ATA Standards

Today what we call the ATA interface is controlled by an independent group of representatives from major PC, drive, and component manufacturers. This group is called Technical Committee T13 (www.t13.org) and is responsible for all interface standards relating to the parallel AT Attachment storage interface. T13 is a part of the International Committee on Information Technology Standards (INCITS), which operates under rules approved by the American National Standards Institute (ANSI), a governing body that sets rules that control nonproprietary standards in the computer industry as well as many other industries. A second group, called the Serial ATA Working Group (www.serialata.org), has formed to create the Serial ATA standards that will also come under ANSI control. Although these are different groups, many of the same people are in both of them. It seems as if little further development will be done on Parallel ATA past the ATA-7 (ATA/133) specification. The further evolution of ATA will be in the Serial ATA form (discussed later in this chapter).

The rules these committees operate under are designed to ensure that voluntary industry standards are developed by the consensus of people and organizations in the affected industry. INCITS specifically develops Information Processing System standards, whereas ANSI approves the process under which they are developed and then publishes them. Because T13 is essentially a public organization, all the working drafts, discussions, and meetings of T13 are open for all to see.

The Parallel ATA interface has evolved into several successive standard versions, introduced as follows:

ATA-1 (1986–1994)

ATA-2 (1995; also called Fast-ATA, Fast-ATA-2, or EIDE)

ATA-3 (1996)

ATA-4 (1997; also called Ultra-ATA/33)

ATA-5 (1998–present; also called Ultra-ATA/66)

ATA-6 (2000–present; also called Ultra-ATA/100)

ATA-7 (2001–present; also called Ultra-ATA/133)

Each version of ATA is backward compatible with the previous versions. In other words, older ATA-1 or ATA-2 devices work fine on ATA-6 and ATA-7 interfaces. In cases in which the device version and interface version don't match, they work together at the capabilities of the lesser of the two. Newer versions of ATA are built on older versions and with few exceptions can be thought of as extensions of the previous versions. This means that ATA-7, for example, is generally considered equal to ATA-6 with the addition of some features.

ATA-1

Although ATA-1 had been used since 1986 before being published as a standard, and although it was first published in 1988 in draft form, ATA-1 wasn't officially approved as a standard until 1994 (committees often work slowly). ATA-1 defined the original AT Attachment interface, which was an integrated bus interface between disk drives and host systems based on the ISA (AT) bus. Here are the major features introduced and documented in the ATA-1 specification:

40/44-pin connectors and cabling

Master/slave or cable select drive configuration options

Signal timing for basic PIO (Programmed I/O) and DMA (Direct Memory Access) modes

CHS (cylinder, head, sector) and LBA (logical block address) drive parameter translations supporting drive capacities up to 228–220 (267,386,880) sectors, or 136.9GB

ATA-1 was officially published as "ANSI X3.221-1994, AT Attachment Interface for Disk Drives," and was officially withdrawn on August 6, 1999. ATA-2 and later are considered backward-compatible replacements.

Although ATA-1 supported theoretical drive capacities up to 136.9GB (228–220 = 267,386,880 sectors), it did not address BIOS limitations that stopped at 528MB (1,024x16x63 = 1,032,192 sectors). The BIOS limitations would be addressed in subsequent ATA versions because, at the time, no drives larger than 528MB had existed.

ATA-2

Approved in 1996, ATA-2 was a major upgrade to the original ATA standard. Perhaps the biggest change was almost a philosophical one. ATA-2 was updated to define an interface between host systems and storage devices in general and not only disk drives. The major features added to ATA-2 as compared to the original ATA standard include the following:

Faster PIO and DMA transfer modes.

Support for power management.

Support for removable devices.

PCMCIA (PC card) device support.

More information reported from the Identify Drive command.

Defined standard CHS/LBA translation methods for drives up to 8.4GB in capacity.

The most important additions in ATA-2 were the support for faster PIO and DMA modes as well as the methods to enable BIOS support up to 8.4GB. The BIOS support was necessary because, although even ATA-1 was designed to support drives of up to 136.9GB in capacity, the PC BIOS could originally only handle drives of up to 528MB. Adding parameter-translation capability now allowed the BIOS to handle drives up to 8.4GB. This is discussed in more detail later in this chapter.

ATA-2 also featured improvements in the Identify Drive command, which enabled a drive to tell the software exactly what its characteristics are. This is essential for both Plug and Play (PnP) and compatibility with future revisions of the standard.

