NBN: Turnbull's "New Technologies"

Summary

Turnbull's "New Technologies" aren't new and aren't economic. Generously, they're "lip-stick on a pig", not useful and really hard to do. We seen many major technology Disruptions in the last 60 - 75 years, Copper vs Fibre is just one.

What do the leading innovators in the IT/Comms sector do when testing? Google run Fibre.

Worse yet, at the height of the Internet Boom in 1997, Paul Allen, co-founder of Microsoft, took a punt in tech-mad & densely populated California on "pico-cell" Wireless with Ricochet. Not even he could make it work. "New" Old Technology like G.fast and "100m Wireless Loops" are technical and economic duds - if they work, they're more expensive than fibre, if they're cheaper, they won't work.

Introduction

Let's first be clear, there is only one technology new to the Customer Access Network (CAN): Optical Fibre. It can be organised in multiple ways: time-shared between multiple (16-64) premises with GPON, frequency-shared with WDM (Wave Division Multiplexing) or simple unshared, active ethernet.

What's "new"? By definition, it is "not yet used". Fibre Optic, not Radio Carriers on Copper like DSL technologies, are real new technologies.

Optical Fibre transceiver technologies

Current CAN max speeds for GPON are 1Gbps, using 2.488Gbps transceivers. Mature decade-old production transceivers are 10Gbps, with Dense WDM, 32 colours/frequencies can share a single fibre.

Since 2010, the next speed evolution, 40Gbps and 100Gbps, has been standardised and these are now in production. The hard work with Optical transceivers is now long-distance (thousands of kilometres) and, strangely, getting signals on and off circuit-boards. It's been decades since very short distances like 100m and speeds as low as 1Gbps have been 'cutting edge' for Fibre transceiver technology.

GPON (1Gbps), 2008, and 10Gbps Ethernet, 2002, have shown themselves to be more than enough for most applications. There is no pressure to increase speed from the user community: CAN Operators and server builders. The drive to build 40Gbps/100Gbps came from backbone and long-distance Network Operators. CAN Operators have multiple, cheap options to increase network speeds: most simply they can adopt volume production WDM 10Gbps transceivers.

DSL Technologies

Turnbull would have the public believe that the interim DSL technologies are undergoing rapid development and these "new" technologies are quite wonderful. This is a ludicrous fantasy. Turnbull is either misinformed & deluded, or informed and disingenuous and/or deliberately deceptive. He describes the evolution of [PDF, wikipedia] ADSL1 (1.1Mhz), ADSL2 (2.2Mhz), VDSL (12Mhz), VDSL2 (12, 17 & 30Mhz) and G.fast (expected 100Mhz) as if it were going somewhere and useful.

Digital Subscriber Line (DSL) technologies are a hack - the copper installed for telephones is low-specification. It's designed to support a 1925 specification and only work up to 4kHz. Turnbull's beloved G.fast, is not yet a standard, will attempt to work at 100Mhz or 25,000 times above the design speed. That's like taking the original 1959 Mini 850cc 25kW engine and asking 625MW from it - that's 10 times more than the 64MW that the largest Rolls Royce turbine can generate.

Scaling up existing Technologies

Could you rework an 850cc engine like that? Possibly, but the technology you'd wrap around it would look awfully like those really high-tech jet engines. It couldn't be reliable or have any longevity - like the minuscule reach of VDSL2 & G.fast, it would have a laughably life. With unlimited funds, you might do it as a bet or demonstration, once and at best for a minute.

This is the reality of Turnbull's "New Technology" - old, slow and horribly expensive. They aren't commercial unless you have a barrier to entry to protect - like the cost of running new Optical Fibre.

Other Technology Disruptions: Smartphones/PC's, Jets in Aviation

We've seen exactly this technology change-over problem many times before. In computing, in 2007 the smartphone, and in 2010 Tablets, started to dominate over PC's (see Asymco chart below).

In aviation, after around 50 years of piston-engines, the era of jet-powered commercial aircraft started in 1952 with the Comet, 1957 the Turbo-prop Electra, 1958 the Douglas DC-8 and Boeing 707. These replaced the piston-engined Constellation, DC-7 and Boeing 377. Wikipedia summarises thusly:

The advent of jet airliners, with the de Havilland Comet, Boeing 707, Douglas DC-8 and Convair 880, rendered the piston-engined Constellation obsolete.

Jet engines, even after only a decade of use and development, beat the pants off the very high-tech piston engines of the day. Manufacturers of piston-engines could've rightly used all the arguments of Turnbull: look at all the great improvements we've made, or are adopting.

In 1955, they could say, like Turnbull in 2013, "these New Technology engines are far superior to anything we've built previously" and like Turnbull be completely right but wholly irrelevant.

The Airlines saw the economics of jets (faster, cheaper fares, higher, safer, more reliable, fewer delays) and demanded manufacturers build them. This was the same changeover that occurred in the Military from 1945 to 1950, where fuel-consumption & operational cost wasn't of primary importance.

Node Economics

There is a crushing economic argument against Fibre to the Node, even for owners of the Copper:
every time you halve the Node-Premises distance, you need to build 2-4 times as many Nodes. Nodes have a minimum cost (power, fibre, cabinet, permits, civil works, ...) and very quickly Node build costs overcome everything else.

In Australia, the rarity of high-density housing, even in Metropolitan areas further raises costs and lowers ports-per-Node. For the Fibre NBN, there's 148,000 street-km serving 11-12M premises: an average of 13m per premise (streets have 2 sides, so 25m house spacing).

