Interesting discussion. There are a few points which might help people understand this a little better.
The concept of the NBN is twofold: (a) separate wholesale service provision from retail service; and (b) replace the ailing and obsolete (so-called) copper network with fibre, where itâs cost effective.
Objective (a) more or less acknowledges the flaw in the Howard governmentâs communications âcompetitionâ strategy, where Telstra was gifted a large piece of public infrastructure in order to inflate its stock value when floated for (partial) public sale. This created a monopoly and crippled competition. The government raised $8 billion from the Telstra float. It cost $11 B to buy back the last mile wires and ducts from Telstra for NBN. In other words, Telstra sold off the most problematic part of its network infrastructure, along with a monopoly right, back to the government. A win for its shareholders.
(b) replacing the (so-called) copper network with fibre is a long term investment. In-ground telecommunication wires were designed for analog phone signals. Digital signals over wires travel as high and low frequency sound signals, which (IIRC) travel electrically as current modulation. Electrical signals over wire attenuate relatively quickly compared to the length of the wire, mainly due to resistance. They attenuate further due to signal leakage (impurities in the insulation, or flaws in it, or joint loss). This attenuation limits the signal speed - much like a toy car pushed along a flat surface slows to a stop due to friction⌠Fibre, OTOH, carries light through a low resistance (transparent and reflective) medium, glass, with a slower rate of signal loss per distance. It takes much less electrical power to drive a light signal over fibre for 10 klms than it takes to send an equivalent electrical signal over wire for the same distance. Fibre in the ground will last (undisturbed) for a century or more, whereas the wires will corrode once the insulator begins to break down after a few decades, give or take.
A single strand of dedicated fibre can carry at least 180 Gbit/sec of data traffic if driven with adequate terminating equipment. In other words, itâs much more future proof. It requires far less electrical power to run a large fibre communications network than it takes for an equivalent electrical comms network. Most of that consumption is in the switching infrastructure.
The biggest cost of upgrading the Australian telecommunications network was the rollout of fibre, especially the last kilometer connections. It is much cheaper to run fibre to every house in an area in one hit than it is to do some here; some there. The Labor plan was flawed, because they tried to do the complete replacement too quickly. There was a shortage of skills required, leading to bodgy, over-priced work, and inevitable schedule delays. Otherwise, it was the right strategy - move everybody to fibre, except for those too widely dispersed to do so relatively economically. The result would have had a much, much longer service life.
The Coalition turned a once in a century capital upgrade into a short term hodge-podge of technology soup that will be obsolete within a decade, and requires a much bigger army of maintenance people to service. Political stupidity. Laborâs solution would have undoubtedly cost more and taken longer than the Coalitionâs crippled approach. However, all that last kilometer wire still needs to be replaced, and all the node equipment has a very, very limited service life. It carries a much higher future liability cost. It is the economics of stupidity writ large.
IBM sells mainframes that range widely in performance. They build them fully configured. You pay for what you ordered. If you need an upgrade, you phone up, an engineer sends a code to the mainframe, and bingo, you now have the next model up, and a hefty new bill. IBM knows that on average, customer needs increase over time. They worked out that it costs more, and thereâs much greater risk of introducing faults with in field physical upgrades - so they ship fully configured, and you pay for what you need. Additional performance is a phone call away. IBM is run by accountants, not politicians.
A data network is something vaguely like a road network, in that both are multiplexed. Data multiplexors can be time division or frequency division, but in either case they allow shared access to a transport network that appears to be dedicated, more or less. Traffic lights provide time division multiplexed access to an intersection. Frequency division is similar to having footpath for pedestrians; a cycle lane for bicycles; a bus lane for public transport; and general lanes for cars, trucks and motorcycles, with barriers that prevent vehicles from changing lanes except at traffic controlled intersections. The analogy gets a bit obtuse when considering end-user performance.
