Wednesday 21 November 2012

NBN: How to fiddle the numbers on FTTN to get "quicker, cheaper, more affordable"

Mr Turnbull keeps pushing his line: the Coalition NBN, presumably an FTTN modelled on the UK, will be "Better Broadband, Cheaper, Sooner, More Affordable". Most recently in a speech to an Innovation Bay dinner.

I will try to deconstruct what the phrase/promise could mean, how that could be achieved and what that would mean to individuals and the Nation over the project lifetime of 30 years.



The Promise
  • Better Broadband
    • Without setting judging criteria, this is meaningless and vacuous: For whom? Where? For how long?
    • By what yardstick will it be "better": bandwidth, install cost, total cost, link quality, availability, upgradability, premium services, ... Or does it come in a variety of colours?
  • Cheaper
    • The intended implication is "lower Capital Expenditure". For whom? For what? When? In total, per-year? For the Government or NBN Co?
    • Allowing external investment by, say, Optus and Telstra and British Telecom would lessen the intensity of Government expenditure, and may be called "Cheaper".
    • It's also very easy to deliver a 'cheaper' project by reducing the deliverable to say 50-66% of Urban premises and no businesses.
    • Does it include the foreshadowed upgrade of services to fibre links to active Nodes (MDSA cabinets)?
    • Mr Turnbull frequently uses the term "economic", in the sense of "cost-effective", when discussing his unspecified NBN Policy.
      • But just what does that mean?
      • The IRR, ROI, Cost-of-Capital and Duration of service radically affect the definition of "break-even" and "cost-effective".
  • Sooner
    • This implies the project finish date will be earlier than a full-fibre rollout.
    • By one day: that's meaningless and insignificant.
    • A total project time of 3-4 years for fixed-line connection of 93% of premises, now that's a promise!
    • Or does it just mean, "we'll shuffle priorities so those whom we deem more worthy/needy will get their 12-25Mbps early so we can declare "Mission Accomplished!" early while we continue to dribble out services for another 10, 15 or even 20 years". 
  • More Affordably
    • Is another way of saying "Cheaper", if the interest on the CapEx, not OpEx & Maintenance, is the major factor in wholesale pricing.
    • Will there be, for the privileged few, a direct Government subsidy, rebate or tax-deduction for their yearly NBN-based charges? That artificially lowers charges and creates notional "Affordability", yet is complete B/S.
      • The nastiest way to provide this "benefit" would be as an Introductory Offer of 12-36 months. Get people on-board, claim the Political Points, then abandon them with no way back.
    • Does it include the premises connection and activation fee?
    • What costs be shifted from the network builder to the subscriber?
    • Of course, without a full GPON FTTP, where costs are fully spread across all subscribers, as a reference or base-line, we've NO way to know what it would have cost.
      • It will be trivial to fudge the numbers to show that GPON services are more expensive, therefore our FTTN services are more affordable. Promise Delivered!
      • The comparisons won't be for equal services, but highly selected.
    • Cross-subsidies (FTTP to FTTN) will be hidden in the accounts by the newly appointed CEO/CFO, sympathetic to the Coalition.
The Delivery
  • The UK's BT/Openreach "Infinity" project has been suggested by Mr Turnbull as his model, but never confirmed. There is plenty of wiggle-room and opportunity for weasel words like, "You made the assumption, I never said that"... With that in mind, I'll proceed.
  • The UK project is for:
    • The company building the FTTN, Openreach, is the wholesaler who already owns the Copper Customer Access Network. It's free for them to connect nodes and take-over customer lines.
      • In Australia, Telstra owns the copper lines to the premises.
      • While NBN Co has a right to use Telstra ducts, pits and premises, it hasn't bought the copper, not the right to use it.
      • Telstra is not a charity or Not-For-Profit Enterprise. Nothing is on record to say they'd agree to sell the copper or lease it under an ULL agreement cheaply.
      • Telstra commented in Feb or May 2012, with the agreement of the SSU and NBN Co SAU they were "spectacularly agnostic" about each Political sides' NBN plans.
    • 16M of 20M premises, only 66%, not 93% of premises. As the NBN Implementation Plan showed in spades, there is around a 10:1 cost differential between the easiest/cheapest premises to pass or connect and the most expensive, before wireless or satellite is used.
      • They are passing 4M premises/year, or 80,000 premises/week, or 320 nodes/week.
      • An additional £600-1200MM is being provided by the Government for out-of-area connections.
    • The nodes notionally cater for 250 premises at 25Mbps, VDSL2, 800m.
      • Australia has a much lower population and premises density than the UK:
        • Only a very few medium-to-high density inner-urban areas will achieve 250 premises/node.
        • Most nodes will only get 32-64 premises per node.
        • Where will the "economic" line be drawn, 4-16 premises per node? higher?
      • For the same rules, now explicitly confirmed as a minimum target by Mr Turnbull, we'll need:
        • 75-90% of the road-distance of fibre laid, compared to the current GPON FTTP.
