The TMS1000: The First Commercially Available Microcontroller

We use a microcontroller without a second thought, in applications where once we might have resorted to a brace of 74 logic chips. But how many of us have spared a thought for how the microcontroller evolved? It’s time to go back a few decades to look at the first commercially available microcontroller, the Texas Instruments TMS1000.

Imagine A World Without Microcontrollers

The Texas Instruments Speak And Spell from 1978 was a typical use for the TMS1000.
The Texas Instruments Speak & Spell from 1978 was a typical use for the TMS1000. FozzTexx (CC-SA 4.0)

It’s fair to say that without microcontrollers, many of the projects we feature on Hackaday would never be made. Those of us who remember the days before widely available and easy-to-program microcontrollers will tell you that computer control of a small hardware project was certainly possible, but instead of dropping in a single chip it would have involved constructing an entire computer system. I remember Z80 systems on stripboard, with the Z80 itself alongside an EPROM, RAM chips, 74-series decoder logic, and peripheral chips such as the 6402 UART or the 8255 I/O port. Flashing an LED or keeping an eye on a microswitch or two became a major undertaking in both construction and cost, so we’d only go to those lengths if the application really demanded it. This changed for me in the early 1990s when the first affordable microcontrollers with on-board EEPROM came to market, but by then these chips had already been with us for a couple of decades.

It seems strange to modern ears, but for an engineer around 1970 a desktop calculator was a more exciting prospect than a desktop computer. Yet many of the first microcomputers were designed with calculators in mind, as was for example the Intel 4004. Calculator manufacturers each drove advances in processor silicon, and at Texas Instruments this led to the first all-in-one single-chip microcontrollers being developed in 1971 as pre-programmed CPUs designed to provide a calculator on a chip. It would take a few more years until 1974 before they produced the TMS1000, a single-chip microcontroller intended for general purpose use, and the first such part to go on sale.

It’s worth taking a moment to consider some of the terminology involved, because in 1974 the some of our current vocabulary was not necessarily in common usage. TI marketed the TMS1000 as a microcomputer because they saw it as an all-in-one computer without the need for extra peripherals. Today we’d consider a microcomputer to be an all-in-one general purpose computer such as the one you’re probably reading this on, following an unbroken line stretching back to the Altair 8800 in the same year, but back then the vocabulary was like the technology; in its infancy. The word microcontroller was in use by then for a computer with self-contained I/O, the Oxford English Dictionary has a citation from 1971 in an IBM technical bulletin, but it seems not to have settled as the universal definition. By comparison the more modern “System on chip” or SoC refers to a general purpose all-in-one computer chip that presents its internal bus to the world rather than a set of I/O lines or peripherals as you would find in a microcontroller.

Just How Simple Does A Microcontroller Need To Be?

The internal architecture of the TMS1000.

The TMS1000 then was the first commercially available microcontroller. But what kind of chip was it? There were four variants in the original range, all sharing the same 4-bit processor with a Harvard architecture, and sporting different numbers of I/O lines and ROM and RAM sizes. The TMS1000 and TMS1200 families had 8192 bits of program ROM and 265 bits of RAM, while the TMS1100 and 1300 families had double those figures. There were versions with high-voltage-tolerant outputs for driving vacuum fluorescent displays, and  they were available in 28-pin and 40-pin packages. Internally it sports an extremely simple architecture by today’s standards, without the banks of registers or pipelining you’d expect from more recent designs. It has none of the vast array of peripherals that you’d find on a modern microcontroller, but its I/O latches are supplemented by a simple programmable logic array. This would have been used as an encoder or decoder, an example in the data sheet uses it as a 7-segment display decoder.

The complex process of TMS1000 software development.
The complex process of TMS1000 software development.

