The Early Historical past of the Microcontroller Half 1

Our son received numerous new toys over the Christmas interval. One among his favourites is a plastic guitar with buttons on the fingerboard, numerous lights, and even a whammy bar. It’s actually enjoyable. At one yr outdated he’s a lot too younger for a pc, however a lot of his toys, together with the guitar, have computer systems inside them. However, after all, we often don’t name them computer systems. As an alternative, they’re ‘microcontrollers’.

A second’s thought tells us that microcontrollers are in all places: in our houses, in washing machines, fridges and microwaves, in our vehicles, and plenty of, many different locations. And a few of these microcontrollers are primarily based on designs which were round for many years. I don’t dare threat opening up the toy guitar – he could be very upset – however I wouldn’t be stunned if the microcontroller inside makes use of an 8-bit structure that was first developed within the Nineteen Seventies. Prompted by my son fortunately taking part in along with his new toy guitar, it looks like a great second to delve into the historical past of the ever-present, however usually invisible microcontroller.

Earlier than we delve into the historical past of the microcontroller, we have to outline what it’s. Wikipedia defines the microcontroller as a ‘small pc’ on a single built-in circuit. This distinguishes it from a microprocessor, which is only a Central Processing Unit on a chip and wishes extra built-in circuits to create a practical pc, and a ‘System-on-Chip’ or ‘SoC’ which is the time period used when the pc isn’t judged to be ‘small’.

It’s maybe shocking that the earliest microcontrollers emerged at about the identical time as the primary microprocessors. One would possibly count on that it’s considerably more durable to construct an entire pc on a single built-in circuit when in comparison with only a CPU. Nonetheless, the earliest microprocessors have been usually utilized in designs which wanted very small quantities of ROM or RAM and which have been extremely price-sensitive. Engineers have been quickly capable of overcome the technical challenges of including a small quantity of reminiscence and enter/output circuitry alongside a processor. And the potential business advantages for the system designer from the decrease prices related to utilizing a microcontroller have been appreciable. Because of this, there was a big marketplace for microcontrollers and most of the main semiconductor producers within the Nineteen Seventies and Nineteen Eighties had their very own line of microcontrollers, bought alongside their microprocessor designs.

Microcontrollers have sometimes had a decrease profile and attracted much less consideration than microprocessors, for 2 causes. First, the necessity to maintain built-in circuit die sizes small while additionally accommodating reminiscence and enter/output, has meant that their CPUs have been much less refined than up to date microprocessors. Second, they’ve sometimes been utilized in methods the place customers can not learn or change their code. So while totally different generations have grown up with entry to ‘residence’ after which ‘private’ computer systems or workstations, and have turn into aware of the architectures of the microprocessors powering these machines, microcontrollers have been locked away, hidden inside calculators, toys, vehicles and home home equipment.

As semiconductor node sizes shrank through the years, it turned potential to make use of present microprocessor designs because the CPU a part of a microcontroller. These CPUs sometimes had increased efficiency, richer software program ecosystems, and have been simpler to program than the sooner designs, which had been designed particularly for early microcontrollers, that they changed. In the present day, most microcontroller designs incorporate a processor both from the 8-bit period (e.g. Z80, 6502), or the 16-bit period (e.g. 68000) or are trendy 32-bit designs (similar to ARM Cortex or RISC-V).

That is the primary put up in a sequence on the earliest microcontrollers. We’ll have a look at a few of the main designs of the time, their historical past and the place they have been used. Though these designs have had a a lot decrease profile than the early microprocessors, they performed a key position in making computing ubiquitous.

Earlier than we get to actual microcontrollers, it’s price discussing the “thought” of the microcontroller. Within the Nineteen Sixties, as increasingly more elements have been crammed onto built-in circuits, it turned more and more apparent that it might quickly be potential to create a ‘pc on a chip’.

