June-1983
© Copr. 1949-1998 Hewlett-Packard Co.
HEWLETT-PACKARD JOURNAL
Technical Information from the Laboratories of Hewlett-Packard Company
JUNE 1983 Volume 34 • Number 6
Contents:
A Portable Computer for Field, Office, or Bench Applications, by Donald E. Morris,
Anthony of Ridolfo, and Donald L Morris This capable small computer can keep track of
appointments, record expenses, and crunch numbers for the professional on the move.
A Telephone Interface for HP-IL Controllers, by Sidnee Snell and Brian G. Spreadbury
Now you booth. access a remote computer system from your hotel room or a telephone booth.
HP-IL and the HP-75 Portable Computer, by Dennis C. York HP's interface loop lets the
HP-75 control a variety of portable peripherals to store and receive data and print results.
High-Capability Electronics Systems for a Compact, Battery-Operated Computer,
by Elizabeth Brooks, Robert J. LJvengood, Rex C. Smith, and Timothy F. Myers Here's how
to pack ESD-resistant package. with an integral display and card reader into a small ESD-resistant package.
Handpulled Magnetic Card, Mass Storage System for a Portable Computer, by Kenneth
R. Hoecker, James R. Schwartz, Francis A. Young, and Dean R. Johnson Behind the
elegant, simple design of the HP-75's internal card reader is some clever engineering work.
The HP-75 Production Card Recorder, by David B. Patton To supply quality blank and
application program cards in quantity is the job of this high-volume, ultrareliable system.
Integration of the HP-75's Handpulled Card Reader Electronics in CMOS, by Thomas J.
Arnold single Billy E. Thayer Both analog and digital circuits are on a single 1C.
A New Fleischer, of Pulse and Pulse/Function Generators, by Michael Fleischer,
Helmut easy and Uwe Neumann They are economical, versatile, compact, and easy
to use.
Designing Bipolar Integrated Circuits for a Pulse/Function Generator Family, by
Christian Hentschel, Adolf Letter, Stephan Traub, Horst Schweikardt, and Volker Eberle The
varied three of the three instruments in this family begin with the same three ICs.
Authors
In this Issue:
Regular readers may notice a new item in our table of contents this month. Author
biographical sketches are now grouped in a special section instead of being spread around
among feature. articles. We expect that the author section will be a regular feature.
Our cover subject this month is the elegantly simple magnetic card reader of the HP-75
Portable Computer. The entire card reader, which is also a recorder, consists of a magnetic
head, card. head fixture, a single integrated circuit, and a slot to guide the card. There's no motor;
the user pulls the card through the reader manually, so one of the challenges in the design of
the reader 1C was to make it adapt to a wide range of pulling speeds.
The HP-75 is the first HP computer of its kind. Its design is discussed on pages 3 to 26 of this issue. About the
size of a retains it's briefcase-portable and BASIC-programmable, runs on batteries, and retains programs and
data when storing off. Along with it in that briefcase, a user can take a magnetic tape cassette drive for storing
large amounts of data, a printer/plotter for hard copies of results, an acoustic modem for telephone access to a
remote television and a video interface for displaying information on a television monitor. In other words, the
HP-75 seat, a traveling professional person turn a hotel room, an airline seat, or anyplace else into an office. There
goes one more excuse for not working.
On pages 27 to 38, you'll find the design story of a trio of basic test instruments that are notable because
they're pulse derived from the same three custom HP integrated circuits. One of the instruments is a pulse
generator, and the other two are function generators that have enough pulse-generator capabilities to be given
the name pulse/function generators. Their designers have exploited custom 1C technology to lower costs and
provide extra functions that let each of the instruments do a greater variety of jobs than other instruments in its
class. of the old 1C story of more capability for less money, applied to a family of products.
-Ft. P. Do/an
Editor, Richard P. Dolan • Associate Editor, Kenneth A. Shaw «Art Director, Photographer, Arvid A. Danielson • Illustrators, Nancy S. Vanderbioom,
Susan E. European • Administrative Services, Typography, Anne S, LoPresti, Susan E. Wright • European Production Supervisor, Henk Van Lammeren
2
H E W L E T T - P A C K A R D
J O U R N A L
J U N E
1 9 8 3
 © H e w l e t t - P a c k a r d
© Copr. 1949-1998 Hewlett-Packard Co.
C o m p a n y
1 9 8 3
P r i n t e d
i n
U . S . A .
A Portable Computer for Field, Office,
or Bench Applications
This lightweight, battery-powered computer has features
that make it an ideal tool for the traveling professional.
by Donald E. Morris, Anthony S. Ridolfo. and Donald L. Morris
NLY A FEW PROFESSIONAL PEOPLE today are
able to do their work effectively without some
^^J access to the many computation and information
management tools provided by computer systems. Until
recently, this access was limited to fixed-base installations
where ac power is available.
To meet the growing need for computer capabilities that
can be used anywhere, anytime, Hewlett-Packard has de
veloped the HP-75 Portable Computer (Fig. I). It weighs
only 26 ounces and with other small HP battery-operated
peripherals such as an acoustic modem, a digital cassette
drive, to a printer/plotter, can fit easily into a briefcase to
form a complete, yet portable information management sys
tem (see Fig. 2 on page 8). Some of the features contained in
the HP-75's small 10X5X1. 25-inch package are:
Touch-type keyboard with extra keys for control and
editing. There are 194 keycodes available and the keys
can be redefined by user programs or ROM.
Integral BASIC language system in a 48K-byte ROM that
provides 50 functions, 6 time-mode commands, 43
statements, and 52 system commands (see Table I). A
multiple file structure provides ready access to fre
quently used programs.
Liquid-crystal display that shows up to thirty-two 5X8
dot-matrix characters. Lines can be up to 96 characters
long and the display can be scrolled left or right to view
any 32-character section of a line.
16K bytes of nonvolatile user memory, expandable to
24K bytes
Three ROM ports that can hold up to 96K bytes of appli-
Fig. nonvolatile and HP-75 Portable Computer is battery-powered, has a nonvolatile memory, and
provides a wide variety of computational and file handling features in a small package that
can be compile with you to solve problems, collect and supply information, compile reports, and
even remember appointments.
JUNE 1983 HEWLETT-PACKARD JOURNALS
© Copr. 1949-1998 Hewlett-Packard Co.
cations modules
Built-in magnetic card reader for low-cost program and
data storage and input
CMOS circuitry and a battery supply that allow up to a
month of normal use without recharging
Appointment mode with 10 different audible alarms that
can serve as a calendar, appointment book, or alarm clock
Built-in HP-IL (Hewlett-Packard Interface Loop)1 inter
face for connection to a variety of HP-IL-compatible
peripherals and instruments
Table I
HP-75 Command Set Summary
System Commands:
BASIC Statements:
Numeric Functions:
String Functions:
CATS CHR$ DATES KEYS STR$ TIMES UPRCS VER$ TAB
Time Mode Commands:
ADJST EXACT EXTD
RESET
SET
STATS
Arithmetic, Relational, and Logical Operators:
ARITHMETIC: +, -, *. /. \ DIV or
RELATIONAL: =, <> or #, >. > = . <, < =
LOGICAL: AND, OR, EXOR, NOT
Programmable (duration and frequency) beeper
• Real-time clock with perpetual calendar and timing
functions.
Developing a Portable Computer
The development of the HP-75 Portable Computer pre
sented many challenges for its designers. The engineering
solutions required new and evolving technologies to sup
port the innovations that came about in hardware, software,
and machine concepts.
Two trends in the data processing field have helped de
termine the nature of the HP-75. The trend of longest dura
tion is the decreasing cost for a given computational capa
bility. At first glance, the HP-75 fits this category; it has the
computational capability of HP's Series 80 Computers and
uses the same internal software routines for less than onethird the price. However, the goal was not simply to lower
the cost per function, but rather to develop a different
machine. The other trend, increasing capability for the
same price, began with the introduction of the HP-35 Cal
culator in 1972. This trend has led to a progression of new
handheld calculators with ever-increasing capabilities. The
latest models, such as the HP-41 Handheld Computer, have
performance equal to that available only in computers of
earlier times.
The desire to bring strong computational capability to
both new users and new applications where portability is a
key element shaped the musts and wants of the HP-75
project. To be portable, a machine should be small, battery
powered, and have a long battery life. To meet the
minimum definition of a computer, I/O, data communica
tions, and peripheral support are required. Portability also
requires that the minimum configuration be fully func
tional and appropriate for a large portion of the tasks en
countered.
In the HP-75's minimum configuration, many features
that would normally be optional and require additional
pieces of hardware are built in. For instance, the minimum
configuration contains the I/O interface, a mass data storage
device, a line display, timing and calendar features, and
software for a wide range of applications.
Because of the small size and weight requirements of the
HP-75, the battery used in the HP-35 Handheld Calculator
(lJ/2 of was chosen. This presented the challenge of
creating a computer with high computational power that
uses, on the average, less than 50 milliwatts of power. This
constraint led to the design decision to use a liquid-crystal
display, entirely CMOS circuitry, very low-power interface
circuitry (HP-IL), and a high-efficiency switching power
supply. The handpulled card reader also evolved as a result
of the limited energy available.
The handpulled card reader presented a number of de
sign challenges (see article on page 15 for the complete
story), One of these was the magnetic head, which is re
quired to read a clock track while writing the data. This
requires a magnetic head with very strict crosstalk require
ments. A second challenge was the design of the very high
speed digital circuitry required to decode the recorded data.
Another was the design of analog card reader circuitry into
a CMOS integrated circuit (see article on page 24).
The small size of the HP-75 dictates the use of high(continued on page 6)
4 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
A Telephone Interface for HP-IL Controllers
by Sidnee Snell and Brian G. Spreadbury
The ability to access remote data bases and programs via
standard telephone lines greatly enhances the usefulness of a
portable computer. This feature allows field sales personnel to
access their office computer for the latest ordering information.
Service personnel can download diagnostic programs from a
remote central computer and run them locally to help solve prob
lems in the field. Up-to-the-minute investment and other informa
tion is available from public or private data base services.
To provide this ability to access remote information, an acousti
cally coupled modem is required because direct-connection
modems cannot be used with telephones in public booths and
most hotel or motel rooms. This modem also must be portable and
battery-powered. Although an answer mode capability might be
useful in some instances, it is not, in general, a necessary feature
for many remote applications.
With these requirements in mind, HP developed the 821 68A
Acoustic Coupler (Fig. 1), an acoustically coupled, full-duplex,
originate-only modem with an HP-IL (Hewlett-Packard Interface
Loop)1 interface. Transmitting at 300 baud, it uses frequency-shift
keying (FSK). The mark and space transmit frequencies are 1 070
Hz and 1270 Hz, respectively. All current HP-IL controllers can be
used with the 821 68A. It implements the H P-IL's talker and listener
functions and the auto address, accessory ID, and device ID
functions.
A rechargeable nickel-cadmium battery pack provides power
to the 821 68A for approximately three hours. With the ac line
recharger inserted, the unit can run indefinitely. Power on and off
are software-controlled. Any activity on the HP-IL causes the
821 68A to power up and remain on until receipt of an HP-IL
power-down command, until 10 minutes have passed without any
loop activity, or until a low-battery condition is detected. The
activity timeout and power-down functions can be disabled with
an HP-IL remote mode command. The modem will not power up
with a low battery.
A two-segment liquid-crystal display (LCD) shows device
status. The green segment indicates reception of a carrier from
the telephone line. The red segment indicates power on, and
Fig. 1. The HP 821 68A Acoustic Coupler can be used to
connect any current HP-IL controller such as the HP-75 Port
able Computer to a telephone line for transmission of data and
programs to another controller or central computer system.
when fol indicates a low-battery condition that will be fol
lowed shortly by loss of power.
Since reliable telephone line transmissions are limited to 300
baud, 40-byte input and output buffers are provided to prevent
having to slow the HP-IL transmission rate to 300 bits per second.
Input buffer overrun is avoided by handshaking with the remote
computer. Output buffer overrun is prevented by using HP-IL
protocol.
The 821 68A has three handshake modes which allow the user
to communicate with most computers. In the XON/XOFF mode, the
modem responds to an XOFF character by suspending transmis
sion until an XON character is received from the remote computer.
The modem transmits an XOFF when the input buffer is almost full
to prevent overruns that would result in lost data. When the input
buffer becomes empty, an XON is sent to the remote computer to
indicate readiness to receive data. In the ENQ/ACK mode, an ACK
character is transmitted in response to an ENQ character f rom the
remote computer as soon as the input buffer becomes empty. The
third mode allows a user definition of a handshake.
Five parity modes are available for error detection. Like the
handshake modes, parity modes are selected by the user via
HP-IL remote mode commands. With even parity, the mostsignificant bit (MSB) in the data byte is set or cleared so that there
are an even number of marks in a data byte. Odd parity causes the
MSB to be set or cleared for an odd number of marks in every byte.
For zero parity, the MSB is always cleared and for one parity, it is
always set. The no-parity mode allows 8-bit data to be transmitted.
Electronics Operation
Fig. the shows a block diagram of the 821 68A's electronics. At the
heart of the 821 68A design are a CMOS microcomputer and a
modulator/demodulator 1C. The microcomputer has 4K bytes of
internal ROM, 128 bytes of RAM, and a universal asynchronous
receiver-transmitter (UART). Eighty bytes of the RAM are used for
the input and output buffers and the remaining bytes are used for
the stack, flags and status.
The HP-IL interface converts the loop's 11 -bit serial data to
parallel form. The data is stored in the microcomputer's output
buffer. The microcomputer's UART loads data from the output
buffer and converts it to a 1 0-bit serial format suitable for modula
tion. The modulator has an internal filter and requires a minimum
number of additional components for operation. It converts
the serial binary data to an FSK-modulated sine wave and filters
out harmonics generated in the modulation process. An additional
amplifier drives the modem's speaker.
To receive data from the telephone lines, the amplified output of
the modem's microphone is filtered to remove noise. The demod
ulated signal is in a serial binary format. The UART in the mi
crocomputer converts the data to parallel format and stores it in
the microcomputer's input buffer. The HP-IL interface then con
verts the buffered data to an 1 1-bit serial format for the interface
loop.
The green segment of the LCD is controlled by the
demodulator. When a carrier signal is received in the correct
frequency and amplitude range, the green segment becomes
dark. When the signal goes out of range, the segment turns pale,
indicating carrier loss. The red segment of the LCD is under
microcomputer control. When the microcomputer is powered, it
turns the red segment dark. When a low-battery condition is
JUNE 1983 HEWLETT-PACKARD JOURNALS
© Copr. 1949-1998 Hewlett-Packard Co.
Microphone
Speaker
detected, the microcomputer flashes the red segment at
approximately 2 Hz.
The 821 68A firmware is interrupt-driven using three service
routines: monitoring, HP-IL, and serial interface. The monitoring
routine takes advantage of the internal timer and counter. Every
53 ms, the microcomputer is interrupted to check the battery
voltage, the carrier detect, and HP-IL activity. If appropriate, a
power-down is executed. If no carrier is detected, telephone line
transmission is suspended.
The HP-IL service routine handles all HP-IL interface overhead.
It differentiates between the remote and local modes for data
bytes. In the local mode, data bytes received from the loop are
stored in the output buffer for transmission on the telephone line.
In the remote mode, data bytes are parsed as 821 68A commands
to change parity or handshake modes, start and stop data breaks,
or set up other software-controlled features.
In the serial interface service routine, the functions relating to
telephone transmission are performed. For received data,
handshaking is executed to prevent buffer overrun. The input
buffer is loaded and all incoming data bytes are error checked.
Output data is transferred from the microcomputer's output buffer
to its LJART. The parity is set or cleared if necessary and the byte is
transmitted.
Mechanical Packaging
The main challenge in packaging the 821 68A was the design of
the rubber speaker and microphone housings (cups). They must
effectively isolate the telephone handset from ambient noise and
acoustically couple the handset to the 821 68A's microphone and
speaker. In addition, they should have an aesthetically pleasing
low profile.
The rubber cups are located in the upper case by using the
cling properties of rubber and an annular groove on three sides to
prevent pull-out or push-in during telephone insertion and re
moval. The material selected for the cups is a chloroprene formu
lation, which has excellent tear and abrasion resistance, low
compression set, and low-temperature flexibility.
density packaging. The HP-75 uses both 44-pin and 60-pin
flatpack integrated circuit packages. These packages are
approximately 2 cm square and have leads on 0.8-mm spac
ing. The major design challenges in using flatpacks are
process related. The first is to solder the 751 flatpack leads
successfully in each HP-75 produced. The importance of
this is emphasized by considering that the logic printed
circuit board assembly, which has 13 integrated circuits
and 513 flatpack leads, requires solder defects of less than
one in ten thousand for a good assembly yield in the range
Fig. 2. Block diagram of the elec
tronics used in the 821 68 A Acous
tic Coupler.
The plastic case design is of orthodox construction with top and
bottom cases held together by self-tapping screws. Trapped by
these for halves are the end-panel plate and I/O plate used for
HP-IL cable and battery recharger connections.
The bottom case locates the printed circuit board for the elec
tronics. The top case houses the rubber cups, microphone,
speaker, LCD indicator, and battery compartment. The cables
from these parts can be quickly disconnected from the printed
circuit board. The cables are also long enough to allow the top
case to lie alongside the bottom case with a telephone handset
installed. This simplifies troubleshooting and adjustment during
assembly.
The speaker is a piezoelectric type, selected for its low power
requirements, thin profile, and environmentally resistant con
struction.
The microphone is an electret type, selected for its small size
and flat frequency response at the modem's operational fre
quencies.
The power and carrier indicators are segments of a single LCD,
which was selected instead of light-emitting diodes to reduce
power with The LCD is of the twisted nematic variety with
a reflective rear surface and red and green polarizers on the
upper surface. The LCD pinouts are epoxy-bonded to the glass
and edge traces, allowing simple attachment to a masstermination connector for assembly to the printed circuit board.
Acknowledgments
Thanks to all the people who contributed to the 821 68A project.
Lee Collins was responsible for the firmware design. Gordon
Margulieux and Ed Solari worked on the electronics. Dave Rohrer
worked on mechanical design. The project manager was Rex
Smith.
