One of the Network Operator Group (NOG) lists I read had a brief flurry of discussion last month about the projected end of support in Australia for the Integrated Services Digital Network (ISDN) — a passing moment that I think bears noting.
ISDN wasn’t the first high-speed network
But it was one of the first widely deployed high-speed networks worldwide.
ISDN’s roots stem from the pre-Internet standards of the CCITT/ITU-T. At its height, ISDN probably had around 25 million subscribers worldwide. Set against the modern Internet this is a tiny number but compared to the 10,000 or less participating people in the pre-1985 Internet, was huge. If classic ARPA research networks built the Internet Protocol into a world-dominant platform, ISDN took telephone wires and made them work at scale, worldwide, for digital data.
Three things must be understood to set this story:
1. The physical platform.
2. The idea of calls and noise.
3. ‘How to encode things’.
1. A world of copper landlines
So much of modern communications assumes high-speed radio and fibre-optic communications; it’s easy to forget that the recent past was about copper. Landlines were considered 50 to 70-year public utility investments in most economies, with a single national telephone company to operate them.
This is the world ISDN emerged into, a world of standard grade physical wires and telephone switches built for long life. ISDN was chosen as a technology because the engineering of the time showed it would work in both the ‘fat’ (E/T series 2Mb or 1.55Mb service for Europe, E series for Asia, and T series for America, with up to 32 64Kb sub-channels and a control channel) and ‘thin’ (two channels of 64Kb and one 16Kb control channel).
The ‘thin’ variety (also called 2B+D for its two data ‘b’ type channels), was engineered to work in typical customer-access-network (CAN) deployments, after a small investment in the phone exchange — the ISDN ‘line card’ that allowed two services to be managed on one physical card. This was a decision that suited all parties; the telco could deploy at scale without losing its existing exchange investment, the card manufacturers could sell volume cards worldwide (at the time, a market of 25M customers was an attractive target!), and the ISDN line speed was enough to satisfy most customers at a minimum 64k (better than twice as fast as a modem, which were typically limited to 28.8k).
Automatic Telephone Exchanges were a significant reinvestment that saw electromechanical ‘strowger‘ exchanges upgraded to ones that operated on dial tones, not pulses. Fundamentally, it was capital investment at scale for an entire economy. This was not something any monopoly telco was looking forward to ‘junking’, at the time.
Copper is a valuable metal, and now that services moved to mobile telephones, fibre optic cables, and satellite dishes, reclaiming all this copper is part of the telecommunications economy — there is a lot of money locked up in the public telephone network.
(If you’re reading NANOG, the list has been discussing reclaim of the copper in the local loop — an unrelated story to the one here, but a strong signal of the global convergence in this underlying story: The copper has value, outside of the telecommunications role it filled for so long.)
The copper wires were laid as cheaply as possible, with the lowest component cost possible, which meant they were designed to do one thing well — send voice. They were not designed for digital data. Retrofitting ISDN demanded engineering investment.
It wasn’t the only alternative. Alongside the engineering of ISDN, Digital Subscriber Line (DSL), was also being engineered at this time. DSL was a bigger departure from the signalling technologies of the day and was ultimately going to require replacing telephone switching gear. This meant its capital cost to the provider was higher, and at the time, riskier.
ISDN represented a more certain path for many telcos, so it was the ‘winner’ in the short to medium term. It usefully married the voice telephony, facsimile, and data world into one composite whole. There was no need to invent new addresses: The International Subscriber Trunk dialling model of phone numbers (E.164) system worked quite well with ISDN signalling. So, for a world predicated on voice calls, ISDN made sense.
2. Calls and noise
A telephone call model is based on a human ear and human voice box at each end. People are very good at discriminating words from noise (the cocktail party effect) and it’s known how much ‘noise’ users can handle before a call becomes untenably bad. This is often called a Quantization Distortion Unit (QDU) and one of the useful determinations of the CCITT was to set a limit on how many QDUs could be tolerated on a copper-wire telephone service.
A rough rule of thumb emerged around the same time. Delivering 64Kbps of digital data with voice encoded into it, gives roughly 1 QDU of noise and distortion. The achievable voice quality and delay is similar (if not better) than speaking down the same line. Comparable work showed this outcome could be achieved with 56Kbps, and by agreeing to use of a 64Kb system, a useful situation emerged — it was possible to connect a 56Kbps system to a 64Kbps system, an American 1.55Mb to a European 2Mb service, with some ‘slippage’ to deal with the extra 8Kbps per sub-channel.
Now, imagine this has all been done for voice, and suddenly a world where people want to convert digital data to sounds via a modem emerges. However, the wires were initially deployed for ‘analogue’ voice, then repurposed for (slower) acoustically encoded dial-up modem signals, and repurposed again for low-rate digitally encoded signals in ISDN. People were now fully online, modem-based services were becoming common, and ISDN was ‘expensive’ by comparison. Partly the reason was that many telephone networks offered untimed local calls (within the local exchange) but as a premium ‘digital’ service, ISDN was time-charged for all calls irrespective of distance. This alone made people favour modem-based connections, but hate the speed. The world wanted more, but ISDN wasn’t going to be the one to deliver it.
