DATAMATH  CALCULATOR  MUSEUM

Rockwell Calculator Integrated Circuits

It must have been a shockwave for a handful of Japanese companies, when Texas Instruments demonstrated in 1967 with the Cal-Tech feasibility study for the first time the application of LSI (Large Scale Integration) chips for portable, battery-operated calculators. American companies not only invented the transistors and ICs (Integrated Circuits) used with electronic calculators in the second half of the 1960s, but they also had a lead in the design and manufacturing of chips using PMOS (p-channel Metal–oxide Semiconductor) technology. On the other hand, Japanese companies like Canon, Casio, Sanyo, Sharp and Toshiba almost completely owned in 1967 the market of electronic desktop calculators and successfully developed, manufactured and sold their products designed around transistors and ICs.

The major manufacturers of electronic desktop calculators acted swiftly and locked into relationships with chips manufacturers capable of designing and producing PMOS LSI devices. In a first wave, around 1969/1970, three successful results of these collaborations hit the market:

The second wave of partnerships between Japanese desktop calculator manufacturers and American chip manufacturers can be identified with:

The winning team in the race to the first "small electronic desktop calculator" was obviously Sharp with the QT-8D using a chipset of five devices jointly developed with North American Rockwell’s Autonetics division: NRD2256, AC2261, DC2266, AU2271 and the clock driver CG1121. Sharp reused this chipset with the ELSI-8 (EL-8), a compact, battery-operated calculator introduced in November 1970, its offspring ELSI-8A and the QT-8B variation of the QT-8D. An upgraded design of the Rockwell chipset supported a User Memory function deployed in the battery-operated EL-8M and the EL-160 small electronic desktop calculator.

First Rockwell chipset produced for Sharp

The 1^st Generation with four LSI chips and one MSI chip still used the "old-style" Rockwell device markings. The complexity of the chips approached 1,000 transistors, commonly known as the threshold for LSI (Large Scale Integration) ICs. In March 1969, Sharp reached an agreement for technological cooperation with North American Rockwell and made an official announcement on the construction of a semiconductor plant. Construction of the semiconductor plant, the Central Research Laboratories, was completed in September 1970. Initially the semiconductor plant in Tenri was working only on the second half of the manufacturing process. It imported LSI chip wafers from North American Rockwell and then placed those chips in packages and finished them as LSIs. In 1972, the plant started operation of the first-half processes and became a facility for complete manufacture of LSIs.

These chips are manufactured in a 15 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Ceramic Quad In-line Packages (QIP) with 42 pins or a 10-pin Metal Header Package (TO-100).

While these chips sport on their Ceramics or Metal can packages the above-mentioned device marking, are the silicon chips inside the packages marked with 1000x – hinting to the original Part Numbering System used with North American Rockwell’s Autonetics division: Sequential numbers for each design, starting with 10001.

Second Rockwell chipset produced for Sharp

The massive contract from Sharp, $30 million US$ in 1969 or around $250 million US$ inflation adjusted in 2025, was Rockwell’s entry into the MOS LSI business, but the contracts with Sharp covered only delivery of these chips till January 1971. Under an agreement between the two companies, Sharp would manufacture them as soon as in 1971 in Japan with technical assistance from Rockwell – jumpstarting Sharp’s entry as provider for MOS LSI Chips.

Rockwell on the other hand understood the opportunities of their PMOS LSI technology and experience to mass-produce chips very competitively for consumer electronics and started the design of their own "Standard Building-Block Circuits for Calculators". This set of modular building blocks consisted of four Standard Building Blocks that needed an additional customer-programmed MOS ROM (Read-only Memory) and interfacing logic for the Input (Keyboard) and Output (Display and/or Printer) Interfaces to complete the electronics of a desktop calculator. The architecture of the Building Blocks allowed even for printing or programmable desktop calculators with 10 to 16 digits of capacity and when the North American Rockwell Microelectronics Company announced on January 13, 1971 the chip set, they dreamed big.

The four Standard Building Blocks or Calculator Chips were nick-named C, A, T, and S based on their functionality:

The C Chip

The C chip serves as master control over other chips in the standard calculator chipset, contains status registers, senses the status of selected registers, the status of flip-flops as selected by ROM address, and the status of keyboard inputs such as the decimal point. Using this information, the C chip generates output commands to coordinate the activities of the other calculator chips.

