Drives are storage media's

A drive is the name name for several types of storage media. There are also storage media,
which are not drives (RAM, Tape Streamers), but on these pages, we will discuss the drives.
Common to drive media is:
A file system can be assigned to them. l
They are recognized by the operating system and they are assigned a drive letter. l
During start up, drives are typically recognized by the PC system software (ROM-BIOS +
operating system). Thus, the PC knows which drives are installed. At the end of this
configuration, the appropriate drive letter is identified with each drive. If a drive is not "seen"
during start up, if will not be accessible to the operating system. However, some external
drives contain special soft-ware, allowing them to be connected during operation. Here some
examples of drives:
Storage media Drive letter
Floppy disks A: B:
Hard disk C: D: E:
CD ROM F:
MO drive G:
Network drive M:
RAM disk O:
On this and the following pages, I will describe the various drive types, their history and
technology. The last two drive types in the above table will not be covered.
Storage principles
Storage: Magnetic or optic. Data on any drive are digitized. That means that they are
expressed as myriad's of 0's and 1's. However, the storage of these bits is done in any of
three
principles:
The physical drive principle Disk types
 Magnetic
 Floppy disks
 Hard disk
 Syquest disks
 Optic
 CD ROM
 DVD
 Magneto optic
 Zip drive
 LS-120 disks
 and others Interface
Individual drives are connected to other PC components through an interface. The hard disk
interface is either IDE or SCSI, which in modern PC's is connected to the PCI bus. Certain
drives can also be connected through a parallel port or the floppy controller:
 Interface Drive
 IDE and EIDE
 Hard disks (currently up to 8 GB)
 CD ROM
 SCSI Hard disks (all sizes) and CD ROM
 ISA (internet)
 Floppy drives
 CD ROM an MO drives connected through parallel port
Let us start evaluating the drives from the easy side:
Floppy drives
We all know diskettes. Small flat disks, irritatingly slow and with too limited storage capacity.
Yet, we cannot live without them. Very few PC's are without a floppy drive. Diskettes were
developed as a low cost alternative to hard disks. In the 60's and 70's, when hard disk prices
were exorbitant, It was unthinkable to use them in anything but mainframe and mini
computers.
The first diskettes were introduced in 1971. They were 8" diameter plastic disks with a
magnetic coating, enclosed in a cardboard case. The had a capacity of one megabyte. The
diskettes are
placed in a drive, which has read and write heads. Conversely to hard disks, the heads
actually
touch the disk, like in a cassette or video player. This wears the media.
Later, in 1976, 5.25" diskettes were introduced. They had far less capacity (only 160 KB to
begin with). However, they were inexpensive and easy to work with. For many years, they
were
the standard in PC's. Like the 8" diskettes, the 5.25" were soft and flexible. Therefore, they
were named floppy disks.
In 1987 IBM's revolutionary PS/2 PC's were introduced and with them the 3½" hard diskettes
we know today. These diskettes have a thinner magnetic coating, allowing more tracks on a
smaller surface. The track density is measured in TPI (tracks per inch). The TPI has been
increased from 48 to 96 and now 135 in the 3.5" diskettes.
Here you see the standard PC diskette configurations:
Diskette size Name Tracks per side Number of sectors
per tracks Capacity
5.25" Single side SD8 40 8 40 X 8 X 512 bytes = 160 KB
5.25" Double side DD9 40 9 2 X 40 X 9 X 512 bytes = 360 KB
5.25" Double side High Density DQ15 80 15 2 X 80 X 15 X 512 bytes = 1,2 MB
3.5" DD DQ9 80 9 2 X 80 X 9 X 512 bytes = 720 KB
3.5" HD DQ18 80 18 2 X 80 X 18 X 512 bytes = 1,44 MB
3.5" XD ( IBM only) DG36 80 36 2 X 80 X 36 X 512 bytes = 2,88 MB
Diskette drives turn at 300 RPM. That results in an average search time (½ revolution) of 100
ms.
The floppy controller
All diskette drives are governed by a controller. The original PC controller was named NEC
An illustrated Guide to disk drives - storage medias.
PD765. Today, it is included in the chip set, but functions like a 765. It is a programmable
chip.
It can be programmed to handle all the various floppy drive types: 5.25" or 3.5" drives, DD or
HD etc.
The controller has to be programmed at each start up. It must be told which drives to control.
This programming is performed by the start up programs in ROM (read module 2a). So you
don't have to identify available drive types at each start up, these drive parameters are saved
in CMOS RAM. The floppy controller reads data from the diskette media in serial mode (one
bit at a time. Like from hard disks). Data are delivered in parallel mode (16 bits at a time) to
RAM via a DMA channel. Thus, the the drives should be able to operate without CPU
supervision. However, in reality this does not always work. Data transfer from a diskette drive
can delay and sometimes freeze the whole PC, so no other operations can be performed
simultaneously.

