The ST-506/412 Interface

    The ST-506/412 interface was developed by Seagate Technologies around 1980. The interface originally appeared in the Seagate ST-506 drive, which was a 5M formatted (or 6M unformatted) drive in a full-height, 5 1/4-inch form factor. By today's standards, this drive is a tank! In 1981, Seagate introduced the ST-412 drive, which added a feature called buffered seek to the interface. This drive was a 10M formatted (12M unformatted) drive that also qualifies as a tank by today's standards. Besides the Seagate ST-412, IBM also used the Miniscribe 1012 as well the International Memories, Inc. (IMI) model 5012 drive in the XT. IMI and Miniscribe are long gone, but Seagate remains as one of the largest drive manufacturers. Since the original XT, Seagate has supplied drives for numerous systems made by many different manufacturers.

    Most drive manufacturers that made hard disks for PC systems adopted the Seagate ST-506/412 standard, a situation that helped make this interface popular. One important feature is the interface's Plug-and-Play design. No custom cables or special modifications are needed for the drives, which means that virtually any ST-506/412 drive will work with any ST-506/412 controller. The only real compatibility issue with this interface is the level of BIOS support provided by the system.

    When introduced to the PC industry by IBM in 1983, ROM BIOS support for this hard disk interface was provided by a BIOS chip on the controller. Contrary to what most believed, the PC and XT motherboard BIOS had no inherent hard disk support. When the AT system was introduced, IBM placed the ST-506/412 interface support in the motherboard BIOS and eliminated it from the controller. Since then, any system that is compatible with the IBM AT (which includes most systems on the market today) has an enhanced version of the same support in the motherboard BIOS as well. Because this support was somewhat limited, especially in the older BIOS versions, many disk controller manufacturers also placed additional BIOS support for their controllers directly on the controllers themselves. In some cases, you would use the controller BIOS and motherboard BIOS together; in other cases, you would disable the controller or mother-board BIOS and then use one or the other. These issues will be discussed more completely later in this chapter in the section "System Configuration."

    The ST-506/412 interface does not quite make the grade in today's high-performance PC systems. This interface was designed for a 5M drive, and I have not seen any drives larger than 152M (Modified Frequency Modulation encoding) or 233M (Run-Length Limited encoding) available for this type of interface. Because the capacity, performance, and expandability of ST-506/412s are so limited, this interface is obsolete and generally unavailable in new systems. However, many older systems still use drives that have this interface.

Encoding Schemes and Problems

    For disk drives, the digital bits are converted, or encoded, in a pattern of magnetic impulses, or flux transitions (also called flux reversals), which are written on the disk. These flux transitions are decoded later when the data is read from the disk.

    A device called an endec (encoder/decoder) accomplishes the conversion to flux transitions for writing on the media and the subsequent reconversion back to digital data during read operations. The function of the endec is very similar to that of a modem (modulator/demodulator) in that digital data is converted to an analog waveform, which then is converted back to digital data. Sometimes, the endec is called a data separator, because it is designed to separate data and clocking information from the flux-transition pulse stream read from the disk.

    One of the biggest problems with ST-506/412 was the fact that this endec resided on the disk controller (rather than the drive), which resulted in the possibility of corruption of the analog data signal before it reached the media. This problem became especially pronounced when the ST-506/412 controllers switched to using RLL endecs to store 50 percent more data on the drive. With the RLL encoding scheme, the actual density of magnetic flux transitions on the disk media remains the same as with MFM encoding, but the timing between the transitions must be measured much more precisely.

    In RLL encoding, the intervals between flux changes are approximately the same as with MFM, but the actual timing between them is much more critical. As a result, the transition cells in which signals must be recognized are much smaller and more precisely placed than with MFM. RLL encoding places more stringent demands on the timing of the controller and drive electronics. With RLL encoding, accurately reading the timing of the flux changes is paramount. Additionally, because RLL encodes variable-length groups of bits rather than single bits, a single error in one flux transition can corrupt two to four bits of data. For these reasons, an RLL controller usually has a more sophisticated error-detection and error-correction routine than an MFM controller.