The A Chip

The A chip includes circuitry for input/output logic control of the data entered into various registers. Specifically, the A chip contains a serial decimal adder, a ROM address register, a two-level stack for return address storage, register status determining logic, X and Y register storage and shift logic, and operand inter-register transfer logic.

The T Chip

The T chip contains the arithmetic registers, the memory registers for four numbers (of length 10, 12, 14, or 16 decimal digits, depending upon the model) in four memory registers and interfacing logic to a keyboard.

The S Chip

The S chip contains logic for auxiliary register Z, exponents, and binary logic. It also contains the exponent register and the master time-base generator. Two hard-wired inputs to the S chip cause the time-base to generate clocking for word lengths of 10, 12, 14, or 16 digits.

ROM Circuits

The ROM circuitry is implementing the microprogram for the set of standard calculator chips. In addition, the ROM contains the 4-phase clock for the calculator. The ROM is available with three different capacities – 256, 384, or 512 words of 17 bits, each. There may be as many as four micro instructions per word.

Input/Output Chips

Depending upon requirements of the particular calculator system, from one to four input/output interfacing chips are necessary.

Programmability

Additional dynamic storage is required for a programmable calculator. From one to three additional chips are needed for those storage, depending upon the lengths of the program (number of instructions) and the type of instructions which are to be implemented.

Register Expandability

A "TT" chip may be substituted for the T chip, and with the addition of one or two “M” chips either eight or sixteen independent full registers may be obtained. Each M chip contains eight memory registers, including decimal point and sign of each word.

As of today, we discovered less than ten designs centered around the C-A-T-S Standard Building-Block Circuits for Calculators and typical designs make use of six or seven PMOS LSI chips. Coincidence or not, these designs look very, very similar to an earlier chipset with six PMOS LSI chips used in Sharp’s CS-363 and CS-242 desktop calculators, with observed Date codes as early as November 1970. We assume that "tit for tat" translated in 1970 into:

Here at the Datamath Calculator Museum we differentiate between customer-specific and generic calculator chips and chipsets and separate in our overview the original Sharp chipset from the C-A-T-S chipset, accordingly.

These chips are manufactured in a 15 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic or Ceramic Quad In-line Packages (QIP) with 42 pins and metal lids.

First Rockwell chipset for Desktop Calculators

We assume that the C-A-T-S Standard Building Blocks are based on the Second Rockwell Chipset produced for Sharp with the Part Numbers 10061 - 10066.
These chips are manufactured in a 15 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic or Ceramic Quad In-line Packages (QIP) with 42 pins and metal lids.

Third Rockwell chipset produced for Sharp

The 3^rd Generation introduced end of 1971 combined the same functionality into just two LSI chips. Starting 1972, Sharp manufactured these chips under license from Rockwell in their Tenri semiconductor plant.
These chips are manufactured in a 10 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic Quad In-line Packages (QIP) with 42 pins and metal lids.

Second Rockwell chipset for Desktop Calculators

We assume that the these chips are based on the Third Rockwell Chipset produced for Sharp with the Part Numbers 10572 and 10573.
These chips are manufactured in a 15 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic Quad In-line Packages (QIP) with 42 pins and metal lids.

Rockwell chips for Desktop Calculators with LC-Display

North American Rockwell pioneered the development of Liquid-Crystal Displays (LCDs) and introduced already in 1972 "almost" single-chip calculator circuits using external display drivers and LCD modules. Rockwell manufactured with these components calculators for various OEMs like Lloyd's, Rapid Data and Sears. Rockwell changed their Part Numbering System in Fall 1972 and used 15xxx numbers for ROMs (Read-only Memorys) and devices integrating ROM-structures. ROM-based calculator chips were consequently renamed from 10xxx to 15xxx numbers.
These chips are manufactured in a 10 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic Quad In-line Packages (QIP) with 42 pins and metal lids.

Rockwell PPS-4 Product Family

Work-In-Progress.

Rockwell PPS-4/2 Product Family

Work-In-Progress.

