Dr John Cioffi on the Viability of 1 Terabit DSL Copper Line Broadband
Broadband ISP speeds of 1 Terabit per second (1000000Mbps) down a traditional copper line? It may sound impossible but that’s what Dr John Cioffi, best known as the “father” of DSL, proposed last year (here) with Terabit DSL technology. In our latest interview we ask John about TDSL’s viability.
At present the vast majority of consumers in the UK connect their broadband ISP services via a form of Digital Subscriber Line technology (e.g. ADSL, VDSL [FTTC] etc.), which sends electrical signals down a tiny bit of twisted pair copper wire in order to deliver an internet connection into your home.
Two of the key people involved with the development of DSL were Joseph Lechleider and Stanford Professor Dr John Cioffi. The latter is a seasoned electrical engineer that has published over 600 papers and holds over 100 patents (many related to ADSL, VDSL, Vectored VDSL, G.fast, DSM, LTE, Massive-MIMO and Wi-Fi). On top of that he’s also the CEO and Chairman of California-based ASSIA.
Suffice to say that if it were anybody else proposing Terabit DSL then we might struggle to take it seriously but that’s definitely not the case with John.
What is TDSL?
At present modern Fibre-to-the-Cabinet (FTTC) style broadband technologies like G.fast can harness 106MHz of spectrum (rising up to 212MHz in the future) and tend to deliver their best speeds at under a few hundred metres from your local street cabinet, with electrical signals travelling inside the copper wire.
Future enhancements like XG.Fast / G.mgfast may even be able to use up to around 848MHz of spectrum but these will suffer even more from signal degradation over distance, meaning you’ll need a very short line in order to get the best performance and some doubt that today’s increasingly “full fibre” (FTTP/H) focused operators will ever adopt it.
By comparison TDSL proposes the radically different approach of using the existing copper wire as a “guide” (waveguide) to help direct a wireless broadband signal in the 100GHz+ millimeter Wave (mmW) band, which could carry huge amounts of data.
One early model predicts that speeds of 100Gbps (Gigabits per second) could be achieved at distances over 300 metres (future G.fast upgrades might deliver c.500Mbps at this distance), while 10Gbps might work at distances over 500 metres and even those on longer lines of 700 metres could potentially achieve symmetric speeds of around 1Gbps.
The fact that such speeds could be achieved, without needing to replace existing copper cables with expensive “full fibre” infrastructure, is a significant incentive for operators to explore the approach. Unfortunately at present TDSL is just a model and ASSIA are keen to drum up support for further research and funding.
On top of that there’s a question mark over the real-world viability of using mmW spectrum in such a way, which is a band that can be easily disrupted (such signals usually don’t travel very far without a dense or powerful network to keep them stable). On the other hand, back in the days of 14Kbps dial-up modems, few could have envisaged that a 1Gbps capable copper technology like G.fast might ever exist.
Dr. John Cioffi is due to talk more about all this on 20th June (15:40) at the TNO Ultra-fast Broadband Seminar in The Netherlands, but before then he’s kindly granted us an interview to discuss TDSL and the challenges involved with its potential adoption.
The Interview
1. Generally when somebody mentions DSL (Digital Subscriber Line) I tend to think of electrical signals travelling inside a metallic copper or aluminium cable. In keeping with that the title of your new ‘Terabit DSL’ (TDSL) solution has perhaps caused some confusion because it actually works in quite a fundamentally different way. Would you mind explaining the core differences to our readers?
ANSWER:
You are correct about DSL to date having used the transmission-line mode of operation of a metallic cable. This is only one of the modes of transmission (there are others that arise in solving the famed Maxwell Equations that largely govern the electromagnetic wave propagation that is fundamental to all wireless and wireline communication). The other “higher-order” modes are often called the “waveguide modes.” Your readers can think of the waveguide modes as “wireless transmission” around the wires, guided by those wires rather than flowing in the wires. The wires’ presence helps the wireless work better in most cases – that is the waveguide.