          • Running fibre along the (guesstimate) 50km of existing ducts and pipes of ~1,250 major exchanges gives 60,000km of Fibre.
          • If 25%
        • and 300-350,000 nodes, with full construction, planning permission and 240V power-supply (and metering) as well as service connection.
          • The TLS-2005 FTTN to 4M premises, 1600m, 12Mbps ADSL2, estimated 20,000 nodes.
          • Others estimate 70,000 nodes, but that's only a scale up from 4M to 12M premises at 1600m.
          • Halving distance to 800m to double data-rate will quadruple or worse the number of nodes.
        • A project time of 5 years will see 60,000 nodes/year or 1,200 nodes/week, four times the construction rate of the UK project!
    • They are only "passing" premises in the first-fix.
      • "Passing" is "Do Nothing" at the premises. It's build a node and connect the wires from the pillar, presumably without destroying current ADSL services.
      • "Connecting" a premise involves line-conditioning (a major expense in the Telstra 2005 plan), cutting the existing copper link and getting a working VDSL circuit, at the guaranteed minimum speed, for the premises and installing a Network Termination Device in the premises, complete with 240V power and battery backup.
        • This is the level that the current NBN Co Plan is rolling out to.
        • It's unclear what the constructors will do if, at connection time, the subscriber line is unable to support the guaranteed minimum speed:
          • refuse the service
          • run new "Cat-5" lead-in and retest.
          • run new copper cable end-to-end, charging full or partial costs.
          • run fibre from the node to the premise, charging full costs. 
      • "Activating" a service is making the connection "live", testing it, install/connect/configure the Customer Premises Equipment, provision the services bought with the various Service Providers and test each service at the Customer Premise.
    • The NBN Co plan is currently for ~120,000 road kms of network to be installed.
      • 75% will be underground, all of which will be in 100mm or 50mm buried conduit,
      • with 50mm lead-in to the premises from connection 'pillar'.
      • 25% will be strung "aerially" on power poles, requiring agreement with Power Distribution companies and payments, including pole replacement/upgrade ($20,000/time). The NSW negotiations have been reported to have broken down.
    • Someone familiar with the TransACT rollout in Canberra, circa 2002 said:
      • The VDSL electronics were very expensive then, and would now be considerably cheaper. Prof. Reg Coutts, a member of the Expert Panel that in Feb 2009 reviewed the Rudd/Conroy FTTN tender, commented that "over half the cost of an FTTN is in the electronics in the Nodes".
      • The cost of installing a node is dominated by the fixed costs of planning and site survey/selection, approvals/documentation, power-, fibre- and copper-connections, and multiple "truck-rolls" for construction, install and test.
        • The cabinet/backplane and electronics installed are the only variable cost.
      • TransACT passed 40-60 premises per VDSL node (a 300m rule), yet they only installed 32 port nodes, possibly only partially 'populated' to further reduce costs.
        • With an FTTN, because of the low initial connections, the nodes can be radically under-dimensioned.
        • Additional nodes can be installed alongside the original node to cater for higher connection rates.
          • Construction costs will be lower because of already established services,
          • but the essential "on-demand" nature of upgrades makes the pre-unit construction costs 2-4 times higher than a "mass manufacture" rollout.
      • In "Phase 1", TransACT budgeted $40MM to pass 55,000 premises (Avg. $800/premise). As they used the power poles of one of their partners, lets assume a zero cost. Cost proportions were possibly, a guesstimate only:
        • Local Fibre loop: 5%
        • Copper Customer Access Network (aerial): 15%
        • VDSL Nodes (50 premises/node: 1,250 nodes): 20%
        • Phone "maxi-nodes" (500 premises/node: 100 nodes): 10%
        • Transit fibre to Central Office:10%
        • Central Office equipment: 10%
        • Backhaul from Canberra to Sydney/Melbourne: 10%
        • Customer Premises lead-in, connection, Network Termination: 20%
    • Any xDSL service with one or very few active modems on shared copper bundles will perform reasonably.
      • As activations increase, not even near 100%, mutual interference and "cross-talk" seriously degrades performance of all services on a node.
      • There is no simple fix for this: it is inescapable physics.
      • xDSL only works because not everyone is using it.
      • If there is 50% take-up, the service will collapse.
    • The sleeper issue is node-upstream capacity leading to congestion and 
      • Initially, quite cheap upstream ethernet interfaces (100Mbps) can be used.
      • Visiting each node to upgrade upstream links is very, very expensive.
      • This is well documented with Telstra's RIM/CMUX deployment (FTTK) in the Gungahlin Experiment.
        • The service provider is not obliged to provide adequate throughput or latency at any time, especially during "busy-hour", only to provide the agreed access line speed.
    • Maintenance costs can be hidden and ignored: they become the result of inefficient practices, incompetence and poor management/execution at NBN Co - nothing to do with Our Fine Plan.
    The Fiddles