In 1970 Gilbert Hyatt filed a patent for a ‘microcomputer structure … that facilitates a totally built-in circuit pc on a single built-in circuit chip.’ Or, in different phrases, a ‘microcontroller’. It took twenty years for that patent to be granted. The Los Angeles Instances reported in 1990 that:

In July, after a 20-year authorized battle that generated an estimated 10,000 pages of paperwork, the U.S. Patent and Trademark Workplace gave Hyatt patent No. 4,942,516 for a “Single Chip Built-in Circuit Pc Structure.” The patent shocked the electronics world.

The LA Instances reported its dialog with Hyatt::

“I didn’t invent the pc, however I got here up with an excellent enchancment,” mentioned Hyatt, a 52-year-old workaholic who prefers to work alone at a modest residence in a cul-de-sac in La Palma. “My work in these days led to the PCs of as we speak.”

Hyatt had technical experience, having graduated from UC Berkeley and labored at electronics conglomerate Teledyne, and had tried to develop his concepts, elevating capital earlier than falling out along with his traders. Nonetheless, in accordance with one former colleague:

Hyatt was a cussed, impractical researcher with out a viable expertise who was obsessive about “paper rights.”

After the 1990 patent approval, Hyatt was capable of declare tens of tens of millions of {dollars} in royalties, from Sony and different main firms. Nonetheless, different corporations continued to problem the patent approval, till, in 1996, a central declare of Hyatt’s regarding the ‘invention’ of the microcontroller was overturned.

“After a five-year evaluation of tens of 1000’s of pages of paperwork, the patent workplace dominated that Hyatt didn’t describe a microcontroller till 1977, in one in every of many amendments often called a continuation.”

Hyatt was later concerned in one other set of authorized instances, this time regarding tens of millions of {dollars} of potential tax liabilities and associated damages, that made their approach to the US Supreme Court docket and would lastly be resolved solely in 2019.

In a 2014 interview Hyatt was quoted as saying:

“I felt that the business had taken my expertise and, not solely hadn’t rewarded me, however hadn’t given me recognition till my patents began to difficulty,” Hyatt mentioned.

The truth, although, is that he did not have entry to the expertise required to create a working microcontroller and possibly wouldn’t have been profitable in creating one if he did.

Main the problem to Hyatt’s patent was Texas Devices (TI). TI began life in 1951 following the reorganisation of ‘Geophysical Service Included’, a Dallas firm that made seismic gear for the oil business and which had expanded into electronics. In 1952 TI bought a licence to make Germanium transistors from Western Electrical, the manufacturing arm of AT&T, whose Bell Labs analysis arm had invented the machine. TI shortly turned a pioneer within the improvement and commercialisation of semiconductors and associated expertise. The primary transistor radio in 1954 was adopted by the primary built-in circuit in 1958 and the, vastly fashionable, 7400 sequence of ‘Bipolar’ Transistor-Transistor-Logic or TTL built-in circuits within the early Nineteen Sixties. By the beginning of the Nineteen Seventies TI was a semiconductor big with revenues of round a billion {dollars}.

Within the late Nineteen Sixties, TI’s consideration had turned to utilizing built-in circuits to shrink ‘desktop’ calculators. Early digital calculators used discrete transistors and so have been heavy and wanted AC energy provides. They have been an apparent business software for more and more highly effective built-in circuits. In 1967, on the finish of a two-year undertaking, a TI staff led by Jack Kilby developed and, was granted a patent on, the primary first ‘pocket’ calculator, often called the ‘Cal Tech’ (maybe surprisingly named after the California Institute of Know-how). ‘Cal Tech’ decreased the a whole bunch of discrete transistors in earlier calculators down to simply 4 built-in circuits. The Nationwide Museum of American Historical past has the primary Cal Tech calculator inscribed with the date (March 29, 1967) when it was introduced to TI’s president, Pat Haggerty.

TI would go in to work with Canon to show the Cal Tech’s expertise right into a business design, often called the ‘Pocketronic’ which appeared in 1970 (which regardless of its title would have been onerous to place into most pockets).

From the angle of 2024, it’s fairly onerous to absorb how huge a deal the emergence of small digital calculators have been within the early Nineteen Seventies. Arithmetic is key to many roles, from primary accounting for small companies to complicated engineering. Previous to the event of the inexpensive, transportable digital calculator, these wanted to be carried out by hand (with variable accuracy) or utilizing an costly and unwieldy, transistor primarily based, digital or electro-mechanical desktop calculator.