Hard disk speeds

There are countless test programs and measuring methods to evaluate the various hard
disks. Don't place too much stock in the sales person's presentation of seek times. Many
hard disks are advertised with a number like 8 ms. That refers to a seek time, which is
measured in milliseconds. There are many different seek times. That makes comparison
difficult. You can measure in terms of: Average track to track speed. How long does it take
the actuator to move read/write heads from one track to another? There are typically 3000
tracks on a platter side. There, a track change could be to just one over. That might take 2
ms. Or, it could be up to 2999 tracks over. That might take 20 ms. On current hard disks, the
average seek time will be between 8 and 14 ms.
Change time between read and write - That takes time also.
Wait time for the right sector. When the arm moves to a track, it must wait for the right sector
to appear under the head. That takes time also. On the average, the platter must rotate ½
revolution, to reach the right sector. This time is directly proportional to the disk rotation
speed. On modern hard disks it usually is between four and eight ms.

When I test a hard disk, I emphasize practical applications. You can take a stopwatch and
measure, for example how long it takes to start Windows 95 or Word 97 (possibly including a
large file). That type of measurement can really tell you something about the hard disk's
performance. However they must be under comparable circumstances. System board, CPU,
and the driver program also influence the results.

IDE
Integrated Device Electronics. Under the IDE standard, the controller chip WD 1003 is
mounted directly on the hard disk, not on the IDE adapter. This means that the conversion to
parallel data is already done on the disk. Because of the short serial cable, this increases the
transfer speed significantly relative to MFM and RLL. IDE is a simple adapter. The adapter
itself contains only amplifying circuits to/from the I/O bus. Therefore it is inexpensive. The
IDE controller does not care whether the hard disk works internally with MFM or RLL coding.
ESDI
ESDI is an improvement over the ST506 standard. An ESDI disk operates on a common 16-
bit AT bus (ISA-bus), but it is better put together than an ordinary IDE. This results in an
almost doubling of the transfer speed between hard disk and controller/bus. ESDI is also
different in many other ways. Among the features are a sector on the hard disk which
identifies its number of tracks, cylinders, etc. This information is usually stored in CMOS.
EIDE
EIDE is the current standard for low cost, high performance hard disks. EIDE stands for
Enhanced IDE. That is precisely what it is. Some manufacturers call it ATA.
All Pentium system boards since 1995 have a built in EIDE controller. That allows the hard
disk to be connected directly to the system board. The EIDE standard is substantially
improved, relative to the old IDE. Here are some of the improvements:
The hard disk size can exceed the 528 MB, which was the IDE limit. l
The hard disk interface is moved from the ISA bus to the high speed bus PCI. l
Four units can be connected on the mainboard, which has two EIDE channels. Each
channel can be connected to a primary and a secondary unit.

Transfer speeds
With connection directly to the PCI bus, EIDE has transfer speeds and disk capacities which
far exceed the older controller principles. EIDE exists in different editions, such as PIO 3,
PIO 4 and Ultra DMA. The latter is the one to choose. The different PIO modes are
significant for transfer speed. PIO 3 can transfer up to 13 MB/sec, while PIO 4 promises 16.6
MB/sec. UDMA promises up to smashing 33 MB/sec. These numbers are theoretical and
they do not hold true in actual use. The fastest actual transfer speed you can experience
from an EIDE disk will be 5-10 MB/sec. That is still good.