ATA-2 was also known by unofficial marketing terms, such as fast-ATA or fast-ATA-2 (Seagate/Quantum) and EIDE (Enhanced IDE, Western Digital). ATA-2 was officially published as "ANSI X3.279-1996 AT Attachment Interface with Extensions."

ATA-3

First published in 1997, ATA-3 was a comparatively minor revision to the ATA-2 standard that preceded it. It consisted of a general cleanup of the specification and had mostly minor clarifications and revisions. The most major changes included the following:

Eliminated single-word (8-bit) DMA transfer protocols.

Added SMART (Self-Monitoring, Analysis, and Reporting Technology) support for the prediction of device performance degradation.

LBA mode support was made mandatory (previously it had been optional).

Added security mode, allowing password protection for device access.

Made recommendations for source and receiver bus termination to solve noise issues at higher transfer speeds.

ATA-3 has been officially published as "ANSI X3.298-1997, AT Attachment 3 Interface."

ATA-3, which builds on ATA-2, adds improved reliability, especially of the faster PIO Mode 4 transfers; however, ATA-3 does not define any faster modes. ATA-3 also adds a simple password-based security scheme, more sophisticated power management, and SMART. This enables a drive to keep track of problems that might result in a failure and thus avoid data loss. SMART is a reliability prediction technology that was initially developed by IBM.

ATA/ATAPI-4

First published in 1998, ATA-4 included several important additions to the standard. It included the Packet Command feature, known as the AT Attachment Packet Interface (ATAPI), which allowed devices such as CD-ROM and CD-RW drives, LS-120 SuperDisk floppy drives, tape drives, and other types of storage devices to be attached through a common interface. Until ATA-4 came out, ATAPI was a separately published standard. ATA-4 also added the 33MBps transfer mode known as Ultra-DMA or Ultra-ATA. ATA-4 is backward compatible with ATA-3 and earlier definitions of the ATAPI. The major revisions added in ATA-4 were as follows:

Ultra-DMA (UDMA) transfer modes up to Mode 2, which is 33MBps (called UDMA/33 or Ultra-ATA/33).

Integral ATAPI support.

Advanced power-management support.

Defined an optional 80-conductor, 40-pin cable for improved noise resistance.

Compact Flash Adapter (CFA) support.

Introduced enhanced BIOS support for drives over 9.4ZB (zettabytes, or trillion gigabytes) in size (even though ATA was still limited to 136.9GB).

ATA-4 was published as "ANSI NCITS 317-1998, ATA-4 with Packet Interface Extension."

The speed and level of ATA support in your system is mainly dictated by your motherboard chipset. Most motherboard chipsets come with a component called either a South Bridge or an I/O controller hub that provides the ATA interface (as well as other functions) in the system. Check the specifications for your motherboard or chipset to see whether yours supports the faster ATA/33, ATA/66, ATA/100, or ATA/133 mode.

ATA-4 made ATAPI support a full part of the ATA standard; therefore, ATAPI was no longer an auxiliary interface to ATA but rather was merged completely within. This promoted ATA for use as an interface for many other types of devices. ATA-4 also added support for new Ultra-DMA modes (also called Ultra-ATA) for even faster data transfer. The highest-performance mode, called UDMA/33, had 33MBps bandwidth—twice that of the fastest programmed I/O mode or DMA mode previously supported. In addition to the higher transfer rate, because UDMA modes relieve the load on the processor, further performance gains were realized.

An optional 80-conductor cable (with cable select) is defined for UDMA/33 transfers. Although this cable was originally defined as optional, it would later be required for the faster ATA/66, ATA/100, and ATA/133 modes in ATA-5 and later.

Also included was support for queuing commands, similar to that provided in SCSI-2. This enabled better multitasking as multiple programs make requests for ATA transfers.

ATA/ATAPI-5

This version of the ATA standard was approved in early 2000 and builds on ATA-4. The major additions in the standard include the following:

Ultra-DMA (UDMA) transfer modes up to Mode 4, which is 66MBps (called UDMA/66 or Ultra-ATA/66).

80-conductor cable now mandatory for UDMA/66 operation.

Added automatic detection of 40- or 80-conductor cables.

UDMA modes faster than UDMA/33 are enabled only if an 80-conductor cable is detected.