You would not have found this chip for sale in single quantities for experimenters, because its on-board mask ROM could only be programmed at the point of manufacture by TI. Thus all coding was performed in a simulator on a time-sharing mainframe operated by TI. This would generate a deck of punch cards which after a very complex debugging and testing process would be used to generate the masks for ROM coding. The microcode could even be modified to order by TI, resulting in possible extensions to the device’s 43 instructions. The mask-programming also means that any TMS1000 you find today will still contain whatever software it was manufactured with; without the original surrounding hardware for context they are of little use beyond a historical curiosity. This doesn’t seem to stop some vendors attaching eye-watering prices to them, but thankfully if you really need a TMS1000 in your collection they can still be picked up for not a lot.

If It Was Everywhere In The 1970s, Why Is It Nowhere Now?

A die shot of a CMOS TMS1000. Pauli Rautakorpi [CC BY 3.0]
A die shot of a CMOS TMS1000. Pauli Rautakorpi [CC BY 3.0]
By the 1980s the world belonged to 8-bit and 16-bit microprocessors and microcontrollers, so other than surviving unseen for a few years as cores for TI’s calculator chips, the TMS1000 series was eventually retired and has since slipped away unnoticed into electronics history. It’s interesting to note that some of its contemporaries are very much still with us, you can still buy plenty of PIC, 8051, Z80, and even 6502 derivatives, yet there were no direct successors to the 4-bit TI processors. The onward march of technology is one culprit for this, but perhaps also the arcane software development played a part in its demise. Those 8-bit CPUs are still with us because anyone could pick up a dev board and an EPROM programmer and start coding, so they attracted a core of developers well-versed in their architectures. By comparison TMS1000 developers must have been few and far between, and certainly not enough to demand successor silicon. There were some special development versions of the processor added to the range that took an external ROM, but by then the market had shifted upwards by four bits.

If you encounter a TMS1000 today it’s most likely that you’ll have a late-70s electronic game in your hands, such as TI’s own Speak & Spell, or Milton Bradley’s Simon. Even these games survived longer than their original processor; you can still buy a modern version of Simon from Mattel, and TI’s line of speech-enabled learning toys kept going into the 1990s. The legacy of this processor is immense though, and can be seen today in every electronic device containing a microcontroller. If you have one, it’s a real piece of history!

Header image: Antonio Martí Campoy [CC BY-SA 4.0]

66 thoughts on “The TMS1000: The First Commercially Available Microcontroller

  1. Like millions of other folks in the digital world, I longed to build a computer for my very own use. Interestingly enough, I chose the TMS1000 (which choice only goes to prove how ignorant I was about building digital circuitry). For all the reasons Jenny mentioned, I didn’t buy any hardware, but that didn’t stop me from thinking about how I’d write software for it.

    My first project was to write a floating-point arithmetic package. I actually got something that (I think) might have worked.

    But then the Altair 8800 came along, and that changed everything.

    1. From Wikipedia:

      “The Intel 4004 is a 4-bit central processing unit (CPU) released by Intel Corporation in 1971. It was the first commercially available microprocessor, and the first in a long line of Intel CPUs.”

      “The TMS1000 is a family of microcontrollers introduced by Texas Instruments in 1974. It combined a 4-bit central processor unit, read-only memory (ROM), read/write memory (RAM), and input/output (I/O) lines as a complete “computer on a chip””

      Seems you and Wikipedia agree.

      1. But wait…

        https://en.wikichip.org/wiki/ti/tms1000

        In September of 1971, TI finished the design for their TMS0100 single-chip calculators. Designs where done by the Texas Instruments engineers Gary Boone and Michael Cochran. Based on their design of the TMX1795, Gary patented the invention for a single-chip processing machine on Aug 31, 1971. On Sep 4 1973, he was awarded U.S. Patent 3,757,306. Building on top of their experiences with the TMS0100 and Boone’s 8-bit microprocessor prototype they went on to design the 4-bit TMS1000 microcontroller series. Boone was later awarded U.S. Patent 4,074,351 for the modern microcontroller.

        After being slightly refined, the chip was released to general market in 1974.

