Wires. Cords. Cables. Coils. Lines. Connectors. Whatever you call 'em, you've probably got plenty of them running between various pieces of beloved hardware. In this wired, wired world of ours, we rely on various cables and connectors to get our technology working in sync, to provide us with internet, with data, with everything from a picture on a display to power. But how many of us really know what's going on in those twisted strands?
To that end, we present to you three common connection technologies - explained, unveiled, and detailed so that you're well versed with the inner workings of your interfaces.
Read on to get the goods on HDBaseT, USB 3.0 and Light Peak!
HDBaseT could replace many of your electronics connectors—HDMI, 100BaseT Ethernet, USB, and power—using a single, lowly Cat 5e or Cat 6 Ethernet cable. The technology streamlines your wiring, permits connections to multiple displays/receivers, and greatly extends your reach with no signal degradation. With this emerging technology, an HDBaseT Blu-ray player can sit 100 meters from a connected TV.
How does HDBaseT fit so many signals into an Ethernet cable? We’ll explain the technology and how it can simplify electronics connections while enabling new features.
Like Intel’s Light Peak, HDBaseT passes data through established electronic signals instead of creating its own. Unlike Intel’s fiber-optic take, HDBaseT uses standard Ethernet cables, so it’s cheap. An added bonus is that many buildings and homes already possess the necessary infrastructure.
Home theater devices such as Blu-ray players rely on a traditional HDMI chip to encode the video. An HDBaseT digital signal processor takes this signal and modifies it to run over Ethernet. The process uses pulse-amplitude modulation (PAM), which encodes and rapidly pulses data at varying levels of voltage. A decoder on the other end turns the information back into a native HDMI signal.
While the PAM technique functions in the same basic manner as in gigabit and 10-gigabit Ethernet and uses the same cables, HDBaseT follows a proprietary modulation scheme. Only the physical cables are used, not the underlying Ethernet packet structure. Even HDBaseT’s 100Mb/s Ethernet gets encoded in this way, although an HDBaseT port can revert to plain Ethernet, skipping HDMI and other features, if you accidentally (or deliberately) plug into a traditional Ethernet network. HDBaseT uses its own physical switches, too, although hybrid versions that interface with standard Ethernet are in development.
Version 1.0 of the HDBaseT specification has already been locked in, and specifies a data rate of about 8Gb/s. Even though HDMI requires 10.2Gb/s, the process still works and leaves room for other signals because it is more selective. Micha Risling, marketing committee chair for the HDBaseT Alliance, says, “More than 3Gb/s for HDMI is for error correction…. We don’t need that because the modulation that we’re using is stronger than that specific error correction.” So, HDBaseT skips the standard HDMI error-correction process.
HDBaseT is capable of an even greater data rate, allowing for gigabit Ethernet or future bandwidth added to HDMI. The process would use the same cables but modulate the signal at a faster rate, requiring a more expensive DSP. The HDBaseT Alliance is planning for that in future iterations of the specification.
Even though the HDMI error-correction overhead isn’t used, the full HDMI 1.4 list of uncompressed video and audio formats are supported, including HDCP and the methods of handling 3D frames. However, the first HDBaseT devices omit a few features that require an HDMI 1.4 chip, such as audio-return channel. Risling says chip provider Valens Semiconductor (where he also works) has new HDBaseT chips ready for consumer electronics companies that are compatible with all HDMI 1.4 features.
While HDBaseT is built around HDMI, its other signal capacities add versatility. Data and control channels include support for USB, IR, HDMI’s CEC, and RS-232. This means that even though your TV and Blu-ray player may be in different rooms, your IR remote can still pause, play, and do everything else. Just like HDMI, these signals are passed from their processors and encoded into the HDBaseT PAM signal.
HDBaseT also layers power into the mix, based on the Power over Ethernet standard (PoE+). Unlike that method, however, which tops out at 25.5 watts, HDBaseT can supply up to 100 watts of power.
The standard was designed to span 100 meters because that’s already the established limit for Ethernet networks. HDBaseT can go farther in some situations, or you can add up to five switches to jump additional 100-meter lengths. Content providers also get a say in your distance; an optional setting could keep you to 100 meters total so you don’t broadcast Blu-ray movies to the neighbors. The HDBaseT Alliance says hardware could activate the restriction based on a DRM flag embedded in the content.
We’re used to consortiums forming, discussing standards for years, then finally shipping hardware. HDBaseT is on a faster track since Valens designed the chip first, and then formed the consortium with LG, Samsung, and Sony Pictures Entertainment in June 2010. Valens started selling the HDBaseT chips to its vendors in almost the same step.
HDBaseT devices from AMX, Crestron, and Gefen already exist, and the HDBaseT alliance hopes that consumer devices such as TVs and Blu-ray players with an all-important built-in connection will come soon. Risling says, “We do expect to see consumer electronics products using HDBaseT in 2011 by more than one vendor.”
The biggest challenge facing the emerging technology is price; currently available signal extenders can cost $700 or more. Valens didn’t detail its specific chip costs, but hopes the costs will quickly come down to a few dollars per device. Risling says, “One of the reasons we have companies such as LG and Samsung and Sony [Pictures] joining the alliance is because they believe it’s doable.”
