Truth: Trial shows why HFC cable is unsuitable for the NBN

99


This article was originally published for Delimiter Members only. In late September 2016, Delimiter ceased publishing new articles. Because of a number of operational and other factors associated with this decision, we subsequently withdrew membership articles from publication. If you would like to see a copy of this article, please contact Delimiter directly with your request. Requests by Delimiter Members will be granted. We will consider all other requests on their merits.


99 COMMENTS

  1. You’ll upset Richard and the LNP “anything but FTTP”cheer squad.

    Not only does this not get close to 100mbps, it was a 20 user trial and the average was only 84.

    As someone who uses HFC I know the pitfalls – during the day it’s 30mbp, at night I’m lucky to get 5.

    Unless they are installing new lines to every home – which would cost between 5-10 billion – how do they expect us to get the speeds advertised.

    When taking into account the new hardware needed for docsis 3.1 – add another 4-5 billion on, it is apparent that HFC and FTTN are not cheaper, faster or sooner!

    • @j likely to upset anyone with even a basic knowledge of the technology or data communications.

      Ovum reports end-to-end Internet throughput (Internet, blog refers to layer-3). NBNCo’s 100Mbps service (all techs) is theoretical peak, deliverable over DOCSIS 3.0.

      A sample size of 20 years (monitored or total?) is too small a sample. 150m HFC users worldwide, 100+mpbs offered in dozens of markets.

      DOCSIS 3.0 & 3.1 support 100mbps VC. Throughput depends on congestion (here we are again). In HFC networks (as GPON) underlying performance is achieved through node splitting.

      1/3rd the CPP of FTTH, 4m added in a few years (would be completed today). Yet the delimiter TRUTH is such technology doesn’t work here.

      • So is the Liberal HFC going to get faster when they add more than 20 users to it?

        If not, what is the relevance of users in other countries.

      • hey Richard,

        you’re right, 100Mbps is theoretically possible here.

        So my question is — why the hell did NBN Co not make sure that it hit that target during its own trial — the trial that was supposed to demonstrate the fitness of this technology for use as part of the NBN?

        Surely it would not have been hard to provision sufficient capacity here to meet the demands of 20 users?

        Are they incompetent, arrogant, or is the Optus network truly a dog?

        • Surely it would not have been hard to provision sufficient capacity here to meet the demands of 20 users?

          You’re assuming they didn’t Renai? That wouldn’t be so bad.

          I’m more worried they did provision it right, and the condition of the equipment/lines is what dragged it down.

        • I’d suggest because to such people, everything is good…

          … oh, umm, except for FTTP, of course.

          Why?

          Well read between the lines, as to why anyone would want, let alone actively promote, second rate technology for them selves, us all here and the rest of Australia.

        • “Are they incompetent, arrogant, or is the Optus network truly a dog?”

          All 3 is my guess.

          The Optus network is a dog because it hasn’t been maintained – this could have something to do with them being a private company and wanting to get the most revenue with the least amount of maintenance (cost) spent.

          During the day my HFC connection hits 30mbps, at night it’s lucky if I get 3mbps – sometimes it’s no usable.

          If private was best we wouldn’t have the issues we now do.

        • Part of me likes to think that they were being 100% honest about the trial and rolling it out exactly as it will be to the eventual end users. Which I guess is a good thing?

      • So…. a 20 household trial of NBN HFC technology is enough for you to deem it successful?

        I wonder how you’d react if Labor policy was designed and legislation implemented after a survey of 20 people was conducted who all said that taxes should be 90% for people earning $50,000 or more per annum.

      • @renai I’m not defending the PR ovum stunt (guess they have to keep their unbelievably large 60 person media team busy).

        Incompetence and arrogance on target.

        However HFC is a very solid last mile technology. It’s application for the NBN will save 4m x $2100 ($8b) in capex, delivery high speed internet many years earlier than FTTH (capturing all the benefits, earlier revenue to covers growing losses). That it wasn’t considered by NBN and their experts an very expensive oversight.

        The evidence HFC technology delivers very high speed internet in many real-world markets (over 150m end users according to Ovum) today demonstrates its suitability.

        NBN is an expensive policy folly. $16b, 7th year and only 15% complete. Reusing existing infrastructure would have seen the fix line network (majority competitively served before) finished today.

        • “It’s application for the NBN will save 4m x $2100 ($8b) in capex, delivery high speed internet many years earlier than FTTH (capturing all the benefits, earlier revenue to covers growing losses). That it wasn’t considered by NBN and their experts an very expensive oversight.”

          hey mate,

          I’ll just point out that the capex argument does not matter, as under all modelling the various NBN models would have all made a return on their investment — the amount you spend does not matter when you’re going to make a profit on that spend.

          As for the ‘years earlier’ issue … so what?

          We’re building a network for the next century here. I’m happy for people to put up with shitty broadband for a half-decade or so if those same people can get the best possible broadband for the next 100 years. Perhaps that’s a little harsh of me … but if your business or your personal life really depends on high-speed broadband, then move to somewhere that has it. Otherwise, you wait until you get the gold class that you can use until 2100.

