PC designers have always sought ways to connect more devices to fewer cables. This reduces the amount of adapter card hardware in the system, so power, space, cost, and maintenance demands are also lowered. In the early 1980s, it became clear that a more versatile and intelligent interface would be needed to overcome the myriad of proprietary interfaces appearing at the time. By 1986, PC designers responded with the introduction of the Small Computer System Interface (SCSI, pronounced "scuzzy"). SCSI proved to be a revolution for PC "power-users" - a single adapter could operate a number of unique devices simultaneously - all "daisy-chained" to the same cable. Where other "low-end" PCs needed one adapter for hard drives, one adapter for the CD-ROM, another adapter for a tape drive, and so on, a system fitted with a SCSI adapter could handle all of these devices, and achieve data throughputs that other interfaces of the day couldn’t even dream of.

Today’s PC industry has changed. Proprietary interfaces are largely discouraged, and the "standardized" interfaces (such as ATA-2 - also known as EIDE) now support a variety of devices while offering low cost and performance levels rivaling SCSI. Yet, SCSI has endured and evolved, and it remains the interface of choice for multitasking and high-end systems. This chapter will examine the inner workings of the SCSI interface, and show you how to deal with installation and troubleshooting problems.

Ideally, peripherals should be independent of the microprocessor's operation. The computer should only have to send commands and data to the peripheral, and wait for the peripheral to respond. Printers work this way. The parallel and serial ports are actually "device-level" interfaces. The computer is unconcerned with what device is attached to the port. In other words, you can take a printer built 12 years ago and connect it to a new Pentium-based system - and the printer will work just fine because only data and commands are being sent across the interface. Very simply put, this is the concept behind SCSI. Computers and peripherals can be designed, developed, and integrated without worrying about hardware compatibility - such compatibility is established entirely by the SCSI interface.

From a practical standpoint, SCSI is a bus - an organization of physical wires and terminations where each wire has its own name and purpose. SCSI also consists of a command set - a limited set of instructions that allow the computer and peripherals to communicate over the physical bus. The SCSI bus is used in systems that want to achieve device independence. For example, all hard disk drives look alike to the SCSI interface (except for their total capacity), all optical drives look alike, all printers look alike, and so on. For any particular type of SCSI device, you should be able to replace an existing device with another device without any system modifications, and new SCSI devices can often be added to the bus with little more than a driver upgrade. Since the intelligence of SCSI resides in the peripheral device itself and NOT in the computer, the computer is able to employ a small set of standard commands to accomplish data transfer back and forth to the peripheral. Now that you understand a bit about the nature of the SCSI interface, the following sections explain some of the important terms and concepts you'll need to know.

At this point, let’s take a look at the evolution of the SCSI interface, and examine the ways in which it has evolved and proliferated. SCSI began life in 1979 when Shugart Associates (you might remember them as one of the first PC hard drive makers) released their "Shugart Associates Systems Interface" (or SASI) standard. The X3T9.2 committee was formed by ANSI in 1982 to develop the SASI standard which was renamed SCSI. SCSI drives and interfaces that were developed under the evolving X3T9.2 SCSI standard were known as SCSI-1, though the actual SCSI-1 standard (ANSI X3.131-1986) didn’t become official until 1986. SCSI-1 provided a system-level 8-bit bus which could operate up to 8 devices, and transfer data at up to 5MB/s. However, the delay in standardization lead to a lot of configuration and compatibility problems with SCSI-1 setups. Table 45-1 compares SCSI-1 specs to other versions.

NOTE: Although SCSI-1 was supposed to support all SCSI devices, manufacturers took liberties with the evolving standard. This frequently led to installation and compatibility problems between SCSI-1 devices which should "theoretically" have worked together perfectly. Today, all existing SCSI-1 adapters should be upgraded to SCSI-2 installations.

Earlier in 1986 (even before the SCSI-1 standard was ratified), work started on the SCSI-2 standard which was intended to overcome many of the speed and compatibility problems encountered with SCSI-1. By 1994, ANSI blessed the SCSI-2 standard (X3.131-1994). SCSI-2 was designed to be backwardly compatible with SCSI-1, but SCSI-2 also provided for several variations. Fast SCSI-2 (or "Fast SCSI") doubles the SCSI bus clock speed and allows 10MB/s data transfers across the 8-bit SCSI data bus. Wide SCSI-2 (or "Wide SCSI") also doubles the original data transfer rate to 10MB/s by using a 16-bit data bus instead of the original 8-bit data bus (the SCSI clock is left unchanged). To support the larger data bus, Wide SCSI uses a 68-pin cable instead of the traditional 50-pin cable. Wide SCSI can also support up to 16 SCSI devices. Designers then combined the attributes of fast and wide operation to create Fast Wide SCSI-2 (or "Fast Wide SCSI") which supports 20MB/s data transfers across a 16-bit data bus. Whenever you see references to "Fast SCSI", "Wide SCSI", or "Fast Wide SCSI", you’re ALWAYS dealing with a SCSI-2 implementation.

But SCSI advancement hasn’t stopped there. ANSI began development of the SCSI-3 standard in 1993 (even before SCSI-2 was adopted). SCSI-3 is intended to be backward compatible with SCSI-2 and SCSI-1 devices. Although SCSI-3 is still not finalized, there are many SCSI devices and controllers that are making use of the advances offered by SCSI-3 development. These early SCSI-3 devices are generally known as Fast-20 SCSI (or Ultra SCSI-3, also termed "Ultra SCSI"). Ultra SCSI uses a 20MHz SCSI bus clock with an 8-bit data bus to achieve 20MB/s data transfers. By using a 16-bit data bus, SCSI-3 offers Wide Fast-20 SCSI (or Wide Ultra SCSI-3, also termed "Wide Ultra SCSI") which handles 40MB/s data transfers.

For the future, the SCSI-3 standard is also proposing Fast-40 SCSI (called Ultra2 SCSI-3 and "Ultra2 SCSI") using a 40MHz bus clock to provide 40MB/s data transfers with an 8-bit data bus. The 16-bit data bus version is known as Wide Fast-40 SCSI (called Wide Ultra2 SCSI-3, or "Wide Ultra2 SCSI") which is supposed to support 80MB/s data transfers. Whenever you see references to "Ultra", "Fast-20", "Ultra2", or "Fast-40", you’re almost certain to be faced with a SCSI-3 setup.

NOTE: You’ll probably encounter a lot of literature using the term "Ultra SCSI", but the use of "Ultra" as a SCSI-3 designator is being actively discouraged because of legal disputes with companies using the term "Ultra" in their SCSI-2 (yes, SCSI-2) devices. As a rule, use the "Fast" or "Wide Fast" terms instead of the "Ultra" terms.

Also keep in mind that SCSI has traditionally been a "parallel" bus - that is, 8 or 16 bits of data are transferred at a time across parallel data lines. SCSI-3 is proposing three new serial connection schemes. You’ll see these noted as Serial Storage Architecture (SSA), Fibre Channel, and IEEE P1394 (a.k.a. "Fire Wire"). These serial schemes will offer faster data transfers than their parallel bus cousins, but are NOT backward compatible with SCSI-2 or SCSI-1.

