In Praise of Cables
- Published
- in Wireless
For most of the last twenty or so years I seem to have started off the year by writing an article claiming that this would finally be the one when wireless data takes off. It’s nice to see things changing: Wi-Fi is finally starting to move outside internet access for PCs and Phone, Bluetooth Smart is appearing in desirable consumer devices and should trigger an avalanche of connected accessories, and smart metering is bringing ZigBee and Wireless M-Bus into homes as a static PAN. That doesn’t mean that there are not still massive unexplored opportunities in M2M, but it’s good progress.
Instead of the obvious call for more, I’d like to look back at the many advantages of cables. As designers rush into wireless, it’s easy to forget what they’re giving up. Wireless offers new opportunities, but only at the expense of many serious compromises. In this brave new world of wireless it’s apparent that some people are forgetting those compromises. In this and the following article I’m going to look at what they are and then address the misconception that wireless standards can be treated in the same way as wired ones, debunking the common misconception that they follow the OSI model.
Once you sign up to the wireless dream, it’s very easy to forget just how good cables are. Unless you’re pushing through extremely high frequencies (in the range that we’d generally consider to be wireless), or very high throughputs, then cabled range within the home isn’t a problem. If you want a cable to go farther you just buy another reel of cable. And for most of the time, throughput’s not much of a problem either. Compared to wireless standards, cables are fast. They support throughputs in hundreds of Megabits and Gigabits per second where most wireless standards struggle to get over a few hundred kilobits.
One of the big differences from cables, which many would say is a key advantage of wireless is topology. This is where wireless becomes non-intuitive, which is both an advantage and a disadvantage. Topology with cables is easy. You get the end of the cable, find a mating plug or socket and plug it in. Instant connectivity – it works. With wireless you have nothing physical to hold or plug into. Wireless is cunningly invisible. To connect it, both ends (and it can be more than two, which is even more complicated) need to start talking to each other. It’s that dreaded term “pairing”, which is where a large percentage of wireless products fail at the first hurdle. It also brings us to the next difference – security.
Security is something that you don’t need to bother about with most cables, because the act of plugging them in gives you physical security. The data that you send stays within the cable and no-one else can access it. Devotees of spy fiction will point out that you can break in and attach covert listening devices, or even use sensitive radio receivers to listen in to the weak electromagnetic field generated by the data flowing along your cable. They’re right, but it’s not likely that your neighbour or the local kids will be capable of doing that. And if they are you’ve probably got more important things to worry about.
With wireless you potentially lose all of that security. The data you’re sending from your laptop to your router, or your smart meter to your display can be captured by anyone else within range. (Unlike cables which are strictly point to point, wireless is generally omni-directional.) That raises two problems – stopping other products masquerading as legitimate device on your network which might send spurious commands, and adds a requirement to encrypt your data, so that if someone else captures some of your transmissions, they can’t decode them.
There is a whole science built up around wireless security, composed of experts who write security algorithms and experts who try to crack them, both for genuine and nefarious reasons. That’s led to something of a security arms race, where to remain secure ever more complex algorithms and authentication schemes are being developed, which require ever more powerful microprocessors to be built into wireless chips. These are there just to maintain the security of the cable-replacement link. Other security, such as the https security the internet uses for credit card purchases has to run on top of these, whether the underlying link is a cable or wireless.
The ever more powerful processors needed to support complex security algorithms bring us to another point of difference – power consumption. It’s a fallacy to say that cables don’t need power – they have an inherent resistance which needs power to send bits of information, although in most cases it’s very low compared to wireless. As you send more data, you need to put in processing effort to shape the pulses going down the cable; you also need to terminate the cable to make it behave more like a waveguide, which again needs power. But you don’t have the overhead of processing for security, nor the fact that wireless is generally transmitting in all directions, even though it’s only receiving in one. (And anyone who says that MIMO is power efficient has been reading and believing too many marketing manuals.)
Then there’s latency. It’s an eternal surprise to engineers starting to work with wireless that when you put data into one end of a link it doesn’t immediately pop out of the other. The IP community is already familiar with the problem, but the magnitude of potential delay can be totally unexpected. Try listening to a film over a Bluetooth stereo headset to discover what lipsynch is all about. Latency can often be measured in seconds.