There was an enormous potential marketplace for cheaper and extra transportable calculators. Their improvement calculators would quickly turn into a race between main US and Japanese semiconductor and digital firms. This, in flip, was a central a part of the broader aggressive wrestle between US and Japanese corporations. For rather more on the calculator ‘race’ see the article ‘The Story of the Race to Develop the Pocket Digital Calculator’ referenced within the additional studying on the finish of this put up.

Growing integration meant that it might quickly be potential to mix the 4 chips within the ‘Pocketronic’ into one. However TI was crushed within the race to create the primary calculator on a single chip in November 1970 by Texas rival Mostek, with the MK6010. “Electronics” journal reported in Feb 1971:

The obvious winner within the race to supply a calculator on a chip has hit the wire.  Mostek Corp., of Carrolton, Texas, is now producing such a chip for Japan’s Busicom Corp. … The 180-mil-square [4.6 mm-square] chip accommodates the logic for a 4 perform 12-digit calculator – greater than 2,100 transistors in 360 gates plus 160 flip-flops.

The MK 6010, like the primary microprocessor, the Intel 4004, was made below a ‘… gross sales association with the Nippon Calculating Machine Co. of Tokyo to be used in Nippon’s line of Busicom calculators.’

It took a yr for TI to reply. In November 1971 they introduced their very own ‘calculator-on-a-chip’, the TMS 1802NC (to not be confused with the RCA 1802 microprocessor, which is a later, unrelated, design). What distinguished the TMS 1802NC was that it was programmable. In response to TI’s advertising and marketing:

The machine is completely programmable, the “program” read-only reminiscence, timing sections, management part and enter/output decoders could be programmed to realize totally different computing traits . This strategy presents most design flexibility at very low price.

When TI mentioned ‘completely programmable’ they meant ‘programmable on the manufacturing facility’, as this system was mounted throughout manufacturing in a ‘masks programmable’ Learn-Solely-Reminiscence. In response to the patent for the 1802:

The programmable learn solely reminiscence and the programmable logic arrays are simply modified by altering solely the gate-insulator masks for the metal-insulator-semiconductor built-in system embodiment throughout the fabrication course of.

In different phrases, calculator designers might pay TI to change this program for their very own merchandise, however this system was mounted for finish customers.

The price benefits of the microcontroller strategy are quickly obvious after we evaluate it with the Intel 4004 microprocessor, which was additionally launched in November 1971. The 4004 price $60 and wanted a number of help chips to create a calculator. Against this, the 1802 bought for $20 in portions of 10,000 or extra (though it was as a lot as $150 for particular person components). The 4004 was significantly extra refined, however the worth distinction made the 1802 a compelling design.

TI was aggressive in setting the $20 worth level for the 1802. As Ken Shirriff has famous, the die dimension of the 1802 was a lot larger than the Intel 4004. With an even bigger die, the 1802 was ready to make use of 7,000 transistors (though see this footnote on transistor numbers), when in comparison with solely 2,300 for the 4004 and a pair of,100 for the MK6010.

In response to TI engineer Gary Boone, who led the early improvement of the 1802, the genesis of the undertaking got here, not from a need to chop prices or compete with Mostek, however relatively to benefit from the engineering sources that TI had accessible and which had been centered on growing customized built-in circuits for a number of TI calculator prospects:

The person necessities differed intimately, however in precept and in total perform, they have been nearly equivalent. So, what goes by way of your thoughts is, “I am bored with doing this. I am working lengthy hours. My household isn’t comfortable. I’ve received to discover a higher means to do that.”

… that is the genesis of significantly the TMS 100 microcontroller chip: it got here out of boredom, excessive demand and a imaginative and prescient of commonalities that have been being inefficiently served by deploying large groups with many chips.