Exploiting the waveguide modes of a binder of twisted pair connections involves a coordinated set of transmitters and receiver, not just one of each. This set of transmitters and receivers is the equivalent of the multiple antennas used in MIMO [Multiple-Input and Multiple-Output]. The coordinated processing allows clean up of the messy spatial interference (called crosstalk) between all the waves propagated around and between the wires. Normal waveguides (like fiber) usually don’t need MIMO and have a single transmitter and receiver (usually – MIMO is starting to be used in fiber also). The twisted pairs need it badly to have a chance at working. If it all can be coordinated well, there is a chance for Terabits/Hz of information to be transmitted over 100 meters. Lower speeds can go further (it looks like 10 Gbps at 500m).
The antennas are very tiny and closely packed in this case because the frequencies used are very highly (100 GHz to 1 Teraherz) where a wavelength is 3 mm down to .3 mm so an entire set of them can be packed closely (and even manufactured as a single module using 3D printing/manufacturing technology).
I hope this helps and was not too technical for your readers.
2. When we first heard about TDSL it sounded like little more than a mathematical model. How far along is this technology in its development and when do you think we might see the first real trials taking place?
ANSWER:
There are some initiatives and many ideas on how to test the waveguide propagation in various situations. There is some US Government interest in funding it at the right labs – I am trying to encourage that funding to the right places to characterize and investigate further.
3. With so many operators now investing in FTTP/H, isn’t TDSL at risk of arriving too late to the party and where do you envisage it being used?
ANSWER:
That’s been said for decades now with twisted pair (too late to the party). Indeed in the mid 1990’s before any even slower DSLs were deployed, venture capital groups would not fund those efforts for the same reason (had to be fiber). The reality is that the replacement of the 1.3B global telephone lines by fiber is a very expensive and time-consuming process. Realistic estimates of cost, based well on the costs today, are that it is AT LEAST an average of $3000/line, so that is $4 Trillion to do it. Its also a very long time as well.
Thus, I’ve tried not to let that “fiber will be everywhere in 3 years” that I have heard regularly since 1989 get too much in the way of the advance of methods that can help. The copper is there and will be for a long time, so we might as well do our best to use it and not get too concerned in its replacement.
Today, many “fiber” connections still have a DSL at the end of them – its just called fiber. If you count these, the global number is roughly 500M DSL connections. Replacing the last drop segment to the consumer residence is the most expensive part of fiber replacement, so it is often not done even when the fiber gets closer, shortens the copper, and thus higher speeds are possible.
Continued on page 2..
Mark is a professional technology writer, IT consultant and computer engineer from Dorset (England), he also founded ISPreview in 1999 and enjoys analysing the latest telecoms and broadband developments. Find me on X (Twitter), Mastodon, Facebook and Linkedin.
« Virgin Media TV Customers Who Take Netflix to Get Single Billing
Tdsl will never happen in UK for the future cos openreach want to get rid of copper in the future.
I thought it was April again when i read this. From the start and the description it sounds promising we then move on to the first question which again sounds promising until you look at the “cross section geometry” image and then ask, hmmm i wonder what happens to that nice perfect, well measure area of cable and gap when you bend the cable.
Question 2 then went on to dash the thing entirely with an evasive answer which basically confirmed the question that this is just one mans theoretical mathematical model/exercise rather than anything that he has actually in any way shape or form been proven to work thus far.
Question 4 and the answer… Again more avoidance to the question, somehow software is going to fine tweak a damaged/patch repaired, non-twisted pair, or even aluminium cable is it? Even though from going back to the first question you need a cable with very specific measurements both to the cable and internal and external insulators.
Question 6 and the response, how can he or anyone argue this is better than switching to fibre cables if he does not know the cost or if it would even be cheaper.
Maybe this would be possible, but i do not see it being cheap, and i do not think FTTP which its not a question of if but when in the future has anything to worry about.
UK phone lines are not twisted pair, please amend.
The UK telephone networks are twisted pair, and voice wouldn’t work over significant distances if they were not. Whilst it’s true that the twisting on PSTN cabling is only lightly twisted compared to the UTP cabling used for Ethernet, it is most certainly twisted pair and it has to be so for noise rejection.
The relevant BT spec for external cabling (including burial) is CW1128
https://www.fscables.com/sites/admin/plugins/elfinder/files/fscables/Datasheets/CW%201128%200.9mm%20Telephone%20Cable.pdf
Every PSTN copper network in the world uses twisted pair.