Even when you’ve got your wireless system to work there’s the annoying problem of robustness, which comes down to available spectrum and interference. That’s not to say that cables can’t suffer from interference. A well place nuclear explosion can create a very disruptive electromagnetic pulse over a short period, but it’s not something you come across on a daily basis in most domestic environments. In contrast different wireless standards that share the same limited spectrum get in each other’s way. Ten years ago, when I first installed Wi-Fi, or 802.11b as it was called then, it worked really well throughout the house and garden. As all of my neighbours installed it, along with baby monitors and TV sender, the performance degraded to the point where it became worse than useless. As a result I’ve just wired my house with CAT-5 for Internet access. It’s a worrying by realistic fact that any in-home wireless system will probably work best on the day you install it, and then get progressively worse as all of your neighbours get one too.
Those are the big physical differences, but there’s a whole host of less tangible ones to consider. Differences that just don’t exist in the world of cables. Let’s start with interoperability. In the cable world that’s mostly down to different plugs and sockets. Of course there’s fine detail like multiway and coaxial cables and the level of screening, but in most cases you can solder a different connector on and it works. In contrast there’s absolutely no interoperability between different wireless standards. Quite often there’s precious little backwards compatibility between the different versions of a single wireless standard. From a consumer viewpoint that means the product you buy today probably won’t work with the one you bought a few years ago. With cables there’s pretty good compatibility – at most you need to change the plug or buy an adapter to get several decades of backwards compatibility. In contrast wireless generally means a complete replacement. Nice work when you can get it, but not something that endears you to the user.
So why the enthusiasm for wireless? Mostly because of the Holy Grail of simplicity that it purports to offer. Simplicity of installation (no wires or holes in walls), simplicity of commissioning (no need to connect cables to inaccessible sockets), simplicity of mobility (you can carry devices around and connect to different things in difference places). But it’s really difficult getting all of that to work.
To put back the basics of a cable – throughput, range, security, latency ease of connection, all of the things I’ve talked about above, is an immensely difficult task. You can specify most cables on a single sheet of paper. In comparison most wireless standards require several thousand pages of technical specification. And even then each standard is a compromise, making its individual choice of what is important for a specific application. Each one takes hundreds of man years of effort to write, test and turn into reality.
Which adds another complication. To make this complexity work requires some highly talented companies and individuals. All of them contribute their ideas and patents into a standard. Which means that all wireless standards involve some form of IP license or certification requirement. That means you either need to pay money to join the standards group if you want to use it in your products, pay someone a license fee, or pay a test house to certify it before you’re allowed to ship it. After which you may discover you’re not allowed to export it, as some of the high tech aspects, particularly around security are subject to export restrictions. Bear in mind that I’m not talking about exotic industrial or military standards here, but everyday consumer standards for products you’d use in the home. In contrast, anyone can make and ship a cable.
But that Holy Grail of wireless connectivity is strong. However difficult it may be compared with the simple cable it paints a wonderfully desirable future free of tangles, along with the promise of a new generation of product design where products are no longer islands, but can communicate with each other in that vision of the Internet of Things.
But it’s important to understand the compromises underlying wireless. To paraphrase the psalm, “and now abide range, throughput, security, these three; but the greatest of these is security”. Well developed, robust wireless standards are neccessary, because without them we can’t attempt to fulfil these requirements. And if any one of these three fail, so will customer confidence. In conclusion, I’m hoping for another year of growth in wireless. But it’s as well to understand the shifting sands we stand on when we make that wish.
I’m not entirely sure about the provenance of the previous comment, but it does make one other useful point, which is that cables are a whole lot better than wireless at carrying analogue signals.
The component cable will improve the quality of your picture. You should also use an optical cable for the audio portion. If you have HDMI connection on your tv and dvd player then I would recommend it. HDMI cable will carry the video and audio signal. Are you also using a surround sound receiver, if so connect it to this then your tv. Hope this will help you out.
Well spotted. And equally true for the original co-axial Ethernet physical layers.
Hi, Nick:
Sorry, minor correction:
“Notably, earlier wired protocols used CSMA-CD, too.” should have been “CSMA-CA” (I’m thinking of token ring).