          The HFC will need to be continually upgraded from the day the rollout is completed, just to keep up with current telecommunications trends. There’s a reason nobody is deploying this bullshit in new installations globally — because it is a legacy technology.

          The NBN company is trying to put a heavy dose of lipstick on this pig, but its own evidence has already shown, in the first 20 user trial, that a pig is still a pig.

          • @renai the capex arguement does matter (cost and revenue critical to profitability). The claim that any model would have made a return is hugely contestable. Models aren’t actuals.

            The original 7% IRR was calculated using the CPP, timelines and revenue presented in their CPs. None of those targets were realised, all substantial negative (from NBNCo profitability perspective). The delays and cost blowouts confirmed in Quigley’s PDF correctly pointing out the MTM SR13 – CP16 blowouts.

            The SR13 significantly downgraded projected IRR’s for S1-S2. Even S6 (MTM) half the original 7%.

            Since then costs for S6 MTM exploded in CP16, revenue hasn’t (today ARPU 4% above original CP projections, other costs exploding).

            The effect of increasing costs well above, revenue on projections is IRR has to fall. My belief (and why you won’t see a Conroy’s NBNCo forecasts through 2022) is even for the cheapest MTM model its below negative, govt already looking for additional levies to make up the expected losses.

            Numbers in CP16 confirming my outlook; $2-3b losses for the outlook covered.

            All Quigley’s CPs (3) have failed, confirmed by actuals. Morrow CP16 under pressure, but never claimed 7% IRR.

          • CP16p16

            “5.9 Long term financial outlook
            In a market as dynamic as telecommunications, the ability for Management or Board to accurately forecast the long term financial prospects is inherently uncertain.

            nbn has a limited factual and operational base for financial projections, with uncertainty in the long term market and competitive landscape, customer usage, Australian Government policy, technology innovation and potential for other disruptive events giving rise to a wide range of possible financial outcomes.

            Using the same long range assumptions as applied in the Strategic Review, the long term financial outlook, based on the Operating Plan extrapolated to FY40, provides an IRR of 2.7% – 3.5%.”

            Chorus in their AR15 identified their weighted average cost of capital at 6.5% (govt equity). Based on this figure (below commercial) NBNCo should lose their GBE status. The ABS should do their duty.

            From experience any company with growing losses projected through 11th year will rarely ever be profitable. No internet upgrade anywhere has carried such losses (for good reason).

          • Richard, you talk about models versus actuals, yet the “actuals” that you use are not those that fits Quigley’s model. Rather dishonest of you, considering the model was based on a ramp up of a FTTH rollout that was instead drastically rolled back. It’s the MTM FTTH model.

            Every actual you use is tainted by this, and as such, makes your argument (or your “outlook” as you call it) about the initial FTTH rollout invalid.

            Trying to link any forecast before Turnbull got his hands on the NBN with today’s figures is … quite simply … incorrect. It’s not the same rollout, and it hasn’t been since the 2013 election, even the FTTH part.

          • @m not true. The numbers presented (again) cover Quigley’s CPs (performance under Morrow improved due to changes to deployment):

            (FY13, FY14, FY15)
            Forecast CP12-15p69 $~22, $~24, $~39
            Actual AR14-15p28 $37.33, $37.34, $40.45 (+4%)

            Premises passed (FY13, FY14, FY15)
            Forecast CP12-15p61 661k, 1681k, 3664k.
            Actual AR14-15p22 227k (-66%), 553k (-68%), 1153k (-66%).

            Premises activated (FY13, FY14, FY15)
            Forecast CP12-15p61 92k, 551k, 1615k.
            Actual AR14-15p22 70k (-24%), 211k (-62%), 486k (-70%).

            Revenue (FY13, FY14, FY15)
            Forecast CP12-15p61 $18m, $120m, $529m
            Actual AR14-15p27 $17m (-6%), $61m (-49%), $164m (-69%).

            You’re wl come to provide any numbers to support your claim.

          • @M even then things like ARPU has been well above the NBN era forecasts for FttP. I will be interested to see that figure for HFC and FttN.

          • Richard … my claim of what? Or is that your standard answer.

            I’ve simply demonstrated that any CP claim you make is invalid, because the previous FTTH deployment is not the same as the current one. Any actuals data you provide is tainted by that change, and is … therefore incorrect. An impossible task really, because in order to do a direct comparison, the original FTTH would have had to continue as planned, which it didn’t.

            So all the FTTH CP data that you provide, any actuals vs forecasts, and the comparisons you attempt to do … are simply flawed.

            The only comparison you can make, is that of the actuals and forecasts made during the MTM and it’s deployment, which as we have all borne witness to, have been pre-loaded by data obtained by reports seeded to deliver results exactly along the lines the Coalition government expected of it.

            Dubious results at best. Flat out lies at worst, and all the shades in between.