There are basically two types of devices on the SCSI bus; initiators and targets. An initiator starts communication when something has to be done, and a target responds to the initiator's commands. The important thing for you to understand here is that this "master/slave" relationship is not a one-way arrangement - an initiator may become a target at some points in the data transfer cycle, and the target may become the initiator at other points. You will see more about this role duality later in this chapter. A SCSI bus can support up to 8 devices simultaneously, but there MUST be at least one initiator and one target in the system. An SCSI host adapter (the expansion card installed in one of the computer's expansion slots) is typically the initiator, and all other devices (i.e. hard drives or CD-ROMs) are usually targets, but that is not necessarily the only possible case.

Many kinds of computer peripherals are candidates for the SCSI bus. Each peripheral offers unique characteristics and applications, but each also requires different methods of control. By adding SCSI "intelligence" to these devices, they can all be made to share the same bus together. The SCSI nomenclature groups similar devices together into specific "device types". The original SCSI standard defines six devices:

• random access devices (i.e. hard drives)
• sequential access (i.e. tape drives)
• printers
• processors
• WORM (write once read many) drives
• read-only random access devices

The SCSI-2 interface adds five more devices to the specification:

• CD-ROM drives
• scanners
• magneto-optical drives
• media changer (jukebox)
• communication devices

As a system-level interface, SCSI requires an operating "handshaking protocol" that organizes the transfer of data from a sending point to a requesting point. There are typically three handshaking protocols for SCSI; asynchronous, synchronous, and fast synchronous. The asynchronous protocol works rather like a parallel port. Each byte must be requested and acknowledged before the next byte can be sent. Asynchronous operation generally results in very reliable (but slow) performance. Synchronous and fast synchronous operation both ignore the request/acknowledge handshake for data transfer only. This allows slightly faster operation than an asynchronous protocol, but a certain fixed amount of time delay (sometimes called an "offset") must be allowed for request and acknowledge effects. The fast synchronous protocol uses slightly shorter signals, resulting in even faster speed. An important point to remember is that SCSI systems can typically use any of these three protocols as desired. The actual protocol that is used must be mutually agreed to by the initiator and the target through their communications. SCSI systems normally initialize in an asynchronous protocol.

There are a number of instances when it would be desirable to allow a target to operate off-line while the initiator is occupied elsewhere. Tape rewind time is just one example. An important feature of SCSI is the ability to disconnect two communicating devices, then reconnect them again later. Disconnect and reconnect operations allow several different operations to occur simultaneously in the system - and is the main reason why SCSI architecture is so desirable in a multitasking environment. It is up to the initiator to grant a disconnect privilege to a target.

The signal wiring used in an SCSI bus has a definite impact on bus performance. There are two generally-used wiring techniques for SCSI; single-ended and differential. Both wiring schemes have advantages and disadvantages.

The single-ended (SE) wiring technique is just as the name implies - a single wire carries the particular signal from initiator to target. Each signal requires only one wire. Terminating resistors at each end of the cable help to maintain acceptable signal levels. A common ground (return) provides the reference for all single-ended signals. Unfortunately, single-ended circuitry is not very noise resistant, so single-ended cabling is generally limited to about 6 meters at data transfer speeds of 5MHz or less. At higher data transfer speeds, cable length can be as short as 1.5 meters. In spite of the disadvantages, single-ended operation is simple and popular because of its simplicity.

The differential (DIF) wiring approach uses two wires for each signal (instead of one wire referenced to a common ground). A differential signal offers excellent noise resistance because it does not rely on a common ground. This allows much longer cables (up to 25 meters) and higher-speed operation (10 MHz). An array of pull-up resistors at each end of the cable help to ensure signal integrity. The problem with differential wiring is that it is more complicated than single-ended interfaces.

When high-frequency signals are transmitted over adjacent wires, signals tend to degrade and interfere with one another over the length of the cable. This is a very natural and relatively well-understood electrical phenomenon. In the PC, SCSI signal integrity is enhanced by using powered resistors at each end of the data cable to "pull up" active signals. Most high-frequency signal cables in the PC are already terminated by pull-up resistors at drives and controller cards. The small resistor array is known as a terminator. Since there is a distinct limit to the number of devices that can be added to a floppy drive or IDE cable, designers have never made a big deal about termination - they just added the resistors and that was it. With SCSI, however, up to 8 devices can be added to the bus cable. The SCSI cable also must be terminated, but the location of terminating resistors depends on which devices are added to the bus, and where they are placed. As a result, termination is a much more vital element of SCSI setup and troubleshooting. As you will see later in this chapter, poor or incorrect termination can cause intermittent signal problems. Later on. you will see how to determine the proper placement of terminating resistors.

Termination is typically either active or passive. Basically, passive termination is simply plugging a resistor pack into a SCSI device. Passive resistors are powered by the TERMPWR line. Passive termination is simple and effective over short distances (up to about 1 meter) and usually works just fine for the cable lengths inside a PC, but can be a drawback over longer distances. Active terminators provide their own regulated power sources which makes them most effective for longer cables (such as those found in external SCSI devices like page scanners) or Wide SCSI systems. Most SCSI-2 implementations use active terminators. A variation on active termination is forced perfect termination (or FPT). FPT includes diode clamps which prevent signal overshoot and undershoot. This makes FPT effective for long SCSI cable lengths.

A SCSI bus will support up to 8 devices. This means each device on the bus must have its own unique ID number (0 to 7) - if two devices use the same ID, there will be a conflict. IDs are typically set on the SCSI adapter and each SCSI device using jumpers or DIP switches. Typically, the SCSI adapter is set for ID 7, the primary SCSI hard drive is set to ID 0, and a second SCSI hard drive is ID 1. Other devices can usually be placed anywhere from ID 2 to ID 6.

Most of the SCSI implementations currently available use single-ended cabling that supports an 8-bit data bus (known as an A-cable). An A-cable is a 50-pin assembly outlined in Table 45-2. There are three major sections to the 50-pin single-ended SCSI cable; ground wires, data signals, and control signals. You will notice that at least half of the single-ended interface carries ground lines. There are 8 data lines (D0 to D7) and a data parity bit (DPAR). Note that SCSI parity is always odd. There are four terminator power lines (TERMPWR) and nine control signal wires. Each signal is explained below:

• -C/D - Control/Data (driven by target) - allows the target device to select whether it will be returning a command or data to the initiator.
• -I/O - Input/Output (driven by target) - allows the target device to determine whether it will be receiving or sending information along the data bus.
• -MSG - Message (driven by target) - allows the target device to send coded status or error messages back to the initiator during the "message" portion of the SCSI bus cycle.
• -REQ - Request (driven by target) - a data strobe signal which allows a potential target device to obtain data on the bus.
• -ACK - Acknowledge (driven by initiator) - a data strobe signal sent in response to the target's REQ signal which informs the target device that it has gained use of the bus.
• -BSY - Busy (driven by initiator or target) - allows a device to inform the bus that the device is currently busy.
• -SEL - Select (driven by initiator or target) - a signal used by an initiator in order to select a target device.
• -ATN - Attention (driven by initiator) - a signal produced by the initiator which informs the target that the initiator has a message ready. The target should switch to the "message" phase.
• -RST - Reset (driven by initiator or target) - a strobe signal that triggers a bus-wide reset of all devices. Usually, only one device produces a Reset signal.