First Rockwell single-chip calculator circuits

Rockwell introduced in 1973 with the 15330 - later renamed to A1030 and even later to A4130 - their first single-chip calculator circuit. The architecture of the 15330 is very similar to Rockwell's PPS-4 (short for Parallel Processing System, 4-bit) system. The development of this groundbreaking calculator chip prepared the design of the later PPS-4/1 (short for Parallel Processing System, 4-bit, Single-chip) system, a Digit Processor competing directly with the Texas Instruments TMS1000 Microcomputer introduced in October 1974. The programmable ROM (Read-Only Memory) of the 15330 is with 512 x 8 Bits capacity much smaller than the TMS1000 ROM with its 1,024 x 8 Bits capacity, while the RAM with 48 x 4 Bits is almost on par with the TMS1000 sporting 64 x 4 Bits.
The 15330 and its derivatives 15332, 15334 and 15336 (aka  A1032, A1034 and A 1036) are manufactured in a 10 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic Quad In-line Packages (QIP) with 42 pins and metal lids.

Later Rockwell single-chip calculator circuits

Work-In-Progress.

First Rockwell single-chip calculator circuits with LED Direct-Drive Capability

Rockwell introduced in June 1974 with the A5300 Product Family the first single-chip calculator circuits with LED (Light-Emitting-Diode) Direct-Drive capability. Earlier chip designs, like Texas Instruments' TMS0100 Product Family, lacked any display drivers and left the choice of display technology and the associated drivers to their customers. Both improvements of the efficiency of LEDs and the design of beefier output drivers allowed at least omitting the Segment Drivers starting around 1973 with designs like the Rockwell 15330. Looking closely into the A5300 reveals a design very similar to Rockwell's PPS-4/1 (short for Parallel Processing System, 4-bit, Single-chip) system, a Digit Processor competing directly with the Texas Instruments TMS1000 Microcomputer introduced in October 1974. The programmable ROM (Read-Only Memory) of the A5300 is with 504 x 8 Bits capacity much smaller than the TMS1000 ROM with its 1,024 x 8 Bits capacity, while the RAM with 48 x 4 Bits is almost on par with the TMS1000 sporting 64 x 4 Bits. As a result is the die size of the A5300 with around 4.0 mm x 3.8 mm significantly smaller than the TMS1000 which measures about 5.1 mm x 5.1 mm. The CPU architecture of the PPS-4/1 was designed with electronic calculators in mind and consequently were the software engineers able to squeeze a fully-featured five-function calculator with Memory into the small program memory of the A5300.
The A5300 is manufactured in an 8 um metal gate PMOS process with enhancement mode transistors used for both gates and loads and using Plastic Quad In-line Packages (QIP) with 42 pins and metal lid.

Later Rockwell single-chip calculator circuits with LED Direct-Drive Capability

Work-In-Progress.

Rockwell PPS-4/1 Product Family

Work-In-Progress.

Rockwell Product Marking and Part Numbering System

The numbering scheme used to mark Integrated Circuits manufactured by North American Rockwell changed multiple times and here at the Datamath Calculator Museum we focus on the timeframe between 1969 and about 1979, the peak of Rockwell's engagement in supplying MOS LSI (Metal-oxide Semiconductor Large Scale Integration) chips for electronic calculators.

Part numbering of Integrated Circuits is a complex topic, and one needs to look into it under different angles. The package of the housing contains one set of information while the silicon die inside the encapsulations holds another set of information. The package, both its top and bottom sides typically bear a manufacturer identification, product identification number, device revision, manufacturing date, manufacturing/packaging site, temperature rating, speed rating, and more. None of this labeling information is standardized and most companies create their own coding scheme. Just with MOS LSI calculator chips, Rockwell changed their part numbering at least twice within three years…

The silicon die itself contains markings from its photomasks that reference to the underlying schematics of the chip, its layout artwork, revision/process version and other information. Usually, this information is encrypted in just a few letters and numbers, real estate on silicon dies is precious.

When North American Rockwell’s Autonetics division surprised in 1969 the industry with the famous Chipset for Sharp's groundbreaking QT-8D and EL-8 calculators, they simply named the five devices on the silicon die with sequential numbers 10001 to 10005. The package chips on the other hand were named NRD2256, AC2261, DC2266, AU2271, CG1121 – with the characters hinting to the functionality of the respective chips, NRD2256 as an example refers to the "Numeric Read-in and Display" Building Block of the calculator brain while the meaning of 2256 is not yet understood. Being an exclusive custom design for Sharp, Rockwell had to mark the packages per Sharp's request. Additionally have these earliest Rockwell chips only a 4-digit Date Code and a small Rockwell logo printed on the metal cap of the 42-pin Quad-In-Line ceramics package. The silicon die of the NRD2256 chip identifies simply as 10001.