Four units
The EIDE interface is not only intended for the hard disks. There are four channels, which
can be connected to four independent units:
Hard disks (must be on the primary channels, which on some system boards have the
greatest transfer capacity)
 CD ROM drives l
 DVD drives l
 LS 120 drives and other MO drives l
 Tape streamers l
EIDE is thus designed as an inexpensive all around interface, which can be connected to all
kinds of storage media. It has a clever auto detect function, which often makes it possible to
connect EIDE units such as hard disks directly to the system board and function immediately.
The BIOS in the PC will find the necessary instructions about the drive via the auto detect
function, and you need not make any adjustments in the CMOS Setup program, as was
necessary with earlier IDE units.

Drives and operating system

The drive must be assigned a drive letter. That is a task for the operating system, which must
be able to recognize the CD-ROM drive. That is usually no problem in Windows 95.
However, the alphabet can be quite messy, if there are many different drives attached. Each
drive must have its own letter. They are assigned on a first come first-serve-basis. The CD-
ROM drive usually gets the first vacant letter after other existing drives, typically D, E, or F.
But the letter can be changed. Once the CD-ROM spins and the operating system (DOS or
Windows) has "found" the CD-ROM drive, data can be read for processing. Now the CD-
ROM works like any other drives. Only, it is Read Only Memory!

About Optic Data Storage

The CD-ROM can be compared to a floppy drive, because the disks are removable. It can
also be compared with a hard drive, because of similar data storage capacity. Actually, a CD-
ROM disk can hold up to 680 MB of data. This equals the capacity of 470 floppy disks.
However, the CD ROM is neither a floppy nor a hard disk! While floppy and hard disks are
magnetic media, the CD-ROM is an optic media. The magnetic media work in principle like
an audio cassette tape player. They have a read/write head, which reads or writes magnetic
impressions on the disk. The magnetic media contains myriads of microscopic magnets,
which can be polarized to represent a zero or numeral one (one bit).
In the optic readable CD-ROM, the data storage consists of millions of indentations burnt into
the lacquer-coated, light-reflecting silver surface. The burnt dents reflect less light than the
shiny surface. A weak laser beam is sent to the disk through a two-way mirror and the sensor
registers the difference in light reflection from the burnt and shiny areas as zero's and one's.

Tracks
Our data consist of bits, each of which is a burnt dent or a shiny spot on the CD-ROM disk.
Music CD's are designed much in the same manner. The bits are not splashed across the
disk, but arranged in a pattern along the track. Without that organization, you could not read
the data.
The platters in hard disks and floppies are organized in concentric tracks. There can be
hundreds of those from center to periphery:
 About I/O units, expansion cards, adapters, etc.

Intro to I/O
This page should preferably be read together with module 2c, 2d, 5b and 5c. The first two
describe the I/O buses and the chip sets. Here we will look at the other end of the I/O buses,
the "exit." There are four I/O buses in the modern PC architecture and each of them has
several functions. They may lead internal and external ports or they lead to other controlling
buses. The four buses are:

 ISA, which is old and hopefully soon disappears. l

 PCI, which is the newer high speed multifunction I/O bus. l
 AGP, which only is used for graphics adapter. l
 USB, which is the new low-speed I/O bus to replace ISA. l

The ISA and the PCI bus both end in a twofold exit: Internal I/O ports (LPT, KBD, COM1,
COM2, EIDE ao.) l Expansion slots in the system board, in which we can insert adapters.

The internal I/O ports

As mentioned, the USB is going to become the main bus for low-speed devices. But so far
we still use the internal "face" of the ISA bus for a range of purposes. At any PC motherboard
you find these:
 The floppy controller l
 The serial ports l
 The parallel port(s) l
 The keyboard controller l
They all occupy IRQ's which is a central part of ISA architecture and a pain in the a... Let us
take a moment to look at these ports and controllers.
The serial ports
Serial transmission means to send data from one unit to another one bit at the time. The PC
architecture traditionally holds to RS232 serial ports. The RS-232 standard describes an
asynchronous interface. This means thatThe serial transfer is limited to a speed of 115,200
bits per second. The cable can be up to 200 meter long.
The serial ports can be used to connect:
 The mouse l
 Modems l
 ISDN adapters l
 Printers with serial interface l
 Digital cameras l
 .... l
These units are connected to the serial ports using either DB9 or DB15 plugs. In a few years
time all these devices will connect to the USB bus instead. data only are transmitted when
the receiving unit is ready to receive them:
The parallel port
Parallel transmission means that data are conducted through 8 separate wires - transmitting
a full byte in one operation. This way the parallel transmission is speedier than the serial, but
the cabling is limited to 5-10 meters. The cable is fat and unhandy, holding up to 25 wires
and the transmission is controlled according to the Centronics standard. Most printer
manufactories use a 36-pins Amphenol plug, where the PC's parallel port holds a 25-pinned
connector. Hence the special printer cable. To the left you se the 25 pin connector, to the
right the 36-pin:
Then parallel port represents the most uncomplicated interface of the PC. It is always used to
connect the printer, but with the bi-directional parallel port (EPP/ECP), other devices have
found their way to this interface. Today you find:
 ZIP-drives
 Portable CD-ROM-drives l
 SCSI adapters l
 Digital cameras l
 Scanners l
all using the parallel port to connect to the system bus.