    The Coalition can achieve their NBN Policy Objective by:

    • Defining "Better" to mean "lower Government expenditure".
    • Defining "Cheaper" to mean "lower Government expenditure" and moving NBN Co to a Public Private Partnership with Telstra, Optus and possibly other players, in return for access to Customer Access Networks and Capital Investment.
    • Offering NO Business services, only residential.
      • "Business" services may be provided by Telstra or Optus as "premium services" over Point-to-Point Fibre.
    • Covering only 6-8M premises FTTN (50-66% vs 93%)
    • Still using a 1600m rule and Multi Service Access Nodes (MSAN) with 3 services available:
      • 12Mbps ADSL2+ for 75% of subscribers
      • 25Mbps VDSL2 for 25%
      • Fibre, at "full cost recovery" for connections
    • Deploying 50-60,000 nodes, not the 350,000 needed for full 25Mbps coverage.
      • Measuring "premises passed" not "premises connected" or "active".
    • Setting a roll-out schedule of 4 years.
      • With "continuous network upgrades", adding nodes at 800m where there was demand and costs low.
    • Declaring $1250/premise to be the "uneconomic" connection level.
      • Fixed Wireless, Satellite or existing ADSL services will be the alternatives.
        • Exactly as "2400 baud modems" were considered acceptable minimum connectivity for Rural and Remote users when ADSL was common in the Cities.
    • Transferring as many costs as possible to the customer at connection time.
    • Claiming, "if people want Fibre Connectivity, they can all have it, if they are willing to pay for it. That is the Free Market in operation. We are not here to subsidise household expenditures, but to provide the infrastructure necessary for our National Economy, which we have done much better and more efficiently than the other mob."
    When the Network becomes popular, causing degraded VDSL speeds, uplink congestion and outrageous latency, NBN Co becomes the whipping boy "due to their incompetence" and they'll be privatised "because that will remove all the fat and make them more efficient".

    No comments:

    Post a Comment

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