The differential SCSI interface replaces most of the ground wires with +signal leads. For example, pin 2 represents +D0, while pin 27 is -D0. These + and - signal pairs are the differential signals. Note that there are still a few ground wires, but the grounds are not related to differential signals as they are to single-ended signals. Just about all of the data and control signals in the differential interface serve an identical purpose in the single-ended interface, but you will notice that the signal locations have been rearranged as shown in Table 45-3. There is one additional differential signal - the DIFFSENS (Differential Sense) line which provides an active high enable for differential drivers. Keep in mind that plugging a differential cable into a single-ended interface (or vice versa) can damage the device, the SCSI adapter, or both.

As you might imagine, wide SCSI implementations will not work with A-cables. A 16-bit cable is needed. Early implementations of wide SCSI used a second cable to provide the extra signal lines, but was quickly abandoned for a single cable assembly (called a P-cable). The single-ended P-cable is shown in Table 45-4. While many of the signals may look familiar, you will notice that there are 68 pins instead of 50 - primarily to support the 8 additional data lines (D8 to D15). Control lines are identical to those in the A-cable. Table 45-5 shows the pinout for a differential 68-pin P-cable.

Now that you have learned about SCSI bus concepts and structure, you can see how the interface behaves during normal operation. Since bus wires are common to every device attached to the bus, a device must obtain permission from all other devices before it can take control of the bus. This attempt to access the bus is called the "arbitration phase". Once a device (such as the SCSI controller) has won the bus arbitration, it must then make contact with the device to be communicated with. This device selection is known as the "selection phase". When this contact is established, data transfer can take place. This part of the article will detail negotiation and information transfer over the SCSI bus.

Devices must negotiate to access and use an SCSI bus. Negotiation begins when the bus is free (BSY and SEL lines are idle). A device begins arbitration by activating the BSY line and its own data ID line (data bit D0 to D7 depending on the device). If more than one device tries to control the bus simultaneously, the device with the higher ID line wins. The winning device (an initiator) attempts to acquire a target device by asserting the SEL line and the data ID line (data bit D0 to D7) of the desired device. The BSY line is then released by the initiator, and the desired target device asserts the BSY line to confirm it has been selected. The initiator then releases the SEL and data bus lines. Information transfer can now take place.

The selected target controls the data being transferred, and the direction of transfer. Information transfer lasts until the target device releases the BSY line, thus returning the bus to the idle state. If a piece of information will take a long time to prepare for, the target can end the connection by issuing a disconnect message. It will try to re-establish the connection later with a new arbitration and selection procedure.

During information transfer, the initiator tells its target how to act on a command, and establishes the mode of data transfer during the message out phase. A specific SCSI command follows the message during the command phase. After a command is sent, data transfer takes place during the data in and/or data out phases. The target relinquishes control to the initiator during the command phase. For example, the command itself may ask that more information be transferred. The target then tells the initiator whether the command was successfully completed or not by returning status information during a status phase. Finally, the command is finished when the target sends a progress report to the initiator during the message in phase. Consider the simple SCSI communication example below:

1. Bus Free Phase (system is idle)
1. Arbitration Phase (a device takes control of the bus)
1. Select Phase (the desired device is selected)
1. Message-Out Phase (target sets up data transfer)
1. Command Phase (send command)
1. Data-In Phase (exchange data)
1. Status Phase (indicate the results of the exchange)
1. Message-In Phase (indicate exchange is complete)
1. Bus Free Phase (system is idle)

Whether you are considering adding SCSI support to your own computer, or planning an upgrade for a customer, there are four essential elements that you must consider; the SCSI peripheral, the SCSI host adapter, the SCSI cable assembly, and the SCSI software driver(s). If any one of these four elements is missing or ill-planned, your installation is going to run into problems.

The first item to be considered is the SCSI peripheral itself. You first need to know what type of device is needed (such as a SCSI hard drive or CD-ROM). The peripheral should be compatible with SCSI-2 architecture. You may also find a growing base of SCSI-3 compliant adapters and peripherals. Each SCSI peripheral device should also have a wide range of available SCSI ID settings. SCSI typically handles 8 IDs (0 to 7) and the peripheral should have the flexibility to run on virtually any ID. If only a few IDs are available, you may be limited when it comes time to add other SCSI devices. Peripherals should support SCSI parity.

NOTE: Ideally, a SCSI-3 host adapter should support SCSI-2 devices. If you have any intention of employing SCSI-3 (Wide/Fast 20 SCSI) devices, be sure to use a SCSI-3 adapter.

SCSI devices are available in both internal and external versions. If you consider an internal peripheral, make sure that there is adequate drive space in the PC to accommodate the new peripheral. Either there is a drive bay available, or an existing device may be removed to make room. If the peripheral is to be an external device (such as a printer or scanner), there should be two SCSI connectors on the device to allow for daisy-chaining additional devices later. All SCSI peripherals other than hard drives will require device drivers. Make sure that the device driver is compatible with the same standard protocol used by the adapter (i.e. ASPI, CAM, or LADDR). This is a serious consideration since peripherals using incompatible device driver standards will not work properly. Finally, try to choose SCSI peripherals that offer built-in cable termination.

The next item to be considered is the SCSI host adapter (often just called a "host" or "HA") that fits in the PC expansion bus (Fig. 45-1). Make sure to choose an adapter that is compatible with the PC bus in use (i.e. ISA, EISA, MCA, PCI, and so on). Bus-mastering 32/64-bit PCI SCSI adapters will provide superior performance if your system will support them. Like the peripheral itself, the adapter should also be designed to support the SCSI-2 standard (or SCSI-3 if possible). Although most adapters are assigned a SCSI ID of 7, the adapter should be flexible enough to work with any ID from 0 to 7. The host adapter will also require a device driver for using devices other than hard drives. Make sure that the host device driver uses the same standard as the peripheral(s) (ASPI, CAM, or LADDR). It is important to note here that the driver standard has nothing to do with the choice of SCSI, SCSI-2, or SCSI-3. It is only important that the peripherals and the adapter use the same driver standard.

Check that you select the proper cabling for the SCSI level you are using. Although SCSI cabling is now highly standardized, some older cables may use slight modifications for particular peripherals (a typical trick used with SCSI-1 devices. Be certain that you know of any specialized cabling requirements when choosing peripherals. Try to avoid specialized cabling if at all possible, but if you must use specialized cabling, you should determine what impact the cabling will have on any other SCSI peripherals that may be installed (or may be installed later). Use good-quality SCSI cables specifically intended for the SCSI level you are using (probably SCSI-2), and keep the cables short to minimize signal degradation.

SCSI cables must be terminated at the beginning (host adapter) and end (after the last device) of the SCSI chain. Try to choose internal peripherals that have built-in terminators. Also try to select a host adapter and peripherals that use the same type of terminator resistor network. SCSI-2 systems use active terminator networks. You will see much more about cabling and termination a bit later in this article.

Device drivers provide the instructions that allow the SCSI host adapter to communicate with the PC, as well as the peripherals in the SCSI chain (or the SCSI bus). The host adapter itself will require a device driver, as will every peripheral that is added. For example, a SCSI system with one CD-ROM will need a driver for the host adapter and a driver for the CD-ROM. Make sure that driver standards (ASPI, CAM, or LADDR) are the SAME for the host adapter and peripherals. The only exception to the device driver requirement (at this time) is the SCSI hard drive which may be supported by the SCSI adapter’s BIOS ROM.