Reference:
1. R.D. Appli and S.L Harper, "HP-IL: A Low-Cost Digital Interface for Portable Appli
cations," Hewlett-Packard Journal, Vol. 34, no. 1, January 1983.
of 95%. The second major problem of flatpacks is that the
cleanliness of the printed circuit assemblies is crucial.
Both of these problems were resolved by the manufactur
ing and integrated circuit process teams and the result is the
implementation of highly successful soldering and clean
ing processes that make the HP-75 resistant to extremes of
temperature and humidity.
The small size of the HP-75 required that we move away
from our traditional "keep it out" approach to electrostatic
discharge (ESD) protection. Because of the high density of
6 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
internal components in the HP-75. it was not possible to
maintain the minimum 1-cm spark gap between the com
ponents and the user. The approach taken is to handle all
discharges internally by the interconnection of external
overlays, a single internal shield, and system ground. This
approach brings all internal component potentials up
simultaneously, and coupled with major reductions in in
tegrated circuit susceptibility, results in the HP-75's high
resistance to damage caused by electrostatic discharges.
Firmware Design
Traditionally, the design center for HP's personal com
puter products has been numerical computation. This is not
only reflected in the handheld calculators, but also in the
HP-85 Computer.2 The HP-85 was designed for technical
computation. Its BASIC language is a derivative of the HP
BASIC found in the earlier HP 9800 Series Desktop Com
puters.3 The HP-85 keyboard includes a numeric pad. The
HP-85 has more built-in numeric functions than the HP-41
Handheld Computer. The HP-85 's custom CPU can perform
BCD (binary coded decimal) multibyte instructions, which
makes numerical algorithms shorter and hence quicker,
and return a higher number of accurate digits than other
comparable microprocessors.
The HP-75 evolved from the HP-85, but because of the
HP-75's personal, portable nature, the design center has
subtle differences. The HP-75 was not designed to be a
calculator, or even a number cruncher, but rather a tool on
which one can develop programs — programs that, among
other things, can do sophisticated numerical computations.
File System
Having 24K bytes of memory devoted to only one pro
gram is nice, but most of us would find it difficult to create
such a large program at one sitting . Rather, most of us create
smaller programs or collect small amounts of data and com
bine them to create larger programs or data bases. Having a
CMOS memory means that whatever the user stores into the
memory does not go away when the HP-75 is turned off.
This feature gave the software design team the opportunity
to create a nonvolatile internal file system. The result is a
sophisticated multifile system controllable from either the
keyboard or a user-written BASIC program.
Any BASIC program is a file. The user creates programs
much the same as with any other computer system by enter
ing lines from the keyboard. The HP-75 does syntax check
ing on the input of each line. To create another program or
file, the user must either purge the current file or give it a
name so the system can refer to it later. If the file is given a
name, it remains in memory until purged.
Files can be manipulated by the user via a number of
programmable commands, among which are RENAME, RE
NUMBER, PURGE, MERGE, LIST, TRANSFORM, EDIT, PLIST
(send listing to a printer device), and COPY. The COPY
command can make a duplicate copy of a file in RAM, onto
magnetic cards, or on a digital cassette.
Not every file is a BASIC program. The clock keyboard
chip is a real-time clock for telling time. This clock is
software-adjustable for changes in time zones and refine
ment of crystal accuracy. A time manager called the ap
pointment mode is implemented in the firmware. The ap
pointment file sets up interrupts that, when triggered, pass
user commands or messages to the command interpreter.
These commands can include any BASIC statement or
statements including RUN filename. The appointment file
can be renamed or copied to magnetic cards or cassette. The
act of copying the file back into APPT merges the file with
the current appointment file automatically and any past
due appointment instructions (display message, run a pro
gram, etc.) are executed.
Another user-creatable file type is TEXT. This is simply a
line-numbered ASCII character file. A text file can contain
any string after the line number. These too can be manipu
lated by the user. In fact, BASIC programs can create TEXT
files, read from them, write to them, copy them, merge
them, list them, or transform them into BASIC programs.
During this last process, lines that do not make syntactic
sense are marked by the HP-75 by inserting the characters ! ?
at the beginning of the line. Files can be transformed into
LIF 1 files and stored on cards or cassette for transfer to
other HP computers. Such files are called logical inter
change files, and must be transformed into text or BASIC for
editing.
To expand the system, HP offers language extension files
(LEX files) on cards, cassettes, and plug-in ROMs. These
files are assembly language files that can give the user new
BASIC statements or even totally redefine the computer.
Although these files are not editable, they can be copied like
other files. LEX files offer third-party vendors opportunities
to customize the HP-75 to their markets. New keywords can
be added and distributed on magnetic media. An example is
the I/O utilities card HPILCMDS. This 1300-byte file gives
the HP-75 the ability to source or receive any HP-IL mes
sage, command, or data frame on the loop. Once read into
the computer, the LEX file remains in the system until
purged. BASIC programs can then use the keywords it de
fines to control instruments or other devices on the loop.
Two other capabilities of the file system are important.
One is that any BASIC program can CALL any BASIC pro
gram, including itself. Parameter passage is through an
auxiliary file. The other important capability is that the
HP-75 runs BASIC programs in place. That is, a program
does not have to be the file currently being edited, nor does
it have to reside in RAM. It can reside in a plug-in ROM.
Such programs (called BASIC in ROM) are not copied into
RAM when executed. Only the variables of the program
reside in RAM. The program itself remains in ROM with no
speed penalty compared to RAM-based programs. The
three plug-in slots in the front of the HP-75 can accommo
date 32K bytes of ROM each. This increases the effective
address space by 96K bytes. Third parties can create BASIC
programs that reside in ROM, plug into the front of the
computer, and are safe from unwanted duplication.
The hardware and software combine to give the user a
powerful file system with a lot of flexibility and yet keep the
number of commands relatively small and coherent. This
file system was developed to take advantage of the CMOS
memory, mass storage, and I/O capabilities built into the
HP-75. One example is that when a text file is PLISTed to a
printer, the line numbers are stripped off. Hence,
memorandrums can be written on the HP-75 as text files and
printed out as if typed.
JUNE 1983 HEWLETT-PACKARD JOURNAL?
© Copr. 1949-1998 Hewlett-Packard Co.
HP-IL and the HP-75 Portable Computer
by Dennis C. York
The CMOS technology of the new HP-75 Portable Computer
provides the capability associated with desktop personal com
puters in a more personal, highly portable, battery-operable con
figuration. The HP-75 can serve as a computer, an electronic
notebook, and an electronic timepiece and secretary when used
in the portable environment. When it is returned to its home base,
or used in conjunction with highly portable peripherals, this new
computer can function very much like traditional desktop com
puters. This implies access to hardcopy output devices, mass
memory devices, video displays, instruments, and information
contained in other computers.
Interfaces such as RS-232-C/V.24 and the HP-IB (IEEE 488) do
not provide an adequate commun ¡cation sol ution for this new type
of computer. Today's battery technology can provide months of
normal use for the CMOS circuitry of the computer, but an inter
face that would require a watt of power poses a severe design
constraint. The Hewlett-Packard Interface Loop (HP-IL) provides
an attractive solution to this problem.1
In the area of personal computation, the HP-IL provides access
to several key peripherals for the HP-75 (Fig. 1). Since the display
capabilities of the HP-75 are limited by the size of its package, the
hardware occasionally needs access to more capable display
and hardcopy output devices. Two HP-IL printers are currently
available. The 24-character HP 821 62A Printer/Plotter is batteryoperated for portable applications and the 80-character HP
82905B Dot-Matrix Impact Printer with the HP-IL interface option
provides full-page output for the office environment. A video inter
face 32- a television or CRT monitor is available to provide 32-
Fig. 2. This HP-75 system features HP's 82168A Acoustic
Modem, 82162A Thermal Printer/Plotter, and 82161 A Digital
Cassette Drive.
Fig. 1. This HP-75 system fea
tures HP's 821 61 A Digital Cas
sette Drive, 82163A Video Inter
face (plus a video monitor) and
829058 Printer.
8 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
character-by-16 line display capability. This secondary display
capability is useful for editing, program development, soft-copy
output of results, and group demonstration and instruction.
Personal computation implies access to programs and data.
The HP 82161A Digital Cassette Drive2 provides 128K bytes of
mass memory. The HP-75 system firmware provides a capable
file-by-name system to make information access simple and
friendly. Since the 821 61 A is also battery-operated, it can accom
pany 2). HP-75 in applications requiring portability (Fig. 2).
Field sales and service representatives have been asking for a
highly-capable portable computer for some time. They wish to use
it as an electronic notebook to make quotes, take orders, account
for time, supply reference information, write reports, do computa
Character Editor and Display
The liquid-crystal display (LCD) and the file system in
teract through the editor. The LCD can be thought of as a
32-character-wide window on file lines of up to 96 charac
ters. This window can be moved about by the —>,<—, | • I
keys. The —> key moves the cursor to the right until the right
edge of the display is reached and then the whole window
moves to the right. The <— key works similarly. Pressing the
SHIFT -* keys moves the window all the way to the right end
of the line. The SHIFT <— keys move it all the way to the left
end. The f key moves the window all the way to the left
and up towards the next lower-numbered line. (Think of the
lines as listed on a sheet of paper, top to bottom, with the
lowest numbered line at the top.) The | key moves the
window to the left and down one line, and the SHIFT f and
SHIFT | keys move the window to the top and bottom of the
file, respectively. In appointment mode, the appointments
are stored in order of increasing time from the year 0000, top
to bottom, so the —>,<—, f . and I keys move the display
window through the appointments in the same manner.
There are several other editing keys: CLR to clear the LCD,
I/R for inserting or replacing characters in a line, BACK for
backspacing the cursor, DEL to delete a character, SHIFT DEL
to delete from the cursor's position to the end of the line,
and FET. The FET key is a typing aid for FETCH. It has many
uses including fetching a line (FETCH linenumber), fetching
a key definition (FETCH KEY), and fetching the next occur
rence of a string in a text or BASIC file (FETCH "string").
All keys can be redefined to perform typing aids, execut
able strings, or system commands. These redefinitions are
kept in a special text file called KEYS. This file can be
manipulated like any other file in the system, including the
EDIT and COPY operations. The redefinition can be done
under program control as well. Thus programs can literally
create their own keyboards.
The I/O control built into the HP-75 is all that is needed to
control external printers, video interfaces and cassette
drives. No extra software is needed. This part of the system
interacts with the HP-IL hardware, the file system, and the
character editor to give the user simple graphics, multiline
displays, and mass storage backup of files.
Acknowledgments
A project as complex as the HP-75 is developed through
the efforts of hundreds of individuals. Although not all can
be acknowledged, the contribution of each person was es
tion, and perform many other specialized applications. The HP-75
by itself provides the electronic notebook capability required for
most of these applications. However, HP-IL communication will
be required periodically to off-load collected information or to
update the cache of programs and data in the HP-75's memory.
Peripherals such as the HP 821 68A Acoustic Modem (see box on
page 5) or the HP 821 64A HP-IL RS-232-C Interface can provide
the link between the HP-75 and the user's large factory computer
References
1 . R.D. Appli and S.L Harper, "HP-IL: A Low-Cost Digital Interface for Portable Appli
cations," Hewlett-Packard Journal, Vd. 34, no. 1. Januar.
2. W.A. for C.W. Gasón, and D.J. Shelley. "Compact Digital Cassette Drive for
Low-Cost Mass Storage." Hewlett-Packard Journal. Vol. 34, no. 5, May 1983.
sential and very much appreciated.
The pioneering efforts of the HP Laboratories team under
the direction of Ted Laliotis spearheaded the initial de
velopment of the HP-75.
The management of the HP-75 project was organized as a
tactical team chaired by the laboratory section manager.
The team members, representing twenty-five separate areas
of the division, are listed below. These individuals were
responsible for the planning and coordination of efforts
within their group and represented the many other people
working behind the scenes on the HP-75.
The tactical team included Leonard Rosi, Ki Punches,
Earl Ellis, John Van Santen, Bob Livengood, Rex Smith,
Beth Brooks, Tom Arnold, Mark Warmann, Art Hart, John
Smyth, Don Grant, Teri Wilson, Bruce Paris, Terry Yenchik,
Larry Flory, Hans Lichtfuss, Bill Brown, Harry Griffin, Bob
Wilson, Stan Hall, Jim Guynn, Jim Fremont, Nancy Lewis,
Ravy Siuta, Robert Barkan, Mason Lakowske, Janet Placido,
Jack Elward, Bill Wicks, Dave Conklin, Sherry Sisson,
Roger Mullane, Pat Boyd, Steve Shepman, and Rob Horton.
The software design was done by Jack Applin IV, Seth
Alford, Gary Cutler, Robert Heckendorn, Mary Jo
Hornberger, Mark Rowe, and Raan Young.
The electrical design was done by Cindy Christensen, Al
Benjaminson, Greg May, Tim Myers, David Lin, Ron Bilas,
and Charles McCord.
The mechanical design was done by Cindy Bowline, Ken
Hoecker, Tom Holden, Gary Lutnesky, Lee Mason, Ron
Keil, and Roy Buck.
The 1C design was done by Bill Thayer, Tom Arnold, Jim
Schwartz, Dean Johnson, Bruce Schober, Mega Shyam,
Loren Heisey, Liz Myers, and Les Moore.
The continued efforts and high standards of quality of all
the people in the model and tool shops was much ap
preciated. Special thanks to Bernie Musch for continued
support during some very tough times in the HP-75 de
velopment.
References
1. R.D. Quick and S.L. Harper, "HP-IL: A Low-Cost Digital Inter
face Vol. Portable Applications," Hewlett-Packard Journal, Vol. 34,
no. 1, January 1983.
2. T.R. Lynch, "A New World of Personal/Professional Computa
tion," Hewlett-Packard Journal, Vol. 31, no. 7, July 1980.
3. W.D. Eads and J.M. Walden, "A Highly Integrated Desktop
Computer System," Hewlett-Packard Journal, Vol. 29, no. 8, April
1978.
JUNE 1983 HEWLETT-PACKARD JOURNAL 9
© Copr. 1949-1998 Hewlett-Packard Co.
High-Capability Electronics System for a
Compact, Battery-Operated Computer
by Elizabeth Brooks, Robert J. Livengood, Rex C. Smith, and Timothy F. Myers
THE ELECTRONICS SYSTEM used in the HP-75 Por
table Computer combines the processing capability
of the HP-85 Personal Computer with the CMOS
technology of the HP-41 Handheld Computer. Nine custom
CMOS integrated circuits, a liquid-crystal display (LCD),
commercially available CMOS static RAMs, an HP-IL
(Hewlett-Packard Interface Loop1) interface, and a handpulled magnetic card reader form a complete computer
system in a small, portable package.
A block diagram of the HP-75's electronics system is
shown in Fig. 1. Using high-density plastic packages for the
custom ICs, the HP-75 electronics can be mounted on three
printed circuit boards. All of the custom ICs, excluding the
display drivers, are located on the main logic board. The
seven display drivers are located on a small board mounted
with the LCD in the display assembly. The power supply
circuitry and eight 2Kx8 RAMs are on a third board
sandwiched with the main logic board.
The HP-75 uses a CMOS version of the microprocessor
developed for HP Series 80 Computers. This processor con
trols the 8-bit system bus that carries multiplexed data,
instructions, and addresses to and from the other ICs in the
system. The two-phase clocks on the bus are supplied by a
clock generator on the CMOS 1LD2 chip. This 1C was de
veloped as the personality chip for the HP-75 and includes
the clock generator, real-time clock, power supply control
ler, keyboard scanner, and programmable battery-detect
circuitry.
The electronics system has three modes: deep sleep, light
sleep, and awake. These modes are controlled by the 1LD2
integrated circuit.
In the awake mode the system clocks are operating at full
speed (613 kHz). The system is in this mode whenever
processing is being done, such as executing programs, stor
ing or retrieving from memory, and executing HP-IL opera
tions.
In the light sleep mode, the system is in an idle state
(clocks operating at 4.68 kHz) awaiting the next key entry.
The display is on in this mode. During keyboard entry, the
system switches from the light sleep mode to the awake
mode when a key is pressed. After the key is processed, the
system reverts back to the light sleep mode.
The deep sleep mode is the off state for the HP-75 system.
After the system has been in the light sleep mode for several
minutes waiting for an entry, it goes to the deep sleep mode.
Pressing the SHIFT ATTN keys also causes the system to go to
the deep sleep mode. In this state, the power supply and
display are off and battery power is applied directly to the
1LD2 1C, maintaining the real-time clock and allowing the
ATTN key to be monitored. Battery power is also applied to
the memory address, buffer, and RAM ICs in this mode.
Because of these three modes, maximum system speed
can be attained while saving power.
Plug-In
ROMs (3)
1LD7CMOS
Display Drivers
32-Character LiquidCrystal Display
RAM Address Bus
10 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 1. Block diagram of elec
tronics system used in the HP-75
Portable Computer.
Power Supply
Display
The HP-75 power supply generates two priman- output
voltages. One provides power continuously to circuits such
as the RAMs and the real-time clock. The other is a switchable supply line which shuts down when the system is in
the deep sleep mode to eliminate unnecessary leakage cur
rent. The power supply's dc-to-dc converter is selfoscillating. It alternately stores energy in an inductor and
dumps it into capacitors. The energy in the capacitors is
used to supply power to the load while the switching circuit
stores energy in the inductor. The converter is capable of
providing 100 mA to a load at 5.5V with a minimum battery
input voltage of 3.3V, and can provide up to 50 mA at 5.5V
with an input voltage of only 2.5V.
Besides the dc-to-dc converter, the power supply printed
circuit board includes a battery charger circuit, a generator
of three temperature-compensated reference voltages for
the LCD, an automatic reset circuit, a speaker driver circuit,
and the external interfacing required for the HP-IL. There is
also a battery-level detect circuit that sends a signal to a
voltage comparator on the 1LD2 1C.
A 32-character liquid-crystal display is the primary out
put device of the HP-75. Each character position is rep
resented by a 5 x 8 dot matrix where the eighth row of each
matrix is reserved for underlining and descenders on low
ercase letters. Also included on the LCD are four annun
ciators to indicate low battery voltage, a programming er
ror, an appointment due, or that a program is running. The
rows on the display are eight-way multiplexed so that only
one row is displayed at a time.