Our wallets ache whenever a new standard takes hold. But if HDBaseT becomes popular, its added features might make home theater component upgrades worthwhile.
Today, we’re starting to see the first motherboards with USB 3.0 support. That support exists in the form of a discrete controller chip, typically the NEC uPD720200; it will be likely be late 2010 or sometime in 2011 before we see USB 3.0 integrated into motherboard chipsets. Still, USB 3.0 is a major leap beyond USB 2.0, so peripheral manufacturers are already announcing products to support the new standard.
First, let’s clarify some terminology. USB 1.0/1.1 was typically just called USB, and supported throughput up to 12 megabits per second. When USB 2.0 arrived, with its 480Mb/s speed, the USB Working Group (www.usb.org) needed a distinguishing name, so that became Hi-Speed USB. USB 3.0 will be called SuperSpeed USB. Got that?
SuperSpeed USB supports maximum throughput of up to five gigabits per second—roughly 10x the speed of USB 2.0. However, USB 3.0 is fully backward compatible with USB 2.0, so you won’t need to throw away your old peripherals as SuperSpeed USB–capable motherboards and systems arrive on the scene.
USB 3.0 manages its backward compatibility with a dual-bus architecture, which operates concurrently with USB 2.0 signaling. It achieves this by adding pins to the USB connector. Note that USB 3.0 is a bidirectional architecture, whereas Hi-Speed USB is unidirectional.
The connector shell for the host system (that familiar flat connector) looks the same. If the host detects no USB 3 connections, it reverts to USB 2.0; otherwise it will run at full USB 3.0 speed.
The interface on the peripheral side—the taller, D-shaped connector—has grown a bit, gaining a slight bulge at the top. Now it looks a little like a skinnier version of the typical Ethernet connector. However, that connector will accept the old style D-shell connector for USB 2 devices. Micro connectors—those pesky, tiny interfaces built into cameras, digital media players, and other smaller devices—are somewhat more problematic. The compatability issue is solved by adding additional real estate to the connector itself, effectively turning it into a double-connector—one for USB 2 devices and one for USB 3 hardware.
There will be two types of micro connectors for USB 3, down from a seeming multitude of small connectors for USB 2.0.
When you start running at 5Gb/s, signaling integrity becomes paramount. So USB 3 will move away from the unshielded twisted-pair cabling used for older USB versions, to shielded, differential-pair cabling. This enables the cable lengths needed for useful peripheral interconnects while maintaining the signal integrity needed for SuperSpeed USB. Cable lengths up to three meters (10 feet) will be supported.
Unsurprisingly, the signaling itself is somewhat similar to PCI Express and Serial ATA, with two differential pairs. Typical USB hot-plug capability will still be supported—a must in the world of easily detachable devices.
One feature that’s going away is the need for polling, which will be replaced by a more interrupt-driven model in which the peripheral will ping the host system, which can then initiate a transfer request. The bus is only active when actually moving data. This improves overall power efficiency, so that battery life remains robust, even at the higher throughput.
In fact, SuperSpeed USB maintains power management at all levels, from the host, through hubs and down the actual physical layers. There’s no broadcasting of packets—they’re only sent when requested. This is managed by asynchronous notifications as to when a device or host link is ready to receive or transmit data. One of the key design parameters was that USB 3.0 links enter a low power state whenever the bus is idle.
More sophisticated power management meant that hubs had to become smarter. Don’t think of a USB 3.0 hub as just a way to add more ports. Hubs will need to monitor upstream and downstream packet to ensure that data is routed to the correct device. Data will need to be buffered until sleeping devices and ports are woken.
A key feature that facilitates power management is Latency Tolerant Messaging (LTM), which allows systems to go into deep sleep states while still maintaining active links to connected peripherals. It’s likely that some peripherals will be more “tolerant” of message delays than others, so the device can actually notify the host system as to what its maximum latency tolerance will be.
One last key feature of note is power delivery. USB 3.0 is rated to deliver 900 milliamps per connection, almost double the current 500mA. That means devices that require higher power now won’t require power bricks. Remember that external optical drive that needed two USB connections—one for power, one for data? That peripheral can now get both power and data off a single connection. More robust USB chargers are also possible.
At 5Gb/s, USB 3.0 is faster than Serial ATA 3Gb/s and almost as fast as SATA 6Gb/s. That makes it plenty fast for external storage, including very fast SSDs. Latencies are likely to be a little higher, but not so much that it will make a difference. That might mean that eSATA’s life is limited, since USB 3.0 will likely be more flexible and offer better power efficiency.
A single USB 3.0 connection will also be roughly the same speed as a single-lane PCI Express 2.0 connection. So external networking, audio, and high-definition video devices that operate in real time become possible. In fact, the first publicly announced USB 3.0 peripheral was the Point Grey USB 3.0 camera, capable of streaming full 1080p video in real time over a SuperSpeed USB connection.
At the same time, don’t expect miracles. If you chain a bunch of devices to a single USB 3.0 port, overall throughput per device will naturally be reduced. Today, we’re seeing systems with eight or 10 USB 2.0 ports; it wouldn’t be surprising to see demand for ports increase in the future. USB 3.0 will also likely become the nail in the coffin of FireWire connections for PCs.