Real-mode device drivers are added by including them in your PC’s CONFIG.SYS and AUTOEXEC.BAT files. One issue to keep in mind when adding device drivers is that drivers use conventional memory (unless you successfully load the drivers into high memory). The more drivers that are added, the more memory that will be consumed. It is possible that a large number of device drivers may prevent certain memory-demanding DOS applications from running. To keep as much conventional memory (the first 640KB in RAM) free as possible, use the DOS devicehigh and loadhigh features to load the drivers into upper memory (from 640KB to 1MB in RAM). Windows 95 uses protected-mode drivers for the host adapter and devices.

SCSI upgrades are not terribly difficult to perform properly, but the subtle considerations and inconsistencies that have always been a part of SCSI implementations can result in confusion and serious delays for you and your customer. The following tips should help to ease your upgrades:

• Add only add one SCSI device at a time. By adding one device at a time and testing the system after each installation, it becomes much easier to determine the point where problems occur. Suppose what happens when you add an adapter, hard drive, and CD-ROM. If the system fails to function, you will have to isolate and check each item to locate the fault. On the other hand, by adding the adapter and testing it, then adding the hard drive and testing it, then adding the CD-ROM and testing it, installation troubleshooting becomes a much simpler matter (although it may take a bit more time overall).
• Record the host adapter’s resources. One of the most difficult aspects of troubleshooting is determining what the configuration of a system is. This is especially important during an upgrade since you must know the interrupts (IRQs), DMA channel(s), and I/O ranges used by other expansion devices in the PC. Any overlap in the use of these system resources will eventually result in a hardware conflict. When you install a SCSI host adapter, make it a point to record its IRQ, DMA, and I/O settings along with the SCSI ID settings of all devices that are installed. Tape the record to the inside of the PC's cover - next time the PC returns for service or upgrade, you'll have the information right at your fingertips.
• Use good-quality cabling. Using the correct terminators and cables can have a profound effect on the performance of your SCSI installation. Good-quality cables and terminators provide electrical characteristics that support good signal transfer. This results in good data reliability between the host controller and peripherals. If cable quality is sub-standard or terminator networks are not correct for the SCSI level being used, the cable’s electrical characteristics and data transfer will be degraded.

The SCSI adapter is an expansion board - much like any other expansion board in your PC. You will need to configure the adapter before installing it. Most SCSI adapters need four system resources; an IRQ, a DMA channel, an I/O range, and ROM addresses. Settings are typically made by changing jumper placement. The user’s manual for your particular adapter will outline precisely what selections are available, and how to change each one. When choosing system resources, be very careful to avoid conflicts with other adapters in your system. Although manufacturers try to avoid conflicts by "pre-setting" the adapter to rarely used settings, you should check for possible conflicts anyway.

You can also set the adapter’s SCSI ID and the SCSI parity. In almost all circumstances, the adapter will use a SCSI ID of 7. Parity is a means of error checking the data passed along a SCSI data path. The problem with parity is that all installed SCSI devices must support it, or parity should be disabled. If you select only peripheral devices that support SCSI parity, you can enable it on the adapter. Record the settings on paper and tape the paper inside the PC's cover. Insert the adapter into an available expansion slot and secure the board properly.

NOTE: New SCSI host adapters (mainly PCI boards) are often Plug-and-Play (PnP) devices, and are configured through the SCSI BIOS and host adapter driver(s) at startup.

It should not be difficult to configure a SCSI peripheral. You are concerned with setting the SCSI ID and SCSI parity. The ID (also called Target ID or Target SCSI ID) can range from 0 to 7. Since the adapter is almost always set at 7, only 0-6 remain. However, SCSI hard drives for AT-compatible machines should be issued IDs of 0 or 1. As a general rule, do not use ID 0 or ID 1 for any devices but hard drives. If you intend to boot your PC from the SCSI hard disk, assign it an ID of 0. PS/2 machines place the bootable hard drive on SCSI ID 6. When assigning IDs in systems with more than one SCSI peripheral, be careful not to use duplicate IDs. Each device must have its own unique ID number.

If the device supports SCSI parity, the setting should be enabled. Keep in mind that to use SCSI parity, all SCSI devices in the system must support it. If even one device does not support it, parity must be disabled system-wide. When another device already in the system relies on SCSI parity (such as a CD-ROM drive), disabling parity to accommodate a new device can render an existing device inoperative. You may need to change your selection of peripheral to one that supports SCSI parity.

Depending on the SCSI device being installed, you may also need to set a Start On Command jumper. Drives draw a serious amount of power during startup. If a large number of devices are trying to draw power, the power supply can be overloaded. A Start On Command option (if available on your peripheral) will keep the device idle until a start command is sent from the SCSI adapter. This way, multiple SCSI devices can be started in a staggered fashion to "spread out" the power load. For external SCSI devices, situate the device close to the computer - SCSI cables tend to be kept short. If the peripheral is an internal device, you should now mount it in an available drive bay.

Once the host adapter and peripheral are configured and installed, you must connect them with a cable. Internal devices are typically connected with a 50-pin IDC (insulation displacement connector) ribbon cable (an A-cable). By placing multiple connectors along the length of cable, daisy-chaining can be achieved with a single connector on each internal device. External devices typically connect to an external 50-pin connector on the rear of the SCSI adapter, and each device offers two connectors which allows daisy chaining to additional devices. Most commercial adapter and drive "kits" are packed with an appropriate cable.

The cable(s) must be terminated. There are internal and external SCSI cable terminators, along with SCSI devices that have terminating resistor networks already built in. The concept of termination is reasonably simple - achieve the desired signal cable characteristics by loading each end of the SCSI "chain" with resistors. If the chain is not terminated properly, signals will not be carried reliably (which invariably results in system errors). For technicians and end-users alike, the trouble usually arises in determining there the "ends" are. A number of examples will help to clarify how to determine the chain "ends".

For a single SCSI drive and adapter as shown in Fig. 45-2, the "ends" are easy to see. One end should be terminated at the host adapter (which usually has terminating resistors built in). The other end should be terminated at the SCSI hard drive (which also usually has terminating resistors built in). In this type of situation, you need only connect the cable between both devices, and verify that the terminators are in place.

When a second SCSI peripheral is added as shown in Fig. 45-3, termination becomes a bit more complex. Suppose a CD-ROM is added with a SCSI ID of 6. The terminator on the existing SCSI hard drive is no longer appropriate - it should be removed, and the termination should be made on the CD-ROM which is now the last device in the SCSI chain. In most cases, a terminator network can be deactivated by flipping a DIP switch or changing a jumper on the peripheral itself. If the terminator can not be "shut off", it can almost always be removed by gently easing the resistor network out of its holder using needle nose pliers. If you remove a terminator, place it in an envelope and tape it to the inside of the PC enclosure. If it is simply impossible to remove the existing terminator on the hard drive, place the CD-ROM between the adapter and hard drive and remove the CD-ROM's terminator (re-arrange the chain). The SCSI host adapter must remain terminated.

So what happens if an external device is used (such as a scanner) as in Fig. 45-4? An external cable connects the adapter to the scanner. Since the scanner (ID 6) and adapter (ID 7) are the only two points in the chain, both are terminated. Most external devices designed for SCSI-2 compatibility allow the active terminator built into the peripheral to be switched off if necessary.