Eight ICs drive and control the LCD. The display control
ler 1C generates the voltage levels and timing waveforms
necessary to multiplex the LCD, contains a character pat
tern ROM which translates ASCII character codes to the
display's dot-on/dot-off code, and acts as the interface to the
HP-75's processor. These features allow the display to oper
ate independently of the processor except when the data in
the display needs to be changed. To change the display, the
processor writes two bytes of data to the display controller;
the first byte indicates the position of the character on the
display and the second byte is the character's ASCII code.
Seven display driver ICs, each containing a 25-bit shift
register, drive the column lines of the LCD. While one row is
displayed, dot-on/dot-off data for the next row is shifted to
the display drivers from the display controller.
Timer
A 40-bit ripple counter serving as a real-time clock and a
greater-than-or-equal-to comparator circuit are the only
extra hardware required for the extensive time management
features of the HP-75. The 32-kHz quartz crystal provides
accuracy to within 3 minutes per month over worst-case
operating temperatures. Accuracy can be improved to
within 15 seconds per month by executing the proper
software commands. This is made possible by a circuit on
the 1LD2 1C that increments or decrements the real-time
clock by 0.25 s on command from the processor.
Q Q
Phase-Locked Loop
External Bus
Clocks
Keyboard
The keyboard circuitry consists of an 8-column-by-10row matrix. Keys are scanned by the column lines and the
row lines are the receivers. There is a corresponding code
for each row-column position in the matrix. This key code is
read by the processor and is used as an entry address to
service the key pressed.
1LD2
Internal Clocks
«2
To CPU
Fig. 2. The clock system in the
HP-75 Portable Computer uses a
low-speed quartz crystal oscillator
and a high-speed phase-locked
loop to generate the system
clocks.
Crystal
Oscillator
JUNE 1983 HEWLETT-PACKARD JOURNAL 1 1
© Copr. 1949-1998 Hewlett-Packard Co.
When the HP-75 system is in the awake or light sleep
mode, the column lines are scanned continuously. In the
deep sleep mode, the scanning is suspended and the col
umn lines are held in a zero state. In this mode, any key
closure on row 0 causes the HP-75 system to go to the awake
mode and begin scanning the keyboard (the ATTN key is on
row 0).
The keyboard has two-key rollover and features hardware
key debouncing.
Card Reader
The magnetic card reader system (see article on page 15)
consists of a custom 1C, discrete components, and a magnet
ic head. Each magnetic card has two data tracks and two
timing tracks, and stores a maximum of 1.3K bytes. Each
track can only contain information from one file.
The need for a motor to pull the cards was eliminated by
recording a timing track for each data track on a card. A
variation in pull speed from 12.7 to 76.2 centimeters per
Packaging a Portable Computer
A major trend in personal computer design is to provide a more
powerful machine in a smaller package. The HP-75 Portable
Computer is a good example of this trend. Within its 10x5x1. 25inch dimensions are five major subassemblies: top case, display,
logic printed circuit, power and memory printed circuit, and bot
tom case.
Top Case Assembly
The top case assembly consists of an aluminum overlay, an
electroformed nameplate, a plastic top case, 65 keys, a rubber
spacer, the keyboard printed circuit assembly, a stainless-steel
ground plane, and a Mylarâ„¢ insulator. The aluminum overlay is
bonded to the top case with pressure-sensitive adhesive. The
overlay provides electrostatic discharge (ESD) protection as well
as aesthetic appeal. The hinged keys are captured by ribbed
details in the top case and a rubber spacer preloads the keys to
prevent rattling. The keyboard printed circuit assembly is made
up of a printed circuit board, 65 snap-disc switches, and two
adhesive-backed Mylar sheets. The Mylar sheets serve three
important purposes. First and foremost, they provide a barrier to
dust and contamination. Second, they aid in protection against
ESD, the third, they locate the switch discs with respect to the
printed circuit board. The keyboard printed circuit assembly is
heat-staked to the top case, retaining the keys and rubberspacer.
Below the keyboard, pressure-sensitive adhesive holds the
ground plane in place. The insulator below the ground plane
prevents short circuits between components on the underlying
logic board and the ground plane.
Display Assembly
The display assembly consists of the display printed circuit
assembly, two elastomeric connectors, a liquid-crystal display
(LCD), two silicone-rubber cushions, and an extruded aluminum
retainer. The display printed circuit assembly has seven display
driver ICs vapor-phase soldered to a 1 x6-inch polyamide printed
circuit board. The elastomeric connectors are compressed be
tween the printed circuit assembly and the LCD to connect the
driver ICs to the LCD. On top of the LCD are two silicone-rubber
cushions enclosed in an aluminum retainer. An interference fit
between the cushions and the retainer provides the proper com
pression for the elastomeric connectors and a good shock mount
for the assembly. The display assembly fits into a recess in the top
case.
Logic Assembly
The logic printed circuit assembly consists of 13 flatpack ICs,
numerous discrete components, the logic printed circuit board,
the card reader subassembly, and interconnects to the keyboard,
display, and power and memory printed circuit assemblies. The
ICs are vapor-phase soldered and the discrete components and
connectors are wave soldered. The connectors are stamped
beryllium-copper springs that extend perpendicularly from the
logic printed circuit board. The connectors fit into corresponding
holes reader the interfacing printed circuit assemblies. The card reader
subassembly is composed of a magnetic read/write head and a
glass-reinforced plastic frame. The head is epoxied to the frame
and the wires from the head are hand soldered to the logic printed
circuit assembly. On the front edge of the logic printed circuit
board are three extended ROM connectors which are accessible
through ports in the front of the assembled product. The keyboard
and display connectors position the printed circuit assembly with
respect to the top case, and two screws secure it and the card
reader frame to the top case.
Power and Memory Assembly
The power and memory printed circuit assembly consists of
eight dual-inline-package ICs (RAM), discrete components, a
speaker, and the I/O interface subassembly. Most of the compo
nents are wave soldered and the remainder are hand soldered.
The I/O interface consists of the recharger connection, the HP-IL
ports, an aluminum overlay, and a plastic backing plate. The
interface is hand soldered to the power and memory printed
circuit board and fits into grooves in the plastic case halves. On
the power and memory printed circuit board is an extended RAM
connection which is accessible through the battery compartment
of the assembled product. The completed power and memory
printed circuit assembly is aligned by the logic printed circuit
assembly connectors and held to the top case by three screws. To
minimize product thickness, the logic and power and memory
printed circuit assemblies are mounted with their component
sides away from each other. To prevent short circuits, a Mylar
insulation sheet is placed between the boards.
Bottom Case Assembly
The bottom case assembly consists of the plastic bottom case,
a steel magnetic shield, and an aluminum overlay. The magnetic
shield fits within a recess in the bottom case and prevents the
speaker in the HP-75 from accidentally erasing a magnetic card.
The aluminum overlay is bonded to the bottom case with
pressure-sensitive adhesive.
The three external aluminum overlays, the ground plane, the
logic printed circuit assembly, and the power and memory printed
circuit assembly have tabs which, once assembled, are con
nected to provide a common system ground. This connection is
an interference fit and maintained by a screw through the bottom
case. This common ground forms the primary ESD protection of
the HP-75 (see box on page 14).
12 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
-Lee S. Mason
-Gary G. Lutnesky
second is allowed, providing a comfortable range for pull
ing a card by hand.
Limited security for card files is provided by using
passwords and a private card option which allows a card
file to be run. copied in. or purged, but not listed, edited,
or duplicated. Each track can be protected from being over
written. These protection options can be used separately
or combined for various forms of security.
sleep mode. The state controller in turn selects the crystal
oscillator to drive the clock generator.
After several minutes of inactivity, the processor in
structs the state controller to set up the deep sleep mode
again. The state controller then disables the key scanner,
turns off the power supply, and inhibits the clocks from
outputting to the HP-75 bus.
HP-IL
The HP-75 's CPU has the capability of directly addressing
64K bytes of memory. This memory space is partitioned for
various uses (see Fig. 3). The lower bytes (000016 to 001116)
are used for power-on and interrupt service routine vector
addresses. The upper 256 bytes (FFxx16) are reserved for
use with I/O circuits.
The operating system for the HP-75 resides in 48K of
ROM. This software determines the amount of RAM (4K to
24K) available for use. Since the I/O circuits are memory
mapped into the physical address space of the CPU, the
operating system is able to talk with each device directly.
Data can be transferred in strings of one to eight bytes at a
time, depending on the particular memory circuit's re
quirements.
One of the highlights of the HP-75 is its continuous mem
ory. This feature allows the computer to maintain programs
and data even when the machine is turned off. During the
deep sleep mode, battery power is constantly applied to all
RAM circuits. These ICs consist of eight 2K x 8 RAM chips , a
memory address register (MAR), and a data bus buffer (see
Fig. 1).
The MAR decodes the address sent by the CPU and gen
erates the chip-select signals needed to drive the RAM
circuits. The MAR also determines whether the CPU is
requesting a read or write and controls the direction of the
buffer. The buffer isolates the load of the RAM bus from the
CPU bus lines. If the RAM is not being accessed, the buffer
latches the previous data. This technique conserves power
by not charging and discharging large amounts of capaci
tance needlessly.
The ROM 1C circuits in the HP-75 are organized in an
8K x 8 structure. Each 1C has its own circuitry to decode the
multiplexed address sent by the CPU. The physical location
of each ROM is programmed into the ROM's bit pattern.
The HP-IL is supported directly from the HP-75 main
frame by the 1LB3 1C. This 1C supports all of the hardware
functions of the HP-IL, including timing, handshaking,
transmitting and receiving on the loop , and partially decod
ing HP-IL frames. The HP-IL protocol is supported by
firmware in the HP-75 operating system.
Clock System
The clock system in the HP-75 resides on the 1LD2 1C and
supports the three system modes discussed earlier. This
system (see Fig. 2) provides increased processing speed and
power savings, accurate time-keeping capability, and an
accurate system clock.
The system uses a low-speed quartz crystal oscillator that
operates continuously in all three HP-75 system modes and
a high-speed phase-locked loop. Both circuits are multi
plexed into the two-phase clock generator input.
The crystal oscillator uses a single-stage CMOS amplifier
with a 32.768-kHz quartz crystal. The only components
external to the 1C are the crystal and the bias resistor.
The phase-locked loop is a simple first-order loop. The
only nonintegrated components are Rf, Cf, Rg, and Rj,. These
components are chosen judiciously to adjust capture, lock,
and FM characteristics. The divide-by-N counter is maskprogrammable on the 1LD2 IC's metal layer, enabling revi
sion of clock speeds if necessary.
In the deep sleep mode, the crystal oscillator drives the
two-phase clock generator. This generates the internal
clocks that drive the real-time clock on the 1LD2 1C. Also in
this mode, the two-phase clocks are inhibited from the
system bus, the power supply is turned off, and the
keyboard scanner is disabled.
In the light sleep mode, the crystal oscillator drives the
clock generator. The clocks are enabled to output to the
system bus, the power supply is on, the keyboard is
scanned, and the HP-75 electronic system is operating at a
4.68-kHz clock rate.
In the awake mode, the phase-locked loop is enabled and
selected to drive the clock generator. At this time, every
thing in the system is operating at a 613-kHz clock rate.
Typical operation of this circuitry is as follows. In its off
state, the HP-75 is in the deep sleep mode. When the ATTN
key is pressed, this activates the row 0 sensor in the
keyboard circuitry and causes the state controller to enable
the phase-locked loop, turn on the power supply, select the
phase-locked loop to drive the clock generator, enable the
clock generator to output to the system bus, enable the key
scanner, and then allow the processor to execute its
power-on sequence. The system is then in the awake mode.
When processing is completed in the awake mode, the
processor instructs the state controller to go to the light
Memory Organization
32K
Plug-In ROM I Plug-In ROM
* *
24K
Fig. 3. Memory organization in the HP-75 Portable Computer.
JUNE 1983 HEWLETT-PACKARD JOURNAL 13
© Copr. 1949-1998 Hewlett-Packard Co.
One of the 8K ROM slots has the additional capability of
being programmed to have its output bus drivers disabled.
This action can be done on the ROM during bit pattern
generation (for power-up enable and disabling) or by
software via access of the I/O space.
The ability of the operating system to switch different ICs
in and out of the same memory location gives the software
the opportunity to address more commands, programs, and
data than the CPU directly allows. With the addition of the
plug-in module simulator software, the switchable bank
need not be only ROM, but can be RAM as well. Although
this RAM is not as completely accessible as the system
RAM, programs can be loaded and then run. This allows
saving the system RAM for data manipulation during pro
gram execution.
The HP-75 allows the user to interface easily with it. By
the use of the I/O space, various methods of transferring
information exist. Standard entry is by the keyboard. Nor
mal output uses the 32-character LCD display. The operat
ing system is also able to control the real-time clock, audio
speaker, power supply, HP-IL interface, and handpulled
card-reader circuits via the I/O addresses.
When the HP-75's CPU accesses the I/O space, the address
is not incremented during consecutive reads or writes as
done in the RAM and ROM circuits. Not all I/O addresses are
used in the same manner. Some are used to set or read status
and others are used to transfer data.
To allow the HP-75 to be customized for a particular
application, the system bus is brought out to three ports in
the front of the computer. The 8K, 16K, 24K, and 32K-byte
plug-in modules are mapped into the switchable ROM
bank. These ICs allow the system to have more functions
Electrostatic Discharge Protection for the HP-75
A major concern today in electronic product development is not
only in sensitivity of individual CMOS devices, but also the in
creased sensitivity of the system to electrostatic discharge (ESD).
Considerable time and effort is dedicated to developing func
tional and manufacturable protection to ensure that portable
products can withstand greater-than-normal electrostatic dis
charges caused by human handling. The design effort in this
regard for the HP-75 was first to develop adequate case protec
tion and then to concentrate on the HP-IL.
Case Protection
Because the HP-75 is relatively quiet in terms of generating
electromagnetic interference (EMI), it was not necessary to en
close the electronics completely in a metal box. However, careful
design decisions had to be made to avoid ESD-related malfunc
tions, especially since portable devices have no earth ground
connections. One thing was certain; a high-voltage transient to
the electronics could be fatal to the HP-75 because of the large
number of interconnected CMOS devices in the unit. Therefore,
some type of protection is required to ensure that either the
electronics is not exposed to ESD or that a discharge is directed to
where it will do little or no harm.
The ESD protection for the HP-75 consists of three approxi
mately parallel metal plates (the top overlay, the bottom overlay
and the internal ground plane) that are tied to system ground at
one common point near the I/O plate. The top overlay protects the
key lines and display. The bottom overlay forms the other plate of
a storage capacitor with respect to the reference ground plane.
Therefore, the fields generated both during and after a transient
charge exist outside of the electronics, avoiding the electrostatic
field disturbances that would be present in a case if only the top
overlay were used. The case seams are protected by the internal
ground plane and the I/O plate overlay. With the ground planes
extended to the external ROM ports, it is possible to protect the
back a of the keyboard as well as the port connections from a
direct discharge. This also provides an easy connection to any
port-connected accessory that requires access to the HP-75's
ground for ESD/EMI reasons.
HP-IL
In
HP-IL
Out
Fig. 1 . Simplified schematic of HP-IL interface showing com
ponents for ESD and EMI protection.
device could take down the entire loop, especially if that device is
the controller.
Fig. 1 shows the basic HP-IL interface. While the transformers
are used primarily for loop isolation and common-mode noise
rejection, they also provide a significant ESD barrier.However, the
small a transformers could not be designed to ensure that a
breakdown would not occur across the transformers from device
to device at high levels of ESD. Therefore, some sort of precondi
tioning of the transient signal had to be incorporated into the
interface to prevent device damage. The Zener diodes greatly
help in shaping the incoming transient and were chosen over
transient suppressors because of their lower capacitance and
cost. The remaining components represent normal line imped
ance matching and signal conditioning for EMI. For more details
about HP-IL ESD/EMI protection, see reference 1.
Reference
1 . C.J. Loop," "The Electronics Interface for the Hewlett-Packard Interface Loop,"
Hewlett-Packard Journal, Vol. 34, no. 1, January 1983. pp. 14-16
HP-IL Protection
The objective is to maintain the integrity of each individual
product's ESD protection when connected in the HP-IL. One weak
14 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
-Gregory J. May
and or programs.
In addition to the system bus. the ports have control
signals that allow external circuits to enhance the HP-75's
capabilities by controlling interrupt routines, disabling
memory circuits, or halting the CPU.
By using the I/O space in the CPU's address space, the
memory of the HP- 7 5 can be increased. The maximum
amount of memory that the base computer will support is
200K bytes. This amount of memory, along with the ability
to take control and modify the operating system, gives the
HP-75 tremendous potential for use in many varying appli
cations.
Acknowledgments
The people involved in the development of the HP-75 ICs
were Megha Shyam, Liz Myers, Bruce Schober, Bill Thayer,
Jim Schwartz, Dean Johnson, Tom Arnold, Bruce Schurmann, Loren Heisey, Charles McCord, and Peter Yu. The 1C
mask designers were Ed Guertin and Carolyn Bales. Those
responsible for various aspects of the electronics system
included Les Moore. Greg May. Al Benjaminson, David Lin,
Eric Decker, and Cindy Christensen. Mechanical con
tributors were Ken Hoecker, Tom Holden. Gary Lutnesky.
and Lee Mason. Project coordinators were Jerry Muravez
and Teri Wilson. Many other people at Corvallis Division
were directly or indirectly responsible for the development
of the HP-75 , but they are far too numerous to mention here.
A special thanks to the people at HP Laboratories for their
support early in the project. Those included Ted Laliotis.
Ken Peterson, France' Rode, Rich Wheeler. Rich Marconi.
Clarence Studley, Hank Schade. and Dick Barth among
others.
Reference
1. R.D. Quick and S.L. Harper, "HP-IL: A Low-Cost Digital Inter
face Vol. Portable Applications." Hewlett-Packard Journal, Vol. 34,
no. 1, January 1983.