One can get a way of the issues that the TI groups had from the structure of the TMX1795 microprocessor that they constructed at across the identical time. Quoting Ken Shiriff:

… Texas Devices did not appear to place a lot effort into the structure, which [Intel engineer Stan] Mazor calls “fairly sloppy strategies” and “throwing some blocks collectively”. Whereas the 4004 and particularly the 8008 are densely packed, the TMX 1795 chip has copious unused and wasted house.

It appears possible that the TI groups merely didn’t have time to refine the design for anyone buyer. They wanted to get it executed and transfer on to the following.

Working with Daniel Baudouin, TI’s MOS Advertising and marketing supervisor, Boone compiled a matrix of buyer necessities, after which got down to create a design that might meet as a lot of these necessities as potential. The 1802 was the end result, designed by Boone and a small staff of round six engineers. Boone additionally labored on the TI 1795X, the identical design that might additionally present the structure for Intel’s first 8-bit microprocessor design, the Intel 8008. Boone’s testimony on his work on the 1802 could be central to TI’s efforts to overturn the Hyatt patent.

There was some debate round whether or not the 1802 design qualifies as a ‘microcontroller’, relatively than a ‘mounted perform’ calculator chip. As we’ve seen, this system ROM within the 1802 may very well be modified, and TI did supply variants by altering that program. The documentation that TI created for engineers utilizing the 1802 (which by then had then been renamed to be the TMS 0102, a part of the TMS 0100 sequence) reveals that it was primarily designed for use in calculators. Most importantly, the enter / output pins of the 1802 have been designed to drive a calculator show and to help studying from a easy calculator keyboard.

Nonetheless, the important thing patent for the 1802, U.S Patent No. 4,074,351, focuses on its use in calculators, however briefly discusses its use in a much wider vary of units:

For instance, the calculator system could also be programmed to carry out meter features similar to for a digital voltmeter ,occasion counting, meter smoothing, taxi-fare meter, an odometer, scale meter to measure weight, and many others.

Maybe a technique of settling the talk is to say that the 1802 was a microcontroller however not a ‘basic objective’ one. In different phrases, it was a programmable pc on a chip however one which had a reasonably narrowly outlined set of functions.

The 1802 was modern in the way it was designed. In response to Boone:

We relied on our personal simulations and our personal software program. In the present day it might be referred to as register switch mannequin simulating. We modeled the [instruction] processing and program and knowledge and enter and output sources on this chip. That principally permitted us to check the design at a excessive degree, earlier than our silicon existed. That was a leap, an innovation in design methodology. 

Wi-fi World reported on the 1802’s structure and capabilities:

The i.c. accommodates an eight-digit b.c.d. arithmetic logic unit; a three-register 182-bit random entry retailer; a 3520-bit read-only reminiscence for holding the programme; and timing, output, and management decoders. Floating-point or fixed-point operation calculations could be carried out and there’s automated round-off of numbers and main zero suppression. Arithmetic and management operations are primarily based on a 4μs single-phase clock system.

In order that’s 182 bits of RAM (organised as 3 x 13 4-bit BCD (Binary-Coded-Decimal) registers plus 2 x 13 bit flags registers), and 3520 bits of ROM (organized as 320 directions x an 11-bit instruction size). It’s fascinating to check this with the reminiscence used within the Intel 4004-based Busicom 141-PF calculator, which had 640 bits of RAM and 8196 bits of ROM. The Busicom provided extra performance, besides, squeezing a calculator into the small reminiscence footprint of the 1802 represents a powerful achievement. It was thought sufficiently noteworthy on the time for the entire program, in each binary and meeting language kind, to be listed in the important thing patent for the 1802 (see the notes to this put up for this patent).

If imitation is the best type of flattery then TI ought to have been flattered. Mostek, who had crushed TI to the primary single chip calculator, adopted that design up with the MK 5020, a clone of the 1802.

