achieve the Coalition tagline, "Better Broadband, Cheaper, Sooner, More Affordable"?

Mr Turnbull would like to have us believe his FTTN is "three to four times cheaper", as he started saying to Leigh Sales on the 7:30 Report Interview.

Which is an ambiguous statement: is that of the full $37.4B project, the $26.7B "fibre and transit network" or the widely estimated $12B construction cost of the Fibre Customer Access Network?

In 2009, one of the seven person review panel that rejected the FTTN tender bids, Prof. Reg Coutts summed the economics up:
“Essentially to go down the FTTN road would mean something in the order of, greater than 50 per cent of the capital being put into digital cabinets in the suburbs," he said. "They then become an obstacle to the final solution… fibre-to-the-premise. Fibre-to-the-node was not a stepping stone to fibre-to-the-premise. In fact, if anything it would put it backwards. The second reason, of course, is in no other market have people proceeded with fibre-to-the-node other than an incumbent. It is a solution that is the right solution for an incumbent that has a copper infrastructure.”
In the CommsDay piece, Mr Turnbull says on the topic of costing an FTTN:
After all, we know the approximate budget and timetable such a strategy entails, because a nationwide Fibre to the Node upgrade has been painstakingly costed, and the logistics of rolling it out carefully analyzed, no less than eight times during the past eight years.
Only, we don't.

I've looked at both the Telstra Aug-2005 and the G9/FANOC proposals: the public versions are very approximate and roughly similar.

The Telstra proposal was only for five major metro areas and then of the 5.4M premises, they cherry-picked just 4M, 12Mbps, 1.5km max, ADSL2 supporting two-thirds with 20,000 nodes and the remainder directly from 450 exchanges. Nodes handled ~160 subs. Build time was forecast at 3 years, cost vaguely set around $3-5B.

The NBN Implementation Study provides much more detail and better estimates. In Exhibit 1.3, they layout the variable cost of deploying fibre to premises: it's not a fixed cost, but a complex curve starting low and with a steep 'knee' after the 80th percentile. This same cost curve is going to apply to an FTTN network, but much worse in lower-density housing areas. It's uneconomic to build a node with 5 ports.

Estiamting from the Telstra proposal with 30,000 nodes (10,000 are in exchanges) and a 1.5km rule:
Halving the distance to 0.8km, and using simplistic geometry, increases the number of nodes four-fold to cover the same area, or 120,000 nodes total, and 110,000 nodes in the field, with 40 subs/node.

That's only for one-third of the NBN: lets scale that up to 12M subs, simplistically assuming it's all linear, which we know for the edges of all networks is untrue and in rural/regional areas with lower housing density is massively optimistic.

We now need three times more nodes: 360,000 with 30,000 subs directly connected from ~1,500 exchanges.

We can check sanity check that figure with NBN Co's rollout plan: 148,000 km "road distance" with 25% aerial (and a total of 206,000 km of GPON). So 110,000 km of cable trenching (at $55/m, $6B)
To pass 12M subs, there are 81 subs/km, or 12.3m per sub. Streets are dual-sided, this might be closer to 25m. Backtracking is also not needed with GPON.

An 800m wire-rule is roughly equivalent to 500m "road distance" because of hierarchical wiring distribution, with our 12.3m/sub, we have 40.5 subs/node. That's close agreement.

The 2002 TransAct network was VDSL1, 100Mbps and 300m. They spent $80M passing 55,000 premises, for an average $1,500/premise. This included new aerial copper ('cat 5') as well as 2 types of node: Analogue Phone and VDSL.

Extrapolating from the 12.3m/premise, we can guess 25 subs/node (VDSL) and 2250 nodes.
If the customer network accounted for 75% ($60M) of the deployment and central office equipment the rest, and nodes 40% of the cost, then they paid $24M for the nodes: $10,000 each ($400/port - half for the line card, half for the cabinet). Which sounds in the ballpark.

Scale that up to 40 port nodes, say $200/port-card and $7,500 for cabinet, $15,000/node and $5,000 to install. $20,000/node or $500/port.

For 12M ports, that's $6B. Half the cost of the GPON Customer Access Network on a most optimistic estimate for the same coverage.