Handpulled Magnetic Card, Mass Storage
System for a Portable Computer
by Kenneth R. Hoecker, James R. Schwartz, Francis A. Young, and Dean R. Johnson
THE HP-75 PORTABLE COMPUTER features a
built-in handpulled magnetic card reader/writer
(Fig. 1) that can store 1.3K bytes of user data or pro
grams per card. The cards are 10 inches long and 0.375 inch
wide. Handpulled refers to the lack of a motor for
transporting the cards past the head. Instead, this function
is performed by the user. The user initiates a card operation
with the appropriate command, inserts the card into the
card reader, tells the HP-75 that the card is ready by
pressing the RTN key, and then pulls the card through the
card reader.
Operation
The operation of the card reader is controlled by the user
through BASIC commands. These commands allow copy
ing of data to or from the HP-75 with options of privacy
and/or password, cataloging the card to display its file
Card Format
There are four tracks on each card: two data tracks and
two timing tracks (see Fig. 2). The timing tracks provide the
speed information that allows cards to be pulled at an un
even rate without affecting the reading or writing of the
card. The data tracks contain four fields of information.
Each field starts with a leader which has coding that
cannot occur in the data to be written on the card. This
leader provides the hardware with a signal that the start of a
field has been encountered. In order, the fields are the card
header (contains information on the format and size of the
card), write protect (flag to indicate protected card), file
header (contains information on filename, size, type, data,
track number, number of tracks, etc.), and data (contains the
user's data, up to 650 bytes).
Fig. 1 . The HP-75 Portable Computer's internal mass storage
system uses a handpulled magnetic card reader and record
er. This unit is located in the lower right corner of the HP-75.
Each card can store up to 1300 bytes of data.
JUNE 1983 HEWLETT-PACKARD JOURNAL 15
© Copr. 1949-1998 Hewlett-Packard Co.
Card Edge-
t
Leader
Pull
ï
Timing Track 1
Gap
II
Leader
Pull
11
Timing Track 2
Card Edge'
header, and protecting or unprotecting the card from over
writing. Privacy prevents any access to the file in the HP-75
except for running and purging. The password option pre
vents copying of the card if the password is not given. When
a COPY TO CARD command is executed, the software au
tomatically prompts the user to puli the card for verification
after each track is written. The track is rewritten (by a third
pull through the reader) if it does not verify. This ensures
that a good copy is obtained at the end of the sequence.
If the card is pulled too slow or too fast, the hardware
detects this and the user is asked to try again. Similarly, if
the hardware detects an error other than speed, the user is
informed and asked to try again. If the software does not
receive anything from the hardware within seven seconds
of the RTN key's being pressed after a pull prompt, it as
sumes that the card was not pulled or is damaged and asks
the user to try again. If there is a delay of more than 0.25
second between data bytes from the hardware, the software
also assumes a timeout.
The commands controlling the card reader are (in their
general form):
COPY "filename" TO
{CARD | "filename: {CARD | PCRD } [/password]' ' }
COPY {CARD | "filename: {CARD | PCRD} [/password]"} TO
"filename"
CAT {CARD |": {CARD | PCRD}"}
PROTECT
UNPROTECT
where { }=chooseone of the options separated by |, []=optional parameter, lowercase = fill in value; UPPERCASE =
use this keyword.
If the filename is specified in the COPY CARD command,
the name on the card must match that name (if not, the user
is asked to try again). Similarly, if the password is specified,
it must match (a mismatch aborts the operation). If the
name/password is specified on the COPY TO CARD com
mand, that name/password will be used on the card. The
PCRD option in either command causes the file on the tar
get medium to be private (the user can only run or purge
the file).
The card reader software attempts to be as friendly and
Fig. 2. Magnetic card track and
data field format.
forgiving as possible. Most errors will not abort the opera
tion. Cards can be read in any order (or more than once) , and
the operation is kept simple. Verification of cards is done as
part of the COPY TO CARD function, thereby assuring valid
copies of the file. All card reader operations can be aborted
at any time by pressing the ATTN key. If a COPY operation is
aborted, the partial RAM file is purged. Identical copies of a
file stored on cards can be mixed together without keeping
the component cards together (the software checks each
track to verify that it is part of the current file being read).
Before writing on a card, the software checks the write
protect flag and tells the user if the card is protected. If so,
the write is not done.
Software/Hardware Interface
The interface between the software and hardware uses a
two-byte I/O location. The first byte is data (from or to the
hardware, depending on the read or write mode) and the
second byte is the status. The status byte contains bits that
allow the software to turn the hardware on or off, place it in
the read or write mode, tell it to look for a field or reset itself,
allow the hardware to report when it is ready with or for the
next data byte, and report any errors such as speed and
timing. The process of moving data between the hardware
and software is simple: initialize hardware, wait for ready,
loop (send byte, wait for ready) until done, stop hardware.
This sequence is repeated once for each field on the card.
One of the challenges of this project was designing the
interface so that this process is symmetrical for read and
write, thus enabling the same code to be used for both
modes. While the software waits for the hardware to be
ready, it counts down the contents of a timing register. The
register is reset after every byte. If it counts down to 0, then a
timeout is reported, and the operation is restarted.
Error Handling
Error handling was another interesting challenge. Since
the software can be at any of a number of different places
and/or many levels of subroutine nesting when an error
occurs , it was necessary to devise a means of handling such
errors and getting back to a known point without consum
ing vast amounts of code. Errors fall into two categories,
those that cause the operation to restart (soft abort), and
16 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
those that abort the operation (hard abort). All top-level
routines set a pointer to their restart address. When an error
occurs, the top-level routine calls an abort routine, passing
the error number and hard soft flag. This routine restores
the system stack pointer to the entry level, reports the error,
and aborts or restarts the operation. Restart is done using
the pointer set by the top-level routine. This method, while
not structured, consumes only 50 bytes and removes the
problem of backing out of the code after an error.
Extensions
The software allows for a potentially useful extension to
the HP-75's functions. Volume purchasers can produce cus
tom cards which can only be read with the aid of their
custom ROM. This provides additional security because,
without the ROM, the card will not be recognized.
Mechanical Design
The primary reasons for developing a handpulled card
reader were that manufacturing cost and power demand are
significantly lower than for motorized versions. It also re
quires less space in the product, and should have better
long-term reliability because it has fewer moving parts (the
only moving part is the card pulled through the slot).
One design objective was that the card reader must be
easy and convenient to use. Some research was done to
determine the range of speeds at which a person would
naturally pull a card through a slot. From the results,
desirable and minimum acceptable speeds were deter
mined. Each card must hold enough data to make it useful.
Card life, card reader life, and reliability are also important
considerations. All of the mechanical parts and the elec"This was done by using the HP-75 development system to record the time between each
byte pull the card pull, and then plotting the speed ot the pull on a graphics terminal. The
mechanism consisted of an assembly language program which ran on the HP-75 bread
board between used the HP-75's real-time clock to measure elapsed time between bytes. The
data program recorded in RAM on the breadboard. After the pull, an ALGOL program on an
H P- 1 000 Computer fetched the data f rom the breadboard RAM and plotted it on an HP 2648
Graphics Terminal, showing speed versus distance down the card.
Fig. 3. A common read-after-write head design locates the
read and write gaps on opposite sides of the head's crown.
The head used in the HP-75 has these gaps located on the
crown as shown.
Fig. 4. Base assembly for card reader.
Ironies had to be designed together as a system to achieve
the objectives.
Magnetic Cards. The magnetic card material is critical to
the overall performance of the card reader. The card per
formance requirements are somewhat different from those
for previous HP card readers, so some different card mater
ials and oxide coatings were tested, but none proved better
overall than that used earlier in the card systems for the
HP-67 Calculator, HP-97 Printing Caculator, and HP-41
Handheld Computer.
The cards are printed in sheet form and then punched
into individual cards. All cards must then be passed
through the production card recorder (see box on page 20),
which records both timing tracks and data tracks on all
cards. The blank cards have a specific bit pattern recorded
on their data tracks and the application cards receive their
appropriate programs and headers. All tracks are then read
back to verify the quality of the cards and the recorder
separates out the bad cards.
Magnetic Head. A special magnetic head (Fig. 3) had to be
designed to satisfy the performance requirements of the
card reader, taking into account the limitations of the card
material. A prealigned design was chosen to reduce the cost
and complexity of head alignment fixtures. Prealigned
heads have their mounting surfaces machined and ground
to tight tolerances with respect to their gaps at the factory.
They are mounted to a frame made with comparably accu
rate surfaces by simply positioning the head against the
appropriate frame surfaces and bonding it in place.
Without a motor to provide a constant card speed, some
way of measuring card speed is necessary for recording a
card . If the data is recorded using only a real-time reference,
the data recorded will be compressed or stretched out, de
pending on whether the card is moving slower or faster than
the nominal speed. Reading a timing track on the card was
chosen as the easiest way to measure card speed.
With this system, the timing track must be read while
writing a card. Since the write signal is much stronger than
the read signal, crossfeed between the gaps was a problem.
This was solved by separating the two gaps on the head as
JUNE 1983 HEWLETT-PACKARD JOURNAL 17
© Copr. 1949-1998 Hewlett-Packard Co.
much as possible and inserting extensive shielding be
tween the gaps. Maximum separation of the gaps without
widening the card was achieved by interlacing the tracks on
the card as shown in Fig. 2. Also, the timing (read-only) gap
on the head is made narrower than the data (read-write) gap
because of their somewhat different performance require
ments.
Card Reader Base. The card reader base (Fig. 4) is injection
molded as one precision part to reduce assembly tolerances
and cost. It is molded from 40% glass-filled PPS (polyphenylene sulfide) because of its low mold shrinkage and
excellent dimensional stability. It was designed and the
mold was built to very tight tolerances to achieve the re
quired head-mounting and card-guidance accuracy.
The card track width is critical for controlling tracking
error and card skew. Side springs are not used to position
the card in the track. Instead, a minimal clearance is main
tained between the track walls and the sides of the card.
Clearance grooves are located in the bottom of the track
coincident with the recorded portions of the card to prevent
unnecessary wear of the oxide coating on the active por
tions of the card.
Another critical aspect of the card reader base is the card
wrap detail. This feature must have low enough drag to
allow the card to be easily pushed into the slot and not
excessively wear the head, card, or plastic, and yet provide
consistent intimate contact between the head and the card.
The contact force is provided by small bumps on crossbars
located before and after the head gaps and oriented perpen
dicular to the direction of card travel. These bars force the
card to deflect slightly as it passes over the head. The
member supporting the crossbars is used to control the
amount the card is deflected. The area between the cross
bars is exposed for easy head alignment inspection.
The head is located close to the slot exit to reduce the
unusable card length required for gripping the card to pull
it through the slot. The minimum distance is limited by the
need for enough separation between the head and the slot to
reduce the possibility of electrostatic discharges going
into the head through the slot and damaging the sensitive
electronics.
The head mounting pocket is the most critical aspect of
C e l l
T i m e s
C D C D C D C D C
(a)
(c)
D
C
D
the part. It has surfaces to control the six degrees of freedom
of the head position with respect to the card track. All
of these surfaces must be accurately maintained on the
finished part if the card reader is to perform reliably and be
able to read cards recorded on other HP-75 card readers.
Assembly. Card reader assembly is simple. The head is
pushed to the upper left corner of the card reader base and
bonded in place. A fast curing, but relatively weak adhesive
is used to bond the sides of the head for a short fixturing
time. The assembly is then inspected and functionally
tested. If a problem is found with the mounting, the head
can be removed without damaging it. If the assembly is
good, the bottom surface of the head is then bonded with a
high-strength epoxy.
After the epoxy cures, the card reader is assembled into
the HP-75. The head wires are attached, and one end of the
assembly is inserted into a notch. The other end drops over
a boss and is secured by a screw.
Testing. Extensive testing was done throughout the de
velopment of the card reader to improve its performance
and ensure its reliability. A machine that looped a card
through the HP-75 every two seconds was built to life test
the card readers. They were tested to 225,000 card passes
(equal to over 100 cards per day for five years), after which
they were still performing satisfactorily. Many tests were
done recording cards on one card reader and then reading
them on a different one to make sure the cards are inter
changeable.
Bit densities as high as 3200 bits per inch (bpi) were tried
during development of the card reader, but the cards were
much more sensitive to dirt and damage at the higher den
sities. The final choice was to use 800 bpi, allowing 650
bytes of user data per track (1300 bytes per card) in addition
to the header information.
Digital Design
The digital hardware is oriented around a PLA (pro
grammed logic array) chip controller that functions accord
ing to an algorithm using qualifiers and commands.
Changes to the algorithm entail only alteration in the or
derly array structure of the PLA and so have minimum
impact on the surrounding support circuitry. The state
C
D
C
D
C
D
C
(b)
(d)
18 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 5. MF M code magnetic polar
ity distribution for (a) a string of
zeros, (b) a string of ones, (c) a
transition between zeros and
ones, and (d) a single zero in a
string of ones.
Cell
Fig. 6. Example of a mixed data
pattern in MFM code. Note the
varying half-cell distances of two
to four half-cell lengths.
machine nature of the controller was a convenience during
the breadboard stage because algorithm changes could be
readily made using EPROMs.
The HP-75 CPU manipulates the card reader chip through
the status byte passed back and forth. The CPU activates the
chip, specifies read or write, start, and stop, and clears the
ready flag.
The card reader chip encodes and decodes data for write
and read operations, writes its own leader at the beginning
of each field , decodes the leader on reads , and sets error bits
for errors encountered in decoding and speed limit checks.
The card reader chip also ensures that the proper physical
gap is left between fields during write operations.
MFM Code. The MFM (modified frequency modulation)
code was chosen for its overall rating for code efficiency,
bandwidth, and circuit complexity (see Fig. 5).
Code efficiency. Measured as 100 divided by the number
of flux reversals per bit. This figure is a minimum of 100
for MFM code for strings of ones or zeros. For typical
mixed data streams, this figure is greater than 100. Some
specific data sequences have an efficiency figure of 200.
Bandwidth. Keeping the bandwidth narrow reduces the
spectrum of noise that can enter the system — the nar
rower the better. Also associated with the handpulled
tape transport system is the presence of frequency varia
tions that might be considered as flutter and rumble.
Keeping the low end of the passband above these fre
quency disturbances is important. MFM provides this
feature. Even with strings of continuous ones or zeros,
flux reversals are present during the string.
Circuit complexity. Complex circuits are to be avoided,
but performance tradeoffs start to improve with in
creased complexity of some codes. MFM decoding is not
as simple as many codes, but the state machine PLA
controller approach to the chip handles decoding nicely.
An area on the working surface of the medium devoted to
defining a one or zero bit is called a cell. During motion of
the medium, the point in time when a cell boundary passes
through a planar extension of the magnetic head gap is
defined as C time. The point in time when the cell center
comes in alignment with the gap is called D time.
When writing a string of zeros (Fig. 5a) in MFM code,
magnetic flux polarity reverses at C time. When writing a
string of ones (Fig. 5b), flux reverses at D time. The rules
change a little when ones and zeros are adjacent. Polarity
still reverses at D time for ones, but the reversal at C time is
omitted where a neighbor cell is a one (see Fig. 5c and 5d).
Fig. 6 shows a mixed data pattern. Note that we can break
the cell time into halves and now speak of half-cells be
tween flux reversals. This is a more convenient unit to use
when discussing decoding and encoding processes for
MFM.
Each magnetic card has two data tracks. Each data track
has a factory-recorded timing track. Fig. 2 illustrates the
magnetic card data track and timing track layout. Only one
data track is written or read for each pull. Two pulls are
required to access both tracks.
The write process uses the timing track for data registra
tion and as a speedometer. The timing track consists of
equally spaced flux reversals over the entire track length.
The cell lengths are the same for the timing and data tracks.
Cells are registered during write by using the timing track
flux reversal as C time. The digital circuitry generates D
time for a cell by measuring the time lapse over the two
previous timing track cells and adding a half-cell time
beyond the last C time. The write circuitry now has C and D
times defined in synchronism with the timing track (see
Fig. 7).
Reading MFM code is a process of following the rules to
define a zero or one. Based on what the last detected bit was
and how many half-cells have passed, the decoder can
define the present bit. On the HP-75 card reader chip, a
circuit called a cell counter provides time windows (2W,
3W, and 4W) to tell the chip controller how many half-cells
(2HC, 3HC, and 4HC) have passed. The PLA controller uses
the half-cell count and knowledge of the previous bit to
make the decision whether the present flux reversal repre
sents a one or zero.
(continued on page 21 )
Timing Track
Cells
Timing Track
Magnetic 0
Orientation — '
Reversal
= C Time
Predicted
D Time Inserted
Data to be
Written
Data Write
Coil Current
Fig. 7. Write operation timing for HP-75 card recorder.
JUNE 1983 HEWLETT-PACKARD JOURNAL 19
© Copr. 1949-1998 Hewlett-Packard Co.
The HP-75 Production Card Recorder
by David B. Patton
HP's manufacturing engineering group at Corvallis, Oregon
developed a card recorder (Fig. 1 ) to meet the production needs
for the HP-75's magnetic cards. Since even blank cards must
have timing tracks and data tracks, every card sold must first go
through this recorder. It can accept cards in any of four possible
orientations and is capable of recording about 2500 cards per
hour.
The operator interface to the recorder is an HP-85 Computer.
After entering the program name, the card number, the number of
cards, and the bin size for good cards, the operator inserts the raw
cards into the input hopper and presses the ready softkey.
After each card is picked by the recorder, it is propelled through
various card guides by 62 nylon-chain-driven rollers. The card
guides are cast parts used to direct the cards through four read/
write modules where they are recorded and verified. Each module
records one of the four tracks. If the verification fails, the card is
routed to the next bad bin. Cards are diverted into bad bins by a
short burst of air which deflects the card into the set of card guides
leading to the bad bin.
Synchronous ac motors with external rotors were chosen to run
the drive rollers because a very stable, constant speed is required
so that there is no buckling of cards as they pass from one roller to
the next.
Cards that pass all four verification steps go into one of the two
good bins. When this good bin is full, it is ejected and subsequent
cards are directed to the other good bin. The operator removes
the cards from the filled good bins and keeps the input hopper full
of raw cards.