The EPP/ECP ports

Today we operate with Enhanced Parallel Port/Enhanced Capability Ports. This method for
bi-directional (half-duplex) parallel communication offers higher rates of data transfer (up to 1
megabyte per second) than the original parallel signaling method. EPP is used for for non-
printer peripherals, where ECP is for printers and scanners. You find the settings for the
printer port in the setup program on the motherboard.
Both port types are parts of the IEEE 1284 standard, which also includes Centronics.
To get the best results all the involved hardware and the operating system has to be
EPP/ECP compatible.
Windows 95 supports IEEE 1284 in its parallel plug-and-play feature. It also supports ECP if
you have a printer and a parallel port with ECP. The printer cable has to be complete with all
25 wires connected.

Adapters
In a stationary PC, adapters are typically printed circuit boards called expansion boards or
expansion cards.
They form a link between the central PC unit and various peripherals. This is the so-called
open architecture.
Typically, adapters provide functions, which are separated from the system board. l
Adapters provide expansion capability to the PC. l There are PC's without expansion slots.
In that case all functions must be built into the system board. You could easily include chips
for graphics, ethernet, SCSI, and sound on the system board. This is not
common in stationary PC's. Portable, laptop PC's have nearly all electronics on the system
board. This is called closed architecture.
A traditional PC has a system board which contains all standard functions (except the
graphics chip). To this system board you can add various expansion cards, which control one
or more peripheral units:

Other expansion board types:

 Internal modem (in lieu of external modem) l
 ISDN adapters l
 Extra parallel ports l
 Video editing boards l
 Special graphics cards, which supplement the usual (3D and MPEG) l
 TV and radio receivers. l
The integrated hard disk controller
In the Pentium based PC, the hard disk is connected to an EIDE controller, which is
integrated on the system board. Likewise, the serial and parallel ports are connected directly
to the system board. This is new. On the 386 PC's, you had to install a special controller
cards (I/O cards) to handle these functions. They are included in the modern chip sets on the
system board. Other functions are not integrated. That includes:

The video controller

You have to install a video card to make the PC function. It would be illogical to assemble a
PC without a video card. You would not be able to see what you are doing, since the video
card governs data transmission to the monitor.
The advantage of this design is, that the user can choose between numerous video cards in
various qualities. A discount store may offer a complete Pentium based PC (without printer)
and with the cheapest video card for $669.-. If the buyer is quality oriented, he would want to
spend an additional $40 to get a much better video card.

The modular PC design

In this way, various expansion boards provide flexibility in assembling a customized PC. At
the same time, various electronics manufacturers are specializing their production:
ASUS and Tyan are good at making system boards. Others, like S3, Matrox, and ATI
specialize in making graphics chips and expansion boards. Olicom make only net boards.
Adaptec make only SCSI controllers and Creative Labs make SoundBlaster sound boards.
This variety of manufacturers offers the consumer wide choices. Your PC can be customized
and configured according to your needs and wallet size.

About the electronics

The adapter is a printed circuit board. They have an edge connector, so they can be inserted
in expansion slots in the system board. The expansion slots connect to the I/O busses. Since
the Pentium system board has two I/O busses, it has two types of expansion slots:
 ISA slots l
 PCI slots l
Typically, on a regular Pentium system board there are three or four of each type. That gives
a total of 7 expansion slots. One expansion board can be installed in each of these. You
simply press the edge connector of the expansion board into the expansion slot. Now it is
connected to the bus.