Suppose both an internal and an external SCSI device are being used as shown in Fig. 45-5. The SCSI host adapter (ID 7) is no longer at an end of the chain, so its terminator should be switched off or removed. It is the internal hard drive (ID 0) and external scanner (ID 6) which now form the ends, so both devices should be terminated. Since both peripherals should ideally support internal termination, nothing needs to be done except to confirm that the terminators are in place and switched on.

Hardware configuration and installation is only one part of the SCSI installation. Software needs to be in order to allow the hardware to interact with your system. The problem with SCSI drivers is that prior to 1991, various drivers were rarely compatible. For example, an adapter and hard drive may have worked fine, but adding a CD-ROM would create havoc since the CD-ROM driver was not compatible with the hard drive or the host adapter driver (or both). After 1991, a set of "universal" driver standards appeared which created a "buffer" between the operating system and hardware which isolated each particular driver from one another. Drivers can now be written for each peripheral without worry of incompatibility so long as the drivers are written to be compatible with the standard.

There are now three competing SCSI standards; ASPI (Advanced SCSI Programming Interface), CAM (Common Access Method), and LADDR (Layered Device Driver Architecture). ASPI is the most popular of the three standards. The idea for compatibility is to select a host adapter and peripherals that support the same standard. For example, if you select a host adapter that uses an ASPI driver, each of the peripherals that you choose must also use ASPI drivers. If you upgrade the host adapter later, you also upgrade the host's ASPI driver - full compatibility should be maintained.

The actual installation process varies little from other software installations. The real-mode driver files for your adapter and peripheral(s) are copied to a sub-directory on the hard drive, then the CONFIG.SYS and AUTOEXEC.BAT files are updated to load the appropriate drivers on system startup. If your particular system commits too much conventional memory to drivers, you can manually optimize your startup files later to load as many drivers as possible into upper memory.

If you intend to use your SCSI system under Windows 95, you’ll need to install protected-mode drivers for the host adapter and devices. Contemporary SCSI host adapters (and many SCSI devices) are compliant with Plug-and-Play operation under Windows 95 - this means Windows 95 should typically be able to identify the SCSI adapter (or newly installed devices) and install the appropriate protected-mode drivers for it. Ordinarily, this process should be automatic, but the following tips may help you handle Windows 95 installations:

• Verify the SCSI adapter and devices in DOS first with its real-mode drivers.
• If Windows 95 does NOT automatically identify the SCSI hardware, you should run the Add New Hardware wizard to register the device(s) and install the protected-mode drivers (remember NOT to let Windows 95 detect devices itself).
• You can use the Add New Hardware wizard to update existing SCSI drivers if new versions become available.
• If your SCSI hardware is not listed in the Add New Hardware wizard, you’ll need to contact the hardware manufacturer(s) and download the correct .INF file and protected-mode drivers. If there are no protected-mode drivers for your SCSI hardware (that would be rare today), you’ll need to use the real-mode (DOS) drivers - this may result in all system drives running in "DOS Compatibility mode", and impair system performance.

As far as the bus is concerned, there is very little that can go "wrong" - wires and connectors do not fail spontaneously. However, it never hurts to examine the wiring, connectors, and terminator network(s) to ensure that the physical connections are intact (especially after installing or configuring new devices). The most likely areas of trouble are in the installation, setup, and operation of the devices residing on the bus.

Assuming that your SCSI devices have been installed correctly, problem scenarios can occur during normal operation. The first indication of a problem usually comes in the form of an error message from your operating system or application program. For example, your SCSI hard drive may not be responding, or the host PC may not be able to identify the SCSI host controller board, and so on.

The advantage to SCSI architecture is that it is reasonably easy to determine problem locations using intuitive deduction. Consider a typical SCSI system with one initiator (a host controller) and one target (i.e. a hard drive). If the hard drive fails to function, the trouble is either in the host controller or the drive itself. When you see drive access being attempted, but an error is generated, the trouble is probably in the drive. If no drive access is attempted before an error is generated, the error is likely in the host controller. As another example, consider a setup with one initiator and two or more targets (i.e. a hard drive and CD-ROM). If both the hard drive and CD-ROM become inoperative, the problem is likely in the host controller card since the host adapter controls both targets. If only one of the devices becomes inoperative (and the other device works just fine) the trouble is likely in the particular device itself.

Of course, these are only common isolation methods, and their effectiveness will depend on the sophistication of the particular system you are working with. There is always some amount of uncertainty in the intuitive approach because it is not quantitative. You can suspect where the trouble is coming from, but you can not prove it. Given the great expense of many SCSI peripherals, it is often unwise to purchase replacement parts based solely on intuitive techniques. To prove the problem's source, you can track communication along the SCSI bus using a specialized SCSI tester. If you perform extensive SCSI testing on a professional level, you may wish to invest in an SCSI bus tester such as Ancot’s DSC-216 portable SCSI bus analyzer. An analyzer can let you track the communication process along the SCSI bus, as well as provide bus speed calculations and command profiling. Once you have located the problem device, you can deal with that device specifically through replacement or repair.

However, specialized test equipment carries a significant price tag - a worthy investment if you have the service volume to justify it, but hardly a reasonable outlay for the casual PC hobbyist. Fortunately, a growing number of contemporary diagnostic software packages are being upgraded with SCSI test capabilities. For example, the PC Technician software by Windsor Technologies can test a limited number of SCSI adapters (such as Western Digital, Adaptec, and NCR) and associated peripherals. SCSIDiag by AMI is a diagnostic specifically designed for SCSI system testing.

The reason for this lack of broad diagnostic software support is simple - SCSI is not supported by the PC motherboard BIOS (where IDE and EIDE are supported). As a result, the diagnostic must be written to handle specific SCSI controllers. The issue to keep in mind when selecting a diagnostic for SCSI testing is that the software MUST be compatible with the SCSI adapter in your system - just because a diagnostic says "SCSI-compatible" does not necessarily make it so for the PC setup you are faced with. As an alternative to commercial diagnostics, you may be able to find small controller diagnostics right on the software disks that accompany the SCSI adapter. You would run the test routine after installation to see that the controller is working, but you can also use it in a pinch for "as-needed" troubleshooting for that particular controller. Check with the manufacturer's BBS or CompuServe forum to find up-to-date test routines for various controllers.

No matter how many precautions you take, you cannot always prevent problems from striking during a SCSI installations or replacements. Fortunately, if you are installing devices one-by-one as suggested, you will have far fewer problem areas to check. Your first diagnostic for a SCSI installation should be the host adapter’s SCSI BIOS initialization message. If you see no initialization message when the system powers up, the problem is likely with the adapter itself. Either it is not installed properly or it is defective. Make sure that the adapter is set to the desired ID (usually 7). Try a new or alternate SCSI adapter. If the adapter provides its initialization message as expected, the problem is probably related to driver installation. Check the installation and any command line switches for each device driver. When installing a SCSI hard drive instead of IDE/EIDE hard drives, you must ensure that any previous hard drive references are "mapped out" of the CMOS setup by entering "none" or "not installed". If pre-existing drive references are not removed, the system will try to boot from IDE/EIDE drives which aren’t there.

