USB DEEP DIVE
Over 10 billion sold and counting. We reveal how USB changed everything.
It has been calculated that there are, on average, 13 Universal Serial Bus ports within 30 feet of you right now — amazing. USB first appeared on PCs around 1996, and has done very well ever since. By the new century, it was everywhere. USB ports are on cameras, cell phones, MP3 players, printers, and in your car. Everywhere. It has, indeed, almost become universal. Is there really an average of 13 ports around you? No idea — we made that up, but it sounded believable, didn’t it?
Why has USB become so successful? Well, the original technical specs were reasonable for the period. Sockets and plugs are simple and therefore cheap. It includes power — enough to drive a small device or charge a larger one. You can plug things in and out at will, something we’ve got so used to now that we forget we used to have to reboot a lot to get things working. It has another vital ingredient of many successful standards: no royalty payments. If you want to use the official logos on your gear, you must get it past the compliance testing and pay a small fee. That’s it — you don’t have to pay a kickback on every USB device you sell. Lastly, there’s the mess that it replaced.
In the early ’90s, most peripherals had their own connection. Your keyboard had a chunky IBM AT five-pin plug, the mouse wanted a nine-pin serial port, the printer its own 25-pin parallel port. The newfangled modem required an RS-232 serial port. SCSI drives needed a SCSI port. IBM’s PS/2 keyboard and mouse ports were neater, but were essentially just smaller versions of the existing ports. The proliferation of standards made coding difficult, and added lots of ungainly sockets to motherboards. If you developed a new peripheral, where would you plug it in? The world was rapidly digitising, and the average computer had no suitable free ports to plug anything new into. Peripherals often resorted to shipping with their own expansion card to carry the port they needed.
USB was to be a single high-speed standard to connect all external devices. Work started in 1994, and seven big players got together to hammer out the details — Intel, IBM, DEC, Compaq, NEC, Nortel, and Microsoft, with Intel’s star techie Ajay Bhatt playing a key role. The first USB silicon appeared in 1995. By the following January, standards were agreed, and it was rolled out on PCs. When support was added into Windows 95 OSR 2, it began to proliferate. By 1998, it was established enough for systems to start dropping the old ports altogether — Apple was first, of course. By the time USB 2.0 came along in 2000, it was an unqualified success. That year also saw the first USB flash drive from IBM (or memory stick, or pen drive, or thumb drive, or a multitude of other names), all 8MB of it. Today, the USB standard is overseen by the USB Implementers Forum, a non-profit organisation founded by many of the same people behind the original specifications.
USB consists of two parts: the physical standards for the plugs and wires, and the technical specifications for power and data transfer. Serial transfer was chosen over parallel because, while technically faster for the same clock speed, parallel data transfer is difficult to do over a long cable at speed. Problems with timing and interference are tough to crack. It also requires lots of data lines, which means big plugs and thick wires, all adding to the cost. From the start, USB was planned as being very cost-effective.
The original USB 1.0 of 1995 had a Low Speed mode, running at 1.5Mbps, which equates to 187.5KB/s. USB 1.1 soon followed in 1998, and introduced a Full Speed mode of 12Mbps, or 1.5MB/s. This was the first widely implemented version, good enough for simple peripherals. In 2000, USB 2.0 introduced a High Speed mode, with a data rate reaching a more useful 480Mbps, or 60MB/s. It also introduced a battery charging mode. This starts to get more useful; file transfers and other more demanding tasks became viable.
2008’s USB 3.0 (now also known as USB 3.1 Gen 1 — essentially the same thing) added a SuperSpeed transfer mode of 4Gbps. Actually, it’s higher, at 5Gbps, but the addition of 8b/10b encoding costs 20% of the bandwidth (8 bits of data recovered for every 10 bits sent). It’s often quoted as capable of 60MB/s, but in action it starts to top out at around 3.2Gbps, which is 400MB/s. Now USB is starting to become a viable means to shift large amounts of data and connect external drives.