    Most of the cheaper disk drives on the market did not have data-channel circuits that were designed to be precise enough to handle RLL encoding without problems. RLL encoding is also much more susceptible to noise in the read signal, and conventional oxide media coatings did not have a sufficient signal-to-noise ratio for reliable RLL encoding. This problem often was compounded by the fact that many drives of the time used stepper motor head positioning systems, which are notoriously inaccurate, that further amplified the signal-to-noise ratio problem.

    At this time, manufacturers are implementing RLL-certifying drives for use with RLL endec controllers. This stamp of approval essentially means that the drive has passed tests and is designed to handle the precise timing requirements that RLL encoding requires. In some cases, the drive electronics were upgraded between a manufacturer's MFM and RLL drive versions, but the drives are essentially the same. In fact, if any improvements were made in the so-called RLL-certified drives, the same upgrades usually also were applied to the MFM version.

    The bottom line is that other than improved precision, there is no real difference between an ST-506/412 drive that is sold as an MFM model and one that is sold as an RLL model. If you want to use a drive that originally was sold as an MFM model with an RLL controller, I suggest that you do so only if the drive uses a voice coil head actuator and thin-film media. Virtually any ST-506/412 drive with these qualities is more than good enough to handle RLL encoding with no problems.

    Using MFM encoding, a standard ST-506/412 format specifies that the drive will contain 17 sectors per track, with each sector containing 512 bytes of data. A controller that uses an RLL endec raises the number of sectors per track to 25 or 26.

    The real solution to reliability problems with RLL encoding was to place the endec directly on the drive rather than on the controller. This method reduces the susceptibility to noise and interference that can plague an ST-506/412 drive system running RLL encoding. ESDI, IDE, and SCSI drives all have the endec (and, often, the entire controller) built into the drive by default. Because the endec is attached to the drive without cables and with an extremely short electrical distance, the propensity for timing- and noise-induced errors is greatly reduced or eliminated. This situation is analogous to a local telephone call between the endec and the disk platters. This local communication makes the ESDI, IDE, and SCSI interfaces much more reliable than the older ST-506/412 interface; they share none of the reliability problems that were once associated with RLL encoding over the ST-506/412 interface. Virtually all ESDI, IDE, and SCSI drives use RLL encoding today with tremendously increased reliability over even MFM ST-506/412 drives.

The ESDI Interface

    ESDI, or Enhanced Small Device Interface, is a specialized hard disk interface established as a standard in 1983, primarily by Maxtor Corporation. Maxtor led a consortium of drive manufacturers to adopt its proposed interface as a high-performance standard to succeed ST-506/412. ESDI later was adopted by the ANSI (American National Standards Institute) organization and published under the ANSI X3T9.2 Committee. The latest version of the ANSI ESDI document is known as X3.170a-1991. You can obtain this document, and other ANSI-standard documents, from ANSI itself or from Global Engineering Documents. These companies are listed in Appendix A.

    Compared with ST-506/412, ESDI has provisions for increased reliability, such as building the endec into the drive. ESDI is a very-high-speed interface, capable of a maximum 24Mbit/sec transfer rate. Most drives running ESDI, however, are limited to a maximum 10 or 15Mbit/sec. Unfortunately, compatibility problems between different ESDI implementations combined with pressure from low-cost, high-performance IDE interface drives have served to make the ESDI interface obsolete. Few if any new systems today include ESDI drives, although ESDI became somewhat popular in high-end systems during the late 1980s.

    Enhanced commands enabled some ESDI controllers to read a drive's capacity parameters directly from the drive, as well as to control defect mapping, but several manufacturers had different methods for writing this information on the drive. When you install an ESDI drive, in some cases the controller automatically reads the parameter and defect information directly from the drive. In other cases, however, you still have to enter this information manually, as with ST-506/412.

