Datamath Calculator Museum DCM-50A (Platform)

So what is it all about this one-of-its-kind DCM-50A Platform?

Texas Instruments introduced the TI-2500 Datamath formally on September 21^st, 1972 and here at the Datamath Calculator Museum we celebrate the 50^th Anniversary accordingly.

When Texas Instruments invented with the TMS1802 – later renamed to TMS0102 - the single-chip calculator circuit and announced it on September 17^th, 1971, they introduced a highly integrated device providing the following features:

The chip was designed to operate with very few external components and connects to a switch matrix keyboard, features fully decoded segmented outputs and needs only simple drivers for the common display technologies. In hindsight most important characteristics of the TMS1802 was its groundbreaking architecture with most of the circuit designed using programming logic array (PLA) techniques, meaning that not only the calculator algorithm in the 320 Words Read-only Memory (ROM) could be defined with changing just a single photo mask in the manufacturing process, but allowing flexibility to adopt with the same manufacturing step:

Texas Instruments marketed this first generation single-chip calculator circuit as TMS0100 family and the famous Bowmar 901B was based on the TMS0103 while the earliest TI-2500 Datamath, Minimath, TI-3000 Deskmath, and TI-3500 are using the TMS0110, TMS0111, TMS0109, and TMS0106, respectively. Even TI’s first "Slide Rule" calculator, the SR-10 introduced in November 1972 features with the TMS0120 a member of the TMS0100 family. Subsequent designs kept the overall architecture with just minor enhancements, paving the way for the TMS1000 Microcomputer introduced in October 1974 with the SR-16 and we know as of today:

The TMS1000 "computer-on-a-chip" grew in the Seventies into a large portfolio with different packages, ROM and RAM sizes, manufacturing processes or additional peripherals and could be considered the most successful 4-bit microcontroller enabling products like Texas Instruments’ very own Speak & Spell educational toy. Its architecture found its way into many later calculators, like the TI-30, TI-35 or TI-55-II.

With many of the single-chip calculator circuits either designed for Texas Instruments' proprietary applications or developed for OEM manufacturers, available documentation of their features is rather scarce and the Datamath Calculator Museum is committed to provide as much information as possible to the calculator enthusiasts and hence we designed the DCM-50A Platform.

Characterization of Single-chip Calculator Circuits

Core idea of the DCM-50A design is providing a generic platform to access all features of a calculator based on Texas Instruments’ single-chip calculator circuits and the schematics and layout of the printed circuit board features accordingly:

Operating physical calculator chips in the DCM-50A platform instead of the actual calculators does not only reveal possibly unused functionality of the devices, it even demonstrates often some peculiarities of the implemented algorithms, too. Some of our first discoveries:

Please learn more about our findings and observations in the dedicated DCM-50A pages:

The Technology Section of the Datamath Calculator Museum features under Calculator Chips "Datasheets" of the mentioned product families.

Reverse-engineering of Single-chip Calculator Circuits

For an even better understanding of Texas Instruments' single-chip calculator circuits, the DCM-50A platform includes some means to peek inside the devices. When Ken Shirriff developed the amazing JavaScript Simulator for the TMS0800, he relied on the United States Patent Application US3934233A describing the TMS0800 architecture and its program code example. While typing the program code from the badly blurred copies of the patent application was tedious and error prone work, took implementing the odd Sinclair Scientific based on the customer-specific TMC0805 a completely different approach: John McMaster de-capped and photographed an actual TMC0805 integrated circuit and the program code, operation masks, single-digit constants and the additional instructions of the instruction sets were visually retrieved from the enlarged die photos and implemented in the JavaScript Simulator, accordingly.

The US3934233A patent application describes in one chapter so-called test circuitry for the TMS0800, basically some shift registers to read out the 320 program words in a TI-specific test setup. Challenging is the timing of the clock and latch signals for these shift registers, the TMS0800 derives a four-phase clock from its internal clock oscillator and some additional external circuitry is necessary to reconstruct the four phases PHI-1 to PHI-4.

Kudos to Ken Shirriff and his team, comparing the electronically dumped program codes of the respective chips and the source codes used with the JavaScript Simulator reveals a perfect match. Okay, to be more precise: The program code from the patent application actually matches the TMS0801 used with the Canon LE-84 and not the TMS0803 used with the TI-1500 and TI-2500-II.

In addition to the described features used to characterize single-chip calculator circuits, the DCM-50A platforms adds certain means to assist in reverse-engineering them:

All signals are made available on two connectors matching the pinout of the Digilent Digital Discovery™, a combined logic analyzer, protocol analyzer, and pattern generator instrument but can be used with any external instrument compatible with 3.3V or 5V signals.

Here at the Datamath Calculator Museum we are using both the Digilent Digital Discovery and a breadboard setup consisting of a MicroNova Mercury 2 Development Board based on the Artix-7 FPGA (Field-programmable Gate Array) interfacing with the DCM-50A Platform and connected to an Arduino Nano Every for housekeeping.

Such a setup allows retrieving the 320 program words of a TMS0800 single-chip calculator circuit in less than 1 second – compared to hours for typing the program codes from a patent application or days for de-capping, photographing, and retrieving the bits from an actual chip.

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

© Joerg Woerner, September 21, 2022. No reprints without written permission.