          • Hi Simon,

            Agreed. ARPU will be very interesting (and IMO not in a good way) for the MTM going forward. This is where Malcolm’s (now Mitch’s) chosen technology will bite their credibility … hard.

            The crying shame is that it’s only going to be after the accident happens can we point to this lesson as an “I told you so” when we’re standing in the wreckage, rather than preventing it from happening in the first place.

          • @m when do you think the new one started (FY13 / FY14?). I could use FY12 but FTTH only launched SEP11.

            It would appear no numbers can demonstrate the success or failure of Quigley’s CPs, not even his own ARs (FY13). Models assumed over actuals (thinking sadly not uncommon these days).

          • “It would appear no numbers can demonstrate the success or failure of Quigley’s CPs, not even his own”

            That’s exactly my point. Cheers.

          • “CP16p16”

            Say no more, here are the current cherry-picked figures form the yes men (of Richard’s and Mal’s choice)…

            No other figures are of any consequence, because these fit what Richard wants them to fit… period, got it, now?

            My dear friend has now (well sometime ago) even omitted the previous go to doc… the much heralded and required at all costs … CBA, as his exhibit A (and superseded them with CP16) since I highlighted that even the more subservient yes men (good old Henry and Co) had to admit, after exhaustingly trying to damn the figures to no avail, that Quigley’s figures were pretty spot on.

            Of course that’s when I was told the CBA (which was until then the gleaming light of MTM wonderment) can’t be believed as it used NBNCo’s (Quigley’s) figures and it was then dropped faster than Tones and his onion?

            Umm yes, ok, CP16 says, so what? It also has massive blow outs in timeframes and costs too. So which do we believe? Just the bits that suit the narrative Richard.

            Ok thought so…

          • @rizz not true; the for casts presented are from Quigley’s CP, the cba supported exactly the same conclusion. CP16 seperate, highlights the uncertainty and the much lower IRR today expected.

            AR’s (Quigley’s and Morrow’s published) all point to the same conclusion. But my fears confirmed, no numbers are acceptable (why they’re never posted). Quigley’s CP IRR predictions accepted as fact regardless of performance. Costs don’t alter the conclusions, doesn’t matter (increase or decrease).

            To see how such economic modelling is prepared the SA pandas are instructive. The then govt announcing the return to the state would be 10 x money invested, estimated at $630m making it one of the most value added business in the entire state. Oh how we laughed. A within a years the zoo was demanding more money (must be everyone else benefited). Victorian govt today claiming similar results for its new train line…

            It’ll never stop, all actuals rejected. Most onto the next spend, no one held accountable.

        • @R “However HFC is a very solid last mile technology.”

          Telstra has proven that users can get in excess of 100Mb (even then there’s contention issues). The issue is that with a pathetic number of test sites (20) why can’t NBN get there! not in the least why spruik such a test to such a degree when it essentially proves they aren’t making it to even a currently deployed standard?

          “NBN is an expensive policy folly. [….] 7th year”.

          Scarey thing is the LNP have wasted 3 of those 7 years with deliberate delay tactics and politics! Couple that with the asbestos in the pits delays and well half that time is just a wasting :(

          • “why can’t NBN get there!” Telstra also only have ~2Mbit upload speeds. As far as I am aware, HFC, like G.Fast, is an aggregate speed which is then divided between up and down channels.

          • R0ninX3ph – The upstream and downstream channels are separated, in DOCSIS3.0 – it’s only in what looks to be Phase2 of DOCSIS3.1 that they’re going to *somehow* support upstream AND downstream in the same spectrum. I haven’t worked out how they’re planning on doing that yet, exactly.

            Typically in DOCSIS3.0 each downstream channel is capable of 50MBit, and each upstream channel is capable of 30Mbit, on the assumption that DS = QAM256 and US = QAM128.

          • Thanks for the clarification Ian, I wasn’t 100% sure.

            Any idea then why, if Telstra has rolled out DOCSIS 3.0, they still have such abysmal upload speeds?

          • Because they’re artificially limiting throughput, as upstream throughput is more difficult to get ‘right’as a result of only ~40MHz of available bandwidth divided between 4-8 channels in that spectrum, vs downstream where you have ~100-1000Mhz of bandwidth available and up to 32 channels, which are also wider (8MHz vs 6.4Mhz)

          • I’m well aware of the cablelabs specs for 3.1 – but haven’t specifically seen how they’re planning on the symmetrical bandwidth upgrade to the pre-existing D3.1 spec.

        • No-one is ever proven wrong in politics because no matter how bad things are, you can always claim they would have been worse under the other guys, and no matter how good things are, you can always claim they would have been better under the other guys.
          No amount of quoting dubious numbers from wildly inaccurate politically motivated reports/plans is going to convince anyone of anything.
          The fact is that the vast majority of new investment in fixed line networks around the world is going into fibre networks. With nbn planning a 4x expansion in the number of premises connected to HFC, they are effectively building a whole new HFC network, a unique undertaking in the world today. If they were comparing like with like – i.e. aerially deployed FTTP with HFC – it seems very unlikely that HFC would be better in any way.

