PAL introduction (was MITS 2SIO serial chip)

From: Michael Holley <swtpc6800_at_attbi.com>
Date: Wed Dec 26 20:28:37 2001

I scanned a few pages from a book I co-authored, I skipped the figures. If
any one wants to see a figure or picture I can put it a web site.
 -----------------------------------------------
Michael Holley
www.swtpc.com
-----------------------------------------------

>From Practical Design Using Programmable Logic
By David Pellerin and Michael Holley
Prentice Hall 1991

The First User-Configurable PLAs

Intersil Corporation and Signetics Corporation, both manufacturers of PROMS,
realized that the structure of National's DM7575/DM8575 mask-programmed PLA
device was ideal for a field programmable device. It became a race between
Intersil and Signetics to bring the first such devices to market. Both of
these companies called their proposed new devices FPLAs for field
programmable logic array.

In the June 2, 1975 issue of EE Times, Intersil announced the IM5200 Field
Programmable Logic Array (FPLA). Soon after, Signetics introduced its
825100. These devices were ahead of their time in many ways. The data sheet
for Intersil's IM5200 FPLA is shown in Figure 2.3.

These first FPLAs were powerful devices (the 825100 is still available,
under the new name of PLS100). They provided a reasonable number of inputs
and outputs, and were flexible enough in their design to be used for a wide
variety of logic applications.

While on the surface the IM5200 and 825100 looked very similar (the 825100
featured two more inputs than the IM5200), they were in fact quite different
in terms of their programming technology. Rather than use the well
understood fusible link technology for their new device, Intersil chose to
use a new type of programmable element that had been developed for their
PROM devices. This programming element was designed to improve the
programming yields over earlier programmable devices. The technology, called
avalanche induced migration, or AIM, utilizes an open base NPN transistor as
the programming element. To program an AIM element, a high current is forced
through the transistor from the emitter to the collector. This causes a
short from the emitter to the base, which leaves the transistor to operate
as a diode.

Unfortunately, the actual devices that rolled out of the Intersil's foundry
turned out to be unreliable, with low programming yields, and the Intersil
devices were not successful in the market.

The Signetics 825100, which utilized the fusible link programming
technology, was more reliable and more successful in the market. Another
factor in the success of the Signetics devices was the efforts of Napoleone
Cavlan who was, at that time, Signetics' Manager of Advanced Products. The
new devices were completely unfamiliar to most circuit designers and
required a much higher level of user education and promotion than the
earlier programmable devices.

Even with a high level of promotion, documentation, and applications
support, most digital designers chose not to use these first PLDs because
the devices were still perceived to be too difficult to use and because of
the risk of relying on the new technology for critical circuit elements.
Even with the reasonable reliability of the Signetics devices, the FPLA had
a relatively slow maximum operating speed (due to the two programmable
arrays), was expensive, and had a poor reputation for testability.

Another factor limiting the acceptance of the FPLA was the large package.
The Signetics part, for example, was contained in a 28-pin DIP (dual inline
package) that was over a half an inch wide. Today, 28 pins is standard for
PLCC (plastic leaded chip carrier) type packages but, in the 1970s, smaller
packages were the norm.

Physical and operating limitations aside, other difficulties were
experienced by designers who used the devices. The challenge in using the
FPLA devices stemmed from a conceptual difference between how the designers
had created circuits in the past and how they were forced to create them
using FPLAs.

Designs that were to be implemented in FPLAs had to be described in an
unfamiliar format that bore little resemblance to the familiar schematics or
Boolean equations that logic designers had used in the past. To design a PLA
circuit and program a device, the user would first convert the design into a
tabular form called H&L. A sample of the H&L format is shown in Figure 2.4.

The data from the form was then entered into a device programmer, such the
Data I/O Model 10 shown in Figure 2.5. This programmer was designed
specifically for the Intersil and Signetics devices and had a CRT and
built-in H&L editor. A major advance that this programmer had over the
previous PROM programmers was its ability to exercise the programmed part
functionally and compare its operation to the results calculated in the
programmer. This meant that faulty or misprogrammed parts could be quickly
identified and rejected.

GE's Associative Logic

Although Intersil and Signetics were the first companies to successfully
market field programmable logic devices, they were by no means the first to
develop them. As early as 1971, General Electric Company was developing a
programmable logic device based on the new PROM technology. The device was
developed by David Greer, then with GE in Syracuse, NY. GE's experimental
device improved on IBM's ROAM structure by providing an internal path for
OR-plane signals to directly reenter the AND plane. This allowed the use of
multilevel logic with no waste of I/O pins.

Late in 1971, an experimental MOS programmable logic device was completed at
the General Electric Research and Development Center in Schenectady, NY by
Gerry Michon and Hugh Burke. Not only did this device feature the improved
PLA-type logic array, but also used the floating gate UV-erasable technology
announced earlier that year by Intel. The GE device (shown in Figures 2.6
and 2.7) was actually the first erasable PLD ever developed, predating
commercially available EPLDs by over a decade. Researchers at General
Electric not only developed the first working FPLAs and EPLDs, but also
described and patented a folded array structure remarkably similar to the
folded arrays that began appearing in complex PLDs nearly fifteen years
later.

In 1974, under the terms of a patent and trade secrets agreement with GE,
Monolithic Memories began the development of a mask-programmable logic
device incorporating the GE innovations. The device was named the
programmable associative logic array, or PALA. The device (MMI part number
5760/6760) was completed in 1976, and could implement multilevel or
sequential circuits of well over 100 equivalent gates. In addition to its
advanced structure, the device was supported by a highly automated design
environment developed by GE. Designs were entered using Boolean equations
and converted automatically to a mask pattern. The system even included a
facility for test vector generation and simulation. While the MMI/GE device
was never marketed, it did serve as a model for PLDs produced by a number of
manufacturers in later years.