IRQ's
When you install an expansion board in a slot, it gets connected to the I/O bus. Now the
board can send and receive data. But who regulates the traffic? Who gives clearance to the
new controller to send data? It would appear that data traffic could soon be chaotic.
To control data traffic on the I/O bus, the concept of IRQ (Interrupt ReQuest) was created.
Interrupts are a fundamental principle in the PC design. There are two types of interrupts:
Software Interrupts are used to call any number of BIOS routines. Hardware Interrupts are
the subject of this page.

Hardware Interrupts
The adapter or unit on the I/O bus uses the interrupt to signal request to send or receive
data. An interrupt signal is like a door bell. The unit signals by applying a voltage to one of
the wires in the bus - an IRQ. When the CPU acknowledges the signal, it knows that the unit
wants send or receive data, or is finished. The advantage of IRQ's is that the CPU can
manage other tasks, while an adapter "massages" its data. When the adapter has finished its
task, it will report to the CPU with a new IRQ. As an example, let us see how keyboard data
are handled. The keyboard send bits, serially, through the cable to the keyboard controller.
The controller organizes them in groups of 8 (one byte). Every time it has a byte, it sends an
IRQ to the I/O bus. The IRQ controller asks the CPU permission to use the bus, to send the
byte to wherever. The IRQ controller reports back to the keyboard controller, giving
clearance to send the next character (byte):

IRQ wires
Physically, the IRQ is a wire on the bus. This wire connects to all expansion slots. Therefore,
regardless of in which slot you install an adapter, the adapter can communicate with an IRQ.
The PC is "born" with 15 IRQ's, but five of them are internal, and can not be used with I/O
cards. We find 10 accessible IRQ's on the I/O busses. Each of those consist of a circuit
board wire, which goes through the entire bus. When you install an expansion card in a
vacant slot, one of the IRQ's is assigned to it.

When a signal arrives on an IRQ channel, that is a message to the CPU. It is told that a unit
wants to get on the bus. Which unit is to be identified through the IRQ number.
Next the unit is admitted to the bus, to send or receive data. When the transaction is
completed, another signal is transmitted to the CPU to indicate that the bus is vacant.
The IRQ's have different priorities, so the CPU knows which IRQ have priority, if two signals
are sent simultaneously.
The IRQ system is guided by a controller chip, like Intel 8259. It can handle 8 IRQ signals
and couple two of them together, via IRQ 2 or 9. All PC's with ISA bus include two 8259
chips.

Plug and Play

Plug and play (PnP) is an industry standard for expansion boards. If the board conforms to
the PnP standard, the installation is very simple. The board configures itself automatically.
These are the minimum requirements:
The PC system board must be PnP compatible. l
The operating system must be capable of utilizing PnP. Currently, only Windows 95 can do
that. l
The adapter must be able to inform the I/O bus which I/O addresses and IRQ's it can
communicate with.
The adapter must be able to adjust to use the I/O address and the IRQ, which the I/O bus
communicates to the adapter.

I/O addresses
Finally, we need to mention how the the CPU finds all these units - adapters, ports. etc. They
all have an address - an I/O port number.
Each unit can be reached through one of more I/O ports. Each port is a byte port. That
means that 8 bits (one byte) can be transmitted simultaneously - parallel mode.
If the unit is on the ISA bus, it handles 16 bits at a time ( words). Then you link two
consecutive ports together, to make a 16 bit channel. If we talk about about a 32 bit PCI unit,
we link four byte ports together to get 32 bits width (called dword).
The PC has a built in listing of all I/O units, each of which has their own "zip code" - a port
address. Since the PC is basically a 16 bit computer, there are 2 in the 16 power possible
addresses (65,536) - from 0000H to FFFFH. They are described in the hexadecimal number
system as 5 digit numbers. Hexadecimal is a 16 digit number system. Digits go from 0 to 9
and continue with 6 letters A - H. Let me show you some examples of I/O addresses:
 About Interfaces: EIDE, Ultra DMA and AGP
What is EIDE?
EIDE is the current standard for inexpensive, high performance hard disks. EIDE stands for
Enhanced IDE and that is precisely what it is. Some manufacturers (like Seagate) call it ATA.
You can think of EIDE as a bus - which is a host controller - which controls it, and you can
connect up to four units.
All Pentium system boards since 1995 have this EIDE controller built into the chip set. That
allows the hard disk and other EIDE units to be connected directly to the system board.
Improvements
The EIDE standard is a great improvement over the old IDE. Here are some examples:
The Hard disk can exceed the 528 MB IDE limit. Currently the largest EIDE disks are 8.5 l
GB and this number keeps increasing.
The hard disk's interface is moved from the ISA bus to the high speed PCI bus. l
Four units can be connected to the system board, which has two EIDE channels. Each
channel can be connected to a master and a slave unit.