ATA-5 includes Ultra-ATA/66 (also called Ultra-DMA or UDMA/66), which doubles the Ultra-ATA burst transfer rate by reducing setup times and increasing the clock rate. The faster clock rate increases interference, which causes problems with the standard 40-pin cable used by ATA and Ultra-ATA. To eliminate noise and interference, the new 40-pin, 80-conductor cable has now been made mandatory for drives running in UDMA/66 or faster modes. This cable was first announced in ATA-4 but is now mandatory in ATA-5 to support the Ultra-ATA/66 mode. This cable adds 40 additional ground lines between each of the original 40 ground and signal lines, which help shield the signals from interference. Note that this cable works with older non-Ultra-ATA devices as well because it still has the same 40-pin connectors.

For reliability, Ultra-DMA modes incorporate an error-detection mechanism known as cyclical redundancy checking (CRC). CRC is an algorithm that calculates a checksum used to detect errors in a stream of data. Both the host (controller) and the drive calculate a CRC value for each Ultra-DMA transfer. After the data is sent, the drive calculates a CRC value, and this is compared to the original host CRC value. If a difference is reported, the host might be required to select a slower transfer mode and retry the original request for data.

ATA/ATAPI-6

ATA-6 began development during 2000 and was officially published as a standard early in 2002. The major changes or additions in the standard include the following:

Ultra-DMA (UDMA) Mode 5 added, which allows 100MBps transfers (called UDMA/100, Ultra-ATA/100, or just ATA/100).

Sector count per command increased from 8 bits (256 sectors or 131KB) to 16 bits (65,536 sectors or 33.5MB), allowing larger files to be transferred more efficiently.

LBA addressing extended from 228 to 248 (281,474,976,710,656) sectors, supporting drives up to 144.12PB (petabyte = quadrillion bytes).

CHS addressing made obsolete. Drives must use 28-bit or 48-bit LBA addressing only.

ATA-6 includes Ultra-ATA/100 (also called Ultra-DMA or UDMA/100), which increases the Ultra-ATA burst transfer rate by reducing setup times and increasing the clock rate. As with ATA-5, the faster modes require the improved 80-conductor cable. Using the ATA/100 mode requires both a drive and motherboard interface that supports that mode.

Besides adding the 100MBps UDMA Mode 5 transfer rate, ATA-6 also extended drive capacity greatly, and just in time. ATA-5 and earlier standards supported drives of up to only 137GB in capacity, which was becoming a limitation as larger drives became available. Commercially available 3.5-inch drives exceeding 137GB were introduced during 2001 but originally were available only in SCSI versions because SCSI doesn't share the same limitations as ATA. With ATA-6, the sector addressing limit has been extended from (228) sectors to (248) sectors. What this means is that LBA addressing previously could use only 28-bit numbers, but with ATA-6 LBA addressing can use larger, 48-bit numbers if necessary. With 512 bytes per sector, this raises the maximum supported drive capacity to 144.12PB. That is equal to more than 144.12 quadrillion bytes! Note that the 48-bit addressing is optional and necessary only for drives larger than 137GB. Drives 137GB or less can use either 28-bit or 48-bit addressing.

ATA/ATAPI-7

Work on ATA-7 began late in 2001 and is still underway at the present. As with the previous ATA standards, ATA-7 is built on the previous standard (ATA-6), with some additions.

The primary additions to ATA-7 include the following:

Ultra-DMA (UDMA) Mode 6 added, which allows for 133MBps transfers (called UDMA/133, Ultra-ATA/133, or just ATA/133). As with UDMA Mode 5 (100MBps) and UDMA Mode 4 (66MBps), the use of an 80-conductor cable is required.

Added support for long physical sectors, which allows a device to be formatted so that there are multiple logical sectors per physical sector. Each physical sector stores an ECC field, so long physical sectors allow increased format efficiency with fewer ECC bytes used overall.

Added support for long logical sectors, which allows additional data bytes to be used per sector (520 or 528 bytes instead of 512 bytes) for server applications. Devices using long logical sectors are not backward compatible with devices or applications that use 512-byte sectors, meaning standard desktop and laptop systems.

Incorporated Serial ATA as part of the ATA-7 standard.

Split the ATA-7 document into three volumes: Volume 1 covers the command set and logical registers, Volume 2 covers the parallel transport protocols and interconnects, and Volume 3 covers the serial transport protocols and interconnects.