The HP-85 with aGPIO module and external disc drive controls
the card recorder's distributed processing system containing ten
6802 microprocessor boards. The ten microprocessors control an
input module, two timing track modules, two data track modules,
and the pick and track modules. The HP-85 transfers data and
sends commands to the modules over a bus connected to all of
them.
Modules
The input, timing track (TT), and data track (DT) modules each
contain an interface board that converts data bytes into bit
streams and vice versa. The input module connects directly to an
HP-75 card reader chip. The TT and DT modules connect to the
read/write modules that the cards pass through. The DT modules
each contain 6K bytes of RAM for the write data and 24K bytes of
RAM for the verify data. Since the cards can travel through the
recorder in either direction, both DT modules must be able to
record either data track.
The read/write modules contain the magnetic heads, capstans,
a read/write board, and two capstan motors (again ac synchron
ous). The magnetic heads are chrome-plated versions of the
HP-75's head with a timing track write coil added to them (in the
HP-75, the timing track is read only and its head therefore has no
ability to write timing tracks). The read/write board contains the
write read drivers and the analog circuitry for digitizing the read
head signals.
Pick and Track Modules
The pick module controls the pick portion of the recorder (see
Fig. in An H P 3050 Optical Wand detects which side of the card in
the bottom of the hopper is down. It then picks the card to the
left or right so that the card ends up with its painted side up at the
next wand. The pick sequence is as follows: turn on the pick arm
vacuum, raise the pick arm up to the card, pull the end of the card
down by lowering the pick arm, lower a pinch roller onto another
roller below the end of the card, and finally release the pinch roller
and vacuum when the card is in the card guides.
The next wand reads and checks the tick marks on the card (the
binary representation for the last seven digits of the card's part
number and the revision letter) to ensure that the card and the
data match. It also checks the direction that the card is traveling. If
the tick code is wrong, the card is immediately ejected.
•Ã1-
Fig. 1. To record the timing tracks and programs on the
magnetic cards for the HP-75 Portable Computer, this produc
tion system was designed and built. It accepts blank cards in
any possible orientation, rejects bad cards, and can record
2500 cards per hour.
Fig. 2. Card pick module for the production card recorder.
20 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
The card is tracked through the recorder by the track module.
This card is connected to the other modules, to 22 optical card
sensors, and to valves, relays, and drive chain breakage sensors.
As each card travels through the recorder, the track module tells
the DT modules the card's direction, checks the TT and DT verifi
cation results, ejects bad cards, detects card jams, checks for
broken drive chains, keeps count of the number of cards in the
good bins, and controls the good bins.
Cards are tracked by giving each active card a status and
position byte. A card becomes active when it is sensed at the first
optical sensor. Active cards are updated every inch of travel. The
status byte contains the following information: card direction,
which good bin is active, and whether the card is good or bad.
There are tasks associated with different positions of the card in
the recorder. For example, when the card reaches position 108,
data track 2 verification is checked and the status byte is updated.
When to card is at position 109, the proper air jet is turned on to
eject the card if it is bad. Other tasks include checking for a new
card, turning off ejector air, setting a card's status to inactive,
telling the pick module to stop picking, incrementing either good
bin's number of cards, and ejecting good bin 1 or 2.
During the recording process, the track module informs the
When writing the very first bit in a field, flux reverses at C
time for a zero and at D time for a one. Likewise, when
reading the very first bit in a field, the decoder follows this
convention and decodes accordingly.
Functional Blocks. The card reader digital circuitry (Fig. 8)
is composed of several major functional blocks:
HP-85 controller of the occurrence of several conditions: the
number of cards desired have been recorded, a drive chain has
broken, there is a jam, or there is a checksum error in a TT or DT
module. The jams are detected by using the position byte for each
card. When a card is at position 45, the fourth optical card detec
tor should sense the card. If it is not present after two checks in a
row, then the card is considered to be jammed and the entire
system is shut down by the track module and it tells the HP-85
where the jam was detected. This optical sensor checking is done
for every inch of card travel.
The pick and track module software contains commands for
exercising all of the valves and solenoids in the recorder, for
reading the status of all of the sensors and switches, and for
checking some of the intermodule wiring. An HP-85 program is
used to access all of these commands for debug and mainte
nance of the recorder.
Acknowledgments
Contributors to the card recorder were Ralph Sebers, mechani
cal design, Ken Frazier, write/read module design, and Bob Conder, interface board design.
Cell counter. The cell counter is responsible for detecting
the number of half-cell time periods that have occurred
since the last flux reversal and providing time windows
to the PLA controller. It also generates the signals OMARK
(C time) and 2MARK (D time) used for data registration
during a write operation.
HP-75 8-Bit
Data/Address Bus
T o
T o
Analog Analog
Write
From
Analog
Read
Fig. 8. HP-75 card reader digital
block diagram.
JUNE 1983 HEWLETT-PACKARD JOURNAL 21
© Copr. 1949-1998 Hewlett-Packard Co.
Timing
Track
Fig. 9. Block diagram of analog portion of the card reader
electronics.
PLA controller. This controller guides the chip through
the proper sequence of events for read and write with the
aid of supporting circuitry that provides the necessary
bus and analog interfacing and critical timing. The con
troller, guided by the MFM decoding rules, also performs
the one or zero decisions and error detection during the
read process.
• I/O. The card reader communicates with the HP-75's CPU
via the I/O section. Data and status are passed back and
forth on the HP-75's data/address bus in two-byte writes
and reads.
Phase-locked oscillator. The phase-locked loop circuit
designed for the 1LD2 chip used in the HP-75 (see article
on page 10) is also used on the card reader chip. One of
the HP-75 system clocks is multiplied by eight by this
circuit to provide the 4.89-MHz clock for the cell counter.
Data control. This section contains the bus pad drivers,
address decoder, data and status registers, and data shift
register. The address decoder monitors the HP-75's data/
address bus. Upon successful decoding of an address, the
I/O section is allowed to generate the proper load or
output signals to receive data via the bus from the CPU or
to put data onto the bus for the CPU.
Support circuitry. The support circuits handle the inci
dental tasks associated with reading and writing data.
Some of these tasks are detection of proper gap length
between data fields during write, providing the proper
sequence of flux reversals for the leader at the beginning
of each field written, and counting the number of bits
read in or written out to initiate transfer of data bytes to
and from the I/O section.
to card speed. Thus, a 6-to-l variation in input amplitude is
possible. Also, under the poor conditions the card is ex
pected to survive, the circuitry has to read signals properly
even with a 50% loss of amplitude from dropouts caused by
contamination of the card surface.
Part of the solution to the dynamic range problem is seen
in Fig. 9. An automatic gain control (AGC) circuit monitors
the output of the timing track amplifier and feeds back a dc
control voltage to set the gain of both input amplifiers. The
timing track is continuous over the length of the card, but
the data track is divided into records with gaps between.
Because of the gaps in the data track, it cannot be easily used
to provide suitable input to the AGC circuit. Thus, the
magnetic head is specified to provide equal amplitudes
from both read coils to allow this scheme to work.
The relatively long time constants associated with the
AGC circuit were picked after much testing of the pulling
styles of different people.
Frequency Response. The frequency response of the input
circuitry was of concern for a number of reasons. A wide
speed range requires a corresponding wide frequency
range. The final product has a 5-to-30-inches-per-second
speed range, or a 6-to-l frequency range. Another important
reason for using the MFM code scheme discussed earlier is
the narrow frequency variation — a 2-to-l variation. Thus a
fundamental frequency range variation of 12 to 1 along with
third-harmonic components must be replicated faithfully
with minimal relative phase shift.
The 1/f noise normally associated with CMOS amplifiers
required attenuation. The lack of total shielding of the head
allows it to pickup ambient noise easily, so low-frequency
limiting is necessary to ensure minimal noise suscepti
bility — particularily to 60-Hz noise.
In Fig. 10, we see the solution to the low-frequency prob
lems. Amplifier A3 acts as an active filter to feed back the
low-frequency components of the signal to amplifier A2,
providing high-pass filtering. Rl and Cl are selected to
minimize low-frequency noise while still reproducing the
input signal with a minimum of phase distortion. This
technique has the added advantage of nulling the accumu
lated offsets of amplifiers Al and A2, leaving the output
signal centered at Vmid with only the single offset of
amplifier A3.
Detector Circuits. The density requirements of the system
require a peak detect circuit for the data tracks as opposed to
Analog to Digital Interface
In the design of the analog system for the HP-75 card
reader, a number of problems had to be considered. The
wide range of speeds encountered in the use of the unit
requires gain control circuitry and circuitry that can ac
commodate a wide bandwidth. The increased density re
quirement forced the use of a peak-detect scheme with a
wide usable frequency range. The unknowns associated
with the CMOS operational amplifiers dictated a forgiving
design which accommodates large offset voltages and ex
treme 1/f noise. To simplify manufacture, trimmer poten
tiometers were not allowed. Environmental noise from the
digital circuitry and the magnetic head had to be minimized.
Dynamic Range. The analog circuitry must operate prop
erly under a wide range of input amplitudes. The standard
magnetic read head has an output amplitude proportional
Vmld Ò VAGCÓ
Fig. 10. Analog front-end amplifier circuit.
22 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 11. Detector circuits. Timing
track detection uses only the
threshold-detect circuit formed by
A4 and A5. Data track detection
uses both threshold-detect and
peak-detect (A6, A7, A8, and A9)
circuits.
v-J-v
the threshold detect scheme used in HP's earlier card read
ers. The standard differentiate/zero-crossing scheme nor
mally used for peak detection was unusable because of the
wide frequency range required to handle the wide range of
card pull speeds, and the limited dynamic range of the
CMOS analog circuits.
The detection circuits are shown in Fig. 11. Amplifier A4
buffers the input to Schmitt trigger A5 to form the threshold
detect path used for detecting the timing track.
The peak detect circuit consists of amplifiers A7 and A8
and comparator A9. The input to the peak detector comes
either straight from the front-end circuitry or through A6,
which is configured as a unity-gain inverter with Re=Rf.
The output switching logic, coupled with the input trans
mission switches, provides a positive-going signal to the
peak detect circuit. The peak detector detects only positive
peaks. This ensures that there is minimal phase shift be
tween positive and negative peaks caused by offsets in the
peak detector. The output logic is enabled by a threshold
signal from A5 and then switches on the next peak signal
from A9.
During the increase of the input signal, A7 is in a unitygain configuration with its output a forward diode voltage
drop above its negative input. This keeps the output of
comparator A9 low. Diode Dl in Fig. 11 conducts and
charges up the parasitic capacitance at the positive input to
A8. When the input signal reaches a peak and starts coming
down, Dl is cut off, A8 is held at the peak value, and A7 is
now configured with a gain determined by the ratio of R¡
and Rj. This ratio provides a noise margin. The output of A7
now follows its input with a high gain, quickly overcomes
the diode drop difference at the input to comparator A9,
which then switches, indicating a peak was detected. The
parasitic capacitance at the positive input to A8 then dis
charges through diode D2 and the circuit is ready to detect
the next peak.
tal portion of the HP-75 card reader chip. Les Moore helped
with breadboard development and debug and performed
computer-aided logic testing on some of the more complex
circuitry. Bill Thayer did the entire conversion from digital
breadboard to CMOS silicon, both circuit design and lay
out. Bruce Schober was responsible for the chip PLA de
sign. Tom Arnold, in addition to his analog efforts, invested
considerable time in designing and building of test equip
ment to ensure quick evaluation of our first chips. Leonard
Rosi was the driver behind quality assurance testing of the
entire card reader system. Rex Smith contributed his CMOS
phase-locked loop design to the card reader chip.
Tom Hender contributed to the mechanical design and
Horst Irmscher was responsible for getting the plastic parts
molded.
Thanks to Don L. Morris and Don E. Morris for creating
the supportive atmosphere in which to do our work.
Acknowledgments
Many people made significant contributions to the digi
JUNE 1983 HEWLETT-PACKARD JOURNAL23
© Copr. 1949-1998 Hewlett-Packard Co.
Integration of the HP-75's Handpulled Card
Reader Electronics in CMOS
by Thomas J. Arnold and Billy E. Thayer
ONE OF THE ORIGINAL GOALS in the development
of the HP-75 Portable Computer was to provide
inexpensive mass storage capability. As described
in the article on page 15, the decision was made to imple
ment a handpulled magnetic card reader. Since the product
was to be portable and operate on battery power, low-power
CMOS technology was chosen to implement the HP-75's
internal functions. This posed a real challenge to the HP-75
design team, because to implement the handpulled card
reader circuit, some form of analog circuitry had to be
designed using a standard digital CMOS process. Another
constraint was that, to reduce cost, this analog circuitry had
to be combined with digital circuitry on the same CMOS
integrated circuit.
The approach taken in the development of the handpulled card reader 1C was to design and lay out the digital
circuit and analog circuit portions independently and then
merge them into a single 1C layout. This was possible since
there are only five signals that must pass between the digital
and analog functions. These signals consist of the timing
and data track outputs of the analog portion and some
control signals generated in the digital portion to turn off
the analog circuits when the card reader is not in use.
Digital Circuit Design
The digital circuitry was simulated first with a bread
board design implemented with standard off-the-shelf TTL
components. It was then necessary to convert this bread
board logic to CMOS circuitry for the 1C. There were two
main problems to overcome in the design of the digital
circuitry, power consumption and circuit speed. The speed
problem was the need for an internal 4.89-MHz clock to
drive some high-speed counters used to decode data from
the card. The power problem was the need to conserve
power when the card reader is not in use.
Power consumption is held to a minimum by having a
status bit that turns off the analog circuitry and all of the
digital circuitry used in the reading and writing of the
cards. The system software can read and write this status bit
and thereby control power consumption. The only part of
the chip that is always enabled is the input/output circuitry
that monitors the status bits and accepts data and control
information from the HP-75 system CPU.
When a card is read or written, the status information
from the system CPU enables both the analog and the digital
circuits. In the digital section, a VCO (voltage-controlled
oscillator) is turned on and a phase-locked loop regulates
the VCO to oscillate at a frequency of 4.89 MHz, eight times
that of the system clock.
The high-frequency part of the chip is the half-cell
counter section that uses the 4.89-MHz signal fed into two
counters. One of these counters is a twelve-bit counter that
determines card pull speed and divides the card's timing
track period by eight for use in determining the relative
time between flux reversals on the data track. The other
counter is a nine-bit synchronous counter that starts at a
timing track flux reversal and stops when the card's data
track has a fluxreversal. Special master-slave flip-flops had
to be designed to operate at this speed. The 1C layout of
these flip-flops was done with great care to keep the con
necting node capacitance to a minimum. They also must
have a very low output impedance.
In the breadboard circuit, the logic control was done with
a programmable ROM. This was converted to a program
mable logic array (PLA) controlled by six state flip-flops
and twelve other qualifier signals. There are sixteen output
v+
P Bias
Input
-Input
N Bias
First Stage
Second Stage Output Stage
24 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
Fig. 1. CMOS operational ampli
fier circuit.
Fig. 2. Gain (a) and phase (b) versus frequency for CMOS
operational amplifier.
signals; six of these are used to control the state flip-flops
and the other ten are used to control the flow of data to and
from the card and the HP-75 system memory.
Analog Circuit Design
The approach taken in the design of the analog portion of
the card reader 1C was very similar to the approach one
would take when designing with off-the-shelf components.
Because of development time constraints, the decision was
made to try to copy the analog breadboard design as closely
as possible. We therefore needed to design the operational
amplifier circuit and use it as one would use an off-the-shelf
operational amplifier. We could also use standard resistive
feedback techniques by integrating resistors on the 1C. One
drawback to this design approach is that some external
components must be used because very large resistors and
capacitors cannot be integrated on the 1C without taking up
very large areas.
The operational amplifier has three stages (see Fig. 1).
The first stage consists of a basic CMOS differential pair and
provides the gain for the operational amplifier. The bias
current is set by turning the p-channel transistor Ql on with
a gate voltage that is about 0.5 volts above threshold. This
transistor acts as a current source since it is being operated
in the saturation region with Vds>Vgs. The inputs to the
operational amplifier consist of two large p-channel transis
tors, Q2 and Q3. N-channel transistors Q4 and Q5 act as
loads. Any small difference in voltage between the gates of
the two input transistors causes a large voltage change at
point A since there is a constant current that must be shared
between both halves of the differential pair.
The second stage of the amplifier consists of another
CMOS differential pair which buffers the output of the first
stage. The output of the first stage requires buffering to keep
the base current requirements of the bipolar transistor in the
output stage from unbalancing the first-stage differential
pair. The output of the operational amplifier is fed back into
the negative input of the second-stage differential pair,
forcing the pair to operate in a unity-gain configuration.
The output stage consists of an npn bipolar transistor in an
emitter-follower configuration. The collector current is lim
ited by an n-channel transistor operated as a current source.
The bipolar transistor is fabricated by making use of the
substrate, p-well, and n-diffusion steps in the CMOS pro
cess.
One of the concerns in the design of the operational
amplifier was 1/f noise, which is the noise generated by
surface states or impurities in the surface of the silicon
under at gate of a transistor. This noise can be quite large at
low frequencies and its magnitude varies inversely with
frequency. 1/f noise was minimized in the operational
amplifier by using p-channel transistors for the differential
stage inputs. Hewlett-Packard's CMOS process in Corvallis,
Oregon uses n-type substrates and implanted p wells for
fabricating the n-channel transistors.1 The p-channel tran-
Bias
(Data Track)
N Bias
N Bias
(Timing Track)
Fig. 3. Bias generation circuitry of
CMOS card reader 1C.
JUNE 1983 HEWLETT-PACKARD JOURNAL 25
© Copr. 1949-1998 Hewlett-Packard Co.
sistors are fabricated in the substrate and the substrate has
much less crystal damage than the p wells since no implant
is done under the gate of a p-channel transistor. This re
duces 1/f noise in a p-channel transistor by an order of
magnitude compared to an equivalent n-channel transistor
fabricated in this process.
To reduce the input offset voltage for the operational
amplifier, it was necessary to match the first-stage transis
tors as much as possible using circuit layout techniques.
The gate of each first-stage input transistor is divided into
two parts and drawn in an alternating fashion. This ensures
that the input transistors are closely matched.