Turnbull's G.fast proposal, for 100m, means they can't cross the street, there can only be 8 premises per node to get anywhere near his claimed speeds. Anyone care to price building one million nodes? Right there is around another $30 billion, a point Turnbull doesn't emphasise, forget the price of those expensive modems you and I will have to buy out of our own pockets.

G.fast can only be economical in tower-blocks [not the average 8 or 9 unit block] and high-density housing areas. Which is just a few percent of our housing market, even in Metropolitan areas.

Turnbull knows, or should know, these figures. That he's not raising the issue of tiny economic roll-out zones when touting his "New Technologies" is well past disingenuous.

Other "New Versions" of Old Technologies

If you think "so G.fast won't work, we could use Turnbull's 'Wireless on every second street pole'", think again. Exactly this sort of Wireless system was tried in the Internet Boom in tech-mad California: Ricochet went bust. It, and modern WiFi, 5G and "4G picocells" have exactly the same scaling and backhaul problems. If its not being done by someone rich and crazy in the USA, it's not technically nor economically feasible.

Google has time, money and talent on its hands to experiment with. They've gone for 1Gbps active ethernet Fibre in US cities. Recently they've flown a proof-of-concept ballon-as-satellite over New Zealand for very remote and inaccessible areas.  When the smartest, richest, goofiest company in the world dismisses all your technology options what do you say if you're Turnbull: Listen to the guys building what we like, everyone else is wrong!

Wrap Up

Is Turnbull the fool he appears to be? (A: No, not in my opinion, but what's his game?)

He can't separate fact from fiction, plays endless word games (like swapping "New" Technology with "Old") and simply omits discussing anything about the sole reason for all these phoney technologies: Cost.

Dealing with yet another Technology Disruption is a straight-forward Technical and Economic decision, not one of Public Policy nor of Political Debate. This Policy intervention hasn't happened before in Australia. There was NEVER a reason for the general public to know, or care, about Nodes, DSL and "last mile" copper and fibre. This is purely a side-show for Political gain.

We're only in trouble now because the Howard Government - to which Turnbull was elected and served as a Minister - failed to act. These people were told very plainly in 2005 by Trujillo that Telstra was a distressed asset (it's Telephone business was in "meltdown") and that a quick-fix was available, Fibre to the Node, but required immediate action, not sitting on their hands as they'd done and continued to do. Fibre to the Node may have worked in 2005, with a 18-24 month rollout.

Abbott and Turnbull, under Howard, are also directly responsible for not structurally separating Telstra before the 2006 T3 float, when they had the chance and it was possible and being called for.

The NBN and this whole "Copper vs Fibre" debate is more than 10 years too late (PSTN phone lines peaked in 2001!) and if Telstra had been allowed to execute the early 1990's plan of CEO Frank Blount, out of normal revenue, then in 2010 the whole Customer Access Network would've been Fibre.

Other Reading

Read what these often sympathetic media commentators say about G.fast.

Tony Brown in SMH (it's only economic if you own the Copper loop):

Offering G.Fast as part of a fiber-to-the-node solution at around 200 metres loop length would be around 70 per cent cheaper than delivering full fibre-to-the-home. (200Mbps at 200m).
Offering G.Fast as part of a fiber-to-the-building solution would be around 30 perhaps cheaper. [It's $1400 to build Fibre + $1000 to Telstra. G.fast may save $800 if you own the copper - that's more expensive for NBN Co. who have to buy the copper.]

Josh Taylor of ZDnet (you really have to push the envelope, and it's not reliable nor long distance);

The International Telecommunications Union is moving toward making G.fast a DSL standard by March 2014. G.fast promises fibre-to-the-premises (FttP) speeds over short distances of copper through a combination of DSL technologies, including pair-bonding, vectoring to eliminate cross-talk on VDSL2, and "phantom mode", which creates virtual pairs between copper pairs.
The high-end speeds of up to 1Gbps can only be achieved over very short distances of copper, between 100 metres and 200 metres.

The Register (it's complex, expensive, short-range and will kill FM radio):

The ITU Study Group 15 meeting gave first-stage approval to Recommendation ITU-T G.9700, designed to protect broadcast services such as FM radio from interference caused by G.fast deployments.
That's an important constraint for the new technology. To achieve its impressive performance over even short distances, G.fast needs more than 100 MHz of RF spectrum, compared to the sub-megahertz RF needed by ADSL and the megahertz spectrum used by VDSL variants.
While the 100 MHz-wide profile enables the very high bitrates, it's also what limits G.fast's transmission distance – and there's another problem the standards setters had to solve.
FM radio stations occupy spectrum from around 87 MHz to around 107 MHz – well within that used by G.fast. And, of course, even if they're underground, the copper plant provides a long antenna. That means any emission from the cables is going to destroy FM radio transmissions.
To avoid this, the kit (or the operator) will have to be able to “notch” around frequencies that are in use in a deployment area.
“Our intention, our goal is to approve the first version of G.fast, a 100 MHz wide profile, early next year,” he says in the video. That will then pave the way for interoperability testing before chipsets become available and manufacturers get busy on equipment. That would enable kit to ship in 2015.
Interestingly, van der Puttin seems to view G.fast as the last gasp for the copper network, saying in the video that the next step is “probably” fibre to the home.

Asymco's chart of the rise and decline in market share of PC's.