The most important feature is the interface directly on the PCI bus. This has given EIDE
transfer speeds and disk capacities, which far exceed older controller principles.
Concurrently, there is a continual development of the protocols, which are needed for the
connection between the units and the EIDE bus.
Four units
The EIDE interface is not designed for hard disks only. There are four channels, which can
be connected to four independent units: Hard disks (which must be on the primary channel.
On some system boards, this has the greatest transfer capacity)l
 CD ROM drives l
 DVD drives l
 LS-120 drives and other MO drives l
 Tape streamers l
EIDE is thus designed as an inexpensive all-around interface, which can be connected to all
kinds of storage media.
Auto detect
The BIOS on the system board has a neat auto detect feature, which often allows EIDE units
to be connected directly and work immediately. The PC start up program automatically finds
the necessary information about the drive via the auto detect function. Sometimes you have
to assist the hard disk installation by activating the auto detect in the CMOS Setup program,
but often it runs by itself. You definitely do not have to key in information about cylinders,
etc., as you had to with earlier IDE units.
What is AGP?
A new bus has arrived on the PC. It is called AGP (Advanced Graphics Port). It is exclusively
designed for video cards AGP will probably not be in widespread use before 1998. Amongst
other things, the system must be supported by the operating system (Windows 98). Likewise,
it is claimed that the system bus
will be raised from the current 66 MHZ to 100 MHZ, to allow AGP to prove its worth. AGP
includes several techniques, of which two are understandable:
PCI version 2.1 with 66 MHZ bus frequency. That is a doubling of transfer speed l
Possibility to utilize system board RAM for texture cache. This will reduce RAM card demand
in connection with the most demanding programs.
One big AGP advantage is that the PCI bus is relieved of work with graphics data. It can
concentrate on other demanding transport duties, like transfer to and from network adapter
and disk drives.
Here you see the AGP-socket at the bottom. It looks like a PCI-socket, but it has been placed
in a different position on the board. In the top you see two (black) ISA-sockets. Then four
(white) PCI-sockets, and then the brown AGP-socket:

About SCSI, USB and other serial busses

The host adapter l

8 units in a chain l
SCSI is intelligent l
About the SCSI standards l
What do you gain with SCSI? l
USB l
IEEE 1394 FireWire l

SCSI (Small Computer System Interface) is high end technology. It is a technology, which
provide means for data exchange among hardware devices such as drives, tape streamers
and scanners. SCSI is especially used in high end PC's such as network servers or just
powerful workstations.
SCSI might be compared to the EIDE interface, which also uses a host adapter controlling
drives. However SCSI have two major advantages to EIDE: A SCSI host controls 7 or 15
devices (using only one IRQ). The SCSI system holds its own computer power, thus freeing
the CPU from workload. If you are critical about your PC power, the SCSI would be worth
considering.
The host adapter
A SCSI system is built around a central, intelligent controller called the host adapter. A host
adapter can control several SCSI units:
Many units on the same host adapter. l
Many types of drives: Hard disks, CD ROM's, MO drives like Zip drives, CD ROM recorders
etc. Tape-streamers (DAT and others),Scanners.
The host adapter has its own BIOS separate from the PC's. When you boot the PC, you will
see the adapter communicating with connected SCSI devices.

8 units in a chain
The regular SCSI 2 system can handle 8 devices (SCSI Wide handles 15). Each device has
to be assigned an unique going from ID 0 to ID 7. The SCSI devices can be internal (installed
inside the PC cabinet) or external. The host adapter is a device itself. Typically, the host
adapter will occupy ID 7.