Steve got here first so I won’t add to what he’s said, but it’s also easy enough to confirm, just look at the copper cable as it enters your property. Twisted. The cables are usually separated out a bit after this, so as to clip into connectors, thus simply looking behind the master socket might not always show it clearly enough.
I should also add that the use of twisted pair cabling in telephony goes back to the very beginning of the technology, as it was invented by Alexander Graham Bell as essential for long distance voice connections. In comparison, early telegraph cabling was open pair and was entirely unsuited to voice.
https://www.electronicsweekly.com/blogs/electro-ramblings/industry-comment/only-connect-twisted-pair-cabling-2013-07/
Yep most UK phone cable is indeed twisted pair. It is important to note there are nowhere near the amount of twists as there are in the likes of CAT5 or CAT6 cable and it does not reject noise as good as either of those.
I personally can not see this working for UK phone cables, for a few reasons, i suspect the wires would need more twists in them than typical phone wire has here, especially for it to work at any significant distance. Phone cables here with regards to noise are affected when it comes to data with something as simple as you still having the old ring wire connected in your property. (thankfully not needed for most modern phones or if you shove a filter on every phone)
Other issues include what was touched on in the opening post, our phone cable network is quite old in many places and (RIGHTLY so before i get nagged) minor damage is often repaired (joined) rather than replaced entirely (it would cost a fortune otherwise) which would no doubt also affect this.
Its another nice idea, which if everything was perfect may work. Unfortunately the real world is not full of non-kinked, shiny, straight cable with perfect twists and gaps in it. Replacing it all with more copper when there is something (ie Fibre) more reliable, which would work at longer distance, cost per metre is similar and install and replace would be a similar job would be madness.
Interesting read as a research project, viable product though or alternative to fibre… NO.
An interesting interview, but it’s ignoring a major component of modern networks which is reliability. Are customers going to accept getting 1/100th or 1/1000th or 1/10000th of the speed when real world conditions cause massive amounts of errors? Because I can tell you customers are very unhappy when their “40-10” FTTC can only reach 20 down due to line conditions/distance. This is going to get worse if you try firing Gbps down the same old piece of copper.
I see so many faults on xDSL circuits, we have customers complaining about slow speeds and packet loss, when they’re overloading their piece of string BT calls a phone line. Fibre just works, and short of a fibre break any errors are usually fixed by replacing SFP/line card or cleaning the fibre.
How much of Openreach’s time is wasted testing lines that are within spec because a customer reports it as “slow”? I see it all the time, and there are SPs that are far larger. I would be prepared to bet that there are huge numbers of man hours wasted having to prove the performance of such unreliable technology. Even if on PON rather than EAD, the number of faults is going to be massively reduced.
Replacing the fibre is expensive, but you have far cheaper costs on maintenance due to the increased reliability.
Not only on maintenance, think running costs. Have you heard of any outlook of electric power getting cheaper? And the more the copper gets squeezed out, the more power it needs, the more active hardware it Needs, the more the line condition needs to be constantly monitored in order to adjust frequencies and so on.
The finances for a proper fiber network are then partially soaked up by the running costs of the not so new network. And when finally everything works, the poor old copper might be corroded away as it is decades over ist sell by date.
And it can impossibly be cheaper to dig away new copper as opposed to new fiber cables just to be able to use technologies invented solely to make use of the old copper lines until replacement.
The cost of Fibre deployment is coming down day by day and is only going to reduce further, please let copper die and rest in peace it had its time,
look to the future and stop getting distracted by these types of concepts, more should be looked at deploying even cheaper and efficient fibre roll-out,
Fibre optics have the potential for “Multi Tera Bits”
This Guy explains it nicely just how fast Fibre can go – almost unlimited,
https://www.youtube.com/watch?v=bbKru71442k
This is way above my head Mark Jackson but I am surprised that someone from Dorset would spell “metre” and “fibre” the american way.
So far as I can see I didn’t spell “fibre” incorrectly, maybe you’re reading John’s answers but I can’t find a mistake in mine? However I will cough to a couple “meter” instead of “metre”, which is clearly a typo as in the same line and elsewhere I’ve used the correct UK spelling. This is of course what happens when interviewing an American while typing very fast 🙂 .