ISPs lease network capacity from NBN. I donât know the granularity of it, but for TCIP traffic (internet) it is pretty much ALL multiplexed across the network. That means NBN sells a portion of its average carrying capacity to each ISP in an area. There are various trunk charge back mechanisms - it has probably changed since I was involved, and those details donât matter. A single strand of fibre can carry a very, very high data speed, depending on the equipment connected at each end (the terminating equipment is the potentially expensive part). A small modem can receive speeds of up to 100Mbits/sec, and drive speeds of 20 or even 40 Mbit/sec.
Over FTTP, it will transmit at its capacity, and then the service will be multiplexed by the ISPâs allocation at the NBN connection point, via software and parameters. The speed that you buy will be a portion of the shared performance the ISP (RSP) has, which will be an average based on an algorithm. Your service can be upgraded to gigabit internet simply by NBNco replacing the terminating equipment, and possibly incorporating some currently unused (dark) fibre into the backbone, by plugging the ends in, and upgrading the terminating equipment in your house.
Over FTTN, the modem is limited by the line length to the exchange (node) which converts between electrical signal and light from that point. The same multiplexing applies from there. FTTP runs over twisted pair, with (IIRC) a pair for send and a pair for receive. The signal overcomes interference by being measure as the relative difference between the pairâŚitâs brilliant for analog phone callsâŚas designed. FTTN is conceptually the same as ADSL. It runs at higher speed, but over shorter distances, running a similar kludge, but more cleverly. Itâs still a kludge. As you run faster and more clever data transmission protocols over twisted pair, from dial-up to ADSL to FTTN (VDSL), at higher speeds, you still get packet retransmissions due to line errors. The longer, and older or more degraded the line, the worse the performance. I took over four hours to upload a 20 MB report to a CBD solicitorâs website over FTTN one night, because FTTN was breaking down. It had to be delivered to the court by 9 am that day. I got to bed at 4:30 am. It was line errors, not contention, that caused the problem. Moisture in the ground; joins in the wires; packet failures galore; upload time-out.
HFC carries signal over (much more reliable and efficient) coaxial cable to a point equivalent to an FTTN node, performing the same conversion. It generally has higher performance and lower data loss, but similar vulnerabilities to FTTN - power requirements. Coax is shielded, and so not as susceptible to interference as twisted pair.
FW operates over radio, with (possibly with limited frequency division multiplexing), and certainly with time division multiplexing - mostly, thatâs done by software. FW is very cost efficient over a wide area with usage low density and good signal. Once it approaches its usage threshold, its performance plummets, a it doesnât matter what you do. Theyâd need to start making direct connections to overcome the RF contention. Mobile broadband is not really different to FW. It works on the same basis - RF traffic to a tower, with a performance plummet point - hopefully well above the number of potential subscribers connected.
FTTC is a variation on FTTN, but with much shorter wire operational distance, so less attenuation, and higher possible speeds.
Back-end (trunk) performance of the network is largely determined by NBN capacity; to a degree by the ISPâs proxy server and comms infrastructure capability; and thereafter, to the international network performance of whatever you connect to.
There will be some differences between the ISP sharing algorithms - some might a higher number of end-user subscribers, and then the software algorithm will queue user traffic for longer to smooth traffic peaks. Think of the number of light changes you wait for in a car at peak hour versus off-peak. What they are doing is sharing a resource.
I have a 50Mbit/sec download; 20Mbit/sec upload service from Internode over FTTP. My modem runs at 100Mbit/sec each way routinely, between it and wherever it measures its signal speed to, so I get more than I pay for, in a manner of speaking.
If you experience broadband issues because you donât have a good enough signal, it might be possible to have an external antenna, or a dedicated, high performance broadband modem in a better position to get the signal. Then run WiFi from there. Youâd have to check with your provider to find out if there are any legal restrictions. A repeater (RX/TX signal amplifier with antennae) might be illegal; IDK.
If you waded through this drivel, I hope it helped.