Rockwell kept the sequential product identification numbering scheme of MOS LSI chips for some time and the number itself gives a good understanding of the timeframe of the chip development. Some examples to illustrate this idea are the first generation of the "C-A-T-S Standard Building-Block Circuits for Calculators" introduced in October 1970 and using the numbers 10061 to 10066, the second generation introduced in Spring 1971 and using 10177 to 10182 and the first chipset for LCD calculators introduced in Summer 1972 (10362).

With the second generation of the C-A-T-S Building Blocks, Rockwell introduced a change of the product identification numbering system and marked ROM (Read-only Memory) devices with numbers starting with 15000 and most other chips with a programmable ROM defining its functionality like Rockwell's early 15332 single-chip calculator circuit adopted the 15xxx product identification numbering. With the introduction of the PPS-4 in Fall 1972, Rockwell changed the product identification numbering of ROM devices again, using a leading “A” character before the number. While the PPS-4 CPU was simply called 10060 and its RAM chip was marked 10532, used the initially available ROM devices A05xx product identification number markings. Rockwell adopted this scheme to some calculator chips, too and the above mentioned 15332 chip was renamed mid of 1973 to A1032 and the 15350 single-chip calculator circuit for example was marked as A1150. During a transition time, the chips used both markings to avoid/create confusion.

These changes are clearly a sign of Product Managers and Marketing Managers having now the responsibility for device marking and not the R&D Teams anymore. In the pursuit to have the best numbering scheme, Rockwell changed the nomenclature for calculator chips with the introduction of the A4800 single-chip calculator circuit again and renamed the 15470 chips to A4000 and the 15350/A1150 to A4350 – finally going from "early designs = low 10xxx numbers" to "high functionality = high A4xxx numbers". The power of marketing.

A typical Rockwell calculator chip sold in Spring 1973 would have these markings on the metal cap of its housing:

15330PC
7312
Rockwell logo

The same calculator chip sold in Fall 1973 would read:

15330
A1030PD
7341
Rockwell logo

And in 1976 the very same chip showed up in a Rockwell 330 calculator with the markings:

15330
A4130CA
7629
Rockwell logo

And in this case the chips are using the same 42-pin QIP (Quad In-line Package), its package code was changed over the course of time from "P" to "C".

In the past years we assembled a list of Rockwell package markings located on early Standard Building Blocks for calculators, single chip calculator circuits, PPS-4, PPS-4/2 and PPS-4/1 devices and single-chip designs for electronic toys – trying to reverse-engineer the logic behind the numbering scheme used in the 1970s for MOS LSI devices. Please find below a summary of our understanding of Rockwell’s early MOS LSI device marking codes as of today and don’t hesitate to get in touch with us with your valuable insights. Thank you.

Early North American Rockwell's LSI (Large Scale Integration) chips using PMOS (p-channel Metal–oxide Semiconductor) technology are using three markings on top of the chip packaging:

PxxxxYZ Product Identification (...1974)
yyww Date code
Rockwell logo Manufacturer Identification

The Product Identification Number xxxx is a unique part number for each device, the Package Type is encoded in a single character Y following the part number and the Photomask Revision is encoded in a second character Z. The leading letter or number P refers to the Process Technology of the chip.

Starting in 1974, North American Rockwell used the first digit of the Product Identification Number to encode the Product Category c and the second character for the Product Family f:

PcfxxYZ Product Identification (1974...1978)

North American Rockwell Product Category changed in 1978 the coding of the Product Identification slightly, omitting the Package Type and introducing a Temperature Range t marking, separated by a "-" and switching the Photomask Revision from letters Z to numbers z:

Pcfxx-tz Product Identification (1978...)