But what we're missing there is the cost headlined in the Telstra-2005 proposal: remediating ('conditioning') the copper lines... And also the $200 boxes, the copper versions of the ONT, attached to every house.

We're past $9B and closing in on the Fibre GPON price. Where is that "three or four times" lower price? You can't achieve that with real-world electronics and labour, unless you're going to provide something else entirely to most people: like fixed wireless.

And rollout-time: Telstra thought 3 years for the 4M easiest subs, 10 years for 12M subs might be realistic. Surprisingly close to the GPON schedule.

There are other concerns that must be factored in:
  • We have a 30-year design life for the network: maintenance dominates the operational cost of an FTTN.
    • For the copper network, we're starting with a degraded and old asset, well past its prime.
    • It will need a lot of maintenance, increasing over time.
    • The GPON network starts new and has lower inherent maintenance costs.
    • Running fibre those last 800m in an Ad-hoc manner is going to be the most costly, inefficient and slow way to do it.
    • Plus there's a triple upgrade cost: the Node card, the fibre run and replacing the Customer Premises Equipment, the ONT. Unless that was over-engineered initially with dual Fibre and Copper interfaces.
    • Copper does catastrophically fail, the conductors can wear out. Think of the 1998 Auckland CBD power crisis. Despite four-way redundancy, the city was shutdown for 5 weeks when all the cables failed together. The lesson - copper conductors can wear out "just sitting there".
  • We start the FTTN with VDSL2, at the end of the DSL product lifecyle:
    • 25Mbps at 800m is possible now, it may increase four-fold in the next decade or two: 100Mbps on ageing, decayed copper, might be possible, but that's pushing it.
    • Compared to 1Gbps out-of-the-box for the GPON network, with an available four-fold upgrade available now.
    • For a 25% cheaper network, you get one that goes 40-times slower. Hmmmm, not a good value proposition.
  • The operating environment for GPON is much better for electronics and people:
    • Would you rather maintain the electronics for 12M ports in 2,500 secure, environmentally controlled exchange buildings with racking designed for ergonomic access, 
    • or in 350,000 nodes sat at ground level and exposed to plants, pests, insects, vandals and climate extremes?
    • FTTN technicians will literally spend most of their time driving around, not fixing faults.
    • The failure rate of electronic components doubles with every 5-degree celcius rise in junction temperate. Unless those nodes are air-conditioned, the service life of the electronics will be severely impacted. The cost and complexity of adding environmental controls, and remote monitoring to all nodes would be astronomical: TransAct uses passive cooling and puts up with shorter service life.
    • What's the cost of providing clean, filtered and stable power, with good backup, in a small box (using ~200W), versus powering 10,000 GPON lines and other equipment?
      • The per-port cost difference is 10:1 to 100:1.
  • Fibre GPON works for many kilometers, VDSL doesn't. GPON can scale to low-density areas, such as regional and rural environments, especially if you have key stakeholders willing to pay the initial connection fee.
    • NBN Co allows "out of area" connection to subscribers if they pay the full cost for it.
    • This initial connection can be shared between a group of subscribers,
    • later on, additional subs can connect and the initial sub is proportionately reimbursed.
Voters and subscribers in regional areas will want Fibre GPON, not an FTTN "with a promise to upgrade". Almost every regional town has a long history of some necessary infrastructure that politicians promise to provide every election, but never have.

Infrastructure Upgrade Promises fly very badly in the country: that dimension alone has the potential to lose the Coalition the election.

If the plan is to move subs over to Fibre within the 30-year service life of the FTTN, why not just do it right from the start?

There are solid technical reasons to build a Fibre GPON NBN from the beginning.
Yes, the initial capital expenditure is higher, but not by very much.

Within a decade, the additional, and increasing, costs of maintaining and replacing/upgrading copper will make it a much more expensive proposition.

Fibre GPON is not a "gold plated" solution, its the only sensible solution both operationally and economically for a new network. It's new technology at the beginning of its product lifecycle, funnily enough like iPads and PC's versus Mainframes, the new technology will do more things, faster and eventually cheaper.

Unless of course, the Coalition plans to scrap the NBN ("pause it") or rollout fixed wireless to 50% of the population. None of which will make the voters happy.

No comments:

Post a Comment

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