I stopped reading the article when it mentioned the replacement cost of an average phone line with fibre: $3000/line. In most areas fibre deployment costs are much lower.
Depends on what your starting point is for that replacement cost, at the cabinet or at the core network etc., and whether you’re factoring in the impact of rural areas on a national average or taking only a peak from rural areas etc. Not clear which model above but obviously a commercial company has a vested interest in giving a figure that looks favourable to the alternative.
The heuristic used in the industry is £2000 as a mean cost for replacement of all copper with fibre in a country, so it’s in the right ballpark. Telcos don’t save money until all the copper is gone so it’s not really valid to just count lower urban costs – that results in a very expensive rural copper network that’s no being subsidised from the USO price of urban copper.
The mentioned “costs” are typically not costs in their usual meaning but prices you would have to pay for it including mark ups, profits, taxes and so on.
But for the Telco an investment/replacement may not be taxed (in the same way) and it would not charge itself a profit as it would be an “out” and an “in” on the same account just creating more work without creating value.
Though in order to justify such charges you can always outsource to an external company ;).
This is interesting read well done Mark J for digging this up. I love a bit of front line research no doubt about it just with my ex-researcher head on. Hadn’t honestly thought about Maxwell equations for 20 years +. It also has many applications that are not DSL related.
As Mark J said if it was anyone else I would just ignore it.
I’m sure it could ultimately be made to work over short distances with perfect copper. But as others have pointed out as it is a wave-guide and changes to the geometry of the wave guide would be fatal to the signal.
Realistically for the UK we cannot afford to let this kind of thing become another stalling excuse for lack of decent connectivity. This is 10+ years away from there being real world equipment to install: if it can be generalised. There is a mathematical model but seemingly no physical research data backing it up.
Also this is never going to be any use for long lines where pure fibre remains the only real answer.
And I agree with @GNewton that the $3000/line for pure fibre is not real world as I am seeing number more like 10-15% of that crossing my desk now.
Ironically this might be better for VM as the coax might be a better wave guide? Although the maths presented here only seems to cover twisted pairs?
Before someone points it out I messed up as I forgot about the £/$ conversion.
I should have said that fibre in the UK is now costing more like £300-400 in reasonable density locations.
£3 – £400 from where to where?
From house to DP or cab or exchange?
£400 might cover the engineers time and materials to get to the hole in the ground but wouldn’t cover the bit to the exchange including any civils involved.
Copper is bad enough, adding wireless to it as well sounds like a nightmare.
Replacing metal with fibre is not the easyest or chepest thing but it’s totally acheivable if we all pull in the same direction and once it’s done hsa hugh benefits on several levels.
I’d much rather see the clever minds of these people working on making fibre deployment easyer faster and cheaper than keeping a dinasour alive.
Why such negativity ?
strategically we should plan for Fibre Optic cable in the UK however each country will be different and the distances vary enormously. Think prairie and outback not Kensington. In addition whilst this research is focused on extension of DSL it may produce contribution elsewhere. There is no reason why such developments cannot contribute in some way to the delivery of data to or around our homes. Just because we implemented DSL in a bad way doesn’t rubbish the technology otherwise we never move forward. Many of the underlying technolodgy we use today were invented or discovered long ago for totally different purposes.
Look where we are today, adsl adsl+ vdsl vdsl2 gfast all offering the upto speed disguised as fiber, we dont need another technology to delay the roll out of fiber,
Obviously this guy has a vested interest in the copper industry and probably has much to gain, let copper lay to rest and move on to full fiber deployment of course this will take a decade or so, this guy keeps saying “fiber deployment is always going to be expensive” dose he factor in the lower maintenance and reliably of full fiber ????
lets move away from copper
Interesting, but I do wonder how well this will work in the real world with multiple cable joints, IDC terminals, etc in the way.
Reading this reminds me of articles where we read about a full cure vaccine or drug for HIV/AIDS but then later on say it’s not quite discovered yet!
This is exactly the same case. Finally in the end he brings in no hope as he is working for a software company.