The ATA/133 transfer mode was actually proposed by Maxtor, and so far it is the only drive manufacturer to adopt this mode. Other drive manufacturers have not adopted the 133MBps interface transfer rate because most drives have actual media transfer rates that are significantly slower than that. VIA, ALi, and SiS have added ATA/133 support to their latest chipsets, but Intel has decided to skip ATA/133 in lieu of adding Serial ATA (150MBps) instead. This means that even if a drive can transfer at 133MBps from the circuit board on the drive to the motherboard, data from the drive media (platters) through the heads to the circuit board on the drive moves at less than half that rate. For that reason, running a drive capable of UDMA Mode 6 (133MBps) on a motherboard capable of only UDMA Mode 5 (100MBps) really won't slow things down much, if at all. Likewise, upgrading your ATA host adapter from one that does 100MBps to one that can do 133MBps won't help much if your drive can only read data off the disk platters at half that speed. Always remember that the media transfer rate is far more important than the interface transfer rate when selecting a drive, because the media transfer rate is the limiting factor.

ATA-7 is still a work in progress, so further changes may come. As a historical note, ATA-7 represents the combining of the venerable Parallel ATA standard and the newer Serial ATA standard under a single specification.
 

Serial ATA

With the introduction of ATA-7, it seems that the Parallel ATA standard that has been in use for more than 10 years is running out of steam. Sending data at rates faster than 133MBps down a parallel ribbon cable is fraught with all kinds of problems because of signal timing, electromagnetic interference (EMI), and other integrity problems. The solution is in a new ATA interface called Serial ATA (SATA), which is an evolutionary backward-compatible replacement for the Parallel ATA physical storage interface. Serial ATA is backward compatible in that it is compatible with existing software, which will run on the new architecture without any changes. In other words, the existing BIOS, operating systems, and utilities that work on Parallel ATA will also work on Serial ATA. This means Serial ATA supports all existing ATA and ATAPI devices, including CD-ROM and CD-RW drives, DVD drives, tape devices, SuperDisk drives, and any other storage device currently supported by Parallel ATA.

Of course, they do differ physically—that is, you won't be able to plug Parallel ATA drives into Serial ATA host adapters, and vice versa. The physical changes are all for the better because Serial ATA uses much thinner cables with only seven pins that are easier to route inside the PC and easier to plug in with smaller redesigned cable connectors. The interface chip designs also are improved with fewer pins and lower voltages. These improvements are all designed to eliminate the design problems inherent in Parallel ATA.

Serial ATA won't be integrated into systems overnight; however, it is clear to me that it will eventually replace Parallel ATA as the de facto standard internal storage device interface found in both desktop and portable systems. The transition from ATA to SATA is a gradual one, and during this transition Parallel ATA capabilities will continue to be available. I would also expect that with more than a 10-year history, Parallel ATA devices will continue to be available even after most PCs have gone to SATA.

Development for Serial ATA started when the Serial ATA Working Group effort was announced at the Intel Developer Forum in February 2000. The initial members of the Serial ATA Working Group included APT Technologies, Dell, IBM, Intel, Maxtor, Quantum, and Seagate. The first Serial ATA 1.0 draft specification was released in November 2000 and officially published as a final specification in August 2001. The Serial ATA II extensions to this specification, which make Serial ATA suitable for network storage, were released in October 2002. Both can be downloaded from the Serial ATA Working Group Web site at www.serialata.org. Since forming, the group has added more than 100 Contributor and Adopter companies to the membership from all areas of industry. Systems using Serial ATA were first released in late 2002.

The performance of SATA is impressive, although current hard drive designs can't fully take advantage of its bandwidth. Three variations of the standard are proposed that all use the same cables and connectors; they differ only in transfer rate performance. Initially, only the first version will be available, but the roadmap to doubling and quadrupling performance from there has been clearly established. Table 9.25 shows the specifications for the current and future proposed SATA versions; the next-generation 300MBps version is not expected until 2005, whereas the 600MBps version is not expected until 2007.

From the table, you can see that Serial ATA sends data only a single bit at a time. The cable used has only seven wires and is a very thin design, with keyed connectors only 14mm (0.55 inches) wide on each end. This eliminates problems with airflow around the wider, Parallel ATA ribbon cables. Each cable has connectors only at each end and connects the device directly to the host adapter (normally on the motherboard). There are no master/slave settings because each cable supports only a single device. The cable ends are interchangeable—the connector on the motherboard is the same as on the device, and both cable ends are identical. Maximum SATA cable length is 1 meter (39.37 inches), which is considerably longer than the 18-inch maximum for Parallel ATA. Even with this thinner, longer, and less expensive cable, transfer rates initially of 150MBps (nearly 13% greater than Parallel ATA/133), and in the future up to 300MBps and even 600MBps, are possible.