Of course, your usual peripherals will continue to work as expected. Keyboards, mice, card readers, and other USB 2.0 devices will still function. You’ll get faster throughput, better power delivery and robust efficiency. It’s no wonder that motherboard and peripheral manufacturers are jumping on the bandwagon.
Make room in your desk drawer for more obsolete cables. Busses that rely on electrical signals might soon be replaced with fiber-optic alternatives that use light waves, if Intel has its way. “Electrical interconnects have some practical limits we’re starting to see,” says Victor Krutul, director of Intel’s optical I/O team. “When you go through a connector, you get reflections and noise; you get electromagnetic interference.”
Intel’s optical Light Peak technology replaces copper wires with laser pulses traveling through fiber-optic cable. Not only does it not suffer from any of copper’s shortcomings, it’s capable of transferring data at speeds up to 10Gb/s in 100-meter lengths. And Intel is already planning to scale the technology to 100Gb/s.
Rather than functioning as an entirely new data bus, Light Peak will serve as a wrapper for existing buses. A single cable will be capable of carrying USB 3.0, HDMI, and other digital signals at the same time. Intel is meeting with consumer-electronics manufacturers to develop a Light Peak standard and to form a consortium, although Intel has already defined much of the technology. We’ll explain how Light Peak works—from the controller chip to the optics and cable—and where USB 3.0 fits into all of this.
Light Peak is a bridging technology that begins and ends with Intel’s 12mm-square controller chip. This silicon sliver can drive two Light Peak ports and performs three primary tasks: enumeration, routing, and power management.
At the enumeration level, a handshake occurs, during which Light Peak devices identify themselves. Plug a Light Peak–enabled HDMI display and USB keyboard into your PC, for instance, and the Light Peak controller will be informed that you’ve connected an HDMI display to port one and a USB keyboard to port two.
From a routing perspective, the I/O controller inside the PC (or other device, as the case may be) transmits native bus signals—USB 3.0, DisplayPort, PCI Express, or what have you—to the Light Peak controller. The controller keeps the data intact, but translates it from electrical current to pulses of light capable of traveling over the fiber-optic cable. A Light Peak controller inside the receiving device—a NAS box, for example—then translates those pulses back into electrical current.
The transmitting Light Peak controller adds a header on top of the data packets, so that the receiving controller can identify them and send them on to the proper device. This all happens at the transport layer, so the OS doesn’t require any Light Peak software to for the hardware to work.
Power management is the third leg of Light Peak stool: Intel expects that Light Peak cables will include copper wire to carry electrical power as well as fiber optics for data. And when a device that’s connected via a Light Peak cable stops sending data—when you MP3 player finishes syncing to your PC, for example—the controller will automatically shut itself off to conserve power.
Photo detectors, lasers, and other components transmit and receive Light Peak signals. The photo detector remains active at all times, looking for data. When a connected device produces a signal using its tiny laser (a 250-micron-square vertical cavity surface-emitting laser, or VCSEL, to be precise), the detector wakens its corresponding controller chip.
Intel will add lasers to reach 100Gb/s speeds. “With optical, you can use wavelength-division multiplexing, or WDM,” according to Krutul, “which is just a fancy word for saying I can put multiple colors of light on a fiber. One can use a prism-like device to mux [multiplex] the different colored lasers into the fiber [at one end], and a second one at the other end to demux [de-multiplex] them.”
Light Peak cables are bi-directional, utilizing a single 125-micron fiber-optic strand to send data and a second one to receive. The cables are 99 percent fabricated from glass; the balance of the material is a doping agent. Intel considered using plastic fiber, but opted for glass because it delivers superior bandwidth. As we’ve already mentioned, a copper strand will carry electrical current to power devices.
Each fiber strand boasts an active area of 62.5 microns, providing ample tolerance to envelop the lasers’ eight-micron beams. Intel specifies multi-mode strands, which means the light waves follow multiple paths as they travel down the strand. Multi-mode fiber delivers higher bandwidth than single-mode fiber (in which light waves travel in a straight path), but the signal can become distorted by the time it reaches the end of the strand. That’s why Light Peak cables will be limited to 100 meters.
Intel hasn’t formally announced any OEM partners that will put Light Peak into consumer devices, although executives from Sony and Nokia have publicly announced their support for the technology. Intel is developing the Light Peak controller chip and has recruited Avago, Ensphere Solutions, FOCI, Foxconn, Foxlink, IPtronics, and SAE Magnetics to build the lasers, optics, and cables.
If you’re wondering where that leaves USB 3.0, which Intel has yet to support in any of its chipsets, Intel says not to worry. “Our work with Light Peak in no way signals a change of our support for USB 3.0,” Krutul says. “We expect that both of them will be in the market simultaneously—maybe even in the same PC.”
Intel had anticipated that Light Peak consumer devices would ship in 2010, but Krutul says the company has revised its target. “We expect companies will start announcing components at the end of 2010, and that we’ll see Light Peak integrated into computers and devices in 2011.”