      2. Wikipedia is absolutely right.

        But if you take a look carefully at the wording, you’ll see that one is a CPU and the other is a microcontroller. A microcontroller contains a CPU, but is a different component.

    1. No they are not!
      Unfortunately they are learning Arduino/ abstraction and have no clue what instruction pipeline is or assembly optimization is. They just grab a bunch of libraries cobble then together and scratch their head when something goes wrong during one specific execution scenario.
      The abstraction is taking away the inner workings which sadly only a handful of kids put an effort to learn. But hey! all of them have “worked” with the same micro controllers!

      1. People who are more than casually interested in computers should have at least some idea of what the underlying hardware is like. Otherwise the programs they write will be pretty weird. (Donald Knuth)

      2. For people who do it for fun, it’s just fine. For the professionals – they’ll get deeper knowledge anyway. Don’t complain about those gosh darn kids not doing it the old way anymore!
        When the transistor arrived, the old folks complained about how they didn’t know how to design tube circuits. And those who used tubes, were bashed by the spark gap crowd etc etc.

      3. I agree – look how much damage to the computing brought C64 with it’s Basic! No kid had to learn about pipelines and assembly! Just type some command and scratch their heads why it’s so slow? And than just to speed things up some of them peek and than poke and use those monitor programs and look where we are now? People using spreadsheets for their home budget instead of writing nice and efficient code in FORTRAN (or at least C)! Madness I tell you…

      4. “The abstraction is taking away the inner workings which sadly only a handful of kids put an effort to learn.” I’m a trained teacher and you have absolutely no idea of the inner working of people. It’s a good thing that kids learn skills. Period. I despise this way of thinking that children should learn programming in assembly or punch cards BECAUSE THAT’S HOW WE USE TO DO IT. Old techies are some of the dumbest, most narrow-minded people I have ever come across.

        1. so you ADMIT that it is THE PEOPLE that are WRONG!

          they should be wired to accept all programs that can work properly and without error, the first time around and without ANY guesses, glitches, or slowdowns.

          this INCLUDES ASSEMBLER SOURCE-CODE CREATION software.

          fix your “people” and turn them into computers, after-all; they sub-conciously learn analog spacial triginometery at age three just to walk; i think they can be forced to manage 64 commands and 64 registers.

          if an old-fashioned rotary-dial telephone can pause a torrent download on windows then your kids can learn assembler.

  2. Worked at a place in the early 80s that used tms9900 to make industrial controllers in applications like a laundry, a main signal box train describer, and an airport terminal flight display/baggage handler. You could get a lot out of a 16 bit microprocessor back then.

    1. The TMS9900, like the IBM 360 had its registers in memory. A task switch involved a BLWP (Branch and Load Workspace Pointer) and R0 .. R15 would have all new contents. I led a project at Tektronix to build the 1980, a video analyzer that was entirely programmable, The incoming video was digitized at 4X the color subcarrier rate and state machines controlled hardware functions in real time.
      Later products with much faster processors followed, but much of the measurement algorithm development still occurred on the 1980 because it used an interactive form of BASIC (Tek Answer Basic).
      Fun days.

      1. What would have been nice is if TI had continued production of the Home Computer’s Peripheral Expansion Box as a box for building industrial control and other embedded stuff. The PEB was already built like a tank with a massively overbuilt, unregulated, linear power supply and a thick steel chassis.

        The Expansion cards each had to have their own onboard power regulation to bring the voltages down to the normal 12V and 5V of the day. An “open” power source like the PEB would easily allow for hardware using any voltage, even quite a bit over the typical 12V and 5V.

        The backplane was a simple, dumb, bus. All lines connected straight through, with pin definitions determined by the interface card connecting to the computer console. That would’ve given a PEB system designer essentially complete freedom for the bus design, to make cards compatible with another system, or deliberately incompatible.

        TI could have switched to a lighter, regulated power supply for the PEB and designed a 9900 CPU card and other peripherals to have a successor or companion to their 960, 980, and 990 mini computers.