We all know this whole discussion is not practical. Copper cannot deliver 1Tbps and we will never have properties that are within a few meters away from the exchange for fast speeds to be delivered. People cannot achieve 17Mbps with full copper nor can they achieve 76Mbps with FTTC. There’s noise margin issues which will also affect the stability of the line when the speed is increased. Also weather interference will also play a part in causing drop out issues. Like others have mentioned Full Fibre will reduce cost maintenance of the network. Fewer customer service/engineering assistants will be needed to maintain Full Fibre compared to the old copper technology.
It just doesn’t work, if it did we won’t be discussing about Full Fibre. This is also like Cloud Gaming OnLive service where the Founder Steve Perlman once promised that this cloud gaming service will fully replace console and PC hardware and deliver streaming gaming services via broadband through a basic toaster computer hardware and we would never have to upgrade our CPU and GPU ever again. It sounded too good to be true but later the service even collapsed as it didn’t work and was deemed unpopular.
“People cannot achieve 17Mbps with full copper nor can they achieve 76Mbps with FTTC.”
Please stop spouting your usual rubbish. Many people (myself included) got/get decent speeds on ADSL2+. I was getting 18 Mb/s on adsl2+ being ~ 700m away from the exchange. On VDSL2 I’m getting the full 80 Mbps sync and am ~ 300m away from the cabinet. Perhaps you forgot to include the word “some” at the beginning of the above statement?
As Chris hints, it’s important to remember that 17Mbps and 76Mbps actually reflect the advertising rule (fastest 10%) and not the capability of the technology itself. VDSL2 lines can actually do a fair bit over 100Mbps at the right distance but Openreach wisely cap at 80Mbps. Likewise ADSL2+ lines can do over 20Mbps in the real-world, although the theoretical capability of 24Mbps is a step too far.
My old ADSL2+ line just about touched 19Mbps as the cabinet is just across the street, while I tend to get near to the max of FTTC. The best speeds like this may not be available to all but millions of premises are still within reach of it.
@Chris & Mark: But the question is can it really hold a stable connection at those max capable speeds?
Many of us forget that there are also noise margin issues and DLM (Dynamic Line Management). I’ve said before as well that I did reach max speeds at 16.60 Mbps in speed tests some time last year. But this connection speed (at least in my case) cannot maintain its stability for more than 2 days. Of-course I live in a high-rise building 800 meters from the Bishopsgate exchange, so it may be different for me.
But I’ve also lived in another leasehold property that was low rise for a few years before returning back to my old property. We have a third property which we rent out to tenants including the other one no one achieved the max speeds.
So far from experience the sweet spot in my case is around Connection Speed 14533 kbps. Connection Time 1213:06:48. 50 consecutive days no drop out with noise margins between 9-11dB respectively.
But try and increase the maximum capable speed at around 19000 kbps router stat (around 16Mbps speed test) you’ll see a dramatic drop in noise margins dropping from 10dB to 3.0dB. As soon as that connection will not maintain noise margins above 1dB the router will re-sync at lower speeds which in my case will drop to 14533 kbps and it will stay there. That is the setpoint and the most stable connection. 50 days no disconnection at 12-13Mbps. But if the connection tries to be set at 15-16Mbps the connection will definitely drop out and the router will re-sync at a lower speed and it will continuously occur without exceptions. I’ve tried with different routers including removing the hidden master socket faceplate and changed micro filters and I also called in a few BT Openreach engineers over the last few years they say there’s nothing they can do as this is the best line performance that can be achieved and there are no issues with my internal wiring either.
Now I don’t have FTTC and even if theoretically I had it at around 800 meters from exchange. The max speed will be 28Mbps. Take a look at this chart https://www.thinkbroadband.com/guides/fibre-fttc-ftth-broadband-guide Distance to cabinet (metres) Estimated downstream connection speed Estimated upstream connection speed Cumulative %’age of premises at this distance 100m 100 Mbps 25 Mbps 5%.