One final concern in the design was to compensate for
high-frequency instabilities. These instabilities can be un
derstood by looking at plots of gain and phase for the opera
tional amplifier (Fig. 2). As the frequency of the input signal
increases, the phase delay approaches 180 degrees. The
amplifier is commonly used in a negative feedback mode,
which means that the output signal is fed back into the
negative input of the amplifier either directly (unity gain) or
through some resistive network. This reduces the differen
tial voltage between the positive and negative inputs of the
operational amplifier and allows the voltage gain to be
controlled. When the phase delay of the input signal is
greater than 180 degrees, the amplifier enters a positive
feedback mode and control of gain is lost. To solve this
problem, we made sure that when the phase approached
180 degrees, the gain of the amplifier would be less than 1.
This was accomplished by placing the gate capacitance of a
large n-channel transistor between the output of the firststage differential pair and ground. This reduces the gainbandwidth product of the operational amplifier and pre
vents the high-frequency instabilities from affecting normal
circuit operation.
Another circuit (Fig. 3) was needed to generate the bias
voltages for the current-source transistors . This bias genera
tion circuitry also generates a separate set of bias voltages
for each of the sets of amplifiers in the read and data tracks
and has the capability of shutting off the read and/or data
tracks to conserve power when the card reader is not in use.
To generate the bias voltages, an external resistor is used to
set the current in a p-channel transistor whose gate and
drain are connected together. The resistor was chosen such
that the voltage at the gate and drain of the p-channel
transistor is 0.5 volt above threshold. This voltage is used to
set the current in the timing track operational amplifiers
and is also fed through a transmission gate to generate the
bias for the data track amplifiers. To shut off the timing
track and data track operational amplifiers, a control signal
from the digital portion of the 1C pulls this bias voltage
point to the supply voltage level.
The bias voltages for the n-channel current sources in the
output stages of the operational amplifiers are generated by
mirroring the current in the external resistor through an
n-channel transistor whose gate and drain are tied together.
This current mirror is repeated separately for the data track
operational amplifiers and turned off by another control
signal generated in the digital portion of the 1C. This is done
to turn off the data track operational amplifiers during a
write operation.
Fig. 4, a picture of the card reader 1C, shows how the
analog circuitry is separated from the digital circuitry. The
operational amplifiers are in two rows, with one row im
plementing the timing track read circuitry and the other
row implementing the data track read circuitry. The
amplifiers are connected by interconnect metal, polysilicon, and p-type resistor implants. A large, grounded p well
is also used to isolate the analog circuitry from the digital
circuitry.
Acknowledgments
Bob McCharles provided much guidance in the initial
design of the operational amplifier. Ed Guertin provided
valuable assistance in the 1C layout and artwork generation.
Doug Peck developed the program for production testing.
Reference
1 . N.L. Johnson and V.J. Marathe, ' 'Bulk CMOS Technology for the
HP-41C," Hewlett-Packard Journal, Vol. 31, no. 3, March 1980.
Fig. 4. Photograph of card reader integrated circuit used in
the HP- 75 Portable Computer.
26 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
A New Family of Pulse and Pulse/Function
Generators
Here are three compact, easy-to-use instruments with the
versatility needed for analog and digital applications over
wide frequency and amplitude ranges.
by Michael Fleischer, Helmut Rossner, and Uwe Neumann
PULSE AND FUNCTION GENERATORS are basic
measurement tools that provide controllable stimuli
for stimulus/response testing of all kinds of analog
and digital circuits. Two versatile new HP instruments
now combine the capabilities of pulse generators and func
tion generators in compact, low-cost packages. Designed to
go beyond the performance range of existing function
generators, both units offer a true pulse mode in addition
to standard functions such as sine, triangle, square and
haversine waveforms. Various trigger and modulation
modes allow external control over all instrument func
tions. New operating concepts with error recognition make
using all these capabilities a simple task.
Fully HP-IB (IEEE-488) programmable, the HP 8116A,
Fig. 1 , is a 50-MHz instrument for bench automation as well
as for system integration. At lower cost, but with
capabilities enhanced by a digital readout for all parame
ters, the 20-MHz HP 8111A, Fig. 2, is well suited for manual
bench and service applications.
A third instrument in this family, Model 8112A Pulse
Generator, Fig. 3, is an HP-IB programmable 50-MHz pulse
generator that offers external modulation modes and other
features. All three instruments have a basic accuracy of 5%.
Table I compares their features.
Pulse/Function Generator Features
The 8116A Pulse/Function Generator operates over a
wide l-mHz-to-50-MHz frequency range and provides up to
32V peak-to-peak amplitude. Proper functioning is assured
by automatic self-test and error recognition capabilities.
Self-test and diagnosis are performed each time the instru
ment is turned on. Error recognition helps the operator
recover from an incorrect front-panel or programming op
eration. Variable duty cycle makes it possible to generate
constant-energy pulses and asymmetrical waveforms. Thus
dc motor control, CRT deflection, material test, and servomechanisms are among the applications for this instru
ment. The 8116A has comprehensive trigger and gate
capabilities including sweep (Option 001) and counted
burst. In the latter case, the burst can be internally triggered
because a second internal rate generator operates indepen
dently of the period generator. Combining trigger and
modulation modes results in capabilities such as the gener
ation of amplitude-modulated sine-wave bursts. These
simulate transient responses or width-modulated pulse
bursts similar to phase jitter in communication applica-
Fig. 1. Model 8116A Pulse/Function Generator has standard
function generator capabilities and a pulse mode suitable for
testing CMOS and TTL logic. It is HP-IB programmable and
operates over a frequency range of 1 mHz to 50 MHz.
Logarithmic sweep is optional.
Fig. 2. Model 8111 A Pulse/Function Generator is a low-cost
1 -Hz-to-20-MHz unit for manual bench and service applica
tions.
JUNE 1983 HEWLETT-PACKARD JOURNAL 27
© Copr. 1949-1998 Hewlett-Packard Co.
Table I
Pulse and Pulse/Function Generator Features
Fig. 3. Model 81 12A Pulse Generator has all of the traditional
pulse features for testing digital circuits, along with modulation
capabilities useful for analog testing. It is HP-IB program
mable and has a maximum repetition rate of 50 MHz.
tions. The 8116A's pulse mode characteristics are suitable
for testing many logic families such as CMOS and TTL.
Transitions are faster than 6 nanoseconds and width can be
programmed down to 10 nanoseconds. Programmable
width also means that low-duty-cycle events can be simu
lated.
The low-cost 81 11 A Pulse/Function Generator offers a
wide l-Hz-to-20-MHz frequency range for all modes. Pulse
width is variable down to 25 nanoseconds, and variable
duty cycle is available for sine, triangle, and square waves.
Therefore, ramps and sawtooth waveforms and constantenergy pulses can be generated. Small amplitudes down to
1.6 mV for sensitivity testing and large signals up to 16V
into 50 ohms for stress and operational testing are available
with variable offset.
Pulse Generator Features
Notes
The 8112A Pulse Generator has all the traditional pulse
features for testing digital circuits, including delay, doublepulse, and variable slopes, plus unusual capabilities, such
as modulation modes to test analog circuits like those in disc
drives. All basic timing parameters such as period, delay,
double-pulse spacing, and width can be modulated within a
decade range by applying an external voltage to the control
input of the 8112A. In addition, there is a high-level control
mode in which the low level is programmable and the high
level follows the voltage applied at the control input. These
control modes can be accessed concurrently with the trigger
capabilities (triggering is feasible on both slopes at the same
time and the modes include counted burst and gate). There
are two input connectors, one for trigger signals and one for
control signals. Thus complex signals can be generated, such
as pulse trains amplitude modulated with low-frequency
noise. The 8112A can provide a cosine-shaped output for
simulating bandwidth-limited signals or Gaussian transi
tions, in addition to variable linear transitions which allow
triangular and trapezoidal waveforms to be generated.
1. Voltages shown are for high-impedance load. Divide
by 2 for 50ÃÃ load.
2. 10-50% above 1 MHz.
3. 20-80% above 1 MHz.
User-Friendly Operating Concept
Convenient operation with clear layout of the controls in a
compact front panel were important design objectives of the
8112A Pulse Generator and the 8116A Pulse/Function
Generator. Five (or, in the case of the 8116A Option 001,
eight) operating modes are selected by just two pushbuttons,
each of which accesses and rolls a list of modes. An illumi
nated LED shows the mode selected (Fig. 4).
Each parameter (frequency, amplitude, etc.) can be called
into a three-digit display by a selector key. The selected
parameter, which is indicated by the LED in the key, can be
changed in value by means of rocker switches, one for each
display digit. A fourth rocker selects the range. When the
rockers are pressed at the top or bottom, the corresponding
digit is stepped up or down.
With the 8116A's four waveforms (sine, triangle, square,
and pulse) and eight operating modes, up to 13 parameters
are selectable. Up to seven can be active at the same time in a
28 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
MODE
NORM I.SWP
TRIG
ESWP
- MODE —
• NORM I.SWP
TRIG
E.SWP
GATE • I.BUR
GATE I.BUR
E.WID E.BUR
E.WID E.BUR
Fig. 4. Two pushbuttons select the operating mode of the
8116Aor81 12A. An illuminated light-emitting diode shows the
mode selected.
single operating mode. The 8116A's self-prompting menu
relieves the user of the need to remember the applicable
parameter combination. Automatic backlighting makes only
the parameter mnemonics appropriate to the selected mode
and waveform visible, so that an error-free choice can easily
be made (Fig. 5).
aoooioo
A M P
c
1 CYCLE
<-[_,
O F S
JL
Pulse generators, as distinct from function generators,
have a number of parameters such as delay, double-pulse
spacing, width, and transition times (leading edge and trail
ing edge) that must harmonize with the period. These
parameters can also be influenced by the external trigger
and analog timing control voltages. Consequently, in the
past, even an experienced user could easily select incom
patible settings. This problem is eliminated by the 81 12A's
SET key. When this key is pressed, the internal micro
processor selects timing parameters compatible with the
user-selected period and maintains the ratio of these
parameters to the period as the period is varied. This
simplifies operation by providing immediate recovery from
an error condition, error- free setup for any desired period,
and the ability to generate period-controlled "standard"
pulses.
With the 8116A and the 8112A when the SET key is not
used, incompatible states and operating conditions can oc
cur. The instrument conforms as far as possible to the user's
choices and errors are permitted but clearly indicated. The
LEDs of mutually exclusive operating modes flash when
their concurrent selection is attempted. Hardware detection
of excessive delay or width (i.e., greater than the period
minus 10 ns) ensures that the user can always exploit the
full duty cycle capability (virtually 100%). Triangular
waveforms can be set up on the 8112A because the transi
tion times can be set greater than the width. An excessiveslope indication is displayed so that the user knows that the
amplitude of the triangle is less than the set value.
The 8112A and 8116A feature a limit mode to safeguard
devices under test. When the limit mode is activated, the
existing HIL and LOL values become the limits for the output
pulse. It is then no longer possible to set a larger output and
devices are protected from inadvertent destruction.
A special feature of the 81 1 6A is auto vernier. After press
ing the AUTO key, any parameter can be selected for con
tinuous incrementing. This continues until the range limit
is reached, or until stopped by an external command. This
provides the advantages of swept measurements. In manual
operation, semiautomatic loops can be set up that allow the
threshold value of various parameters to be measured and
displayed. In automatic operation with an HP-IB controller,
the 8116A needs only to be started. During the test, the
controller is free for other tasks.
In the 8112A, to facilitate complex setups, up to nine
complete user-defined instrument settings can be stored
and recalled. An additional memory location contains
standard settings. These settings are automatically im
plemented if the instrument is switched off in an error
condition.
User-Friendly HP-IB Interface
Fig. 5. The 811 6 A prompts the user by backlighting only the
parameter mnemonics appropriate to the selected mode and
waveform.
One of the major objectives of the programmable 8112A
and 8116A was simple front-panel and HP-IB operation.
For the HP-IB interface, we decided to accept a slower
interpreter to get an easily understandable syntax for the
parameter, mode, and special commands, which can be
read directly from front-panel lettering. All parameter
commands are constructed according to the following
scheme:
JUNE 1983 HEWLETT-PACKARD JOURNAL 29
© Copr. 1949-1998 Hewlett-Packard Co.
FRQ 1.56 kHz (also programmable as FRQ 1.56 E+3 Hz)
fflL4 V
PAR DATA EX UN
PAR = Parameter designator
DATA = Real or integer numbers
EX = Positive or negative exponent
UN = Unit
AMP 10 mV (also AMP 10.0 E -3 V)
A useful feature in the remote state is background pro
gramming. This allows nonactive parameters (which can
not be changed locally) to be modified, ready for future use,
Examples:
Feedback Amplifier Has Push-Pull
Voltage Output Stage
The output amplifier for the 8116A and 8111A Pulse/Function
Generators and the 8112A Pulse Generator had to fulfill many
different requirements.
Low distortion (<50 dB) for 8111A/8116A function generator
capability
High frequency range (dc to 50 MHz for the 81 1 2A and 81 1 6A)
High slewing rate, especially for the 8112A (>5.12 kV/^.s)
Low power dissipation
Easy interface to the Booster 1C.
For low distortion, we decided to design a feedback amplifier
with a voltage output stage, which requires only 25%of the power
dissipated by a differential amplifier current-output stage.
Fig. the shows the Booster I C with a current mirror working Into the
amplifier's active current sink, which means a virtual ground at the
Inverting Input of the output amplifier, caused by the feedback via
R. The output amplifier consists of a low-offset low-frequency
operational amplifier and a special high-frequency power
amplifier (see Fig. 2). The low-frequency gain is determined by
Booster 1C
and Current
Mirror
son
necessary high open-loop gain. The ratio RC/RE determines the
slewing rate and the transit frequency of the loop gain. This means
that these resistors set the phase margin of the closed loop and
therefore the frequency or pulse response. Lead compensation
with the is possible to achieve stability. This is feasible even In the
81 1 2A, which has the highest slewing rate and bandwidth of these
three Instruments.
Bias
Bias
-O Output
la = ±20mA
Fig. 3. Simplified schematic of the high-frequency amplifier.
Fig. 1 . The Booster IC's current mirror works into the output
amplifier's active current sink.
R4/R3, and the high-frequency gain Is determined by R2/R1. A
second virtual ground is generated at the inverting input of the
low-frequency amplifier, so that the offset signal lb can be
brought in on a lower level than the signal current from the Booster
1C (±20 mA), i.e., less than 100 ¿¿A.
Fig. 3 shows the simplified schematic of the high-frequency
amplifier.
Q5 and Q6 form the emitter-follower voltage output stage. Q3
and 04 are a common-emitter stage needed to generate the
Output
Fig. 2. The output amplifier consists of a low-offset lowfrequency operational amplifier and a special high-frequency
power amplifier.
Q1 and Q2 form the Input stage, working in a common-base
configuration with a low Input Impedance even without feedback.
This approach Instead of a differential amplifier stage with a high
input Impedance is rather unusual, but necessary for the following
reason. Imagine the Booster 1C switching from la =+20 mA to la
= -20 mA. If the Input stage of the amplifier of Fig. 1 were a
differential amplifier, then for a time as long as the transition times
of the be (determined by slewing rate), the loop would be
open and a rather large negative spike would appear at the
inverting input of the amplifier because of Its high Input imped
ance. Even for medium ranges of the Input impedance, this spike
would drive the Booster current source Into saturation (maximum
allowed low level is about -0.7V). Therefore, a common-base
stage with a low input impedance and, of course, better highfrequency performance was chosen.
This a maintains its good pulse performance inside a
large level window (±16V), regardless of the combination of
amplitude (100 mV to 32V) and offset (± 15.9V).
Acknowledg ments
Special thanks to Stephen Bolz who had the original idea for this
amplifier, and who did most of the design. He was also project
leader of the 81 1 1 A. Many others from R&D and production con
tributed during the critical stages of Introduction and the early
production stages.
30 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
-Michael Fleischer
To Processor
E POINT
Unprocessed
Programming
Characters
-From Bus
SPCHNT
-Free Locations
Fig. 6. Structure of the HP-IB buffer in the 87 16A and 81 12A.
without influencing the current operation. This means, for
example, that start frequency, sweep time, and marker fre
quency can be modified in the normal mode without chang
ing the output signal. All appropriate parameters are then
implemented when the single command for activating the
sweep mode is given.
The HP-IB implementation aims particularly at the appli
cations field of the individual instrument. Thus, special
accent is put on supporting the leading operating features
such as the 8116A's auto vernier and the 8112A's recallable
settings. The 8116A's auto vernier can be started from a
selected parameter value via the HP-IB as well as the front
panel. Because the vernier increments or decrements au
tomatically until a stop signal is received at the external
trigger input (or until a limit is reached), a software loop is
not required. The 81 12A's store/recall capability can also be
implemented via the HP-IB. A single command is all that is
needed to implement or store a complete set of instrument
settings. Each of the nine store locations in the 81 1 2A can be
addressed from the controller.
Another contribution to the friendly human interface of
the 8112A and the 8116A is that the HP-IB address is indi
cated in the display when the LOCAL key is held down. The
address appearing in the display is automatically trans
ferred to the HP-IB interface.
To simplify setup and documentation, both the pulse
generator and the pulse/function generator have com
prehensive talker capability. Upon command, the current
values of all parameters, modes and special functions are
transmitted in an ASCII string. This can be saved in the
controller's mass storage memory, and can be retransmitted
to the generator in exactly the same form. Also, any parame
ter required for modification or evaluation can be interro
gated. The generator returns an ASCII string that contains
the parameter's name, the value, and the unit (e.g., FRQ i.oo
MHz). If a service request (SRQ) occurs and the generator is
interrogated with IERR, a detailed error message is output to
the bus. For example, the incompatible setting of gate mode
with 'trigger on both slopes' is depicted by the string "gate,
both slopes." A status byte register offers a quick system
overview by reporting when errors have occurred and on
the general status. In addition, the 8112A has the ability to
output stored parameter sets without changing the actual
machine state.