And for that matter, with the prevalence of switches, rather than hubs in the wired networking world, the dramatic reduction in the size of collision domains (the smallest area a single data “conversation can take place) to basically one device has improved wired data transmission enormously. The collision domains for wireless (at least as far as interference goes) are as large as their entire transmission area, yet another advantage wired data networks have. To paraphrase your article, wired networks go where you put them; wireless networks go, well, just about in every direction (hopefully including the intended one).
Hi, Nick:
CSMA is actually *Carrier* Sense Multiple Access — that’s where the source determines if anyone else is “speaking” at the moment. What that source would do next depends on if it is wired or wireless (both mediums use CSMA, with slightly different rules). In wired, the current generation uses CSMA-CD (Collision Detection), which allows the source and target to determine that a collision has occurred, and what to do about it. In wireless, it uses CSMA-CA (Collision Avoidance), and wireless is essentially half-duplex.
Notably, earlier wired protocols used CSMA-CD, too.
There’s no question that wired networks (802.3) have a *far* greater bandwidth that wired (802.11) networks, though.
Brett is correct, wireless per se does not introduce latency. My point is that wireless standards do. That’s because of two reasons: access and processing. Access latency occurs where a wireless standard is operating in shared spectrum. With most cables, nobody else shares the cable (unless it’s a bus system like RS-485 or the old 10Base-2 coax Ethernet). That means you don’t need to worry about your data colliding with anyone else’s. In shared spectrum that’s not the case, so wireless standards introduce techniques to make sure that a transmission does not clash with that from another device. The most common way is to listen for activity before transmission. If another radio signal is detected, the unit backs off, introduces a delay and then tries again. That’s a scheme called Collision Sense – Multiple Access (CSMA). In a very busy spectrum it adds delays. How long these are depends on the way each standard implements this, but it can reach seconds. That may not matter, as for non-synchronous data it may be hidden by higher layers. Other standards take a different approach, using very short packets, hoping that they will slip between other transmissions in the spectrum, with retry schemes to send any packets that fail to get through. For a designer the impact is that they do not know exactly when the data will arrive at the far end of the link.
The other type of latency is the delay which is introduced is from processing. A common example of this is where an analogue signal is converted to digital and then back again to analogue at the far end. This may be done to decrease the bandwidth, or to allow it to be carried by a digital protocol. The codecs employed in these devices are again subject to compromises. These balance the time it takes to encode and decode the data, the accuracy of conversion and the power consumption, which depends on the processing complexity. For small battery powered devices, these compromises may introduce several seconds of delay, because in this case power consumption is the most important factor in that compromise.
None of this needs be a problem. In most cases consumers are unaware of the latency. But it can catch designers out. Lipsynch – the delay between an audio signal and its video complement is one of the best known one. This needs some clever techniques for introducing delays to the video by determining the offset and artificially synchronising the two streams. But it can rear it head in much simpler products. I’ve seen designs for wireless light switches that confuse consumers by introducing a perceptible delay between pressing the switch and the light coming on. All of this can be designed around, but engineers need to recognise the issue, particularly the fact that it depends on location (as the level of interference I different in every home) and will vary over time. Which brings me back to the most important observation for any wireless designer – test your products in real locations. They may work very differently in the hands of the customer from the way they do on the workbench.
On the point about Wi-Fi, the issue is that I can now see around thirty access points. Most of my neighbours appear to use these for streaming video, which is a popular use in the UK due to the BBC’s excellent iPlayer service. The result is that the 2.4GHz spectrum in my road is a bit like a motorway at rush hour. That’s not Wi-Fi’s fault – I’m a great believer in Wi-Fi. But without some cooperative cell planning between access points (which will come in the future) I’ve reached the point where cables are more reliable.
Many of the assertions above regarding wireless are simply incorrect. Wireless has no greater latency than do wires, and is subject to exactly the same laws of physics. In fact, wired communications may accurately be said to be “wireless in an expensive tube.” As for the minor difficulties the author has encountered with his Wi-Fi: These have more to do with bad government management of the spectrum than with any inherent limitation of wireless.