The 1802 was quickly utilized in calculators from each the US and Japan. TI itself used it in its personal first calculator design, the TI-2500 Datamath, in July 1972. One of the vital artistic makes use of of the 1802, although, got here from the UK. Squeezing further capabilities out of low-cost {hardware} was Clive Sinclair’s speciality. Attracted by the low worth of the 1802, he assigned the duty of growing a brand new calculator to Chris Curry. Curry has instructed the story of flying throughout the Atlantic on New 12 months’s Day 1972, nonetheless hung over from Sinclair’s New 12 months’s Eve celebration, to gather three samples of the 1802 from a TI manufacturing facility in Texas.

Returning to the UK, Curry constructed the ‘Sinclair Government’ calculator across the 1802. The 1802 was constructed utilizing a PMOS fabrication course of and so was rather more energy hungry than later CMOS primarily based designs. The Sinclair staff dramatically decreased the facility consumption of the Government by supplying energy to the 1802 in 1.7-microsecond pulses, with the capacitance of the chip retaining knowledge in periods when there was no energy. This prolonged battery life to round 20 hours while utilizing small ‘listening to support’ batteries, which in flip allowed the Government to help a ‘slimline’ kind issue with a thickness of simply 9 millimetres.

The low worth of the 1802 enabled the Sinclair Government to be launched with a worth of £79.95 (cheap for the time, however nonetheless equal to over £1,000 or $1,600 in 2024). Its modern design and decrease price made it an enormous success and allowed Sinclair to construct a calculator enterprise that, at its peak, was making over 100,000 models a month.

This wasn’t the tip of Sinclair’s innovation utilizing TI calculator chips. In 1974 he launched the Sinclair Scientific, utilizing one of many 1802’s successors, the TMS 0805, which might present simply what was potential with simply 320 directions by cramming scientific and trigonometric features into that tiny house. Ken Shiriff has reverse-engineered this system to indicate the way it was executed, and constructed a simulator for the calculator, emulating what Ken calls TI’s ‘loopy’ 11-bit opcodes..

In a twist of historical past, Chris Curry would later go on to discovered Acorn Computer systems, the agency that developed the unique ARM structure. Later variations of the ARM structure would turn into probably the most fashionable processor designs within the 32-bit microcontroller market. TI’s adoption of the ARM structure and use of ARM-based TI System-on-Chip designs in early 3G cellphones from Nokia and others could be a key step in Arm’s journey to market success. TI would additionally present ARM with a future CEO, Warren East, when it closed its UK manufacturing plant in Bedford in 1994.

The TMS 1802 and the opposite members of the TMS 0100 sequence have been profitable within the calculator market, however there have been alternatives for TI to develop the usage of the ‘computer-on-a-chip’ into different functions the place its low price could be a compelling proposition. So the TMS 0100 sequence was adopted by the TMS 1000 sequence, which was designed for use each in calculators and in a wider vary of management functions. Launched in 1974, the TMS 1000 used a 4-bit processor structure and would promote for as little as $2 in quantity.

Let’s have a look at the structure of the TMS 1000 sequence. It’s fascinating to see what simplifications have been essential to allow it to suit into such an affordable machine.

The very first thing to stress about each the TMS 0100 and TMS 1000 sequence, which distinguishes them from microprocessors, is that they can not entry exterior RAM or ROM. The fundamental TMS 1000 supplied 1,024 x 8 bits of ROM and 64 x 4 bits of RAM (or as much as double this quantity of RAM and ROM for some members of the sequence) and that’s it. In case your design wanted extra reminiscence than this then you definately’d wanted a microprocessor primarily based system, with the additional prices and complexity concerned. With no choice to entry exterior reminiscence, a lot of the pins on the TMS 1000’s bundle can be found to learn and write indicators from exterior {hardware}, similar to a keyboard or show. 

With so little RAM it wouldn’t make sense to attempt to load packages into RAM so the TMS 1000 sequence, just like the TMS 0100 sequence, are Harvard structure designs : instruction and knowledge reminiscence are fully separate.