Terminators in both ends

The last unit in both ends of the SCSI chain must be terminated. This means that there must
be resistors (jumpers or switches) attached to two of the units.
If you only use two devices, you do not have to worry about it. The host adapter is one end of
the chain and the other device is the other end. With three or more units you have to take
care of termination:

SCSI is intelligent
SCSI is remarkable in having an intelligent protocol, which assures maximum utilization of
the data transfer. The basis of SCSI is a set of commands. Each individual device holds its
own controller, which interprets these commands.
All commands within the SCSI system are handled internally, meaning the CPU does not
have to control the process:

About the SCSI standard

SCSI stands for Small Computer System Interface. It is intended as a universal interface,
defined and designed in 1982 by NCR and Shugart Associates. It exists in numerous
variations. Here you see some of the more significant editions:

Standard Year `Bus speed `Bus width Max. bandwidth

SCSI-1 1986 5 MHZ (Asynchronous) 8 bit 5 MB/sec
Fast SCSI-2 1990 10 MHz (Synchronous) 8 bit 10 MB/sec
Fast-Wide SCSI-2 1992 10 MHZ (Synchronous) 16 bit 20 MB/sec
Fast-Wide UltraSCSI-20 1994 20 MHZ (Synchronous) 16 bit 40 MB/sec
SCSI-3 1996 ? ? 80 MB/sec

SCSI-2 is the 16 bit standard from 1990. It is found in Fast and Fast-Wide-SCSI. Today.
there are many SCSI standards. Among others, you can come across SCSI-20 and SCSI-40,
which refers to the bus speed. The last one is also called SCSI-3, but its standard is not
finalized. The SCSI standard seem to have its own life with plenty of new development.

What do you gain with SCSI?

Expensive but good: SCSI makes the PC a little more expensive, but much better. That's
all. The advantages are, that on the same PC you have free access to use many units and
good hard disks:
It is easy to add many high end accessories, such as DAT-streamers, CD-ROM recorders,
MO drives, scanners, etc.
You can use SCSI hard disks. l
You can use CD-ROM drives on SCSI , where they perform a lot better than on IDE. l

Advantages of SCSI hard disks

SCSI hard disk are generally of higher quality than other disks. EIDE disks come in various
qualities from different manufacturers. However, even the best EIDE disks cannot compete
with the best SCSI disks. Typically, good SCSI disks come with a 5 year warranty. They
come in larger capacities than the EIDE disks and they are faster. At 5400, 7200 or 10.000
RPM they have shorter seek times. They also have a bigger cache. Another advantage is the
large number of accessories, which can be attached. If you buy a 4 GB SCSI disk today, you
will guaranteed need additional disk storage in a few years. Then you just add disk number
two to the SCSI chain, and later number three. The system is more flexible than EIDE, where
you can have a maximum of four units incl. CD-ROM. The SCSI hard disks can also adjust
the sequence in the PC's disk read commands. This allows to read the tracks in an optimal
sequence, enabling minimal movements of the read/write head. Quantum calls this
technology ORCA (Optimized Reordering Command Algorithm). It should improve
performance by 20%. Finally, the SCSI controller can multitask, so the CPU is not locked up
during hard disk operations, which you can experience with IDE. SCSI hard disks can
achieve substantially larger transfer capacity than the IDE drives, but they have the same
bottle necks: the serial handling of bits in the read/write head, where the capacity is highly
dependent on the rotation speed.

Booting from SCSI disk

If the hard disk has to be booted, traditionally it has to be assigned ID 0. If the SCSI
controller has to control the hard disk, then the PC CMOS setup must be modified, so the
(IDE) hard disk is not installed if not both types of hard disks are installed.The operating
system will find the host adapter after start up and BIOS will be read from the hard disk
through the adapter. New BIOS's allow a choice of booting from either IDE or SCSI disk.

USB
The USB (Universal or Useless Serial Bus) is a cheap, slow bus running up to 12 Mbit/sec.
Just as FireWire it is an open royalty-free specification. The USB holds up to 127 units in one
long chain. Units can be plugged an unplugged on the fly very easily. Here you see the
plugs, the two small ones, number two from the left:
There will be problems with USB in the beginning, since many motherboard manufactories
produced their own versions of the port before at was fully standardized.
USB is only supported by Windows 95 OSR2.1, but with Windows 98 we shall really use it.