3. Encoding the market
A parallel technical development, DSL, used a higher frequency range than normal voice signalling, and could be operated alongside a normal phone (with suitable splitters to break out and isolate each kind of signal). It had been tested in labs and in the field for some time. It was understood that by moving to very high-frequency signalling, digital signals could be sent down the existing trunk lines and customer access network quite well.
But this demanded signal processing power, and at scale. The signal processing power was needed to encode and decode the higher radio frequencies, insert error correcting codes to handle noise, and manage complex parallel signals (using many frequencies at once) to increase bandwidth. So, to deploy DSL, an entire industry had to be willing to move over to new chips, new encodings, and new exchange equipment. For the return on investment to make sense, there had to be economies of scale.
This is why Asynchronous DSL (ADSL) was chosen. It limits the bandwidth going in one direction, to allow more of the bandwidth to be sent in the other direction, and limits crosstalk and noise. For services that primarily involve fetching data from the world and sending mostly smaller amounts back out, this made sense. For Internet home use, this was a perfect fit. All that was missing was a willingness to recapitalize the exchange and deploy the DSL head-end equipment. The surge of Internet users in the domestic market, combined with the dotcom boom of capital entering the telecommunications industry and market deregulation worldwide provided that motive: 25,000,000 ISDN users was nothing in a world of hundreds of millions of online home users.
It also helped that another technology, Asynchronous Transfer Method (ATM) was being deployed for high-speed networks and provided chipsets at sufficient scale to be viable, if the same encoding was ‘repurposed’ for use in the DSL deployment. This reduced the capital exposure enough to ensure worldwide manufacturing capabilities could supply the necessary equipment.
Unlike ISDN, the sales model now expanded from ‘the telco provides everything’ to ‘the telco can provide everything, but users can buy their own too.’ This was much closer to the domestic consumer experience with dial-up modems; people were free to buy their own ADSL modem and take it with them between houses. And, as the market in telecommunications liberalized, people were free to ask for a different company to provide the ADSL service by tapping into the exchange line at a splitter and breaking out the signals to their head-end unit, not the primary telco’s unit. Markets of ISPs emerged that were overlays, not independently running wires, paying for access to the national wire network to sell competing services.
The end of the line
ISDN represents the end of a digital data network lineage of the early 1960s, 1970s, and 1980s. High Level Data Link Control (HDLC) had its roots in IBM networks and was adopted as Link Access Procedure Balanced (LAPB) in the X.25 world, and finally as Link Access Protocol D-Channel (LAPD) in ISDN. This nexus of HDLC, X.25 and ISDN drove ISDN to worldwide acceptance as a telco-ITU mediated standard technology, respecting the telephone network at scale.
DSL represents the newer form of digital data signalling. It uses far higher line rates, and far more complex digital encoding methods, which are now common in copper, fibre, coaxial, and mobile radio systems. However, it doesn’t reference the older world of telephony except within the limits of what the mobile world defined in the GSM series of 2G, 3G, 4G, and 5G standards for encoded voice over digital signalling.
Read: All the Gs series.
Ultimately, DSL drove out ISDN because it makes more sense where voice is not the predominant service. The splitter is almost obsolete now because so few people care to have a voice service in their home. Copper wire is most likely relegated to ‘naked’ service delivery of digital data, or even repurposed into a complex network of local loop, in-the-street digital signalling, and back-end coaxial/fibre delivery, to allow very high bit-rate DSL (VDSL), offering 100+Mb services.
As one star sets, another rises. Vale ISDN!
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.
Hi, this article is extremely informative and interesting! Although there are indeed a lot of technical jargons which makes it rather challenging g for a person from non-technical background (me!) to thoroughly enjoy without having to google multiple times while reading.
There are even times when google can’t help me! Would you please explain what you mean in the last paragraph by ” Copper wire is most likely relegated to ‘naked’ service delivery of digital data”? It seems like you’re using ‘naked’ with a specific technical meaning but as you might have guessed, googling naked didn’t do me any good with understanding it 😉
Hi Manju, thanks for your feedback and kind words. I will try to keep my jargon to a minimum, or at least explain the terms better inside the blog in future.
“naked” in the context of copper line services, means that there is no intention of running old fashioned “plain old telephone service” (sometimes called POTS) and so the lines can be used purely for digital service delivery, without needing to use filters, to separate the digital signal from the voice signal. If you still had a handset attached to the line, you might well hear truly awful noise if you picked it up, and you’d probably break the digital subscriber line (DSL) signalling being sent down it. So when people commission a “naked” DSL service they disconnect all the handsets, and only have the DSL modem plugged in.
Often times, this DSL modem has a Voice over IP (VOIP) component, and you can plug the handset back into the DSL modem: its reading Voice packets in IP, and converting them to speech, and vice-versa when you speak into the handset or tone-dial a number.