MMI's PAL

Although the Signetics devices enjoyed some success, PLDs didn't really gain
widespread acceptance until the late 1970s, when MMI introduced the PAL
device. After working with GE on the PALA device, MMI's first effort in the
design of their devices was to reproduce the Intersil and Signetics FPLAs. A
few hundred copies of the Signetics 825100 were produced for internal
evaluation, but these devices were never released. From this experience, and
MMI's earlier experience with GE, the PAL device was born. The new family of
devices was announced in the summer of 1978.

The project to create the PAL device was managed by John Birkner and the
actual PAL circuit was designed by H. T. Chua. Birkner had come from
Computer Automation, Inc., where he had developed a 16-bit processor using
80 standard logic devices. His experience with standard logic led him to
believe that user programmable devices would be more attractive to users if
the devices were designed to replace standard logic. This meant that the
package sizes had to be more typical of the existing devices, and the speeds
had to be improved. The new devices that resulted from this thinking were a
breakthrough and a huge success in the market.

The PAL devices utilized the now mature and reasonably reliable PROM fuse
technology and featured only one programmable array. This combination
resulted in a device with much faster operation than the earlier FPLAs.
Programming of the devices was simple since they were implemented in
industry standard packages and used well understood PROM fuse technology.

One factor in the success of the PAL devices was the high level of customer
support offered by MMI in the form of applications and user documentation
that served to demystify the design process. The PAL Handbook, written by
Birkner himself, provided a conceptual bridge between the discrete logic
methods of the past and the high-level design methods of the future.

One indication of the success of the PAL is that, ten years after the
introduction of the first 161_8 and 1688 PALS, these devices still made up
the majority of all PLDs used even though there were over 200 unique PLD
device types available at that time. In recent years, more flexible devices
such as the 22V 10 have begun eroding the 16L8/16R8 monopoly, but even these
devices are based in large part on the original Birkner/Chua design.
Similarly, the PALASM language remains a widely used method of PLD design
description. Most, if not all, of the current PLD design tools can trace
their heritage back to PALASM.

Another factor in the success of the PAL was the level of programming
support. The earlier FPLAs had suffered from the same programming problems
that plagued early PROMS; unreliable customer-built programmers and little
understood programming algorithms.

John Birkner and his crew were determined to make the PAL a success, so they
worked very closely with programmer companies to ensure that reliable
programming would be accessible to the device users. The first PAL
programmer was developed as a joint effort between MMI and Data I/O and
actually utilized two PALS as part of its construction. This meant that the
first prototype PAL programmer had to be bootstrapped by emulating the
function of its own PALs with PROMS and some additional TTL devices.

The first few years of PAL production weren't without their problems,
however. Production couldn't keep up with demand, and production yields
suffered. In addition, the programming algorithms (the specific programming
voltage and waveform specifications) hadn't been finalized, leading to often
unacceptable programming yields.

The PAL shortage had a serious impact on end-users of the time. Tracy
Kidder, in his book Soul of a New Machine, chronicles the development of the
Data General MV8000 computer. The designers of the MV8000 had gambled on the
new PAL devices, and were soon feeling the effects of limited supply. Kidder
describes a plaque awarded to Data General engineers working on the
redesigned Eclipse computer, another large design that used a number of PALS
in its construction. The plaque was named the PAL award, but where a
commemorative device should have been, there was instead an empty socket.

Digital Equipment Corporation also felt the effects of limited PAL supplies
when designing their new VAX 730 computer. To overcome the supply problems,
DEC's semiconductor liaison suggested to MMI that they should produce a mask
programmable version of the PAL for large quantity production. This was
done, and the resulting product was named the HAL for hard array logic. HAL
devices corresponding to all of the PAL devices were produced. These devices
were mask programmed as specified by DEC and other customers.

The HAL devices weren't without their own problems; in one early production
run, MMI neglected to properly label the many HAL silicon wafers that were
being produced for the VAX. According to Paul Franklin, then with MMI, it
became necessary to carefully examine every wafer with a microscope in order
to determine exactly which HAL devices they were composed of. The PAL
survived its early teething pains and exploded in the market, earning MMI
millions of dollars. The HAL turned out to be a marketable product as well
and is still available.

For their efforts, MMI rewarded both John Birkner and H. T. Chua with unique
bonuses-new cars every year. The October, 1982 issue of National Geographic,
in an article about silicon valley's glory years, pictured Birkner and Chua
with their fourth such annual bonuses-a Mercedes and a Porsche.

Another factor in the success of the PAL devices was PALASM, which stands
for PAL Assembler. PALASM was a computer program written by John Birkner
that converted design descriptions composed of Boolean equations directly
into programming data for a specified PAL device.

Boolean equations have been used since the early days of logic design (long
before the integrated circuit was conceived of) to express logic functions.
Boolean algebra is at the core of any university course in digital logic, so
it's a form of representation that is more familiar to engineers than
tabular forms.

PALASM was a simple computer program written in FORTRAN. The entire program
required only six pages of FORTRAN source code and was published in the PAL
data book. A typical PALASM design file is shown in Figure 2.8.

The simplicity of the PALASM language made it possible for programmer
manufacturers to implement the language directly in the programming
hardware, as was done in the Data I/O LogicPak, the Structured Design
programmer, and other similar PAL programmers.

-------------------------------------------------------------------
Received on Wed Dec 26 2001 - 20:28:37 GMT