Be aware that faulty SCSI ID settings can result in system problems such as "ghost" disks - disks that the system says are there but that can not be read from or written to. Some peripherals may also not work properly with the ID that has been assigned. If you have problems interacting with an installed device, try the device with a different ID, and make sure that there are no two devices using the same ID. Don't be surprised to find that certain types of cables don’t work properly with SCSI installations. Make sure that everything is terminated correctly. Also be sure that any external SCSI devices are powered up (if possible) before the PC is initialized. If problems persist, try different cables. An quick-reference checklist is shown below:

• Check the power to all SCSI devices (make sure that the power supply has enough capacity to handle all of your attached SCSI devices).
• Check the cable to all SCSI devices. It should be a good quality cable which is attached securely to each device.
• Check the orientation of each connector on the SCSI cable. Pin 1 must always be in the proper orientation.
• Check the SCSI ID of each device. Duplicate IDs are not allowed.
• Check that both ends of the SCSI cable are properly terminated, and that the terminators are active.
• Check the SCSI controller configuration (IRQQ, I/O, BIOS addresses, and so on). Verify that the SCSI controller is not conflicting with other devices in the system.
• Check SCSI host adapter BIOS. If you’re not booting from SCSI hard drives, you can often leave the SCSI BIOS disabled. This will also simplify the device configuration.
• Check the CMOS Setup for drive configurations. When SCSI drives are in the system and IDE/EIDE drives are not, be sure that the drive entries under CMOS are set for "none" or "not installed".
• Check the PCI bus configuration in the CMOS Setup. See that the PCI slot containing the SCSI host adapter is active, and is using a unique IRQ (usually named IRQ A).
• Check for the real-mode drivers under DOS. If you’re working under DOS, see that any needed driver(s) for the host adapter and non-HDD device(s) are installed in the CONFIG.SYS and AUTOEXEC.BAT files.
• Check for the protected-mode drivers under Windows 95. If you’re working under Windows 95, see that any needed protected-mode drivers for the host adapter and SCSI devices are installed.
• Try remarking-out real-mode drivers if problems occur only under Windows 95. Real-mode SCSI drivers can sometimes interfere with protected-mode SCSI drivers. If the SCSI system works fine in DOS, but not in Windows 95, try temporarily disabling the DOS drivers in your startup files.

Even the best-planned SCSI setups go wrong from time to time, and SCSI systems already in the field will not run forever. Sooner or later, you will have to deal with a SCSI problem. This part of the article is intended to show you a variety of symptoms and solutions for many of the problems that you will likely encounter.

Symptom 45-1. After initial SCSI installation, the system will not boot from the floppy drive. You may or may not see an error code corresponding to this problem. Suspect the SCSI host adapter first. There may be an internal fault with the adapter that is interfering with system operation. Check that all of the adapter's settings are correct and that all jumpers are intact. If the adapter is equipped with any diagnostic LEDs, check for any problem indications. When adapter problems are indicated, replace the adapter board. If a SCSI hard drive has been installed and the drive light is always on, the SCSI signal cable has probably been reversed between the drive and adapter. Make sure to install the drive cable properly.

Check for the SCSI adapter BIOS message generated when the system starts. If the message does not appear, check for the presence of a ROM address conflict between the SCSI adapter and ROMs on other expansion boards. Try a new address setting for the SCSI adapter. If there is a BIOS wait state jumper on the adapter, try changing its setting. If you see an error message indicating that the SCSI host adapter was not found at a particular address, check the I/O setting for the adapter.

Some more recent SCSI host adapters incorporate a floppy controller. This can cause a conflict with an existing floppy controller. If you choose to continue using the existing floppy controller, be sure to disable the host adapter’s floppy controller. If you’d prefer to use the host adapter’s floppy controller, remember to disable the pre-existing floppy controller port.

Symptom 45-2. The system will not boot from the SCSI hard drive. Start by checking the system's CMOS setup. When SCSI drives are installed in a PC, the corresponding hard drive reference in the CMOS setup must be changed to "none" or "not installed" (this assumes that you will NOT be using IDE/EIDE hard drives in the system). If previous hard drive references have not been "mapped out", do so now, save the CMOS Setup, and reboot the PC. If the problem persists, check that the SCSI boot drive is set to ID 0. You will need to refer to the user manual for your particular drive to find how the ID is set.

Next, check the SCSI parity to be sure that it is selected consistently among all SCSI devices. Remember that ALL SCSI devices must have SCSI parity enabled or disabled - if even one device in the SCSI chain does not support parity, it must be disabled on all devices.. Check the SCSI cabling to be sure that all cables are installed and terminated properly. Finally, be sure that the hard drive has been partitioned and formatted properly. If not, boot from a floppy disk and prepare the hard drive as required using FDISK and FORMAT.

Symptom 45-3. The SCSI drive fails to respond with an alternate HDD as the boot drive. Technically, you should be able to use a SCSI drive as a non-boot drive (i.e. drive D:) while using an IDE/EIDE drive as the boot device. If the SCSI drive fails to respond in this kind of arrangement, check the CMOS setting to be sure that drive 1 (the SCSI drive) is "mapped out" (or set to none or not installed). Save the CMOS Setup and reboot the PC. If the problem persists, check that the SCSI drive is set to SCSI ID 1 (the non-boot ID). Next, make sure that the SCSI parity is enabled or disabled consistently throughout the SCSI installation. If the SCSI parity is enabled for some devices and disabled for others, the SCSI system may function erratically. Finally, check that the SCSI cabling is installed and terminated properly. Faulty cables or termination can easily interrupt a SCSI system. If the problem persists, try another hard drive.

NOTE: Later SCSI host adapters use BIOS which allows SCSI drives to boot even WITH IDE/EIDE drives in the system. In such a configuration, the "Boot Order" entry in CMOS Setup will determine whether A:, C:, or SCSI will be the boot device.

Symptom 45-4. The SCSI drive fails to respond with another SCSI drive as the boot drive. This typically occurs in a dual-drive system using two SCSI drives. Check the CMOS Setup and make sure that both drive entries in the setup are set to "none" or "not installed". Save the CMOS Setup. The boot drive should be set to SCSI ID 0 while the supplemental drive should be set to SCSI ID 1 (you will probably have to refer to the manual for the drives to determine how to select a SCSI ID). The hard drives should have a DOS partition and format. If not, create the partitions (FDISK) and format the drives (FORMAT) as required. Check to be sure that SCSI parity is enabled or disabled consistently throughout the SCSI system. If some devices use parity and other devices do not, the SCSI system may not function properly. Make sure that all SCSI cables are installed and terminated properly. If the problem persists, try systematically exchanging each hard drive.

Symptom 45-5. The system works erratically. The PC hangs or the SCSI adapter cannot find the drive(s). Such intermittent operation can be the result of several different SCSI factors. Before taking any action, be sure that the application software you were running when the fault occurred did not cause the problem. Unstable or buggy software can seriously interfere with system operation. Try different applications and see if the system still hangs up (you might also try any DOS diagnostic utilities that accompanied the host adapter). Check each SCSI device and make sure that parity is enabled or disabled consistently through out the SCSI system. If parity is enabled in some devices and disabled in others, erratic operation can result. Make sure that no two SCSI devices are using the same ID. Cabling problems are another common source of erratic behavior. Make sure that all SCSI cables are attached correctly and completely. Also check that the cabling is properly terminated.