The instruction set of the TMS 1000 sequence introduces an fascinating characteristic when in comparison with extra acquainted microprocessor designs. It’s ‘programmable’. It comes with a ‘commonplace’ instruction set which we’ll describe in a second. Nonetheless, directions are, partially, applied by what is called a ‘Programmable Logic Array’, one other type of ‘mask-programmable’ ROM, which could be modified for particular person prospects. From the TMS 1000 sequence Information Guide:

The programmable instruction decode is outlined by the instruction PLA. Thirty programmable-input NAND gates decode the eight bits of instruction phrase. Every NAND gate output selects a mix of 16 microinstructions. The 16 microinstructions management the arithmetic unit, standing logic, standing latch, and write inputs to the RAM.

For instance, the “add eight to the accumulator, outcomes to accumulator” instruction could be modified to carry out a “add eight to the Y register, end result to Y” instruction. Modifications that take away an instruction that’s not used fairly often are fascinating if the modified directions save ROM phrases by rising the effectivity of the instruction repertoire.

After we have a look at the structure of the TMS 1000, we are able to see measures taken to cut back the variety of transistors wanted for the design, to maintain manufacturing prices down. Only a few examples:

• Paged RAM : The TMS 1000 has a 4-bit accumulator (A), a 4-bit reminiscence pointer register (Y) and a 2-bit register (X) that gives the higher two bits of any RAM tackle. For instance of how reminiscence entry works in follow, there’s an instruction ‘TAM’ that transfers A to the reminiscence tackle X * 16 + Y. So the TMS 1000 implements segmented or paged reminiscence addressing with solely 64 ‘nibbles’ or RAM.

• Paged ROM : Instruction addresses are additionally segmented with an tackle being derived from a 6-bit program counter plus, in its increased 4 bits, a 4-bit ‘web page register’. Applications keep in a 64-byte ‘web page’ until they department to a different web page.

• Single degree ‘stack’ : There was no stack, however a register was supplied to retailer this system counter on a subroutine name, permitting for only a single degree of calls.

• Pseudo-Random Program Counter Increments : This system counter doesn’t enhance by one after executing every instruction however as a substitute modifications to a worth in a predetermined, however random-looking sequence.

The final instance might look significantly unusual:

The Program Counter is applied as a suggestions shift register relatively than as a binary counter. Which means the logically consecutive program reminiscence places don’t reside bodily consecutive in this system reminiscence.

A suggestions shift register would have wanted fewer gates than a binary counter. The mapping of logical location to bodily location is given within the desk under (from Adam Osborne’s 4 & 8 bit Microprocessor Handbook).

So, for instance, after executing an instruction at location #37 the following instruction to execute is at location #2F.

I’m certain for a lot of readers the TMS 1000 instruction set appears to be like infuriating! Ease of programming misplaced out to retaining the price low! Nonetheless, I actually admire the creativity of the TI staff to find methods to maintain the TMS 1000 small.

The TMS 1000 sequence quickly discovered its means into a variety of units, with TI usually ‘consuming its personal pet food’ by utilizing the chips in its personal merchandise. TI helpfully supplied a schematic of how the TMS 1000 is likely to be utilized in a easy ‘terminal’ machine with a keyboard and 11-digit show.

Famously, TI used it in its personal ‘Communicate and Spell’ instructional toy. The Communicate and Spell would play a synthesised set of 200 phrases and invite the kid to spell them on a keyboard. A TMS 1000 microcontroller managed the operation of the toy, studying the keyboard and dealing with the show. The onerous work of voice synthesis, although, was executed by one other TI chip, the TMS 5100, linked to 2 128 kbit ROM chips.

The recognition of TI’s microcontrollers quickly attracted competitors. Because it turned potential to implement extra refined designs on a single low-cost built-in circuit, the 4-bit structure of the TMS 1000 turned an pointless constraint. Some rivals would construct 4-bit designs, while others would transfer to extra highly effective 8-bit architectures. We’ll have a look at these rivals and what occurred subsequent in Half 2 of the ‘Early Historical past of the Microcontroller’.

Earlier than we go away the TMS 1000 fully although, though it’s not manufactured, there’s an open-source implementation of the design accessible from the OpenCores web site. As is usually the case, fashionable expertise by no means fully disappears.

After the break, the references, hyperlinks to extra ‘additional studying’ on early TI microcontrollers, together with unique documentation and key patents.