USB will replace the mess of cables and plugs we today use for:
 Keyboard l
 Mouse l
 Joystick l
 Loudspeakers
 Printers l
 Modems and ISDN-adapters l
 Scanner and camera l
All these units - and lots of others - will be connected using one single plug at the PC. The
keyboard may hold a hub, so other USB-units is connected here. Each unit holds at minimum
two plugs, so they all can be daisy chained:
All units have a firmware identification code, that communicates with Windows 98. The unit
must have a power feed (could be minimum 100 ma) to be recognized by the USB controller
and Windows 98. If one unit fails this way, Windows shows an ! on yellow background to
signalize that something has to be done. This could be to unplug other USB devices to
increase the available power in the chain. Many hardware manufactories today produces
their modems, cameras and scanners in versions with two way interface. The device
connects traditionally using a COM-port - or you use the USB.

The Hub
We should be able to connect 127 units all together. An important Unit is the hub, as we
know it from the ethernet. The USB hub may be found in the keyboard, but probably we will
use little, powered boxes holding 8 USB connectors. Five of these hubs can be daisy-
chained, providing connection for 36 units. USB hubs can for convenience be placed on the
backside of monitors, in scanners also. The USB cabling can deliver 500 mA of power. This
sufficient to feed a keyboard or other low-powered units. But it is not enough for multiple
units, therefor we will need powered hubs. We shall also find COM to USB converting hubs.
A box will house four DB9 connectors serving as COM5, 6, 7 and 8. They all connect to the
PC via one USB port. This way serial devices can connected without the IRQ puzzle we
often experience nowadays.

Shared USB-units
An other interesting aspect is that USB allows shared peripherals. This means that two PC's
can share an USB-unit. Or you may even use the USB for a low-priced network connection?

IEEE 1394 FireWire

The next technology is called FireWire. It does not look very much like the SCSI we know,
but is a further development being a serial high speed bus. The interface IEEE 1394 has a
bandwidth of 400-1000 Mbits per second. It handles up to 63 units on the same bus. The
units can be plugged and unplugged hot - meaning you do not have to power-down the PC.
The Firewire is expected to replace:
 Parallel Centronics port l
 IDE l
 SCSI l
 EIDE (later on) l

Entertainment
The first versions will be used for digital audio/video-electronics like:
 Digital cameras and cam-orders
 DVD drives l
 Scanners l
FireWire comes from Apple but it is an open standard which can be used for free. Hence all
mayor hardware companies has adapted it. Especially the entertainment electronic industry
(Video/games/television) have great hopes with FireWire. It will connect all types of digital
electronics with the PC and this way open up for a much more modular design. Since
FireWire is advanced and yet claimed to be cheap & simple, the communications protocol
can handle a lot of other units like:
 Network controllers
 Hard disks, CD-ROM drives l
 Printers l

Two modes
The FireWire standard operates with two modes.
Asynchronous as other busses. This means that operatings across the bus is controlled
using interrupt signals. The bus reports to the host when a task is fulfilled. ISO synchronous.
In this mode data is being transferred at a steady preset speed – continuously and without
any supervision from the host. This opens up for data-streaming useful for video or
themultimedia presentation. The FireWire is an peer-to-peer interface. This means that data
can be transferred between two units attached to the bus without supervision from the PC.
FireWire has 64 bit address bus. Compared to SCSI each unit does not need an unique ID,
they are dynamically configured "on the fly". Neither does the bus have to be terminated. All
together a lot more simple than SCSI. One of the problems with SCSI has been the limitation
on distance between the units. FireWire can hold up to 16 units in the same "string" and
there can be up to 4,5 meters between two units. The first implementations of FireWire will
connect it to the PCI bus using the new PIX6-controller, which will be a part of one of Intel's
new chip sets. I think it will last 2-3 years before we really see this new technology in the
market. But it will be worth waiting for it, it opens up for new world of inter connectivity
between TV, PC, video end all other types of electronically gear.

USB and FireWire - serial busses of the future

In the future a PC will hold only two I/O busses - both serial:
 USB for all low speed gear.
 FireWire for high speed I/O as disks, video
With FireWire and USB motherboard and software configuration will be vastly simplified. I
imagine one driver for each bus covering all units on the bus. No setup of IRQ, DMA and I/O-
address - great. The PC will thoroughly become a modular setup of Plug And Play units!