Next, suspect that there may be a resource conflict between the SCSI host adapter and another board in the system. Check each expansion board in the system to be sure that nothing is using the same IRQ, DMA, or I/O address as the host adapter (or check the Device Manager under Windows 95). If you find a conflict, you should alter the most recently installed adapter board. If problems persist, try a new drive adapter board.

Symptom 45-6. You see an 096xxxx error code. This is a diagnostic error code that indicates a problem in a 32-bit SCSI host adapter board. Check the board to be sure that it is installed correctly and completely. The board should not be shorted against any other board or cable. Try disabling one SCSI device at a time. If normal operation returns, the last device to be removed is responsible for the problem (you may need to disable drivers and reconfigure termination when isolating problems in this fashion). If the problem persists, remove and re-install all SCSI devices from scratch, or try a new SCSI adapter board.

Symptom 45-7. You see a 112xxxx error code. This diagnostic error code indicates a problem in a 16-bit SCSI adapter board. Check the board to be sure that it is installed correctly and completely. The board should not be shorted against any other board or cable. Try disabling one SCSI device at a time. If normal operation returns, the last device to be removed is responsible for the problem (you may need to disable drivers and reconfigure termination when isolating problems in this fashion). Try a new SCSI host adapter board.

Symptom 45-8. You see a 113xxxx error code. This is a diagnostic code that indicates a problem in a system (motherboard) SCSI adapter configuration. If there is a SCSI BIOS ROM installed on the motherboard, be sure that it is up-to-date and installed correctly and completely. If problems persist, try replacing the motherboard's SCSI controller IC, or replace the system board. It may be possible to circumvent a damaged motherboard SCSI controller by disabling the motherboard’s controller, then installing a SCSI host adapter card.

Symptom 45-9. You see a 210xxxx error code. There is a fault in a SCSI hard disk. Check that the power and signal cables to the disk are connected properly. Make sure the SCSI cable is correctly terminated. Try repartitioning and reformatting the SCSI hard disk. Finally, try a new SCSI hard disk.

Symptom 45-10. A SCSI device refuses to function with the SCSI adapter even though both the adapter and device check properly. This is often a classic case of basic incompatibility between the device and host adapter. Even though SCSI-2 helps to streamline compatibility between devices and controllers, there are still situations when the two just don’t work together. Check the literature included with the finicky device and see if there are any notices of compatibility problems with the controller (perhaps the particular controller brand) you are using. If there are warnings, there may also be alternative jumper or DIP switch settings to compensate for the problem and allow you to use the device after all. A call to technical support at the device's manufacturer may help shed light on any recently discovered bugs or fixes (i.e. an updated SCSI BIOS, SCSI device driver, or host adapter driver). If problems remain, try using a similar device from a different manufacturer (i.e. try a Connor tape drive instead of a Mountain tape drive).

Symptom 45-11. You see a "No SCSI Controller Present" error message. Immediately suspect that the controller is defective or installed improperly. Check the host adapter installation (including IRQ, DMA, and I/O settings), and see that the proper suite of device drivers have been installed correctly. If the system still refuses to recognize the controller, try installing it in a different PC. If the controller also fails in a different PC, the controller is probably bad and should be replaced. However, if the controller works in a different PC, your original PC may not support all the functions under the interrupt 15h call required to configure SCSI adapters (such as an AMI SCSI host adapter). Consider upgrading the PC BIOS ROM to a new version - especially if the PC BIOS is older. There may also be an upgraded SCSI BIOS or host adapter driver to compensate for this problem.

Symptom 45-12. The PCI SCSI host adapter is not recognized, and the SCSI BIOS banner is not displayed. This often occurs when installing new PCI SCSI host adapters. The host computer must be PCI REV. 2.0 compliant and the motherboard BIOS must support PCI-to-PCI Bridges (PPB) and bus mastering - this is typically a problem (or limitation) with some older PCI motherboard chipsets, and you’ll probably find that the PCI SCSI adapter board works just fine on newer systems. If the system DOESN’T support PPB, it may not be possible to use the PCI SCSI adapter. You can try an ISA SCSI adapter instead, or upgrade the motherboard to one with a more recent chipset.

If the system hardware DOES offer PPB support and the problem persists, the motherboard BIOS may still not support PPB features as required by the PCI 2.0 standard. In this case, try a motherboard BIOS upgrade if one is available. If the problem continues, either the board is not in a bus mastering slot, or the PCI slot is not enabled for bus mastering. Configure the PCI slot for bus mastering through CMOS Setup, or through a jumper on the motherboard (check your system’s documentation to see exactly how).

Symptom 45-13. During boot-up, you see a; "Host Adapter Configuration Error" message. In virtually all cases, there is a problem with the PCI slot configuration for the SCSI host adapter. Try enabling a IRQ for the SCSI adapter’s PCI slot (usually accomplished through the CMOS Setup). Make sure that any IRQ being assigned to the SCSI adapter PCI slot is not conflicting with other devices in the system.

Symptom 45-14. You see an error message such as "No SCSI Functions in Use". Even when a SCSI adapter and devices are installed and configured properly, there are several possible causes for this kind of an error. First, make sure that there are no hard disk drivers installed when there are no physical SCSI hard disks in the system. Also make sure that there are no hard disk drivers installed (i.e. in CONFIG.SYS) when the SCSI host adapter BIOS is enabled. HDD drivers aren’t needed then, but you could leave the drivers in place and disable the SCSI BIOS. Finally, this error can occur if the HDD was formatted on another SCSI controller which does not support ASPI, or uses a specialized format. For example, Western Digital controllers only work with Western Digital HDDs. In this case, you should try a more generic controller.

Symptom 45-15. You see an error message such as; "No Boot Record Found". This is generally a simple problem which can be traced to several possible issues. First, chances are that the drive has never been partitioned (FDISK) or formatted as a bootable drive (FORMAT). Repartition and reformat the hard drive. If you partitioned and formatted the drive with a third-party utility (i.e. TFORMAT), be sure to answer "Y" if asked to make the disk bootable. A third possibility can occur if the disk was formatted on another manufacturer’s controller. If this is the case, there may be little alternative but to repartition and reformat the drive again on your current controller.

Symptom 45-16. You see an error such as; "Device fails to respond - No devices in use. Driver load aborted". In most cases, the problem is something simple such as the SCSI device not being turned on or cabled correctly. Verify that the SCSI devices are on and connected correctly. In other cases, the SCSI device is on, but fails the INQUIRY command - this happens when the SCSI device is defective, or not supported by the host adapter. The device may need default jumper settings changed (i.e. the drive should Spin up and Come Ready on its own). You may find that the SCSI device is sharing the same SCSI ID with another device. Check all SCSI devices to verify that each device has separate SCSI ID. You may have the wrong device driver loaded for your particular device type. Check config.sys to make sure the correct driver is loaded for the drive type (i.e. TSCSI.SYS for a hard disk, not a CD-ROM).

