Saturday, 27 July 2013

NBN: Defining "Fast Broadband", answering "Why Fibre?"

Politicians bandy around fine-sounding terms like "very high speed broadband", "very fast speeds" and "superfast broadband"trying to make distinctions or spin away the weaknesses of their own proposals.

We don't have a language police, so can't stop them speaking meaningless rubbish. Nor is there an accepted definition of "Fast Broadband". Al Gore's Information Superhighway, didn't pop-up in mid-1990s, the term was around for nearly 20 years by then. As an author noted in 1993, "in the 1980's, 56kpbs was fast" and in 193 they were talking mega-bit (1Mbps).

Previously, I've included a discussion of perceived "speed" and attempted a scalable definition:

  • Fast broadband can be defined as "Actions in my application seem to happen without delay". 
  • Fast Internet was, is and will continue to be, "I don't notice waiting for the next thing to happen".
The proponents of extracting the last ounce of value from the telephone copper network, designed only for 4kHz bandwidth, not the 17Mhz-30Mhz required by VDSL2, are always challenging the desirability of an Optical Fibre Customer Access Network, the "local loop".

Their question is: "What (current) applications need Fibre?". My answer:
  • Anything that has potential for unlimited bandwidth demand: video streaming, audio, still images, downloadable movies, video conferencing, tele-health, Geographical Information Systems (GIS), gaming and backups/archival. Optical transceivers at 100Gbps are already in production. In the laboratory multi-Terrabit (Tbps) systems are being developed.
  • Anything that requires a consistent Fast Broadband experience, no matter what demand is placed on it. Affordable, production Fibre transceivers already scale out more than a 1,000 times.
  • Anything that can benefit from Moore's Law driving cost-per-Bit/second down by 1,000 or more times in the next 10 years alone.
Copper solutions meet none of these requirements:
  • Copper has inherent limits that are already limiting services provided. While twisted-pair copper has been demonstrated at 1Gbps over short distances, it's up against three problems:
    • Physics: radio frequencies in copper lose power more quickly at higher speeds. You can go faster, but the cost is, you sacrifice distance. It looks exponential when you plot it.
    • Electronics: you can extract higher bit-rates from existing systems, but at an exponentially increasing rate. To get 10% more speed, you need double the processing and power.
    • Economics: to achieve 1Gbps to most premises, even at modest distances, would mean complete recabling: replacing all the copper. The whole Copper-v-Fibre debate is about avoiding recabling.
  • To provide wide-scale access-rate increases as demand increased, solely with existing copper, means costs increase extraordinarily fast:
    • Extensive upgrades of the DSL electronics at both ends.
      • You have to double the price of the electronics to get a modest rate increase.
    • Radically increasing the number of nodes, to decrease loop length.
      • To halve the loop length and double, or better, access rates means quadrupling the number of nodes.  Not only do you need much more expensive electronics, you need to quadruple your investment..
    • The construction & power-supply overheads, or 'civil works', of nodes stays roughly constant: although you build nodes around one-quarter the size, they cost 90-95% as much. Pushing copper to higher speeds removes the economies of scale and increases deadweight - overhead that doesn't make revenue.
    • You will have to "remediate" (tech-speak for replace) a larger and larger fraction of the existing copper to achieve promised higher rates over ever shorter distances.
      • The whole Copper-v-Fibre debate is about avoiding recabling.
  •  The reverse of Moore's Law applies to DSL. We know what Copper twisted-pair designed for Data Networking looks like: it's the blue Cat-5/6 cable you use on your computer. That's a much superior specification than your 1925-compliant telephone cabling, and even then it's only specified to 100m. Ever since the first version of ADSL, the price of each successive generation of faster DSL has gone up and the useful distance has shortened at a frightening pace.
    • The marginal cost of extra speed, the increase in price for an extra 1Mbps of line access rate, is extremely high and increasing at an accelerating rate.
    • This is the reverse of Moore's Law: price per transistor goes down while compute power increases.
    • The Price/Performance gap between Fibre and Copper is already 100:1, depending how you calculate it. This will only increase, but at an accelerating rate as faster Copper costs more and higher-speed Fibre costs less.
We know from 3G/4G Mobile Phones and ADSL that are higher speeds become available and volume charges ($$/GB) declines, usage increases. This is the economic definition of Price Elasticity of Demand.

The Data Networking market is confirming it has high elasticity, just as Telephony had before it. Drop the price a little and demand for the service increases, a lot. The cost structure of Telecommunications is dominated by Fixed Costs: Interest and Depreciation, then network operations and maintenance. The costs, to a Telco, of a service are barely related to what you use, most of their costs come from just building and running the network: anything they can do to increase total revenue increases their profits.

If a Telco drops prices by 5% and increases usage by 10%, they've just made 4.5% more profit. If they drop price by 10% and get 20% more usage, they earn 8% more profit while spending nothing.

We know from 30 years, 1965-1995, of International Telephone experience with O.T.C. that these economics are correct. People love to talk! The Telephony market was highly elastic: people are both price-sensitive and they get used to spending more money on phone calls.

The Silicon Revolution, and Moore's Law, drove down the cost of International circuits steadily for decades. O.T.C. reduced the real price per minute of its phone calls for three decades, until merged into Telstra. They became immensely profitable on the basis of Price Elasticity and Moore's Law.

The same conditions applying now for Data Networking, over Fibre, not Copper.

There is a demonstrated high elasticity in demand for Internet access and Fibre is the only technology riding Moore's Law to lower input costs. These are the economic and technical drivers that Fibre, and only Fibre, can ride.

No comments:

Post a Comment

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