    2. I used the SBP9900, in a military project in the 1980’s, this was an integrated injection logic version of the TMS9900 and they cost a fortune! The development kit filled a room! I recently saw the prototype of the unit in an air museum, aah, happy days.

  3. Let us not forget the COP402 and its relatives. They were 4 bit parts who found themselves into a lot of things. Not surprisingly I have a batch of them, programmed of course, and the COP420 part who is wearing the name of the company it was made for. It has the part number on its back.

      1. Well the two I have are the COP411 and the COP420. And the COP420 for a 4bit part is slightly more complex and almost too difficult for a hobbyist to use. The technology salvage place was completely clueless as to what they were.

  4. I remember the days when we had to design a “TTL Grave” to get things done. Not to mention the manufacturing of standard PCBs that was totally out of reach of the hobbyist. Perfboards and a Verowire Pen were the tools of the time.

    You kids can’t imagine how much the more senior crowd appreciate the immense capabilities we have at hand today. You could hardly dream this up in the eighties.

  5. The TMS1000 is also emulated well in MAME, which emulates the majority of TI Speak and Spell, Speak and Math, etc. titles, as well as a whole host of games which used the TMS1000, such as Simon.

  6. Stop, already, with the stupid, infantile word games.

    This horse was beaten to death when people argued about the fact that the original IBM PC was not a computer, but a microcomputer. Do any of you remember that particular piece of tech-wordplay, “I’m-smarter-than-you-are”, insanity? Well, that’s exactly what this is.

    All of everyones’ “appeals to authority” (wikipedia; a particular manufacturer’ nomenclature; data sheets; etc.) in order to justify what you consider to be the correct terminology all boils down to what some INDIVIDUAL THOUGHT sounded “…about right…”. Absolutely nothing more.

    To use a characterization we used to use in the semiconductor industry–I don’t care if it’s “…a piece of hammered horse-shit…”,
    if it can be programmed, it’s a computer: mini-, micro-, peta-; CPU; you name it, but don’t expect me to be impressed by your erudition in claiming that something is NOT a microcomputer because it’s a–harUMPHHH–microCONTROLLER.

    Pure, simple unadulterated obfuscation and posturing–completely unnecessary, and absolutely, one-hundred-percent confusing to the very people who look to this venue for clarification, elucidation, and education.

    Clarification, elucidation, and education this ain’t.

    1. I think the distinction between XXXController and XXXProcessor is clear enough, and widely enough accepted – An XXXController has built in peripherals e.g. general purpose I/O pins, counter, timers etc, whereas an XXXProcessor does not (including the “Central” type of Processing Unit).

      No word games – every XXXController has a built in XXXProcessor, but not the other way round. An XXXController can form the heart of an XXXWave Oven, DVD player, etc. etc. _without_ any additional address bus decoding logic, data bus latches etc. exactly because of the built in GPIO pins, timers etc. An XXXProcessor cannot.

      1. There has been a semantic shift on many EE terms. For example, a 6502 wouldn’t be called a microcontroller today. But on introduction it was described and intended as such. The term didn’t imply “integrated program memory”. Only later a distinction was made. You see similar, sometimes subtle, shifts in terms as “TTL”, “RISC”, “PC”, “emulator”.

        1. Yes. I was going to say something.

          The history is well known, the Intel 4004 designed to be used in a calculator, the 8008 intended for use in a terminal. I had a National book about using one of the early microprocessors to replace logic.

          I don’t know about industry, but hobbyists looked at them and saw “computer” rather than controller. All of the microprocessor companies sold single board computers, which was mostly to introduce people to tye microprocessor, and thus sell the ICs, just incidentally giving us simple computers.

          In parallel with the home computers were lots of “embedded controllers” . Thet weren’t general purpose computers, but dedicated to controller. Some surely used the tms1000, some used microprocessors that included ram and some I/O, as they became available.