    The ESDI's enhanced defect-mapping commands provide a standard way for the PC system to read a defect map from a drive, which means that the manufacturer's defect list can be written to the drive as a file. The defect-list file then can be read by the controller and low-level format software, eliminating the need for the installer to type these entries from the keyboard and enabling the format program to update the defect list with new entries if it finds new defects during the low-level format or the surface analysis.

    Most ESDI implementations have drives formatted to 32 sectors per track or more (80 or more sectors per track are possible)--many more sectors per track than the standard ST-506/412 implementation of 17 to 26. The greater density results in two or more times the data-transfer rate, with a 1:1 interleave. Almost without exception, ESDI controllers support a 1:1 interleave, which allows for a transfer rate of 1M/sec or greater.

    Because ESDI is much like the ST-506/412 interface, it can replace that interface without affecting software in the system. Most ESDI controllers are register-compatible with the older ST-506/412 controllers, which enables OS/2 and other non-DOS operating systems to run with few or no problems. The ROM BIOS interface to ESDI is similar to the ST-506/412 standard, and many low-level disk utilities that run on one interface will run on the other. To take advantage of ESDI defect mapping and other special features, however, use a low-level format and surface-analysis utility designed for ESDI (such as the ones usually built into the controller ROM BIOS and called by DEBUG).

    During the late 1980s, most high-end systems from major manufacturers were equipped with ESDI controllers and drives. More recently, manufacturers have been equipping high-end systems with SCSI. The SCSI interface allows for much greater expandability, supports more types of devices than ESDI does, and offers equal or greater performance. I no longer recommend installing ESDI drives unless you are upgrading a system that already has an ESDI controller.

The IDE Interface

    Integrated Drive Electronics (IDE) is a generic term applied to any drive with an integrated (built-in) disk controller. The IDE interface as we know it is officially called ATA (AT Attachment), and is an ANSI standard; however, IDE can roughly apply to any disk drive with a built-in controller.

    The first drives with integrated controllers were hardcards; today, a variety of drives with integrated controllers are available. IDE can support up to 504MB in the DOS OS and up to 100GB in Windows 95 and up.  In a drive with IDE, the disk controller is integrated into the drive, and this combination drive/controller assembly usually plugs into a bus connector on the motherboard or bus adapter card. Combining the drive and controller greatly simplifies installation, because there are no separate power or signal cables from the controller to the drive. Also, when the controller and the drive are assembled as a unit, the number of total components is reduced, signal paths are shorter, and the electrical connections are more noise-resistant, resulting in a more reliable design than is possible when a separate controller, connected to the drive by cables, is used.

    Placing the controller (including endec) on the drive gives IDE drives an inherent reliability advantage over interfaces with separate controllers. Reliability is increased because the data encoding, from digital to analog, is performed directly on the drive in a tight noise-free environment; the timing-sensitive analog information does not have to travel along crude ribbon cables that are likely to pick up noise and insert propagation delays into the signals. The integrated configuration allows for increases in the clock rate of the encoder, as well as the storage density of the drive.

    The IDE connector on motherboards in many systems is nothing more than a stripped-down bus slot. In ATA IDE installations, these connectors normally contain a 40-pin subset of the 98 pins that would be available in a standard 16-bit ISA bus slot. The pins used are only the signal pins required by a standard-type XT or AT hard disk controller. For example, because an AT-style disk controller uses only interrupt line 14, the mother-board AT IDE connector supplies only that interrupt line; no other interrupt lines are needed. The XT IDE motherboard connector supplies interrupt line 5 because that is what an XT controller would use.

    ATA-2 is an extension of the original ATA (IDE) specification. The most important additions are performance enhancing features such as fast PIO and DMA modes. ATA-2 also features improvements in the Identify Drive command allowing 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 is often called Enhanced IDE (or EIDE). EIDE is technically a marketing program from Western Digital. Fast-ATA and Fast-ATA-2 are similar Seagate-inspired marketing programs, which are also endorsed by Quantum. As far as the hard disk and BIOS are concerned, these are all different terms for basically the same thing.