      • So Richard if 20 to small but the ofcom report of 20 user getting with in 10% of the speed on FTTN was enough.

  2. Go Renai, at least someone in the media writing about this stuff!!

    You must be a thorn in Turnbull’s bum :)

    • Heh I suspect Turnbull doesn’t pay that much attention to Delimiter any more ;) Fifield, however, would certainly be taking note of my work — especially when I am the only journalist sitting in the Senate Press Gallery during the NBN debates, working away on my laptop ;)

    • Once again, Renai won’t be getting Christmas cards from the Libs or NBN this year.

      Wouldn’t it be wonderful if the magical NBN unicorns surprise Renai in a year or two with a retail product that exceeds 100Mbps for all HFC customers, hence making him eat humble pie. (Oh gosh, I really crack myself up sometimes!!)

  3. Surely you won’t have more then 20 people using the internet at the one time?
    I see HFC as a building a good community, you get to know all your neighbours when you politely ask them their internet time is up and time for someone else to have a go.
    Because we all deserve a fair go (depending on your usage time of course)

    • This comment made me laugh out loud.

      I can actually imagine this kind of thing happening in 1990 or so. Reminds me of living at home. Parent picks up the one, no dial tone. “Renai — ARE YOU ON THE INTERNET?”

      • “picks up the one, no dial tone. “Renai — ARE YOU ON THE INTERNET?””

        Always hated getting disconnected when someone picked up the phone, living in a sharehouse at the time the only calls we generally made was for Taxi’s and Pizza and neither would accept a mobile phone call :|

        Ah the golden days of CU Internet

    • What’s wrong with that? It worked for those holiday time-share places that were popular in the 80’s.

      What did the PMG (phone company for those too young to remember) call this kind of service in the old days – a party line. Apparently it was popular for quite a while until the government ran individual wires to everyones house.

    • I, too, used to be on Telstra HFC and was seeing great speeds. How is it actually possible that the NBN company can’t match these speeds after upgrading the electronics in Optus’ HFC cable networks???!

      • They are basically matching these speeds the total speed is the upload and download combined.

        I would give anything to have 30-40 mbps upload

        • “They are basically matching these speeds the total speed is the upload and download combined.”

          Ah, interesting point.

          However, that’s not what the NBN company promised they would do … they are supposed to be committed to 100/40.

          • Whilst the 87/33 and 122/2 are very similar for aggregate throughput, that isn’t how DOCSIS works. (yet)

            D3.0 has discrete upstream and downstream channels, and bonding of throughput only occurs on the bonding groups of each upstream and downstream separately. It isn’t a “combined” throughput like how 54Mbit Wifi-G is both upstream and downstream throughput. It operates in FDD mode.

          • Ian,

            Yes, you’re correct, but there’s nothing stopping NBN Co from reallocating spectrum from downlink channels to be repurposed for uplink instead. Yes, it would be ‘non standard’, but given the size of their project all the major vendors would be willing to reconfigure their hardware to spec.

          • Sure, that would be possible, however you would require a full redeploy of amplifiers, modems, and custom firmware on the CMTS as well.

            Currently, most CMTSs only support upstream channels up to 85MHz, which would actually be a limiting factor due to hardware (duplexers etc) inside the device itself.

            If you’re talking DOCSIS3.1 with synchronous up and down, that technology hasnt left the lab.

            And you’re back to node sizes realistically no bigger than 50.

            It really *really* isn’t as simple as just reallocating channels. There’s hardware limitations, software limitations and actual electrical limitations that need to be worked around.

            How you’re going to amplify upstream signals when it “shares” the same bandwidth ad downstream is a problem I’ve not seen solved yet, unless you remove the need for any amplifier at all.

          • I thought they *were* replacing just about everything other than the coax? Not certain either way, that was just my assumption given existing Optus HFC is demonstrably not fit for purpose.

            Also remember the sky’s the limit in terms of what they’ll be willing to spend upgrading – HFC is an ideological alternative to FTTP, not a technical nor economic one. If they have to replace everything including substantial sections of the HFC coax they will – remember about half the premises within HFC areas don’t have existing cable connecting them, so that’s already a hell of a lot of coax being rolled out. In 2016/17. Coax cable. What a fricken joke.

          • If you’re talking DOCSIS3.1 with synchronous up and down, that technology hasnt left the lab.

            Arris have modems out now that support 5Gbps down and 2Gbps up, so while it’s not synchronous, it probably doesn’t need to be :)

          • Jesus, even at QAM4096 on the upstream, i could only squeeze in 1500MBit into 60MHz of bandwidth, assuming perfectly clear no-noise upstreams. You’d need 45dB MER to even get close to that.

            I’m guessing they’d be using atleast QAM16384, to be able to get that bandwidth to multiple modems concurrently. To get QAM16384 to be reliable, you need atleast 49dB MER.