          But others just used 8080s and 6502s and whatever. In 1979 or early 1980 I knew someone who got a contract to update a local record pressing plant, and he used the intersil 6100, which was compatible with the PDP-8. Maybe because it was CMOS, but likely because of the PDP-8 angle. I was never sure if he hired someone to write software first, or chose the cpu first.

          So a lot saw “microcontroller” use, even if they weren’t as dense as what came later.

          1. ” I was never sure if he hired someone to write software first, or chose the cpu first.”

            When I worked in aviation weather research (1990’s) the FAA was building a huge ($1+Bn) Air Traffic control system using 1980’s hardware. (80286 procs and the like).
            Our group was one of those tasked with writing software for part of the system.

            My boss (among others) were trying to convince the FAA to design/write the software they needed _FIRST_, and once that was ready, go and buy modern equipment that would run it.
            For such a humongous system, the software took multiple years to write, while the hardware was quickly getting outdated.

            IIRC, it was the “FAA Modernization Program”, and it was a $1Bn project that was already $1Bn over budget and years behind schedule before DOT Secretary Federico Pen~a shut it down.
            (IMO, it was the only worthwhile thing he accomplished.)

    2. I was idly thinking I’d like to filter comments that start with “I think you’ll find” :)

      In this context I think it’s an important distinction to make, because as someone who’s made control computers using Z80s etc as well as PICs, Atmels etc there is a HUGE difference in the task involved. They’re different components, even if they can do a similar job.

      My past isn’t in the semiconductor industry, despite being an electronic engineer and lifelong hardware hacker I’ve spent most of my career in the more techie side of the publishing industry. The job I did before going freelance and then working for Hackaday was by far one of the coolest places I’ve worked because of the breadth of knowledge among my colleagues, the Oxford Dictionaries. There I encountered people for whom the history of language is as much a passion as tech is for us, and I gained an appreciation for the history of language when followed as a science. Everyone has their pet theories and hills they’re prepared to die on in language, but what sets lexicographers apart is their willingness to work only in the light of evidence gathered by painstaking research. Thus mentioning an OED citation is entirely appropriate when giving a potted history of a word, which I think was worth devoting a paragraph to in order to set the scene.

      Of far more interest is the device itself, which I hope everyone will agree is a somewhat unsung seminal component in the history of semiconductors.

    3. Hear, hear. Reminds me of the guys that insist that centrifugal force doesn’t exist. There is only centripetal force. Bah. It may impress folks at cocktail parties, but it’s 100% useless information.

  7. So .. why doesn’t some enterprising soul craft an emulator? Sure, in 197x you’d have to have your code masked into ROM (oh gawd that’s practically as cumbersome as an ASIC), but in an emulator, that wouldn’t be such a problem now, would it. Then we could devise our own alternate pasts ..

  8. Kind of a tangent, has anyone ever seen a logic diagram for the insides of a TMC0981, the heart of the TI-30 of yore? I guess because TI still sells a ’30, maybe they try to keep a lid on that, but surely they’re not still putting 1970s chips in the fresh crop of 30s!

    I’ve always been curious how the 30 worked on the inside. That and the TI-57. I understood gates and flip flops well enough that when I got hold of a 7400 data book, it became self evident how you could paste blocks of registers and buffers and what not together and produce something like a 6502 or 8080 – but I never had that same kind of ah-ha moment for the calculators. So it’s for academic reasons, really, that I’d like to know how they did it. Haven’t had any luck goggling it, hence my appeal for clue here

  9. BigTrak was cooler than Simon by far. Wonderful battery sucking toy recently re-created by Zeon. Doubting that TI takes paper tape roll programs and would have to brush off the dies; probably doesnt have original MCU. Magnetic coupled differential was also very cool. Hack worthy toy.

      1. There’s no might to it. Original Big Trak hasTMS1000. Theres a teardown in ur own backyard. There were several revisions but all had same MCU. The Stereo phono jack ‘trailer hitch’ looks alot like TI cheapo serial port doesnt it? Still havent got the new one.