Serial ATA uses a special encoding scheme called 8B/10B to encode and decode data sent along the cable. The 8B/10B transmission code originally was developed (and patented) by IBM in the early 1980s for use in high-speed data communications. This encoding scheme is now used by many high-speed data-transmission standards, including Gigabit Ethernet, Fibre Channel, FireWire, and others. The main purpose of the 8B/10B encoding scheme is to guarantee that there are never more than four 0s (or 1s) transmitted consecutively. This is a form of Run Length Limited (RLL) encoding (called RLL 0,4) in which the 0 represents the minimum and the 4 represents the maximum number of consecutive 0s in each encoded character.

8B/10B encoding also ensures that there are never more than six or fewer than four 0s (or 1s) in a single encoded 10-bit character. Because 1s and 0s are sent as voltage changes on a wire, this ensures that the spacing between the voltage transitions sent by the transmitter will be fairly balanced, with a more regular and steady stream of pulses. This presents a more steady load on the circuits, increasing reliability. The conversion from 8-bit data to 10-bit encoded characters for transmission leaves a number of 10-bit patterns unused. Several of these additional patterns are used to provide flow control, delimit packets of data, perform error checking, or perform other special needs.

The physical transmission scheme for SATA uses what is called differential NRZ (Non Return to Zero). This uses a balanced pair of wires, each carrying plus or minus 0.25V (one-quarter volt). The signals are sent differentially: If one wire in the pair is carrying +0.25V, the other wire is carrying —0.25V, where the differential voltage between the two wires is always 0.5V (a half volt). This means that for a given voltage waveform, the opposite voltage waveform is sent along the adjacent wire. Differential transmission minimizes electromagnetic radiation and makes the signals easier to read on the receiving end.

A 15-pin power cable and power connector are optional with SATA, providing 3.3V power in addition to the 5V and 12V provided via the industry-standard 4-pin device power connectors. Although it has 15 pins, this new power connector design is only 24mm (0.945 inches). With three pins designated for each of the 3.3V, 5V, and 12V power levels, enough capacity exists for up to 4.5 amps of current at each voltage, which is ample for even the most power-hungry drives. For compatibility with existing power supplies, SATA drives can be made with either the original, standard 4-pin device power connector or the new 15-pin SATA power connector, or both. If the drive doesn't have the type of connector you needed, adapters are available to convert from one type to the other.

The configuration of Serial ATA devices is also much simpler because the master/slave or cable select jumper settings used with Parallel ATA are no longer necessary.

Serial ATA is ideal for laptop and notebook systems, and it will eventually replace Parallel ATA in those systems as well. In late 2002, Fujitsu demonstrated a prototype 2.5-inch SATA drive. Most 2.5-inch hard drive manufacturers are waiting for mobile chipsets supporting SATA to be delivered before officially introducing mobile SATA drives. It is expected that during 2004 many of the mobile chipsets will incorporate SATA.

MTBF

MTBF is the "mean time between failures," that is, the average elapsed time that passes before a failure occurs in a batch of drives under intense test conditions.

The initial inclination of those unfamiliar with the spec is to interpret it as the average expected lifespan of a single drive and either dismiss it as meaningless (bad) or assume this means their favorite drive will last decades (bad).

The SCSI-standard MTBF rating of 1.2 million hours, for example, does not mean that the average unit will run for 137 years before it fails. Likewise, the other extreme would be 1.2 million drives operating for one hour before one failed- equally unlikely.

Rather, MTBF is a product of a large quantity of drives (numbering in the hundreds or low thousands, perhaps) and the number of hours that such a batch runs before experiencing a failure. If a manufacturer places 1000 units to the test and on average manages to squeeze 50 days of operation out of the batch between each individual unit failure, that firm's product has achieved an MTBF of at least 1.2 million hours.

Sometimes, acceleration factors are used for calculating the MTBF of a drive. These are derived from standard statistical methods. Running the test at elevated temperatures beyond "normal," for example, will reduce the time needed to meet a certain number of test hours. Of course, the accuracy of this acceleration factor has a large effect on the final derived MTBF number.

That said, there's obviously some room for maneuver left in a "sweet spot" where firms may attempt to achieve target MTBF by either using more drives or more hours- there's no set unit count or elapsed time that we may regard as standard.

MTBF should be regarded as a minimum statement of reliability by the manufacturer. These days, no manufacturer will spec an enterprise-class drive below 1.2 million hours. Likewise, no firm will bother with MTBF less than 400,000 for a desktop-class disk. It is this consistency, rather than the spec's lack of meaning, that allows one to gloss over these claims.

Cadangan dan teguran dipersilakan kealamat zubir@pc.jaring.my