previous | contents | next
As technology has evolved, the fundamental building blocks of computer systems have become more complex. Standard building blocks have evolved from circuit-level components (e.g., discrete transistors and diodes), to gates and flip-flops (SSI, or small-scale integration), to register-transfer-level chip sets (MSI, or medium-scale integration), to PMS-level chip sets (LSI, or large-scale integration). When semiconductor technology had evolved to the point that a whole processor or a whole computer could be implemented on a single chip, a conceptual barrier was breached. No longer were computers expensive and to be used in a centralized facility requiring constant attention. Rather, computers could be used to replace random logic in dedicated applications. Not only would these single-chip processors provide more logical power, but they were also cheaper and easier to use than random logic.
Thus the microprocessor1 recreated the evolution of computer architectures that had previously been blazed by mainframe computers. The architecture evolution occurred over a shorter time frame, since the experiences of the mainframe manufacturers were available to study and technology was evolving rapidly, allowing new machine generations every 2 years. The availability of low-cost processing power has allowed increased functionality in existing products (e.g., computer peripheral controllers, intelligent terminals, electronic scales, traffic control, and instruments) and has opened entirely new application areas (e.g., calculators, video games, appliance controllers, home computers, word processors, point-of-sale terminals). Indeed, the point has been reached where a semiconductor vendor can sell more processors in a single month than even the most successful minicomputer and/or mainframe manufacturer has been able to sell in the lifetime of a computer.
Sections 1 and 2 of Part 3 explore the two smallest classes of computer systems: monolithic microcomputers and microcomputers. Monolithic microcomputers are single-chip systems incorporating the processor, program ROM, variable RAM, and perhaps dedicated I/O. The intended market is low-cost and high-volume applications such as hand-held calculators, watches, video games, automobiles, and appliances. These architectures may be specialized and limited to the functional characteristics of the application. Microcomputers, on the other hand, are single-chip processors (or processors with a small number of chips) requiring external RAM and ROM chips. These are usually faster and more powerful than monolithic microcomputer systems, since the off-chip placement of memory and I/O frees gates for more complex instructions and wider data paths. As time passes and technology allows more gates per chip, microcomputer architectures evolve into the monolithic microcomputer class.
This section will illustrate the monolithic microcomputer class via the Texas Instruments TMS 1000, intended for hand-held calculator applications, and the General Instruments PIC1650, intended for control applications. Finally, the selection closes with a brief discussion of program-compatible monolithic microcomputer families. Heretofore the family concept was reserved for only the largest computer classes.
Texas Instruments TMS1 000
Because of their limited requirements, hand-held calculators were one of the first consumer-oriented products to take advantage of the emerging LSI technology. After the initial, ad hoc designs, calculators have been implemented as stored-program computers. Today, most calculators consist of a single LSI computer chip plus a display. And the adoption of LSI technology was very rapid. For example, the TI-2500 calculator introduced in 1974, had 119 parts, of which 82 were electronic in nature. By 1976, the TI-1200 consisted of only 22 parts, of which only the calculator chip (a complete computer with programs) and display were electronic.
This single-chip integration is made possible by the tightly specified user environment. Input (e.g., keyboard or magnetic card) and output (e.g., LED display, printer, or magnetic card) options are limited. Further, the input language (i.e., function per key) is also fixed.
Chapter 34 details the LSI MOS chip used in the Texas Instruments SR-16 calculator. The chip contains a microcomputer complete with a program ROM having 1,024 eight-bit words; a temporary storage RAM: input (from keypad); output (to control keypad scan and LED display); and an oscillator (clock).
The TMS1000 chip was designed to span a range of hand-held calculator products (from four-function up through simple memory calculators). Since the chip had to be customized with the ROM program appropriate to a product, other programmable features were included to improve the chip's flexibility. This programmability was provided by two programmable logic arrays (PLAs).
The output PLA converts five bits into twenty 8-bit output patterns in order to conserve program space. These patterns are specified by the calculator designer and represent different output patterns on a seven-segment display.
The second PLA is for instruction decoding. The chip provides 16 microinstructions, such as "gate register Y to ALU." Twelve of these microinstructions form the fixed instruction set. The remaining instructions can be formed by combining any of the 16 microinstructions (a feature found in early minicomputers like the
1Micro- here means physically small and does not necessarily imply that the processor is microprogrammed.
previous | contents | next