      1. It’s the place on the deck of a ship where the radio equipment is kept and used. Originally it was a retrofit, since radio first went into existing ships. Later, they became not a shack, but incorporated into a new design Of course by then the equipment wasn’t spark gaps.

        There is recent talk about sending a sub down to the Titanic, remove the top of the radio shack, and extract any equipment. I suspect the radio shack is chosen because it’s accessible.

  10. I’m curious–

    Let’s consider Hackaday’s favorite “computer”: the Raspberry Pi–

    1) Does Hackaday consider the Raspberry Pi 4 a computer, a microprocessor, a microcontroller, some in-between chimera? What, precisely?

    2) Does Hackaday consider the BCM2711–used in the RPi4–to be a processor, a CPU, a computer, a microcomputer, a microcontroller? What, precisely, does Hackaday consider the BCM2711 SoC to be? What do you consider it to be?

    3) If the BCM2711 is truly a microprocessor / microcontroller / CPU / full-blown processor, why can it not be programmed in assembly language (if it COULD be, one wouldn’t be limited to bit-banging at a pathetically low rate), as can be any OTHER microprocessor / microcontroller / CPU / processor / mainframe? (If you even want to dare answer this question, then tell ALL of us where we might get a low-cost, or free, Assembler for the device).

    Careful, now: personal opinions absolutely won’t do–including completely wrong statements as to your ability to program the RPi in native assembly language. Only logical, technical, objective answers based on sound electrical engineering principles and precedent are allowed.

    1. YOU might not know how to program the Broadcom chip in assembly. I don’t either! But the people who wrote the compiler surely do. Is the C compiler open-source? Perhaps there’s some files there, or nearby, that tell a little bit about the workings. Or perhaps you could write a short program in C, then disassemble it. We know it’s basically ARM, right? Try disassembling the drivers for some of /dev.

      Yeah there’s lots to bitch about the Ras Pi from an open source point on view. The Pi itself is a computer, the CPU is a SoC, and whatever HAD or you and me, or anyone else who isn’t slinging millions of dollars around, thinks, doesn’t matter.

      1. The problem with the RPi is not inept programmers or a broken compiler; it’s about the fact that the boot code in the Broadcom chip is proprietary. They don’t WANT anybody tinkering with it.

        A good assembly-level programmer can absolutely write assembly-language code for his application code; the CPU is an ARM chip, and the instruction set is very straightforward. But you get no visibility into the chip, and must abide by the system calls Broadcom has defined.

    2. 1, It’s a computer.
      2, It’s a CPU with a built in MMU and a bunch of peripherals on an internal PCIe interface
      3, ARM assembly. Look on the bare metal development section of the Raspberry Pi Foundation’s forums for the IO addresses, or just have a peek in the Raspbian source code. A quick glance suggests the GPIO pins can be driven at about 12.5 MHz, but your coding may vary.

  11. 4-bit processors are notoriously awkward. You’d think they’d have minimal and easy instruction sets, but for commercial reasons they’re full of arcane op-codes and operations. Here’s a summary of the TMS1000/1200 instruction set.

    https://docs.google.com/document/d/1LW1kuvZlXua_Ml4OHB8efh2l3jkhgfvNxr79-e2sh78/edit?usp=sharing

    Some striking features are:

    1. There’s no bitwise logical operations (but you can clear or set individual bits), the ALU just adds and subtracts.
    2. There’s only an indexed addressing mode M(X,Y) which accesses 1 of 64 nybbles and X can only be loaded with a literal or inverted (though Y can be auto-incremented).
    3. There’s specific instructions for adding 1, 6 or 10 to the accumulator, but no other immediate adds!
    4. Mnemonics are really hard to learn, you’d think they’d all be simple 3 letter things, but you get crazy stuff like: TAMZA, YNEA and A6AAC.

    So, doing anything is pretty challenging which makes products like Speak and Spell all the more impressive!