  4. The idea that someone actually paid for the linked ovum report is very disturbing. If you haven’t read it, don’t bother; here’s what it says :
    * HFC equipment that works in the rest of the world also works in Australia.
    * If you have low contention ratios (20 end users in the trial !) then you will get good (close to 100/40 Mbps) performance.

    Peering through the fog of nbns “transparency”, it appears that HFC is only cheaper than FTTP because they are comparing an aerial HFC deployment with high split ratios to an underground deployment of fiber with lower split ratios (and possibly also with higher redundancy).

    From the ovum report :
    “Within the trial area there are 18,881 premises and a potential HFC base of 4,500 premises (all lead-ins are aerial)”
    It’s unclear what this means, but it appears that most premises in the area do not currently have an HFC connection. i.e. to convert to HFC nbn on DOCSIS 3, they will have to replace all the active equipment, build new lead-ins to most of the premises, and run fiber down half the streets to support node splitting. Of course, doing this will eliminate the cost advantages of HFC.

    • As far as Telstra goes, if you have Foxtel cable, then you can get HFC over the same cable.

    • Doing ad hoc upgrades like this, particularly given the cost of HFC cable and related equipment, would be significantly higher than a full scale FTTP rollout with the efficiencies resulting from economies of scale. The $4k+ cost per premises of FTTP can only be derived from low volume deployments.

  5. So I’ve flicked through the report, and if that’s what was provided to NBNco, i’d be asking for my money back. My clients require reports FAR more detailed in both numbers as well as analysis of the information.

    Some quick figures, which may help some people who may not have the expertise in this area.

    The *maximum* users per node is typically around the 1000 mark. In reality, you’ll see less than that in order to avoid congestion on the CMTS. So what would those figures need to be?

    Lets assume they are using 32 DOCSIS channels, each capable of 50Mbit each (QAM256 @ 8Mhz EuroDOCSIS). That’s an absolute maximum of 1.6GBit per CMTS for Downstream. Upstream is a bit more painful, but lets assume they have 8 upstream channels (typical is 4 currently). Each upstream channel can be up to 6.4MHz wide, and if used at 256QAM results in 40MBit per channel. Aggregate upstream is therefore 320Mbit.

    Now, you can increase your downstream modulation, if your SNR is good enough. QAM256 requires a minimum of about 34dB, regardless of upstream/downstream. You’ll need 39dB for QAM1024, which DOCSIS3.0 is capable of (These figures based on error rates of no more than 1.0E-6 and running SCDMA not ATDMA). That would increase your downstream channels bandwidth to 100Mbit.

    So lets assume, best case, they’ve cherry picked the best quality coax network for this trial. The design has been done well, the SNR figures are favourable, happy days. 40dB+ on the downstream and 35dB+ on the upstream. Let me tell you – that’s no easy feat.

    That gives us 3.2Gbit of downstream on the DOCSIS system, with each modem likely channel bonding at 8×4, giving 800Mbit/160Mbit. Keep in mind, we’re only providing 100/40 via NBN, but that isn’t the only reason for the channel bonding – it also allows for better dynamic allocation of bandwidth across all devices on that channel bonding group.

    Personally, i’d find it surprising that the reason for NBN to be providing their own modems for the trial was to ensure operation at QAM1024 on the downstream. Most current modems only support up to QAM256. Typical CMs currently run 8×4 bonding by default. You can also purchase higher quality CMs that operate at 16/8 and 32/8, typically used for commercial/business services.

    So obviously with these figures you can see that at 1:1 CIR you’ll ‘fit’ up to 32 CMs per CMTS for the downstream, but only 8 CMs on the downstream, if you’re guaranteeing 100/40 per CM. Lets be realistic though, you’re not going to get 1:1 CIR, contention is probably going to be closer to 10:1 or even 15:1, if you want to guarantee that it is less contended than typical RSP contention ratios. NBNCo don’t really want to risk being seen as the ‘bottleneck’ in this scenario!

    So at 15:1, those numbers become 480 on the downstream and 120 on the upstream. But that’s based on the idea that you’re going to have the same contention ratios on both upstream and downstream. Pretty unlikely, realistically. You’re only ever going to see about 10% of your downstream bandwidth being used for upstream – So if you’ve got 1000Mbit of 1:1CIR providing services to users, you’ll only see about 100Mbit used on the upstream, on average. As a result, you can increase your contention ratio for your upstream considerably, almost by a factor of ten. So lets say on the upstream, your contention ratio is 150:1 – that gives us 1200 subscribers!

    Ok, so upstream bandwidth isn’t an issue? Well no, it’s not that simple – You’ll see the numbers above in large networks. But when you’re talking about less than 1000 users, as we are going to see on a CMTS, the economies of scale don’t apply quite as evenly. Lets drop those above figures by half (picked a figure out of thin air), and you’re looking at a much more realistic figure.

