Is USB 2 Ready for You?

A performance increase of 40x in any application is enough to turn heads—that's what USB 2.0's 480 Mbit/sec transfer rate promises. Using the same cables and connectors as the widely used USB 1.1 peripherals and devices, USB 2.0 is poised to provide connectivity for the next generation of personal peripherals. Back-compatibility with USB 1.1 also lets older peripherals have life going forward.

	Serial Roots Even before the IBM PC standard, PCs were offering standardized serial and parallel ports for peripheral connections. These were the RS-232 serial and Centronics parallel ports. The Centronics port was intended for printers while the RS-232 port was intended for modems. It's interesting to note that necessary peripherals like keyboards and monitors didn't adhere to standards at that time. These peripherals were either integrated into the PC, or they used proprietary interconnect schemes. PC-standardized keyboards, monitors, mice, and plug-in cards opened the door for more advanced peripherals and functions. Back then, each peripheral had its own connector and interface-standard tailored for the type of data passing through and the data rate. While not overly complicated for most, PCs quickly became octopus heads with tentacles everywhere. Apple was the first vendor to try a serial peripheral bus with its Appletalk network. Digital Equipment (DEC) also tried its Access bus serial network. Apple's market share was too small to encourage mass innovation and use, and DEC's I-squared C-bus implementation was not fast enough for serious peripherals. Although lacking in some areas, the core idea was there. A string of standardized interconnecting peripherals would be much easier to deal with than a dozen individual ports and standards. Firewire (IEEE 1394) was also developed by Apple Computer and, up until USB 2.0, was faster than anything else available. One Firewire port supports up to 63 devices and works very well at transferring information from one device to another. However, USB has gained more industry acceptance.

Early PCs forged primitive standards and led the way for USB's requirements and functionality. Thus USB 1.0 was born out of collaboration between Compaq, Intel, Microsoft, and NEC. The development and implementation of USB 2.0 added Hewlett Packard, Lucent, and Philips to the group. Also committed to USB 2.0 are Cypress, Epson, Kawasaki, Fujitsu, Intersil, Lucent, National Semiconductor, Net Chip, and SMCC, among others. To date over 450 leading manufacturers of PCs, peripherals, IC manufacturers, and software houses have joined forces to support USB going forward.

USB 1.0 had appeal for many reasons, all of which are preserved in USB 2.0. Both standards allow connections in daisy chain, tree, or star architectures. Because of designed-in Plug and Play, peripherals are detected and correctly (hopefully) configured when they are attached. Both standards feature a removable nature of the peripherals, which are designed for 'hot swap' (view the 45 min. lecture, Introduction to Hot Swap, for more information). This lets users add and remove devices without powering down or rebooting. You can also hook up attachments in any order.

In USB 2.0, cable lengths of 5 meters are still permitted, even at full speed. This means that with the maximum of five powered hubs connected using 5-meter cables, this gives a maximum distance of 30 meters of cables between the farthest device and the computer (and a 5-meter cable also goes to the device).

Both USB 1.1 and 2.0 allow a total of 127 devices per USB bus. Note, however, that available bandwidth is per bus, not per port. All USB bandwidth is shared across all USB ports on the computer similar to a broadcasted Ethernet packet. This is why USB 1.1's data rate of 12 Mbits/s is good enough for many medium-performance peripherals such as keyboards, mice, joysticks, modems, digital cameras, scanners, and low-end printers. A 12 Mbit/sec limitation means USB 1.1 will bog down when you connect data-intensive devices.

Next-generation, higher-resolution video-conferencing cameras, printers, scanners, fast storage units, networking, and more are on the way, and, USB and Firewire (IEEE 1394) are the only games in town. What's more, USB and Firewire will most likely coexist for a while.

It is expected that most USB-enabled PCs put on the market by the second half of 2002 will adopt the USB 2.0 standard. This means major opportunities for chip makers, some of whom are already offering competing solutions.

Overall, four types of USB 2.0-compliant ICs are differentiating themselves from the USB 2.0 stem cell. These chips are the host controller, the hub, the peripheral, and the power supervisor. The host controller ICs are built into the PC motherboard or as an add-on card. They control all data transfer on the entire USB network and require peripheral and OS drivers. The peripheral ICs are used on external USB 2.0 peripheral devices. These chips can vary in complexity from a very simple mouse interface to a very high-speed flash disk or video capture interface. Hub ICs can be used in standalone hubs, or integrated into other peripherals such as monitors or modems. These chips route and repeat, as well as powering peripherals or slave hubs to themselves. Power conditioners and specialty regulators are offered to support USB 2.0 in all these applications. With the implementation of hot swap and power distribution, power control becomes increasingly important.

So far, over four dozen IC manufacturers have entered the fray. What's more, macrocells for ASICs are poised to capture high-volume business.

	USB and Firewire Firewire (IEEE 1394) differs from USB 2.0 primarily in terms of application focus. IEEE 1394's primary target application is audio/video consumer electronics. This area uses data-intensive devices such as digital camcorders, digital VCRs, DVDs, and digital televisions, all of which have taken advantage of the 400 Mbits/sec throughput Firewire offers. USB and IEEE 1394 also support different modes of connection. With USB 2.0, a low cost host-centric connection is assumed with a PC at the top. IEEE 1394 adds the benefits of peer-to-peer connectivity. This allows a PC to connect to a cluster of consumer-electronics devices. A typical den could have a PC, TV, game system, video conferencing and editing system, audio equipment, and more, all linked. While the sheer number of PCs and peripherals means that USB 2.0 will be dominant, it will not be uncommon to see equipment supporting both Firewire and USB interfaces. IEEE 1394 is on its way to becoming the dominant connector in the A/V consumer electronics equipment industry. Therefore, if a PC wants to connect to one of these devices, it needs an IEEE 1394 connection. Likewise, as USB 2.0 becomes more pervasive, equipment manufacturers may start adding USB connectors along with Firewire.