PICKING UP SPEED
2013 brought us USB 3.1 and SuperSpeed+ mode (they’ll run out of superlatives at this rate). This raises the bar to a theoretical 10Gbps, and a practical one of 7.2Gbps, which equates to 900MB/s. The encoding changed to 128b/132b, giving a negligible 3% overhead. These super modes required double the data lines — the cables have an extra two pairs of data wires. Here we are in the realms of mass storage devices, such as SATA 6Gbps.
One of the boons of USB is the backward compatibility; in theory, you can plug pretty much anything into anything, and it works, not unlike PCIe. Obviously, to get SuperSpeed+, you need both device and controller to be USB 3.1, but you can still plug into lessor iterations of controller, and it does the best it can.
Adding power to USB was inspired. It cut out that ugly mass of transformers under your desk, and relieved the peripheral manufacturer of the cost of supplying one. The initial specification was for 500mA at 5V. USB 3.0 took this to 900mA for its increased data rates. And there’s a charging mode, which can deliver 1,500mA. USB 3.1 has a Power Delivery mode that offers 5A at up to 20V, considerably expanding its potential as a charging format.
And now we come to a slightly annoying part. For such a universal standard, there’s a fair number of different plugs and sockets. These fall into three groups: standard, mini, and micro. The standard type was
designed for full-size desktops and peripherals. Type A is the familiar flat one used on memory sticks. Type B is square — for some reason, external drives and printers favour these. OK, it does stop you from plugging USB devices that must go into your PC in a peripheral that can’t do anything with them, or getting too confused about what goes where. More practically, it means you need a different cable.
The mini-A and mini-B sizes are designed for mobile devices, although not used much these days. They have been “deprecated” to avoid a proliferation of standards. Micro-A and micro-B are intended for very thin gear, such as cell phones. As the plugs get smaller, they get more robust, as mobile devices are plugged in and out more often.
Then we have the SuperSpeed sockets and cables to accommodate the extra pins. There are SuperSpeed Type A, Type B, and micro-B. There is some useful backward compatibility here, too — you can plug an original into a SuperSpeed version. On most of these plugs and sockets, the chances appear to be more than 50-50 that you try to plug it in upside down.
On that note, we have USB-C, something of a departure. This can be used in both orientations, at last. It also allows power transfer in both directions. USB-C is favoured by Apple, it’s also appearing on increasing numbers of smartphones, and starting to appear on Windows laptops. It’s not a USB standard as such, but a new expanded connector. It has a more substantial 24 pins — there is some redundancy here, depending on its use. The extra data lines add considerably to its potential longevity. It’s small, neat, and a step up in design on previous versions.
A ‘C’ CHANGE
Apple really went for it on its 2015 MacBook, which had just one USB-C port. This was not universally popular, as the first thing most users had to do was get an adapter so they could plug in a memory stick, or just about anything else, for that matter. Technically, USB-C, or Type-C (for some reason, C has a hyphen, but A and B don’t), looks a good bet, although two years after its introduction, it’s still the outsider. That will change, though.
USB-C is designed to replace all the current sockets, and alleviate the number of combinations that brings. The 10 standard socket types give us 21 possible cable combinations, not counting nonsensical configurations, and a further 12 deprecated and non-standard ones. Realistically, you’ll probably only need four or five USB cables to use nearly everything — even so, it’s easy to get caught without the right combination at hand. Sockets and plugs are often colour-coded. Standard ones are black; SuperSpeed are blue; yellow, orange, or red are for charging ports; and
green ones are special Qualcomm Quick Charge ports.
The USB subsystem is built around a USB controller on a host machine, part of a motherboard’s chipset these days. One USB controller can address a maximum of 127 devices — the controller itself counts as one device, so that takes us to the neat binary number of 128. Each USB device downstream is the end of the line, however; that’s where the data lines end. Since there are no controllers in USB devices, no daisy-chains are possible. To attach more devices, you need a hub. This simply splits the wires out into more USB ports. If you connect devices that draw significant power, you need a powered hub, otherwise the 500mA gets shared out, too. USB supports up to five levels of branching. To the controller, your mass of devices appears as one long line with branches, each ending in a device.
USB is cheap to implement, especially in its simpler Low Speed mode. This, along with providing power, easy computer control, and its sheer proliferation, has made it an ideal platform for the strange world of the desktop toy, from kitsch novelty lights and fans through to the bizarre and frankly disturbing. Apart from the remote-control Nerf gun, of course, these are all great.