    1. Speak And Spell must have had it’s dicionary in a separate ROM. Actually that would be handy back then cos they could do international versions with just a ROM chip swap rather than a whole new run of TMS’s needing masked up. Weird the best way to do things swaps around when you have certain limits imposed on the technology. Yeah, respect to everyone who had to program this bloody odd chip! Just the literature I’ve read online makes it out to be very odd. It has help for some things, like driving a 7-segment display, yet goes out of it’s way to be weird for other things.

      Is the “Add 6” instruction something to do with BCD? I can’t quite see it but I suppose TI’s chip designers are smarter than I am.

      I thought programming the Atari 2600 was baroque, weird, and arbitrary. And it is! But the TMS makes the Atari’s 6502, complete with a video chip you had to feed graphics to for each scanline, as it was being drawn, look like a nice day programming in LOGO to draw a lovely square.

    2. Actually, the TMS1000’s instruction set is quite powerful, if not a bit arcane. This is a shame, since, as pointed out below, T.I. had it totally within their power to make the mnemonics much more ‘remember-able’.

      For example, there are actually fourteen immediate-add instructions (including “add 1”–“IA / IAC”); not just ‘add 1’, ‘add 6’, and ‘add 10’. These are, respectively, “IA / IAC”, and then “A2AAC” through “A14AAC”.
      These last 13 could have just as easily have been assigned the mnemonics of “A2” through “A14”, with no loss of clarity (they DO have the disadvantage of forcing the programmer to have to actually remember something. And as for things similar to “TAMDYN”–well. it–and those–are simple abominations which were completely avoidable and totally unnecessary.

      see

      http://www.nyx.net/~lturner/public_html/TMS1000ins.html

  12. Most people feel that there is something “magic” about the generation of the mnemonics which comprise the assembly-language instruction set of a computer (mainframe; mini-; mili-; micro-; nano-; femto-; atto-. Use whatever prefix makes you feel, and seem, important).
    Nothing could be further from the truth; the mnemonics associated with a particular operation of a particular conglomeration of hardware–which comprise a new computer-hardware design–is strictly at the whim (quite literally) of the designer of that computer or, more than likely, the manager of the project.

    When Lockheed Electronics designed their extremely powerful MAC-16 minicomputer at the same time the L-1011 wide-body jet was being designed and “pre-sold”, we accused LEC of pandering, because they assigned their assembly-language instruction for negative-binary arithmetic (a one cycle instruction to generate a negative number in the accumulator) the mnemonic “TWA”–for “TWos-complement-the-Accumulator”. You can’t get a more memorable mnemonic than that–and that’s precisely what “mnemonic” means.

    TWA bought a lot of L-1011s; probably the first- or second largest buyer.

    If you are considering a particular computer and find that it has anything other than a rational set of assembly-language mnemonics, run like hell; that’s an indication of hubris–and self-importance–on the part of the designer(s). It’s VERY easy to make it EASY. It’s also a very good indication of the types of interactions you’re likely to have with the manufacturer, down the road.

  13. It should be pointed out that, while the TI Speak & Spell used a member of the TMS1000 family (one with 2K words of instruction rom and 128 nibbles of ram, vs the 1K and 64 in the diagram of your article), that was only handling the keyboard and display. The voice synthesis was handled by a separate TMC0281 chip which was groundbreaking for its time, establishing the category we now refer to as DSP (Digital Signal Processor). The speech data was stored in two separate ROM chips (16K bytes each) connected to the synthesizer chip. The speech content was encoded using linear predictive coding, specifying several oscillator/sound sources and filter coefficients to model sound generation. The synthesizer chip implemented digital filters applied to the sources using these coefficients to generate the output stream to a D/A converter. The digital filtering involved on-chip multi-stage pipelined hardware multiplication, at a density not seen in consumer products to that point.
    (I find it amusing that one of the google searches turned up “How To Pronounce TMC0281: TMC0281 pronunciation”)

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.