    240 on the downstream to ensure sufficient bandwidth for downloads and 600 on the upstream. Downstream of course is the limiting factor here for bandwidth, however upstream signals are far more intolerant to noise – There’s both higher levels of ambient noise at upstream bandwidths, but also a far higher number of devices on the upstream, which all adds up to more noise.

    SO that’s what possible on current DOCSIS 3.0 systems, if you push them to the *absolute* limit. The biggest limit on what is possible is going to be NOISE, realistically, not bandwidth.

    What does this trial with 20 users show? Absolutely nothing. Noise will be low with a node split this small, bandwidth could be supplied with typical 32 downstream channel CMTS running at QAM256, not even QAM1024.

    I would like to see the raw data for this report, as i don’t believe that they are doing 33MBit on average per user. That would require 16+ US channels @ 6.4Mhz @ QAM256. Upstream on DOCSIS3.0 sits in 5-85Mhz, but typically below 20MHz is unusable due to ambient noise. 65Mhz left over for channels, that gives you just under 10 channels of upstream available. Even best case, 80MHz wide divide by 6.4Mhz channels = 12 channels. Some DOCSIS3.0 devices only operate on 5-65MHz, reducing that down to about 40MHz usable, or 6 Channels. You could run some QAM16 or QPSK channels below 20MHz if you wished.

    It is simply not possible to have every user on average use 33Mbit upstream on a single CMTS. Downstream at a peak of 87Mbit on average is possible, on the assumption that you’re using 16 channels @ 256QAM @ 8MHz wide EuroDOCSIS channels. Just. And i can tell you, running significantly more than 4×6.4MHz channels on an upstream will stress the upstream RF amplifiers heavily, introducing more noise in the system. Personally, i’ve run 12 x 3.2MHz channels across a single upstream band, but had to reduce it to 1.6MHz channels and QAM16 for it to operate within noise limits.

    The *only* way I could see these figures as accurate, is if they are *actually* peak throughput within that 6 hour timed period, per CM. ie, if an average of peak of 87MBit was achieved, lets record it as if every subscriber sustained 87MBit across that whole period, rather than just peaking at that speed at some point over that 6 hour period.

    Either way, this ‘pilot’ is nothing more than NBN using their preferred vendor (Arris?) on DOCSIS3.0, and not even to the point where DOCSIS3.0 would be pushed to its limits, with the exception of upstream bandwidth possibilities.

    NB. Upstream of 256QAM is officially not supported on DOCSIS3.0, but it does exist and it can be used. Current models CMs do support it. It typically isn’t used however due to noise as the limiting factor, ie, needing 35dB MER minimum to operate error free @ 1.0E-6.

    • Great Post Ian!

      iirc Optus use DOCISIS and it’s Telstra that use EuroDOCSIS, what difference does that make to your numbers?

      • DOCSIS uses 6MHz wide channels for downstream, whilst EuroDOCSIS use 8MHz wide channels. That’s about as far as it goes.

        Reason being that EuroDOCSIS was design to “operate” nicely with DVB-C used in Europe, whilst 8VSB is used for TV networks in the US.

          • Per Mhz, yes.

            QAM256 @ 6MHz = ~37Mbit, which is conveniently 75% of an 8MHz wide channel.

        • You have the same argument as Matt Rath from Facebook, apologies if you’re not him i just remember him selling the same argument to Malcolm’s facebook page about GPON instead of FTTN. He was obviously ignored as we found it’s his staff that man his social media comments.

          • If you’re replying to me… Then no, i’m not Matt Rath, not come across the name ever.

            If you are replying to me, i’m not really ‘selling’ any argument. Just putting some figures out for people to do with as they wish.

    • Always good to see this kind of analysis in the comments. I’m a software guy so I’m learning a lot more about networks with this NBN debacle.

    • Yeah, i’ve not worked out (but also haven’t looked into it a lot) how they’re planning on doing that. Node sizes would HAVE to be small enough to not require any amplification of the DOCSIS channels from the CMTS, from what i can work out so far.

      Currently, there’s lots of issues with just rolling out D3.1 on a D3.0 network. Plenty of hardware in the way between the CM and the CMTS.

    • They are talking about transmitting and receiving on the same frequencies at the same time. There is a bunch of academic work on this. The basic problem is that your own transmitter drowns out the signal that you are trying to receive. To overcome this, they use specially designed hardware to try and provide isolation and they measure the tx/rx leakage and use knowledge of the tx signal waveform to attempt to cancel it from the rx signal. It all gets very messy and complicated.
      There are no doubt some situations where doing this kind of thing makes sense, but I wouldn’t bet on it ever taking off on a large scale. Cable networks are a dead-end technology and I suspect the big cable companies will provide better performance beyond DOCSIS 3.1 by gradually migrating customers to fiber.

      • The same thing is done on WiFI – it’s effectively an enhanced version of CSMA. Probably throw in some TDMA style time slicing into the mix to try and reduce hidden node issues at the same time.

        I’m sure it’ll work, because TDD already works with WiFi. FDD works better, is all, so long as you’re not needing symmetrical throughput.