Process Technology (P):

Product Identification Number (xxxx):

The Product Identification Number xxxx went over the course of the 1970s through four iterations:

Package Type for Quad In-line Packages (Y):

Photomask Revision (Z,z):

Temperature Range (t):

Product Category (c):

Note: The Product Category numbering scheme has some exclusions due to grandfathering Axxxx numbers from the PPS-4 Product Families and renaming of early calculator chips from their 15xxx to Axxxx numbers. We identified as of today:

Product Family (f):

Electronic Calculators:

Misfits:

Electronic Games:

PPS-4/1 Product Family:

Rockwell PMOS Process Timeline

Rockwell Quad In-line Package

When Sharp introduced in October 1969 with the QT-8D the World's first small electronic desktop calculator, it was using a Chipset of five LSI (Large Scale Integration) Integrated Circuits jointly developed with North American Rockwell’s Autonetics division. Four of the five chips were packaged in 42-pin Quad In-line Packages (QIP) with their prominent staggered footprint. Rockwell stayed this at first glance unusual package committed for most, if not all, of its later calculator chips and due to a technology exchange with Sharp it quickly gained popularity in Japan. General Instrument / Sanyo adopted the package with a smaller pin count, NEC added a larger 64-pin variation and Rockwell added a 52-pin version to its portfolio. As of today we know variations of the QIP with 36 pins, 42 pins, 52 pins and 64 pins.

Rockwell introduced the 42-pin package in 1969 with a hermetically sealed white ceramic body, gold-finished bottom and top metal caps and gold plated pins. In 1972 the design was cost-optimized, using a 42-pin lead frame with tin plated pins, a grey plastic encapsulation and tops and bottom metal caps. In 1977 the final design was of the 42-pin QIP was introduced, omitting the metal caps of the housing and mimicking the setup of standard Dual In-line Packages (DIP). This style of plastic-encapsulated lead frames was introduced for DIP already in the 1960s and Texas Instruments used 28-pin and 40-pin DIP for calculator chips already in 1970, leading to the question:

It is all about power distribution on the chip! Early calculator chips in the 1970s were manufactured in PMOS (p-Channel MOS) technology and the substrate of the small silicon dies needs to be connected to the most positive voltage, usually labeled V_SS. The PMOS output transistors of these chips are "high-side" switching, meaning they conduct to V_SS when activating a digit or segment output. In typical calculator applications, the output currents of one output are in the range between 1 mA (external digit and segment drivers) and about 5 mA to 10 mA (integrated segment drivers). When displaying an "8" with the decimal point lit, the output currents of the eight segment drivers might add up to around 50 mA, tenfold higher than the typical supply current of such a calculator chip.

The semiconductor process technology of the 1970s did allow for only one metal layer to route all the signals on the silicon chip and special care needed to be taken to connect the substrate properly to V_SS and distribute the V_SS pin to all segment driver output transistors. One neat trick in the toolbox of the team creating the layout of the chip topology was using the exposed die paddle of the lead frame as an additional layer to route the high-current V_SS power path. Looking for example at an overlay of the actual pin connection of the TMS0850 single-chip calculator circuit and its chip layout helps to explain this approach:

In this example the V_SS power distribution is supported with a "short cut" the under the silicon chip through bonding the 12 o'clock and 6 o'clock pads of the die to the exposed die paddle. Looking at a typical lead frame of a 28-pin DIP shows the thin paddle along the center axis of the package of the plastic-encapsulated lead frame.

The earliest LSI MOS chips still used ceramic packages and their design had a huge advantage with respect to power distribution compared to these plastic-encapsulated lead frame setups: The metallized surface of the cavity for the silicon die could be accessed from all sides, not only from the 12 o'clock and 6 o'clock positions, allowing for much higher flexibility in the routing the high-current V_SS power path.

North American Rockwell made in the 1970s proper use of the additional degree of freedom in the distribution of the V_SS power path and even omitted in some designs the power ring of the metal layer, potentially reducing the chip size and hence increasing manufacturing yield and chip cost. With the introduction of the cost-optimized QIP in 1972, they added a rather large, circular chip carrier under the 42-pin lead frame allowing not only to connect any V_SS pin of the lead frame to the substrate but even to connect every position of the silicon die to the metal disc.

Looking for example at an overlay of the actual pin connection of the B5000 single-chip calculator circuit and its chip layout explains Rockwell's approach:

Note: The B5000 is a single-chip calculator circuit with LED Direct-Drive capability and has the output transistors of its digit drivers D1 to D9 connected to Pin 39 (V_DD).

If you have additions to the above article please email: joerg@datamath.org.

© Joerg Woerner, March 1, 2025. No reprints without written permission.