by Steve Kolokowsky and Steven P. Larky, Cypress Semiconductor
Universal Serial Bus (USB) is an open interface standard that has been included on PC motherboards since late 1997. In 1998, the iMac was introduced, featuring only two I/O interfaces—USB and Ethernet. The first versions of the USB spec (1.0 and 1.1) supported maximum transfer rates of 1 Megabyte per second, with a raw bit rate of 12 Megabits per second. Windows 98 provided the first operating system support for USB with built-in drivers for a range of devices, including mice, keyboards, and speakers. In April, 2000, the USB Implementers Forum (USB-IF, comprised of Microsoft, Intel, Philips, HP, Compaq, NEC, and Lucent) released version 2.0 of the USB specification. This release added a new "high-speed" transfer rate. High-speed USB is 480 Megabits per second, 40 times faster than the fastest rate offered in USB 1.1.
Mass storage applications, including hard-disk drives, CD-RW drives, DVD drives, and flash cards are ideal applications for USB 2.0. USB has provided a standard interface for portable devices for years, but the half-gigabit speed of the new 2.0 standard removes any speed bottleneck at the USB interface. Copying files from an internal drive to a USB drive is actually faster than copying between two internal drives sharing a UDMA interface.
USB 2.0 improvements over USB 1.1 seem to have been developed with disk-drive applications in mind. The packet size was increased from 64 bytes to 512 bytes, which is conveniently the same size as a hard-drive sector. Also, the signaling speed increased to almost half a gigabit per second, the same speed as UDMA/66. This provides enough bandwidth to keep up with a 350× CD-ROM drive or the fastest hard drives on the market today.
The best part about using a USB 2.0 disk drive is the ease of use. USB was designed from the ground up to be a true plug-and-play interface. All you do with a USB 2.0 drive is plug it and start to use it. If you plug a 2.0 drive into an older USB host, it is backwards-compatible. If you plug an old USB device into a 2.0 host, it works just fine. This ease of use belies the complex command translation and the number of standards that must be traversed as data move from the Operating System (OS) to the disk drive itself. In the case of a USB 2.0 disk drive, we need to "peel the onion" to work through the layers in order to reveal the steps that make the end-user experience so smooth and successful.
At the outermost layer is the device driver within the OS. To eliminate the need for vendors to write device drivers, the USB Implementors Forum (USBIF) fostered the development of several Class Specifications. A class spec is an OS-independent definition of how a particular USB device talks on the bus. This abstract definition allows OS vendors to create a single driver that will work with all scanners or disk drives, for example. The USB Mass Storage Class Specification has been implemented as a standard part of many operating systems. Microsoft added support to Windows in Windows 2000 and Windows Me, continuing on with support in the latest PC operating system, Windows XP. Mac drivers have been available since OS 9.6. Even Linux support is available now. In addition, many hardware vendors have made support available for legacy operating systems such as Windows 98.
Peeling into the next layer, we need to examine the commands that pass through the Mass Storage Class driver. The Mass Storage Class Spec uses several command-block specifications from the Small Computer Systems Interface (SCSI) standard for all devices, including UFI for floppies, RBC for flash devices, and SFF 8020 for CD-ROMs. These command sets support device 32-bit LBA addressing, enabling disks or disk arrays up to 2 terabytes (200,000 times larger than the hard disk in the original PC XT!). These SCSI commands are sent over a single pair of USB endpoints that are shared by commands and data.
Each transaction starts with a CBW (Command Block Wrapper) that holds the command, direction, expected transaction length, and a unique tag. Once the device has received the CBW, the host starts the data-transfer phase of the command. The data-transfer phase could be terminated by the host (in a successful transfer) or by the device (in an unsuccessful or unexpected transfer). After the data-transfer phase of the command, the device sends a CSW (command status wrapper) to the host to indicate success or failure and the number of bytes that remained in the transfer when the data phase ended.
The vast majority of disk drives used in PCs are IDE (Integrated Device Electronics) drives. Since the Mass Storage Class Specification only uses the SCSI command set, the next layer we have to cross is the command translation needed to support IDE devices. Each SCSI command must be interpreted by the USB-ATA bridge and split into register writes to the IDE LBA registers, sector count registers, and command register. This translation step greatly increases the complexity of the bridge solution. Bridge solutions vary in the number of commands that they support and the accuracy of their SCSI translation.
The innermost layer is the electrical interface to the disk. Fortunately, disk-drive manufacturers have maintained an excellent level of compatibility with their interface standard, the ATA/ATAPI spec, maintained by the T13 subgroup of NCTIS. This standard uses a common 40-pin cable to support speeds from 3 Mbytes/second up to UDMA/33 (33 Mbytes/second). An 80-pin cable with a 40-pin connector provides backwards compatibility and increased speeds up to 133 Mbytes/second. This standard interface is used in all of the USB-to-mass-storage bridges on the market today.
Despite the number of layers described above, the electrical engineering issues involved in building a USB 2.0 Mass Storage device are fairly minimal. The main components designed by the OEM are the disk-drive power supply and the ATA-to-USB translation board, commonly called a "tailgate" board because it sticks out from the disk drive like the tailgate of a pickup truck. The disk drive is typically the same ATA interface model used in PCs, although some drives are available with built-in USB support. The selection of the USB-to-ATA bridge device is critical. For applications with fixed requirements, a fixed-function device—such as the In System Design ISD-300—is a great choice. For applications that may require field upgrades or special features, use a programmable solution such as the Cypress CY7C68013 (EZ-USB FX2™). The FX2 is an 8-bit microcontroller that is capable of doing UDMA/66 to USB 2.0 bridging. This device has 16 I/O pins available for additional features when running the ATA
interface. Excellent reference design kits are available for both of these chips.
The USB 2.0 standard is moving from the lab into mainstream use. PCI hosts that support USB 2.0 appeared in retail shops in April, 2000. Operating system support for USB 2.0 is available in Windows XP and Windows 2000. Computers and motherboards with on-board USB 2.0 support will soon become the standard, replacing the ubiquitous USB 1.1 support available today. Combining end-user ease of use with class-leading performance, USB 2.0 disk drives will build on the existing USB 1.1 host controller base and will gain momentum as USB 2.0 becomes the truly Universal Serial Bus.