        • No, it is not done on WiFi, although people are trying.
          They are not talking about TDD, they are talking about transmitting and receiving at the same time. At any given instant, WiFi either transmits, or receives, but it doesn’t do both.

          • Gotcha.

            I guess the point is, as they’re using ODFM and wide slices of bandwidth (192MHz for D3.1), there are no channels as such.

            Either way, a nightmare for any return path amplification, hence my comment about small node sizes required to remove any requirement for US amplification.

  6. So HFC and fttp users are well catered for ,but a bit of a crock for fttn users eh? What a mess.

    • It’s sounding like HFC premises will be just as screwed as FTTN, maybe even more so if they stick with D3.0!

  7. On a more serious note, does anyone know what the performance benchmarks will be before NBN re-segment a cable because it’s congested, or are you all just laughing at me right now?

    • In theory, it should be when you’re seeing something like 85% peak utilisation, but in reality, i think it’ll be closer to 100%, for an extended period.

  8. It was always inadequate. So after the billions in new hardware they give you less. Talk about a downgrade. will still go down when it rains.

  9. Their 1gbps over cable is a scam because they will have to take back channels from Foxtel. And probably only one person can get that at a time.

    • They might force foxtel to transmit over the internet – although for this to work they would have to have working infrastructure.

      It would be funny if it wasn’t so sad.

  10. An average of 84Mbps is certainly not a result we were looking for from a new $50 billion network. This was meant to be a big improvement to the already fairly high speeds on HFC.

    This is obviously a fraud. The technical practicalities of their alternative don’t work out, so the Coalition is just putting rubbish out there on purpose because they *really* want to offload it to Telstra/Optus/TPG to do it. Immature behaviour from people who claim to know how to do government “properly.”

  11. Worth stating, that they capex cost for the headend(cmts etc) of this, would be upwards of 30-40k… Consider that with a node size of 20…

    • Considering the size of the HFC expansion they are planning (1million to 4 million premises) do you think they’ll save any money in capex compared with covering those areas with aerial GPON ?

      • That all depends on the contention they allow on the links (as a direct result of the node sizes).

        I doubt we’ll ever see D3.1 with synchronous throughput any time soon, but IF (and it’s a massive if) they can get D3.1 as it stands today out and working, even at just 4096QAM and 256QAM, they’ll have enough bandwidth to improve on their test results.

        If their test results were actually on D3.1 and using 4096QAM (or even 1024QAM), i’m extremely disappointed with the results. I’ve achieved the same thing just on D3.0. In reality, the results are only twice as good as D2.0, which is a sad state of affairs.

        Considering the pure upgradability of GPON infrastructure (the cables themselves, and to a slightly lesser degree the OLTs), nothing will ever work better than FTTP GPON, with the exception of point-to-point fibre from someones house to the POI.

        FTTdp is a compromise technology, but it is probably the best option to do it cheaper for now. Not a lot cheaper, but it does remove the requirement for in-room installs if using g.fast or similar. At short loop lengths (under 200m, which should be the hard limit maximum loop length allowed for FTTdp), you can get sufficient bandwidth i would say for the next 10 years, but still allows for a relatively easy upgrade to FTTH at a later point, for a much smaller price.

        Any line faults requiring new cables to be run should be replaced with fibre, in the above scenario, regardless. This should be done at the cost of the wholesaler, which should be legislated.

        We need to stop this short-term thinking… at the very least look medium term (10-15 years).

        As to your original question before i got a little side-tracked ;)
        They’ll be likely using distributed CMTS technology for the coax side of the network, which has the ability to use either GPON (or 10GPON, etc) or active (be it fibre or copper) ethernet. It makes sense to use GPON style technology in that scenario, so they’ll be laying out the aerial (or underground) fibre for it anyway.

        The only reason they don’t want to do any sort of FTTH technology is because it requires in-house installation of equipment, which is a big cost of the overall job. At the end of the day, it all comes down to node sizes. This is why FTTdp is being considered as the ‘upgrade’ to FTTN. The reality is, it should be being rolled out now, instead of FTTN in current deploys.

        And of course, it goes without saying that if we were actually thinking about long-term plans, we should be doing FTTH for everyone – where the cost saving measure comes from not supplying the ONT to the user, but allow the ISP to sell a compliant GPON ONT instead, or let the user pick one up from Harvey Norman. The only issue that then remains, is the battery backup system, which should really only be made available free to those who require an always-on phone for medical reasons, whilst others must pay for it as well.

        By doing this, you’re removing the requirement for a long-duration in-house install, by simply providing a single fibre port in the home, leaving everything else up to the homeowner. No power needing to be messed about with, no devices to install, no cabling needed outside of the fibre outlet.

        THAT is how we should have cost-saved on this project. Not by changing the technology type to include FTTN, HFC/DOCSIS, FTTdp, and any other kind of technology mix to be thrown on top of what is already being considered.