Okay so at 100 meters to cabinet that will be 100Mbps with FTTC. But you have to be within 100 meters! At 800 meters 800m 28 Mbps 10 Mbps 80%. And at 1500m 15 Mbps 4 Mbps 98%
See what I am saying? That’s FTTC statistics. That’s why my area is on a plan to be upgraded to FTTP. For those max FTTC speeds to be achieved as discussed with BT they say they would have to install another green cabinet closer to my property for FTTC to be supported and receive appropriate speeds.
I only replied to the bit where you said, “People cannot achieve 17Mbps with full copper nor can they achieve 76Mbps with FTTC,” which isn’t correct as millions of premises are within the sweet spot for that (I’m one of them).
Like I said above.. “at the right distance” is the key phrase, although you could add other caveats like ISP congestion, slow WiFi and poor home wiring as external factors to consider in each property. But none of this is really related to the article, which is proposing a very different approach.
“But the question is can it really hold a stable connection at those max capable speeds?”
Yes, provided you live reasonably close to the exchange and have a clean/error free line then you should easily get > 15 MB/s on ADSL2+. Just because YOU cannot, doesn’t mean EVERYONE else cannot.
“But try and increase the maximum capable speed at around 19000 kbps router stat (around 16Mbps speed test) you’ll see a dramatic drop in noise margins dropping from 10dB to 3.0dB”
How would you increase your ADSL2+ speeds? By moving the exchange closer to your home? Unless you meant playing around with the noise margins on your router?
Copper based DSL services are far from perfect but they can certainly give you a stable connection provided the line conditions are met.
@Chris: How it works is basically rebooting the router either from the router settings http://192.168.0.1/sky_diagnostics.html or turning the router on and off causes the speed to go up to 17000-19000 kbps which in real world speed tests it would deliver between 15-16 Mbps. I do get between 15-16 Mbps if I try to reboot the router. But within a few hours to a couple of days the internet disconnects and the speed gets re-synced back to a lower speed!
This vicious cycle continues after every reboot. Noise margins cannot be maintained steadily high enough to maintain the higher speeds. The speed does not correct itself overtime without rebooting the router because DLM assumes line is not stable enough to handle the higher speeds despite the fact that the line is able to pump those speeds.
If I leave the router as it is, like I have done so last 51 days no single drop outs. No issues with drop out since the internet connection is re-synced at a lower speed and noise margins are revved up.
No doubt if the right conditions are met and there are no line errors like in yours and Marks case you’ll have a solid experience. Distance indeed is a very important factor.
To quote from this article “If it all can be coordinated well, there is a chance for Terabits/Hz of information to be transmitted over 100 meters. Lower speeds can go further (it looks like 10 Gbps at 500m).”
This is where I am expressing my concerns. As I live in a high-rise block and most properties in urban areas are situated in high rise buildings this will equate to roughly over 100 meters extra distance of copper cables travelling through depending on which floor of the building you live in. The risk of line errors increases as distance is also increased making 100% copper or FTTC not a viable solution for all customers. Those such as myself with poorer line quality/distance to exchange would best be served with FTTP as this is the best permanent long term solution that will not be affected from line errors or distance.
I cannot see this project ever leaving the theoretical stage. It had been published quite a while ago and has yet failed to show how cable could keep the required geometry in order to allow function.
At least the dug away cables are underlying various mechanical influences as there would be weight/pressure of the soil. Which itself is different from location to location. Compaction through vibration from traffic or works nearby, bends in the cable and so on are also factors worth considering.
I entirely forgot to mention, that a repair in the cable almost certainly would render it completely unusable for the desired frequencies.
The project is a complete fantasy. Perhaps Dr John Cioffi wants to make himself look popular.
As I mentioned previously to have such a stable line under copper for this to work for the masses is 100% impossible. It may work for the very few but you need short cables that travel over very short distances.
This quote gives it away “deliver their best speeds at under a few hundred metres from your local street cabinet, with electrical signals travelling inside the copper wire.”
Also like I mentioned previously any line interference’s etc will cause internet disruptions. You cannot force modern high speed technology into 120 years old tech copper, it’s a bottleneck. And then to carry the rest of the signal via wireless transmission will potentially bring in new problems.
It may be interesting as an idea or a research development project but it does not bring any hope of it becoming a success! If it did a discovery like this would’ve been invented long ago and nobody would’ve taken fibre optic seriously.