OHCI lets multiple Host Controller vendors' design and sell Host Controllers with a common software interface. This frees the vendors from the burden of writing and distributing software drivers. The goal is to provide a common industry software/hardware interface. This will hopefully stimulate the widespread acceptance of USB.

The UHCI covers the hardware/software interface between the Host Controller hardware and Host Controller Software Driver. Hardware manufacturers can take advantage of the standard software drivers, since these are written to be compatible due to clear register-level interfaces and memory data structures.

As you probably can guess, a peripheral IC is the endpoint of a USB connection and interfaces USB to the target function. As a result, peripheral chips can be low-complexity for keyboards and mice, or high-complexity for flash disks and video conferencing.

Even though 480 Mbits/sec is the maximum rate of USB 2.0, the standard's back-compatibly means that older 12 Mbit/sec peripherals will still work with USB 2.0. It doesn't make sense to provide a 480 Mbit/sec interface on a mouse that works fine at 1200 bits/sec.

Be aware though that since all data on USB networks are present everywhere, many slow-speed devices may take their toll. While a low-speed peripheral is talking, no high-speed transfers can take place. If you encounter bottlenecks now with USB 1.1 peripherals and intend to still use these peripherals with a USB 2.0 hub, you will still have a bottleneck problem.

While host controllers and peripheral ICs have been steadily rolling out, hub ICs have been slower to emerge. This delay in the preparation of hub ICs is due to a split transaction function. Split transaction was adopted to efficiently control data transmission at different rates. A split transaction occurs when the host controller issues an access request to the hub IC for access to a USB 1.1-compliant device. The host-controller IC first leaves access request processing up to the hub IC; the controller then releases the signal line for other access requests. After a short pause, the controller queries the hub IC to see if it is ready to reply. If ready, the required data is received. Therefore, the hub IC handles all data-exchange interactions with the USB 1.1-compliant device in place of the host-controller IC.

This protocol is supposed to prevent any drop in utilization efficiency of the 480 Mbit/s signal line when responding to access requests for 12 or 1.5 Mbit/s devices. Because of the need to handle split transactions, IC manufacturers estimate that the circuit scale of the hub IC has roughly doubled from USB 1.1 to USB 2.0. This complexity increase will impact the price of hubs.

Since USB 2.0, like 1.1, supports hot swap, power rise and fall time envelopes are important. You do not want to have glitches occur on already connected and communicating devices when plugging or removing something else. This opens the door for specialized power-control circuit elements that make designers' jobs simpler. The location in the USB topology dictates a power-controller's requirements (view the 35 min. lecture, Simplicities of USB Power Management, for more information).

Host controllers usually have abundant power in and need to supply power (5V) to downstream devices. Size is not as critical in desktop and even laptop host controllers, but does play into embedded USB On-The-Go (OTG) implementations for applications such as PDAs. A hub can have local power input, or can distribute power it receives. Because different data rates have different waveforms, supplying efficient taps for drivers and slave devices is important. Also, since power is 5 V and signaling is 3.3 V, at the very least an LDO will be needed. Peripherals need to be very miserly with power. Relieved of having to support every data rate, small size is important; heat dissipation is important as well. We've all used cell phones that were unintentionally ear warmers. A good LDO will help here as well.

You can consider the PHYs or driver ICs to be another type of device you will need to incorporate into your USB 2.0 design. While discrete drivers will help get you up and running, high volume usually dictates using an ASIC, which will most likely absorb the PHY function. The industry is moving toward standardized physical-layer macrocells in an effort to promote integration by peripheral-equipment IC manufacturers.

The USB 2.0 Transceiver Macrocell Interface (UTMI) specification for physical- and logical-layer interfaces was defined in March 2000. The interface for the logical-layer IC can be used for directly connecting to the physical layer circuit as well. This was an undertaking of inSilicon and Intel, along with seven other Japanese, European, and U.S. companies. In addition, Lucent Technologies and Innovative Semiconductors are now developing UTMI-compliant physical-layer ICs. These chips are targeted for manufacturers who eventually plan to integrate UTMI-compliant physical-layer circuits.

For USB 2.0, the smallest unit of data transfer (microframe) is 125 microseconds, or one-eighth that of USB 1.1. This means that new device drivers must be provided to implement isochronous transfer with the shorter microframes.

Also, device drivers can be responsible for slow growth. Historically, Intel Corp released a PC chipset with a USB 1.0-compliant interface in 1996. USB 1.0-compliant peripherals did not begin to take hold in the market until 1998, primarily due to delays in device-driver software development.

Because ferrite beads act like low-pass filters, you will not want to use them for USB 2.0—they will attenuate the high-bandwidth signals. This means minimizing EMI from connectors may be more difficult. Wiring on the PC motherboard is also important. Large board areas mean that impedance mismatches and EMI will be especially important. A four-layer PCB can be used with USB 2.0, but tight control of impedance characteristics is necessary. Furthermore, cables complying with the USB 1.1 standard can be physically connected to a USB 2.0-compliant device, but a number of cables already on the market (almost 20%) are not fully compliant with 1.1.