Sunday, 28 April 2013

NBN: Estimating Node costs and numbers

This topic is too hard to research in fine detail, so this is all I'll do on it.

The presumed node cost of $8.1 billion, 9M lines at $900 each, is in the same range of the 2009 Telstra proposal. The cost of electronics in the nodes is only a small fraction of whole FTTN cost. While the electronics have followed Moore's Law and improved greatly in Price/Performance, the other costs are dominated by labour and haven't.

  • the proportion of the brownfields network where it needs replacement with direct Fibre, and
  • the number of nodes required, as the overheads are high.
There's a calculation at the end of this piece where 154,000, not 68,000 nodes would be required.
At $30,000/node, that's $2.5 billion extra. At $55,000, just the nodes cost $8.5 billion with up to another $8 billion is network costs.

These numbers aren't reliable, they are "sanity checks" only. Without seeing the supporting figures, I think $8.1 billion for an FTTN is optimistic, but not impossible.

This ACCC FOI document  via Grahame Lynch, contains a good summary of various proposals.
A spreadsheet of various cost estimates over time, see the 'nodes' tab.

Grahame asserts a node will cost $30,000, complete but without line cards, the 2006 Allen Consulting figure is $50,000 for infrastructure and $43,000 for the node, or $93,000, three times his estimate.

68,000 nodes = $2 billion
9M line cards = $750M

There's an estimate that from Analysis/Mason that 81,000 km of fibre will be needed for nodes. This will cost $20-$25,000/km to rollout:
Node loop fibre = $1.6-$2 billion

NBN Co noted that the transit network, upstream from the local Exchanges, will cost $4 billion all up. So 70% of it would cost $2.5 billion or more.

Reconfiguring the copper network: $500M (guess without source)
Rehabilitation/Remediation of copper network. In 2005 this was estimated at $4.7 billion for only 4M services. I don't find that estimate reliable. Later Telstra does come up with $12 billion for 90% of premises. This work entails a whole lot of work removing RIMs/CMUXs, loading coils, bridge taps (whatever they are) and improving the copper & joints.
I'd allow a modest $3 billion for 72% coverage, because poor copper will be replaced by Fibre.

So this back-of-envelope ballpark figure is:
$2.75 billion for nodes + $2 billion local fibre loop + $2.5 billion transit + $3.5 billion for work on the copper:
 $10.75 billion, which is close enough to $8.1 billion. (within 20%)

An extract from a previous piece:
  • Full Node Costs, both number and Cost-per-node are unstated in the plan. Nor is the critical determining factor, the distance rule stated: Is it 400m, 800m or somewhere in-between?
    • "60,000" nodes have been spoken about,
      • maybe at $30,000/node + $75/line-card + my guess $3,000 for 10Gbps GBIC's (4/node: dual uplinks and 2-per-end). 
      • $1800M + $650M + $180M
      • ~= $2.5 billion
    • There's also 81,000 km of fibre to be laid and jointed at $25,000/km
      • $2 billion
    • And a share of the $1.7 billion transit network and PoI's
      • Plus extra for Exchange switches and other equipment.
      • 1,000 exchanges with $250,000 of switches and ancillary equip.
        • $250M, minimum
    • Scaling up from 1500m to 800m and onto 400m is not simple:
      • whilst the area per cell is notionally "radius squared", four times as many cells are needed when reducing the distance by half, in practice other effects come into play:
        • For 1500m to 800m, I'll assume a 3.5 ratio allowing unserviced areas.
        • For 800m to 400m, I'll assume 2.1 ratio to allow for majority "same run".
    • The best data we have are the 2006 Analysis Mason estimates  of 38,457 nodes (20723 for "Big 5 cities" + 17734 for a full FTTN)
      • This was for 12Mbps ADSL2+ at 1500m with 33% services direct from exchanges, costing only 40% of lines off Nodes.
      • Without a 40% improvement from network reconfiguration
      • Nor with the 28% FTTP coverage in the Coalition Plan
        • or 62% of FTTN coverage via Nodes.
    • Fewer services can be directly connected to cheap exchange lines as distance reduces by geometric area:
      • 1500M: 33.3% = 66.66% node-equivalents in the field
      • 800m: 8.3% = 91.7% in field
      • 400m: 2% = 98% in field
    • Using the 2006 Analysis figures, 57,685 node-equivalents were needed.
      • For 800m:
        • 57,685 * 3.5  (scale factor) * 0.6 (reconfig) * .917 (field) * 0.62 (fibre)
        • = 68,872 nodes or $2.9 billion with previous cost estimates
      • For 400m:
        • 57,685 * (3.5 * 2.1) * * 0.6 (reconfig) * .98 (field) * 0.62 (fibre)
        • = 154, 567 nodes or $5.7 billion with previous cost estimates
        • Tesltra commented on the FANOC proposal, cutting the copper and adding new nodes would cost $1 billion extra at least, plus massive disruptions. That was for only "the Big 5 cities", add 50% national.
    • The Coalition does promise that in the second upgrade, to 50Mbps in 2017-2019:
      • only 90% of FTTN premises will be guaranteed 50Mbps.
      • Is that an 800m rule with vectoring, a software upgrade, or
      • through a 400-50m rule?
        • Transmission lines reduce 'exponentially' with distance.
        • If 25Mbps is available at 400m, only 20Mbps at best is available at 500m and 6Mbps at 800m. Distance losses are much higher with VDSL than ADSL2 and 1, so attainable rates will be lower.
        • $3.5 billion difference in 800 vs 400m.

No comments:

Post a Comment

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