Symptom 45-17. You see an error such as; "Unknown SCSI Device" or "Waiting for SCSI Device". The SCSI hard disk has failed to boot as the primary drive - check that the primary hard disk is set at SCSI ID 0. Make sure that the drive is partitioned and formatted as the primary drive. If necessary, boot from a floppy with just the ASPI manager loaded in config.sys and no other drivers, then format drive. It may also be that the SCSI cable termination is not correct (or TERMPWR is not provided by the HARD DISK for the host adapter). Verify the cable terminations and TERMPWR signal.

Symptom 45-18. You see an error such as; "CMD Failure XX". This typically occurs during the FORMAT process - the "XX" is a vendor-specific code (and you’ll need to contact the vendor to determine what the error means). The most common problem is trying to partition a drive that is NOT low-level formatted. If this is the case, run the low-level format utility that accompanied the SCSI drive, then try partitioning again. If you’re suffering a different error, you may need to take other action depending on the nature of the error.

Symptom 45-19. After the SCSI adapter BIOS header appears, you see a message like "Checking for SCSI target 0 LUN 0". The system pauses about 30 seconds, then reports "BIOS not installed, no INT 13h device found". The system then boots normally. In most cases, the BIOS is trying to find a hard drive at SCSI ID 0 or 1, but there is no hard drive available. If you do not have a SCSI hard drive attached to the host adapter, then it is recommended that the SCSI BIOS be disabled.

Symptom 45-20. The system hangs up when the SCSI BIOS header appears. This is usually caused by a terminator problem. Make sure that the SCSI devices at the end of the SCSI chain (either internally or externally) are terminated. Check all device IDs to make sure that they are unique, and also check for system resource conflicts (i.e. BIOS address, I/O address, and interrupts). You may also need to disable the Shadow RAM feature in the CMOS Setup.

Symptom 45-21. The SCSI BIOS header is displayed during system startup, then you get the message; "Host Adapter Diagnostic Error". The card either has a port address conflict with another card, or the card has been changed to port address 140h and the BIOS is enabled. Some SCSI host adapters are able to use the BIOS under port address 140h, so check for I/O conflicts. You may need to reconfigure the SCSI host adapter.

Symptom 45-22. When a VL bus SCSI adapter is installed, the system hangs at startup. Chances are that the VL SCSI adapter is a bus mastering device and requires that the VL slot support full 32-bit bus mastering. Most VL bus systems have either "slave" slots and/or "master" slots. The SCSI adapter must be inserted into a "master" slot. If you are not sure if the system supports bus mastering or if you have a master slot, contact the system manufacturer.

Also, the slot that the SCSI VL card is inserted into must be a 5 Vdc slot that operates at 33MHz or less. The VL bus speed is typically set through a jumper on the motherboard. It should be set in the <=33MHz position. The motherboard may also need to be set for write-through caching. This may be set in the motherboard’s CMOS setup utility, or it may be configured via a jumper on the motherboard (if there is both a CMOS setting and a jumper, be sure they are both set the same way).

Symptom 45-23. When upgrading a VL bus system CPU to a faster model, the system locks up with a SCSI VL card installed, or won’t boot from the SCSI HDD. Most likely there is a DMA or other timing discrepancy between the SCSI adapter and the VL local bus. The SCSI adapter probably works fine on VL bus systems running up to 33MHz. Faster CPUs can increase the VL bus speed beyond 33MHz. Above this 33MHz speed, variations in motherboard, chipset, or CPU design may cause the SCSI adapter to function intermittently or to fail. In some cases this problem can be resolved:

• The motherboard may have jumpers that govern the VL bus speed - be sure that the VL bus speed jumper is set in the <=33 MHz position. This may also be set in the motherboard’s CMOS setup.
• In the CMOS setup, you can disable the CPU external cache, or change the caching method to write-through instead of write-back.
• The internal cache on some CPUs may cause the VL SCSI adapter to hang as well. Try disabling the CPU’s internal cache.
• Reducing the CPU speed may be necessary to allow the SCSI adapter to function reliably.
• Try disabling the system’s "turbo setting" during the boot-up sequence, then re-enable the turbo setting after the system has booted.

Symptom 45-24. The VL SCSI adapter won’t work with an "SLC" type CPU. VL SCSI adapters often refuse to run with "SLC" type CPUs because the SLC uses 16-bit architecture rather than 32-bit at the VL bus. Some VL SCSI adapters will run in this configuration, but it is rare. Use an ISA SCSI adapter instead of an VL adapter in this circumstance.

Symptom 45-25. When running the Qualitas 386MAX memory manager software on ISA or VL systems with an SCSI host adapter, the system crashes when booting. 386MAX is known to cause problems with SCSI systems, and you’ll need to adjust the 386MAX command line. Do not allow 386MAX to load during boot up, then include the key NOIOWRAP on the 386MAX command line. This will allow you to boot with 386MAX loaded.

Symptom 45-26. When installing an EISA SCSI adapter and running the EISA configuration utility, you see an; "EISA configuration slot mismatch" or "board not found in slot x" error. This error is caused by the fact that your board is not completely seated in the EISA slot. You can verify this by booting to a floppy diskette, and running the DOS Debug command. After typing Debug, you will receive the debug prompt (a dash). Then type "i (space) Xc80" where "X" is the EISA slot where your board is physically installed. If a "04" is returned, the board is correctly seated and the problem lies elsewhere. If "FF" is returned, the board needs to be pushed down further. Power down your system before re-seating your board.

Symptom 45-27. You can’t configure an EISA SCSI adapter in Enhanced Mode. You get the error; "Unable to initialize Host Adapter" or the system hangs after the SCSI BIOS scans the SCSI devices. These errors are usually limited to motherboards which do not support LEVEL INT triggering. These chipsets (such as the Hint and SIS) require a few modifications be made to the host adapter’s EISA configuration (.CFG) file. Make the following changes to the !ADP000X.CFG file:

CHOICE = "Enhanced Mode"

FREE

INT=IOPORT(1) LOC (7 6 2 1 0) 10000B

LINK

IRQ=11|12|10|15|14|9

SHARE = "AHA-1740" (Change to: SHARE = NO)

TRIGGER = LEVEL (Change to: TRIGGER = EDGE)

INIT=IOPORT(3) LOC(4 3 2 1 0) 10010B | 10011B | 10001B | 1010B | 10101B | 10000B

(Change first zero in each binary number to a one; Example: 10010B = 11010B)

Another option is to download the latest .CFG file for your SCSI adapter card (i.e. ASWC174.EXE). Reconfigure the card with new .CFG file and select edge triggered IRQ.

This concludes the material for Chapter 45. Be sure to review the glossary and chapter questions on the accompanying CD. If you have access to the Internet, take a look at some of the SCSI system resources listed below:

Adaptec: http://www.adaptec.com

AMI: http://www.megatrends.com

Ancot: http://www.ancot.com/

Fibre Channel Association: http://www.Amdahl.com/ext/CARP/FCA/FCA.html

Quantum: http://www.quantum.com/src/

SCSI FAQ at http://www.cis.ohio-state.edu/hypertext/faq/usenet/scsi-faq/top.html

SCSI guide: http://www.delec.com/Tech_Links/SCSIGuide/

SCSI Trade Association: http://www.scsita.org/

SCSI-2 spec: http://abekas.com:8080/SCSI2/

Symbios articles: http://www.symbios.com/articles/articles.htm

Symbios specs: http://www.symbios.com/x3t10

Western Digital: http://www.wdc.com