        • ok, but to expand from 1 to 4 million, won’t they still need in-house installation of the cable modem for most premises ?

          • Not the cable modem – but the leadin service for DOCSIS, yes.

            I’m assuming that the CM will be supplied by the user, potentially via the ISP.

            The cabling component will be the most expensive out of all of it, as you’ll be running aerial RG11(i assume) from the house to the nearest power pole, or underground if that’s how its run in the area.

            The cable itself isn’t expensive, same as FTTH – the cable is quite cheap. It’s the labour costs to run it that are the issue.

          • Ian, wouldnt it be RG6 Quad shielded? RJ11 is a lot more expensive and quite hard to run due to it’s greater thickness?

        • We should count ourselves lucky that we have such knowledgeable people frequenting Delimiter, but Ian, do you know how depressed all that makes me feel? :)

          • Haha, sorry buddy… Just trying to help people out with some of the darker details about what we’re paying for.

          • Q – would you prefer to have all the knowledge with which to understand your situation and the ramifications of the events surrounding you, or remain ignorant and only have to react to events as they affect you on a day-to-day basis, but be powerless to control or affect anything including making strategic decisions about your own circumstances? Knowledge isn’t just the ability to know what is bad and why, it is also the ability to appreciate what is good – great art (whether that’s painting or music or writing) is only truly appreciated by those who understand the history and the technicalities deeply; while the layman can appreciate it for its subjective beauty, the more one understands and has deep experience of the medium the greater their recognition for the skill, artistry, creativity etc of the artist. Ergo, the greater your knowledge, the better your appreciation. The same goes for all aspects of life – the creationist has awe at the unknowable awesomeness of their chosen deity, but the scientist can comprehend the tremendous complexity and arbitrary chance required for the vanishingly unlikely outcome of not just our existence, but all existence, and the infinite number of vanishingly unlikely circumstances combined that led us to be here.

            We are living through a period in history that will be looked upon by future generations as an abject failure of humanity to collectively turn the tide of gross stupidity and selfishness that will lead to untold hardship for the vast majority of future people. Conservative government is the accelerant to the fire of self serving stupidity. It is only through understanding what is going on and helping others to open their eyes that we can have any hope of a prosperous future.

            So don’t get depressed, get angry. These people are stealing our money and our future, they’re stealing our health and our access to health care, our chance at future prosperity, business opportunities and job opportunities, they’re stealing our rights to access the ideas of others and our ability to protect ourselves from rich foreigners and corporations who want the right to barge into our lives and then sue us when we try to say no, they’re stealing our privacy so they know what we’re doing and saying, our right to collaborate with each other and the right to oppose them (whether it’s our right to have a say, our right to expose them or our right to protest when nothing else seems to be getting through).

            Arm yourself with knowledge, the more knowledge the more you will comprehend just how comprehensively we’re being screwed, and the angrier you will get and (hopefully) want to change something.

        • >>where the cost saving measure comes from not supplying the ONT to the user, but allow the ISP to sell a compliant GPON ONT instead<<

          But then you loose the benefit of being able to buy services from different providers on separate ports on a level playing field infrastructure managed by NBN. That has to be an advantage for a competitive market in an NBN operated NTD.

  12. The purpose of the pilot was to demonstrate that real life RSPs could actually deliver service to real life end customers. That was a success.

    It was not, and the text of the report makes clear, that it was not a definitive speed test of the HFC access network and describes the reason for the average speed being below the theoretical max.

    If we are to have a debate about whether a remedial upgrade to an existing HFC network is worth doing in the immediate term as we move towards FTTP then lets inform that debate with relevant facts and not the miss-reading that has happened here.

    • All well and good Andrew, but when the bit at the top of the nbn™ article starts with:

      nbn has concluded its HFC pilot in Redcliffe, QLD, with Retail Service Providers delivering end-users speeds of up to 100Mbps download and 40Mbps upload – our sights are now set on launching commercial HFC services in June.

      and the only two sub-headings in the article are:

      Great speeds achieved*
      and
      More speeds ahead

      it’s pretty easy to see why folks thought they were testing for speed.

      Oh, and if anyone was wondering what the astrix linked to in the article, it was this:

      *nbn provides services to its wholesale customers, telephone and internet service providers, and does not provide services directly to end users. These speeds and speeds achieved by nbn in a trial (including trials at a test facility) are not necessarily reflective of the speeds that will be experienced by end users. End user experience, including the speeds actually achieved over the nbn™ network, depends on the technology over which services are delivered to your premises and some factors outside our control like equipment quality, software, broadband plans and how the end user’s service provider designs its network.

      :)

      • >>and the only two sub-headings in the article are:

        Great speeds achieved*
        and
        More speeds ahead

        it’s pretty easy to see why folks thought they were testing for speed.<<

        From the article maybe… but not if they read and understood the Ovum report. You pretty much make my point.

  13. Yep, who knows why the people that instigated the report didn’t actually talk about the report, and only focused on the speed? Those crazy nbn™ guys!

Comments are closed.