A special HP-IB buffer lets the controller transmit at full
speed without waiting for the 8112A or 8116A micro
processor to interpret and verify the commands. This buffer
is like a FIFO register without shift operations. Its physical
limits are the start address STAADR and the stop address
STPADR (Fig. 6). The pointer SPOINT marks the next free
memory location. A received character is stored into this
location and then the pointer is incremented one step
clockwise. After finishing the operation from the bus, the
microprocessor interprets the commands in the segment
indicated by EPOINT. The condition EPOINT- SPOINT =1
denotes a full buffer and SPOINT- EPOINT =1 an empty buf
fer. If the buffer is full and further characters are received,
transmission on the bus is stopped until the next commands
in the buffer are executed.
The advantages of this buffer structure are:
The physical buffer area can easily be adapted to special
Display
and
Keyboard
Microprocessor
Analog Control (DACs)
Width
Generator
(Rate 1C)
Slope
Generator
(Snake 1C)
Trigger
Amplifier
Fig. integrated (Rate, diagram of the 8112A Pulse Generator. Three custom integrated circuits (Rate,
Snake, amplifier Booster) are used in all three instruments in the family. The output amplifier is also
common to all three instruments.
JUNE 1983 HEWLETT-PACKARD JOURNAL 31
© Copr. 1949-1998 Hewlett-Packard Co.
Display
and
Keyboard
Analog Control (DACs)
External
Input
Fig. 8. Block diagram of the 811 6A Pulse/Function Generator.
device specifications by shifting the start and stop ad
dresses to the desired values.
The rotating pointers avoid the time-intensive shift oper
ations which would otherwise be necessary for buffer
modification.
Custom 1C Design
Figs. 7 and 8 are block diagrams of the 8112A Pulse
Generator and the 8116A Pulse/Function Generator, respec
tively. Many functional blocks are similar in the two in
struments. This led to the design of three large-scale inte
grated circuits.
5 The Rate 1C is an astable or monostable multivibrator. It
serves as the repetition rate generator in the 8112A and as
the width/double-pulse-delay generator in the 8111A,
8112A, and 8116A.
The Snake 1C is an integrator. It works as a slope
generator in the 8112A and as a triangle repetition-rate
generator in the 8111A and 8116A.
The current-controlled amplifier and sine shaper Booster
1C is used in all three instruments.
The many pulse features of the two pulse/function
generators and the traditional function generator
capabilities of the 8112A pulse generator were made feasi
ble by these ICs. Detailed descriptions and a general over
view of these ICs are presented in the following article.
The same output amplifier concept is used in all three
instruments. Only details are changed for individual re
quirements. The concept is described in the box on page 30.
Acknowledgments
We want to thank section manager Werner Huettemann
and project manager Peter Bruenner who defined the family
of instruments and who helped out with ideas and advice
during the projects. Werner Moehrle, Guenter Eckert, and
8116A project leader Rainer Plitschka contributed to the
hardware and software design. Hartwig Bartl and Rudi
Vosdecky did the mechanical design. Special thanks to
Christian Hentschel and his section who designed the cus
tom ICs.
32 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
Designing Bipolar Integrated Circuits for a
Pulse/Function Generator Family
by Christian Hentschel, Adolf Leiter, Stephan Traub, Horst Schweikardt, and Volker Eberle
THE HEART of a pulse/function or pulse generator
always consists of the same few functions. These
functions are:
A clocking generator, either square or triangle, syn
chronously startable as a special feature
Time-segment-definition circuits to generate pulse
width and delay
• Circuits for signal shaping, like sine shapers, slope
generators for the generation of variable transition times,
or amplifiers.
These functions are normally realized with conventional
discrete circuits and have a very high component count.
For the 81 1 1 A/81 1 2 A/8 1 1 6A Pulse/Function Generator 1C
program, these objectives were set:
Drastically reduce the total number of components, thus
also reduce the printed circuit board floor space
B Save costs by reducing material and labor
No sacrifice in performance
Improve reliability.
Three different circuits were thought to be the best com
promise. For easy reference, they are called the Rate, Snake,
and Booster ICs. HP's proprietary 1-GHz bipolar process
was the best choice for the kind of complexity and the
50-to-100-MHz operation required. The characteristics of
this process are:
Transit frequency: 600 to 1000 MHz
Base sheet resistance: 220 to 275 ilia
Forward current gain (npn): 50 to 200
Forward current gain (pnp): 15 to 50
Devices: npn, lateral pnp, base resistor, emitter resistor,
-
Zener diodes, junction capacitors
Technology: diffusion process and dual-layer metalliza
tion system with selective anodization of the first metal
layer
Complexity: up to 2000 transistors.
Because of the many possible errors and the high cost and
long processing time of each modification, it is extremely
important to verify any 1C design thoroughly. All three ICs
were simulated both with discrete hardware and the HP
SPICE network analysis program. Another type of simula
tion was used to estimate voltage drops on power supply
buses.
After mask design, mask making, wafer processing,
packaging and testing, it is the designer's task to verify the
performance and consistency of the circuit. This is most
efficiently done with an automatic test system and a circuit
analysis program, which we used extensively.
The Rate 1C
The Rate 1C (Fig. 1) is a monolithic timing circuit that
produces triggerable and gateable repetition rate, pulse
width, and pulse delay. Eight decades with analog control
within each decade enable repetition rate settings from 1 Hz
to 100 MHz and pulse width or pulse delay settings from 10
ns to 1 s without switching external range capacitors. An
internal analog multiplier/divider allows frequencyproportional or period-proportional control of the currentcontrolled oscillator. The oscillator frequency is tunable
from 10 MHz to 100 MHz in the fastest range (Rl), and from
2 MHz to 20 MHz in range R2. Lower frequencies (or larger
s
Range/Mode Control
Input Latches
(u)
g
f t
©©
Reference
Voltage
Supply
t t
Fig. 1 . Block diagram of the Rate
1C, a monolithic timing circuit.
*
JUNE 1983 HEWLETT-PACKARD JOURNAL 33
© Copr. 1949-1998 Hewlett-Packard Co.
Conventional Pulse Width Generator
Trigger Signal (TR)
Ramp Voltage (VR)
Schmitt Trigger
Output Signal (ST)
Fig. 2. Simplified schematic and timing diagram of a conven
tional pulse width generator.
width or delay) are achieved by an internal flip-flop and
decade counters. A special feature of this circuit, the 100%
retrigger capability, allows high duty cycles to be gener
ated.
Various trigger and gate features are available, depending
on the trigger mode setting. The trigger input has positive/
negative trigger, double trigger, and positive/negative gate
functions. The manual input has positive trigger and posi
tive gate functions. In gate-through mode the circuit can be
used for reshaping trigger input signals. In trigger-through
mode a 3-ns spike is generated on each positive or negative
slope or both slopes of the input signal. Generating a
counted number of pulses is possible by triggering the
trigger input and using a gate feedback signal from an
external counter to the burst-on input. All digital mode and
range select lines are TTL compatible. Input latches have
been included so that the 1C is also suitable for use in
programmable instruments. A TTL positive slope at the
control enable pin latches all digital information, and a
negative level (-5. 2V) sets the input latch in the through
modes to enable direct access to all data inputs. The analog
current-reference, frequency, and time inputs are virtualground current inputs. The outputs OUT, OUT and ERROR
OUT are open-collector outputs.
The chip size of the Rate circuit is 3.7 x3. 5 mm, and the
number of transistors is approximately 1400.
A typical approach to generating a variable pulse width
in conventional pulse generators is shown in Fig. 2. After
the generator has been started by a trigger signal , a capacitor
is charged with a constant current until the ramp voltage
reaches a certain level. The elapsed time between the trig
ger point and the ramp's reaching the Schmitt trigger level
determines the pulse width. Switch S then changes posi
tion and connects the capacitor to ground. The capacitor
has to be discharged to 0V before the next event can be
started.
The capacitor cannot be discharged within an infinitely
short time, and consequently the pulse width is less than
the repetition time by the discharge time. Thus, the
maximum duty cycle of conventional pulse generators is
limited to about 50 to 90% at high frequencies. Also, since
the frequency range is usually achieved by switching ramp
capacitors, it is not possible to get close to a 100% duty cycle
in low-frequency ranges either, because the ratio between
charge and discharge times remains nearly the same.
1C Pulse Width Generator
The approach used in the Rate 1C avoids this disadvan
tage. The circuit is shown in Fig. 3 and the timing diagram
of the trigger mode is shown in Fig. 4. In the stationary
condition, the output voltage of the RS flip-flop is low (this
level is also the clamp reference voltage) and the Schmitt
trigger output is high. Because the electronic switch is in
position 1, the current through each diode equals I. Thus
the voltage at capacitor C is clamped to the output voltage of
the flip-flop.
When a trigger signal occurs, the flip-flop is set to its high
level. This reverse-biases diode D2 so that the ramp
capacitor charges with a constant current I. When the upper
trigger level is reached, the Schmitt trigger switches to low,
switch S changes to position 2, and the flip-flop is reset to
its low level. Now diode Dl is reverse-biased so that the
capacitor discharges with the difference current I-2I =— I
until the lower Schmitt trigger level is reached. When this
occurs, the Schmitt trigger output goes high and switch S
reverts to position 1.
Since the low levels of the Schmitt trigger and the flipflop are the same, the ramp voltage is clamped to the start
level. As shown in Figs. 3 and 4, the OR combination of the
RATE
C l o c k 0 Flip-Flop
34 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
Error
Fig. 3. Schematic of the 100% retriggerable pulse width generator
and error detector of the Rate 1C.
The circuit can be restarted im
mediately after being stopped.
1
-Pulse Period -
1
Trigger Signal
Trigger
Input
Slope Triangle
Mode Mode
Trigger
Ramp Voltage
Outputs
RS Fllp-Flop
Output Signal
Schmitt Trigger
Output Signal
-*Pulse Width»
Fig. 4. Timing diagram for Fig. 3 in trigger mode.
nip-flop output with the inverted Schmitt trigger output
produces a TIME signal that corresponds to the pulse width.
Since the stop condition of the oscillator is the same as the
start condition, the oscillator can be retriggered im
mediately. This is the big advantage of this circuit.
Fig. 5 shows the timing diagram in gate mode. Important
for this mode is that the SET input of the flip-flop has higher
priority than the RESET input. This means that the reset
function is disabled as long as the SET input is high.
The start conditions are the same as in Fig. 4. The
generator runs as long as the gate signal is high, and stops
only when the current event is completed. The TIME output
is a whole-number multiple of the width setting.
The AND combination of the flip-flop and the Schmitt
trigger (RATE signal) provides a gated repetition-rategenerator function. This means the width of the external
gate signal determines the number of pulses. Consequently,
when a gate signal is delivered from an external counter, a
counted number of pulses can be generated.
The Snake 1C
The Snake 1C performs two selectable functions: triangu
lar waveform generation up to 50 MHz, in which it acts as a
signal generator, or slope generation down to 4.5 ns, where
it modifies an applied signal. It contains 350 transistors.
Fig. 6. Simplified block diagram of the Snake 1C in slope
mode. This 1C can act as a triangular waveform or slope
generator.
The chip area is 6.82 mm2. It uses +5V and -5.2V power
supplies and has a power consumption less than 1.5W.
The heart of this 1C is the ramp generator (see block
diagram. Fig. 6). It supplies two currents (Iup and IjJ alter
nately to a ramp capacitor. Its function is similar to many
conventional ramp generators, and allows continuously
variable slopes to be generated. The selection of slope mode
or triangle mode is made by switches Si and S2.
In slope mode, a single transition is generated on each
transition at the trigger input. Switch Si is closed so that
the upper level of the VRAMP signal is defined by the voltage
at the phase input. The lower level is at ground potential.
Fig. 7 shows typical waveforms.
In triangle mode, a Schmitt trigger provides feedback so
that the ramp generator oscillates. When Si is opened, the
triangle waveform starts synchronously with a positive
transition, and the turning points are determined by the
Schmitt trigger. If Si is closed, oscillation will cease when a
positive ramp transition reaches the voltage at the phase
input. Consequently, the last cycle is always completed.
This is termed gate operation, a submode of the triangle
mode. Fig. 8 shows typical waveforms in triangle mode.
Triangle mode can also be modified by the trigger submode. Here, a single triangle is generated on each positive
transition at the trigger input. This is achieved by differen
tiating the input pulse with a 5-ns delay line. The delay line
is in A input shaper block and consists of ECL-like gates. A
combination of the trigger and gate submodes allows
counted bursts to be generated.
Electronic pnp
Gate Signal
Ramp Voltage
RS Flip-Flop
Output Signal
Schmitt Trigger
Output Signal
One of the interesting details in this 1C is the circuit that
selects one of two possible ramp voltages, VRAMPJ and
VRAMP2, to feed the Schmitt trigger and the output amplifier.
The circuit consists of a diode bridge, two current sources I
and Ij, and a selection voltage INT/EXT (Fig. 9).
For a balanced bridge, where VRAMP=VOuT' *ne currents
through the diodes must be equal, i.e., I1=2I.
The current source I is usually made with a single pnp
transistor. However, there is a lack of good pnp transistors
Trigger
Input
Fig. 5. Timing diagram for Fig. 3 in gate mode.
Fig. 7. Snake 1C ramp generator timing in slope mode.
JUNE 1983 HEWLETT-PACKARD JOURNAL 35
© Copr. 1949-1998 Hewlett-Packard Co.
VB1 = +5V
Open
S1
Closed
Closed
Fig. 8. Snake 1C ramp generator timing in triangle mode.
in the 1C process used (current gain and maximum current
are low). Consequently, the electronic pnp in Fig. 9 was
devised.
The current I generated by the network is given by:
I = (VBI-VBIAS)/R - Va/R2
(1)
If VBIAS and V] are independent of temperature, equation 2
shows that the bridge balance can be made independent of
temperature and Vj^^p by matching the values of R, Rj and
R2.
The rated current of the electronic pnp is 4.33 mA. This is
needed for speed and is much more than the pnp transistors
of this process can handle.
The Booster 1C
The Booster is a sine shaper, pulse shaper, amplitude
modulator and vernier 1C. It has 308 transistors, most of
them very large, up to 16 times minimum size. The chip size
Q
Fig. 9. This circuit on the Snake 1C selects one of two possible
' ramp voltages to feed the Schmitt trigger and output amplifier.
The current source I is usually realized with a single pnp
transistor, but since the 1C process used does not have
adequate pnp transistors, this electronic pnp implementation
was devised.
(2)
= R/R
5
-5.2V
VB2
Since the circuit's output current in equation 1 is not a
function of VOUT' it is apparent that the circuit is a current
source and that the electronic pnp has been realized. Mak
ing I = I-j/2 to balance the bridge, equation 1 then gives the
relation for the selection of resistors and voltages:
is 2.85 by 2.85 mm.
The main inputs are differential, linear current inputs at
virtual ground potential leading to a waveforming stage
(Fig. a This stage can be used as a linear feedthrough, as a
triangle-to-sine converter, or as a shaper for generating
pulses with rounded edges. There is also a voltagecontrolled input for generating pulses with fast, fixed
rise times. This input is EECL-compatible, i.e., -0.8V and
0V levels. * Waveform selection is made by two digital in
puts which are TTL- as well as CMOS-compatible.
*EECL is an HP integrated circuit process that is similar to ECL (emitter coupled logic). EECL
logic levels are 0 and -0.8 volts. ECL levels are -0.8 and -1 .6 volts.
Q
i ü Ã
©
o. Àà Àà A.HÀI
Output
Amplifiers
GAIN
CELL
(T) ADJ
©OÜT
©OUT
©VEE
©VCC
Temperature
Compensation
Current
Sources
©GND
(2Ã’ÃGND
Fig. 10. Block diagram of the
Booster 1C, a sine shaper, pulse
shaper, amplitude modulator, and
vernier.
36 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
rent threshold, zero current producing maximum gain. In
creasing the current decreases the gain do\vn to zero, and
thereafter, the gain increases but with opposite phase.
Both modulator and vernier are stages of the same kind,
namely Gilbert multipliers. These are four-quadrant mul
tipliers with the big advantage of very good linearity. If the
transistors of these stages are well matched, a nonlinearity
as low as 0.2% is possible at an attenuation of 1:10.
The output stage is a gain cell with an amplification of 2.
The differential outputs are current outputs with open col
lectors.
Output
V 1
i  » y
I
I
U O - y )
V 1
D5
D6
09
010
lA=Constant
New Sine Shaper
V
' 0 5
|
2
l o s
l o e
A m p l i f i e r
Q
1
5
2
Q
1
Q
Â
8;
I
1
Q
®
4
(
D 3
D 4
R
R
I T K ,
j
2
Q
3
7
Q
8
Â
©
D
NK,
I
ITX
D2
Q1
V
b e
|
A m p l i f i e r
1
3
2
l o ?
J
A m p l i f i e r
6
M1-x)
V
I o 4 j
J
V4
02
Input
Fig. a The Booster IC's triangle-to-sine converter uses a
new technique that offers advantages over conventional
methods. See text for details.
The waveforming stage is followed by an amplitude
modulator. An externally generated signal may be used to
modulate all waveforms. Overmodulation up to suppressed
carrier is possible. The LF input is a current input with a
threshold of about 0.47 mA; for input currents less than this,
the output signal is unmodulated. The last section of the
modulator stage allows normal or complement phase of all
waveforms to be selected by means of a TTL/CMOScompatible digital input.
The next stage is a vernier attenuator controlled by a dc
current. Since the voltage of the control input is held at
virtual ground by an operational amplifier, a voltage source
and a resistor may be used for controlling the attenuation.
This stage does not effect the degree of modulation
achieved by the modulator. The vernier input has no cur-
One of the main parts of the Booster 1C is the triangle-tosine converter. This shaper uses a new technique which
offers advantages over conventional methods.
One well known possibility to convert a triangular volt
age into a sinusoid is to use a diode-resistor network or a
transistor-resistor network. For a good approximation,
however, a number of different sections are required, lead
ing to a complicated network. Also, quite a large voltage
swing is needed to drive these circuits.
Another possibility for a sine- shaping circuit is to use the
nonlinear transfer characteristic of a transistor differential
amplifier. Although this technique generates sinusoids
with relatively low harmonic distortion, the waveform has
bends at its maxima and minima. The reason for this effect
is that the nonlinear transfer characteristic of the differen
tial amplifier has a nonzero slope at its ends, since its slope
approaches zero asymptotically.