    SCSI (pronounced "scuzzy") stands for Small Computer System Interface. This interface has its roots in SASI, the Shugart Associates System Interface. SCSI is not a disk interface, but a systems-level interface. SCSI is not a type of controller, but a bus that supports as many as eight devices. One of these devices, the host adapter, functions as the gateway between the SCSI bus and the PC system bus. The SCSI bus itself does not talk directly with devices such as hard disks; instead, it talks to the controller that is built into the drive.

    A single SCSI bus can support as many as eight physical units, usually called SCSI IDs. One of these units is the adapter card in your PC; the other seven can be other peripherals. You could have hard disks, tape drives, CD-ROM drives, a graphics scanner, or other devices (up to seven total) attached to a single SCSI host adapter. Most systems support up to four host adapters, each with seven devices, for a total 28 devices! Some of the newer SCSI implementations allow for 15 devices on each bus.

SCSI-1 and SCSI-2

    The SCSI-2 specification essentially is an improved version of SCSI-1 with some parts of the specification tightened and with several new features and options added. Normally, SCSI-1 and SCSI-2 devices are compatible, but SCSI-1 devices ignore the additional features in SCSI-2.

    Some of the changes in SCSI-2 are very minor. For example, SCSI-1 allowed SCSI Bus parity to be optional, whereas parity must be implemented in SCSI-2. Another requirement is that initiator devices, such as host adapters, provide terminator power to the interface; most devices already did so.

SCSI-3

    SCSI-3 is a term used to describe a set of standards currently being developed. Simply put, it is the next generation of documents a product conforms to. See the section "New Commands" later in this chapter.

Fast and Fast-Wide SCSI

    Fast SCSI refers to high-speed synchronous transfer capability. Fast SCSI achieves a 10M/sec transfer rate on the standard 8-bit SCSI cabling. When combined with a 16-bit Wide SCSI interface, this configuration results in data-transfer rates of 20M/sec (called Fast/Wide).

Fast-20 (Ultra) SCSI

    Fast-20 or Ultra SCSI refers to high-speed synchronous transfer capability that is twice as fast as Fast-SCSI. This has been introduced in the Draft (unfinished) SCSI-3 specification and has already been adopted by the marketplace, especially for high-speed hard disks. Ultra SCSI achieves a 20M/sec transfer rate on the standard 8-bit SCSI cabling. When combined with a 16-bit Wide SCSI interface, this configuration results in data-transfer rates of 40M/sec (called Ultra/Wide).

Fast-40 SCSI

    Fast-40 SCSI is a future revision of SCSI-3 (mentioned earlier in the chapter) capable of achieving a 40M/sec transfer rate.

Wide SCSI

    Wide SCSI allows for parallel data transfer at a bus width of 16 bits. The wider connection requires a new cable design. The standard 50-conductor 8-bit cable is called the A cable. SCSI-2 originally defined a special 68-conductor B cable that was supposed to be used in conjunction with the A cable for wide transfers, but the industry ignored this specification in favor of a newer 68-conductor P cable that was introduced as part of the SCSI-3 specification. The P cable superseded the A and B cable combination because the P cable can be used alone (without the A cable) for 16-bit Wide SCSI.

    A 32-bit Wide SCSI version was originally defined on paper as a part of the SCSI-2 specification, but has not found popularity and probably never will in the PC environment. Theoretically, 32-bit SCSI implementations would require two cables: a 68-conductor P cable and a 68-conductor Q cable.

Fiber Channel SCSI

    Fiber Channel SCSI is a specification for a serial interface using a fiber channel physical and protocol characteristic, with a SCSI command set. It can achieve 100M/sec over either fiber or coaxial cable.

Termination

    The single-ended SCSI bus depends on very tight termination tolerances to function reliably. Unfortunately, the original 132-ohm passive termination defined in the SCSI-1 document was not designed for use at the higher synchronous speeds now possible. These passive terminators can cause signal reflections resulting in errors when transfer rates increase or when more devices are added to the bus. SCSI-2 defines an active (voltage-regulated) terminator that lowers termination impedance to 110 ohms and improves system integrity.

02/15/2001