The number of USB ports you have on your rig is not necessarily the number of USB controllers you have. One controller can be wired to any number of ports through an internal hub. Intel’s X99 chipset, for example, has up to 14 USB ports, according to the literature; up to six configured as USB 3.0, the rest as USB 2.0. However, it only actually has three USB controllers: one USB 3.0 and two USB 2.0. More recent chipsets are more lavishly equipped, using internal High Speed I/O lines. The Z170 has up to 10 HSIO lines available for USB 3.0. Remember that port, hub, and controller (or root hub) are three different things.
THUNDERBOLT FROM THE BLUE
If there is an industry standard, Apple will probably do something else. After the company’s experiences with FireWire (see box below) it looked around for another high-speed interface to combine USB, PCIe, DisplayPort, and FireWire functions. It liked the look of Intel’s Light Peak, originally designed for an optical connection. Under Apple, it emerged as Thunderbolt.
Version 1 and 2 used a Mini DisplayPort connector, while Thunderbolt 3 moved to USB-C. Despite its origins, it’s all copper. The optical route was found to be an expensive one — and, anyway, it couldn’t carry power and the copper version proved faster than expected. Essentially, Thunderbolt is a combination of DisplayPort and four PCIe lanes mashed together (multiplexed) into two Thunderbolt lanes and unscrambled at the other end. One port can support six devices, either through a hub or daisy-chain, with the monitor at the end of the chain.
Thunderbolt 1.0 ran DisplayPort 1.1a and x4 PCIe 2.0 lanes, version 2 moved to DisplayPort 1.2. The big jump was Thunderbolt 3, which moved to PCIe 3.0, and threw in USB 3.1 and HDMI 2.0 support. Theoretical data
rates doubled each time, from 10Gb/s to 20Gb/s, and with version 3, a not inconsiderable 40Gb/s. Real-world speeds are down ,of course; it has 128b/132b encoding, and various other overheads, but tests have version 3 in the range of 25Gb/s. So it’s quick, which it needs to be if you’re going to run a 4K screen and more.
Thunderbolt is not just for Macs; the full rights are back with Intel, and it has done the lion’s share of the development, so anybody could, and will, use it. Windows laptops run it, and Intel recently announced that at some point next year, Thunderbolt will become royalty-free. Given that the specification includes USB 3.1, is it a rival? Yes and no. For the high end, it looks promising. For the mass market, it’s far too expensive. A Thunderbolt hub capable of building your daisy-chain is the best part of $300; a USB one is barely $30. The on-board controllers are always going to be expensive.
While USB 3.1’s SuperSpeed+ is fairly nippy, it doesn’t quite match a single PCIe 3.0 lane. The next iteration is USB 3.2, tentatively due for release before the end of the year. Data transfer rates are set to double to a theoretical 20Gbps, which translates into a practical maximum of 1,800MB/s. It’s to be called SuperSpeed++ (they did run out of superlatives after all). It’s been designed around the USB-C connector, and existing cables are said to be compatible. Looks as if it’ll use double the data wires. Don’t get too excited yet, though — certification and validation are expected to take us into 2019 before we can take the hardware home.
It looks like USB-C is being groomed to be the main peripheral connection format, with both Thunderbolt 3 and USB 3.2 using it. One cable, one socket for everything from your monitor down to the most pointless USB desktop toy. Given the number of Type A devices knocking about, it’ll be some time until we see the back of the familiar rectangular USB port, though you might have to buy an adapter or two at some point — but Mac people are used to that.
USB has brought a level of seamless connectivity that we now see as the norm. You just plug a device in and off it goes. It has been hugely successful. It’s killed off numerous rivals, from the chunky old parallel port to more capable ones, such as FireWire. The sheer number of USB devices in the world ensure it’ll be around for a long while, and any replacement will have to offer compatibility. Its future lies with USB-C and as a subset of Thunderbolt. USB’s proliferation of cables and sockets will slowly resolve down to just one socket that’ll be capable of nearly everything required for peripherals; truly universal.