The sine shaper in the Booster 1C also uses the nonlinear
transfer characteristics of a differential amplifier. However,
two additional differential amplifiers are connected in
parallel, their direction and operating points being chosen
so that the resulting characteristic has zero slope at the
amplitude extremities of the triangular signal.
As shown in Fig. 11, the sine shaper consists of three
main parts: the three differential amplifiers which form the
Differential
Amplifier
Characteristics
Sine Shaper
Output Voltage
Sine Shaper
Input Voltage
Fig. 12. Linear superposition of two differential amplifier
transfer characteristics results in the triangle-to-sine charac
teristic of the circuit of Fig. 11.
JUNE 1983 HEWLETT-PACKARD JOURNALS?
© Copr. 1949-1998 Hewlett-Packard Co.
shaper itself, an input multiplier, and an output amplifier.
The multiplier's input terminals are fed with the current Igx
and IE(l-x), respectively, where IE is constant and x is the
time- variable fraction of the input current, which is triangu
lar in this case. The purpose of the input multiplier is to
convert the temperature-independent input current into a
temperature-dependent one and compensate the tempera
ture dependency of the three sine-shaper differential
amplifiers. The current source IT is designed to deliver a
current that is the required function of the absolute temper
ature, so that the output currents of the input multiplier are
now IT(I-X) and ITx, respectively. Thus the voltage drop
between nodes A and B in the sine shaper is also a function
of the absolute temperature.
Differential amplifier 11 is the main waveform shaper
and differential amplifiers 12 and 13 correct the transfer
characteristic as already mentioned. Currents Ig and I g have
the desired sinusoidal waveforms but are still dependent on
temperature. The output amplifier eliminates this tempera
ture dependence and produces the output currents IAy and
Fig. 12 illustrates how the resulting characteristic of the
Authors
June 1983
Donald E. Morris
Don Morris studied electri
cal engineering at the Uni
versity of New Mexico
(BSEE degree in 1 967) and
Oklahoma State University
(MSEE degree in 1968 and
PhD degree in 1 970) before
': / Z ••••• joining HPin 1970. Hewas
/ it^^l I project manager for the
9825A Computer (about which he coauthored an
earlier HP Journal article), production engineering
manager, and later R&D section manager before
assuming his current responsibility as product as
surance manager. Don has taught digital logic at
Oregon State University and math and program
ming at local vocational education facilities. He has
his own twin-engine airplane and flies it whenever
he can. He is married, has two sons, and lives in
Corvallis, Oregon.
sine shaper is generated. VAB is the triangular input voltage
of the sine shaper. IQS— Iq4 indicates the transfer charac
teristic of differential amplifier 11. Zero slope for the cir
cuit's transfer characteristic can be achieved by linear
superposition of the transfer characteristics of differential
amplifiers 12 and 13, Igy-Igs and IQS-IQBThe resulting transfer characteristic of the circuit is
Is~ïs> also shown as a clean sinusoidal waveform in
Fig. 12.
The maximum deviation from an ideal sine wave is
±0.22%. Harmonic distortions are about -60 dB. However,
because of process tolerances, the mean value of the har
monic distortion observed in production is about —54 dB
with a standard deviation of ±2 dB for a temperature range
from 0°C to 55°C. Using the 1-GHz LSI process, the circuit
is suitable for frequencies up to 100 MHz.
Acknowledgments
Thanks to many people within HP's Santa Clara Division.
We particularly want to thank Santa Clara's mask designers
and product engineers for their cooperation, and Jim Grace
for many useful discussions.
Donald L. Morris
I Don L. Morris received a BS
degree in mechanical en| gineering in 1971 from
Bradley University and a BS
| degree in electrical en
gineering in 1973 from the
_ ,- ^ University of Illinois. After
1 "7- V^^B joining HP in 1973, he
\- -Jr. worked in both production
and R&D. He worked on the 82143A Printer and
coauthored an article about it in an earlier issue of
the HP Journal. Don was the HP-75 production
manager before leaving HP recently. A native of
Lincoln, Illinois, he lives in Corvallis, Oregon, is
married, and has two children.
Anthony S. Ridolfo
Born in Great Falls, Montana, Tony Ridolfo received
his mathematics education
alWabashColle9e(BAdegree in 1966), Ohio University (MS degree in 1968),
and Iowa State University
(PhD degree in 1975). After
* teaching mathematics and
computer science for a while as a college profes
sor, he joined HP in 1976 and has programmed
38 HEWLETT-PACKARD JOURNAL JUNE 1983
© Copr. 1949-1998 Hewlett-Packard Co.
firmware for a number of HP's calculators. He was
project manager forthe HP-75 Computer's operat
ing system before moving to marketing systems
engineering. Tony is a member of the American
Mathematical Society and the Society for Industrial
and Applied Mathematics. He serves on a local
math textbook adoption committee and is treasurer
for a local Boy Scout troop when he is not busy
refereeing youth soccer games, playing bridge or
golf, or skiing. He is married, has three children,
and lives in Philomath, Oregon.
Brian G. Spreadbury
! Brian Spreadbury joined
HP in 1976 with several
I years of experience in aviaI tion system design. He has
contributed to the mechan
ical design of the HP-85
i Computer's tape drive
¡(about which he coauthor
ed an earlier HP Journal
article), the 821 63A Video Interface, and the
821 68A Modem. Born in Aldershot, England, Brian
studied mechanical engineering at Bournemouth
College, England and received the H.N.C.-Mech
degree in 1962. He enjoys cross-country skiing,
fishing, canoeing, and hiking. Living in Corvallis,
Oregon, he is married and has two sons.
Sidnee Snell
:-ee Snell joined HP in
1 979 after earning a BS de
gree in electrical engmeermg at the University of
California at Davis. Besides
:ntributions to the
82168A MOOT
worked on low-cost plotter
hardware and firmware.
Sidnee is vice president of the Oregon chapter of
the Society of Women Engineers and lives in Corvallis. Oregon. She is married, owns two cats
enjoys gourmet cooking.
Dennis C. York
Dennis York is a project
manager for applications
software at HP's Corvallis,
Oregon facility. With HP
since 1 973, he was respon
sible for application
software used in the HP-65,
HP-67, and HP-41 Cal
culators, contributed to the
HP-41 's firmware, and was HP-IL product man
ager. Dennis has a BS degree in mechanical en
gineering awarded by Oregon State University in
1 972. He ¡s married, has three children, and lives in
Corvallis. He enjoys playing softball and basketball
when not checking out new computer programs.
Elizabeth Brooks
Born in Laramie, Wyoming,
Beth Brooks grew up near
Mt. Vernon, Virginia. She
received a BA degree in
electrical engineering from
Rice University in 1975 and
joined HP shortly after.
While working as an R&D
engineer at HP, Beth con
tinued her studies and received an MS degree in
electrical engineering from Stanford University in
1977. She was an R&D project manager for the
HP- 75 Computer, then a product engineering man
ager, and now is involved with increasing engineer
ing productivity. She is married (her husband is a
project manager at HP) and lives in Corvallis, Ore
gon.
! X C.
I * , Smith
smun
Rex
A native of Roseburg, Ore
gon, Rex Smith attended
Oregon State University,
earning a BSEE degree in
- Â » 1973, He also has an MSEE
. degree from Stanford Unit versity awarded in 1976.
J; Rex has been with HPsince
1973 and has contributed
to the design of the 4910G Open and Split Locator
and the HP-01 Watch. He was project manager for
the 821 68A Modem and 821 63A Video Interface
and managed the initial architecture and 1C de
velopment for the HP- 75 Computer. He currently is
managing portable data communications de
velopment. He served four years in the U.S. Navy
(primarily on Orion aircraft) and is involved in
church leadership. Married and the father of three
Computer. Bom in Astoria, Oregon. Ken spent
children, he lives in Corvallis. Oregon. His interests
include classical piano, watching college basket-
married, lives in Corvallis, Oregon, and enjoys bas
ketball, water skiing, camping, foreign travel, and
: aliening, and farRobert J. Livengood
eight yearts of his childhood in Brazil. He is
bridge playing.
Dean R. Johnson
-- HPsince 1969, Bob
Livengood has contributed
to a number of HP's
products — frequency
-"s. digital multime!.ers, and portable comput
ers such as the HP-75 — in
the areas of digital bipolar
i MOS ICs. He ¡s a proj
ect manager for portable computers at HP's Corval
lis, Oregon facility. Bob is a graduate of the Univer
sity of Missouri (a BSEE degree in 1969) and Col
orado State University (an MSEE degree in 1974).
He is married, has three sons, and when not in
volved with youth and church activities, spends his
free time landscaping his yard. Born in Brush, Col
orado, he now lives in Albany, Oregon.
Timothy F. Myers
• Tim Myers joined HP in
^ 1 979 after receiving a BSEE
degree from Montana State
University. He worked on
display and memory cir..- cuits for the HP-75 Com
puter with HP Laboratories
_l personnel and now is pro1 duction engineer for the
HP-75 at HP's Corvallis, Oregon facility. He is mar
ried and has several interests— photography,
holography, golf, softball, skiing, and exploring
Oregon's back country. A native of Montana, he
now lives in Corvallis.
Francis A. Young
Raan Young was born in
Pullman, Washington and
earned his BSEE degree in
1979 at the University of
Washington. He contrib
uted to the HP-75 Com
puter's operating system
and applications software
and ¡s an author of a paper
on bar-code translation for the HP-41 Handheld
Computer. Raan is a member of the IEEE, Living in
Corvallis, Oregon, he ¡s married and is interested in
electronics, music, photography (including video
techniques), and house remodeling.
Kenneth R. Hoecker
I A grad uate of 0 regon State
I University with a BSME de
gree awarded in 1972, Ken
Hoecker came to HP in
1977 after serving in the
U.S. Army for two years and
doing postgraduate study
in mechanical engineering.
He was a production en
gineer for the HP-01 Watch and joined R&D as part
of the original design team for the HP-75
Before joining HP in 1979.
Dean Johnson did digital
design of micropro
cessor-based products. At
HP. he is concerned with
implementing analog sys
tems in CMOS technology.
Raised in the Midwest.
I Dean attended the Univer
sity of Minnesota where he earned a BEE degree in
I973 and an MSEE degree in 1 979. He is a member
of the IEEE and active in his community's Big
Brother program. Living in Corvallis, Oregon, he
enjoys mountain climbing, hiking, and playing racquetball.
James R. Schwartz
Jim Schwartz started at H P
in 1971 after completing the
requirements for a BSEE
degree at San Jose State
University. He left HP in
1973 and rejoined the
company in 1977. He now
isanR&DengineerforHP's
i iw. -A. Corvallis, Oregon facility.
Raised in Chico, California, he lives in Corvalliswith
his wife and two soms. Until he was idled by an
injury, his outside interests included tennis and
fast-pitch softball.
David B. Patton
Dave Patton contributed to
the design of several pro
duction tools for the HP-75,
including the production
card recorder, before as
suming his current R&D
I. W position working on porta» ^—^ *) ble Pinters. He graduated
â„¢ - Ã" ,<'i ' from Purdue University in
1978 with a BS degree in electrical engineering
technology and worked for Purdue's aerospace
sciences laboratory before joining H P in 1980. Mar
ried, he lives in Corvallis, Oregon, and is interested
in church activities, music, gardening, motorcy
cles, bicycling, and skiing.
Thomas J. Arnold
Born in Manitowoc, Wis
consin, Tom Arnold at
tended the University of
Wisconsin at Madison
where he received a BS
degree in elec t r i c a l e n
gineering. He came to HP in
1 979 as a product engineer
and now is an R&D project
manager. Tom ¡s the author of a paper on CMOS
latch-up and lives in Corvallis, Oregon. He is in
terested in home computers and motorcycles and
enjoys downhill skiing and camping.
JUNE 1983 HEWLETT-PACKARD JOURNAL 39
© Copr. 1949-1998 Hewlett-Packard Co.
Billy E. Thayer
I Bill Thayer began work at
HP in 1964 after serving in
the U.S. Air Force for ten
years. He worked on some
of HP's first transistorized
instruments and the printed
circuit boards in HP's first
desktop calculator, the
9100A. Bill helped design
the NMOS chip set used in the 9825A Computer
before moving to HP's Corvallis, Oregon facility to
design CMOS ICs. He designed parts of the
keyboard/clock and card-reader chips for the
HP-75 Computer. Born in Rushville, Nebraska, he
now lives in Corvallis. He is an elder in the Presbyte
rian church, married, and has three sons. He enjoys
water and downhill skiing and camping in the Ore
gon outdoors.
Michael Fleischer
| A native of Essen, Michael
: Fleischer joined HP in
i Boeblingen in 1980 shortly
after receiving his En
gineering Diploma from the
Ruhr University in Bochum.
After working on several
hardware problems, he
joined the 81 12A team and
was responsible for completing the timing and
slope sections and the output amplifier. Michael is
lecturing in electronics at a Stuttgart college. He is
single and likes to spend his spare time backpack
ing, traveling in his camper, and playing the guitar.
Helmut Rossner
Shortly after receiving his
H electronics engineering
degree in 1969 at the En
gineering School in
Esslingen, Helmut Rossner
joined the HP pulse
generator design team at
I Boeblingen. His first projI ects were the 801 2A and
801 3A Pulse Generators. Later, he codesigned var
ious programmable data generators and precision
pulse generators. In the new series of pulse
generators and pulse/function generators, he was
project leader for the 81 1 2A. Hel mut is married and
has two daughters. He is keen on all kinds of tour-
ing, whether by bicycle, on skis, orón foot, but most
of all he is an enthusiastic mountaineer and is a
member of the German Alpine Club.
in seven patents. Stefan is married and lives in
Boeblingen. He has two sons, plays tennis and enjoys
outdoor activities.
Uwe Newmann
Adolf Leiter
Uwe Neumann earned his
Diplom Ingenieur (FH) at
the Engineering School in
Constance in 1978 and
joined HP GmbH as an R&D
engineerthesameyear. He
worked on the 8161A's
software and was later re
sponsible for the software
development of the 8112A Pulse Generator and
8116A Pulse/Function Generator. Uwe is married,
has a daughter, and enjoys walking, traveling and
cross-country skiing.
I Before joining HP, Adolf
Leiter studied electrical
engineering in Munich. He
joined HP's Boeblingen In
struments Division in 1979
| as an R&D engineer and
' took over the responsibility
for the Rate 1C design from
Stefan Traub. He is pres
ently working at HP Laboratories in California. Adolf
is single and is living in Cupertino, California at the
moment. He enjoys tennis and skiing and outdoor
activities in general.
Christian Hentschel
•PSBKM^M Before joining HP in 1972,
• Christian Hentschel was ¡nBPH! volved at the Institute of
IB •• Theoretical Electrotechnol^ ogy, Aachen University, in
research on thin-film, highfrequency lumped circuit
elements, a topic of his doc
toral dissertation. At HP's
Boeblingen Instruments Division, he worked on the
8082A 250-MHz Pulse Generator, was project
leader for the 8080A configurable 1 -GHz/300-MHz
Pulse/Data Generator, and subsequently became
technology section manager. Christian is married
and enjoys photography, tennis, sailing, windsurf
ing, and other sports.
Horst Schweikardt
¿t?-- After finishing his elec^— . *^SH|^t trotechnical studies at the
A University of Stuttgart,
K _^ _n y| Horst Schweikardt joined
Kf **, ifj*' HP GmbH in 1972. He was
I involved in the develop• •v-x· ^^^ ment of low-cost oscilloI /• • scopes, was project leader
H^V Jr ^^1 for the 1223A Storage
Oscilloscope and 21 4B Pulse Generator, and then
joined the 1C team where he worked on the Rate
and Snake designs. Horst is married, has a oneyear-old daughter and enjoys bowling, photog
raphy, and mountain hiking.
«r,
Stefan Traub
1 Stefan Traub joined HP in
1969 with previous experiI ence in R&D. He worked as
a hardware development
' engineer on a variety of proj, ects, including medical instrumentation and oscillo
scopes. He was project
leader for the 1 5239A Dis
play Unit for medical monitoring and the 1 222A Oscil
loscope. He contributed to the design of the Snake 1C
and the Booster 1C and was project leader for the Rate
1C until 1979. His work has resulted
Hewlett-Packard Company, 3000 Hanover
Street, Palo Alto, California 94304
H E W L E T T - P A C K A R D
J O U R N A L
Volker Eberle
^^^^^^^ Volker Eberle earned his
1 Diplom Ingeniuer at the
Technische Universitat,
Stuttgart in 1975. He then
i joined HP, starting on the
8092A Delay Generator, a
module of the 8080A Con
figurable Pulse/Data
' Generator. During the final
phase of the 8080A development, he became proj
ect leader, guiding the system into production. He
then moved on to the Booster 1C as project leader.
Volker is married and has two daughters. His lei
sure activities are gardening, model railroading,
and playing the accordion.
Bulk Rate
U.S. Postage
Paid
Hewlett-Packard
Company
JUNE 1983 Volume 34 . Number 6
Technical Information from the Laboratories of
Hewlett-Packard Company
Hewlett-Packard Company, 3000 Hanover Street
Palo Alto, California 94304 U.S.A.
Hewlett-Packard Central Mailing Department
Van Heuven Goedhartlaan 121
1181 KK Amstelveen, The Netherlands
Yokogawa-Hewlett-Packard Ltd., Suginami-ku Tokyo 168 Japan
Hewlett-Packard (Canada) Ltd.
6877 Goreway Drive, Mississauga, Ontario L4V 1MB Canada
CHANGEOF ADDRESS:
0200C2Q7G7&&4bLAC£CAOU
MR C A bLACKBURN
JOHN HÃœPMNS UNIV
APPLIED PHYSICS LAb
JOHNS HOPKINS RD
L A U R E L
M U
2 0 7 0 7
To change your address or delete your name from our mailing list please send us your old address label. Send
changes to Hewlett-Packa'rd Journal. 3000 Hanover Street. Palo Alto, California 94304 U.S.A. Allow 60 days.
5953-8512
© Copr. 1949-1998 Hewlett-Packard Co.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement