Happy 45th Birthday to the IBM PC and Model F/XT

45 years ago today, on 12^th August 1981, the computing landscape changed forever. IBM launched the Personal Computer, the vessel that kickstarted x86’s popularity in computing at offices and homes alike. It was, of course, not a feat IBM achieved by itself – the IBM-compatible “clone” market is what helped keep the ecosystem prominent in the long run. But I like to think a big part of its success was its keyboard. Not the perfect layout as most may agree, but certainly a competent, well-built keyboard – the Model F/XT. Thus to celebrate, in the most appropriate way for Admiral Shark’s Keyboards, we will look at this keyboard, where it came from, and what it became. Of course, with context as needed, including what led to the IBM PC, and both who and what was behind it.

Before the PC, IBM released products adjacent to a personal computer, and toyed with the idea of a personal computer throughout the 1970s. The PC’s type number, 5150, places it in the lineage of the 5100 Portable Computer, and the 5110 and 5120 Computing Systems. 5100 in turn is an evolution of a prototype computer known as SCAMP (Special Computer APL Machine Portable), which was developed at the IBM Palo Alto Scientific Center in 1973.^[4] William C. Lowe, an executive in the IBM General Systems Division (GSD) at the time, was considered instrumental in foresting this prototype. The 5100 was touted as a “personal computer” by some in the media, but it was a very expensive and not accessible to many people. It was notable for being portable though (for 1975), emphasised as IBM’s smallest computing system to date at that point.^[5]

Tommy R. Hardy, a now-veteran IBM designer, conceptualised some interim ideas for a personal computer, including the “Yellow Bird” in 1976 and the “Aquarius” in 1977, both vaguely wedge-shaped in a form factor not unlike the later Commodore VIC-20, 64, etc., albeit bigger. Aquarius even made it to a pre-production prototype, but ultimately dismissed due to lack of confidence in its use of bubble memory modules. When Atari’s 8-bit computers launched in 1979, IBM also considered aquiring Atari and utilising the 800 as a basis for the PC. Hardy also conceputalised a pearl-white, IBM-branded 800 to this end,^[6] but this idea proceeded no further. The next big thing, though, is Datamaster.

It would be impossible to have a discussion about the PC and its keyboard without talking about the IBM System/23 Datamaster, an influential experience for IBM that aided them in the PC’s development and its keyboard. The 5322, the original “desktop” (all-in-one) Datamaster, was released in July 1981. It was followed by the 5324 “floortop” (tower) Datamaster, announced on 18^th May 1982.^[10] They were Intel 8085-based microcomputers intended to run business applications written in BASIC.

The development for Datamaster began in February 1978, and its hardware was ready by July 1980. However, as recalled by IBM engineer David J. Bradley, who worked on Datamaster, its launch was delayed by a year and to just a month before PC thanks to IBM deciding to develop the Datamaster’s BASIC more in line with the IBM System/34 midrange computer’s implementation of it.^[11] The 5322’s styling is attributed to Tommy R. Hardy via its U.S. design patent (275,285), which was filed on 4^th February 1980. The design at that stage was largely like the finished product, though the keyboard layout is slightly different, with extra keys and a non-stepped Enter key. If design patents for 5324 exist, I was not able to find them during my research.

The Datamaster keyboard establishes the basis for every other keyboard discussed today. It is presently the earliest confirmed Model F keyboard implementation to make it to market, which would soon expand into an entire family of IBM flagship keyboards prominent for the first half of the ’80s. They all use capacitive buckling springs, a keyswitch design by Richard Hunter Haris (U.S. utility patent 4,118,611), where pressing a key compresses a metal coil spring underneath in a way that it eventually “catastophically” buckles, which in turn pivots a small plate made of a capacitive material atop a capacitance-sensing PCB called a “pad card”.

The 5322 keyboard is a complete sub-assembly (CSA) that sits inside the host Datamaster. The keys are arranged in a layout derived from the Model B-based IBM 5251/5252 Display Station Typewriter Keyboard, which was IBM’s most immediate and common pre-Datamaster keyboard design. The aforementioned pad card bears a 12×8 key matrix and an Intel 8048 as its controller. The keyboard uses a parallel-style interface where an inverted 7-bit scancode is sent across separate scanbit lines at once and together. Like its layout, its scancodes are also derived from the 5250 family. On the receiving end, according to RetroAND (bitspassats.com), the Datamaster has an 8255 Programmable Peripheral Interface configured for mode 1 input. As the keyboard is completely internalised, it connects directly to the 5322 motherboard (CPU planar board) via an 8×2 header.

As the 5324 is a distributed system, its version of the keyboard is discrete, connecting to its host via a DB-25-terminating cable. Chronologically, this was released after the PC, so it derives its styling from the PC’s keyboard, but with a considerably enlarged cover set and slightly longer riser feet. It may sometimes be confused with the IBM 5291/5292 keyboard also discussed later, but the 5324 keyboard has the largest bezel between them, to the point it sometimes gets nicked the “bezelmaster” Model F keyboard. The 5324 keyboard reuses the 5322’s keyboard CSAs. But the 5324 keyboard adds a “keyboard adapter card” before its external cable. As explained to me by RetroAND, it has positive buffers and double negations for the control signals, likely to reduce noise. In effect, it prepares a design that was originally integrated into something, to be a distributed design that requires a longer cable.

Roll back the clock to July 1980: Lowe – who was involved with SCAMP – pitched the idea of a small, low-cost business and consumer computer to IBM CEO Frank Cary, who subsequently gave him a month to develop a prototype and to launch the finished product in a year. It was dubbed “Project Chess”. Lowe was promoted soon after, so Chess fell to Philip Donald (Don) Estridge.^[17][18] The team consisted of about twelve engineers, who were famously able to develop the PC outside of IBM’s typical processes, ultimately allowing them to meet the one-year deadline, rather than complete it in 2 to 5 years like most other IBM systems of the time.

The earliest sketches of what became the PC, from 10^th August 1980, were only relatively recently published. They show that a lot of what became the PC were nailed early on, including its use of the Intel 8088, five expansion slots, two floppy diskette drives, use of ASCII instead of EBCDIC, and a detached Model F-based keyboard. There are some notable differences, of course, including the display controller, the power supply unit’s location inside the PC, and that it originally specified 8″ floppy diskette drives instead of 5¼”.^[19] But all in all, it is still recognisably the PC.

To achieve their goal, the team was also assisted by prior experience with developing Datamaster, and using readily available components. The PC’s use of an Intel CPU, various Intel supporting controllers and timer, and 8-bit ISA bus were derived from Datamaster’s use of similar technologies. The choice of an 8088 with 20-bit/1MB address space was largely informed by woes with Datamaster’s 8085 and its 16-bit/64KB address space. The choice of 8088 over 8086 was down to savings that could be had by needing to support just an 8-bit data bus instead of 16.^[11]

As you know, the IBM 5150 Personal Computer was released 45 years ago today. Its importance cannot be understated. Its open architecture and use of the 8086-sibling Intel 8088 established one of the most popular and long-lived computer ecosystems. For better or worse nowadays, the combination of it and a Microsoft-sourced operating system is still relevant thanks to this. To me personally, learning about it and tinkering with its descendants are the backdrop to my technical interests and knowledge. The PC was joined by the IBM 5160 Personal Computer XT (PC/XT) on 8^th March 1983. Over the original PC, the PC/XT adds a built-in hard drive and 3 expansion slots,^[20] otherwise being a very similar design. Whilst there are many factors as to why the PC succeeded, it was certainly aided by the inclusion of a competent keyboard.

The IBM Personal Computer Keyboard is often colloquially referred to as the “XT keyboard” or “Model F/XT”, using its association with the PC/XT to differentiate it from the later IBM Personal Computer AT Keyboard (“Model F/AT”). Components of the F/XT can also be referred to as being “XT”. At its core, the F/XT has a keyboard assembly much like the Datamaster’s, but it adopted a new serial-based protocol retroactively called IBM PC Mode 1 (also known as the “XT keyboard interface”). It emits 9-bit data packets facilitating IBM scancode set 1. The keycap nomenclature was changed to ASCII style instead of EBCDIC. In the aforementioned sketches, the interface and nomenclature were decided from the beginning, but it was also supposed they may cost-save the design by eliminating the leftmost function key bank,^[19] which of course did not come to fruition.

Two major variants of the original F/XT are known, both with their own nuances and differences or similarities to the Datamaster keyboards. They are referred to as Type 1 and Type 2.

The original design, Type 1, is very rare today. I have only seen examples with “81” date codes on either its chips or stamped somewhere inside the cover set, so I imagine it was phased out quickly in favour of Type 2. What is interesting is that IBM chose a different approach to Datamaster for its electronics. It has a 24×4 key matrix, and the controller is split into a capacitance-sensing driver and demultiplexer on the pad card, and a separate daughterboard with an Intel 8048 as its controller. The reason for this design may be explained by the IBM 5291/5292’s existence later.

A telltale sign from the outside is that its original cable’s 180° 5-pin DIN has no black plastic jacket like Type 2. The keyboard also has a reset line, which the IBM PC Technical Reference implies was not used^[23] and is entirely dropped by Type 2.

The Type 2 design is what most F/XTs actually are, and certainly by the time of PC/XT, the only one available. As already mentioned, their DIN plugs have a plastic jacket and lack a reset line. But much has changed internally too, perhaps anarchonistic in some ways. It reverts to using a 12×8 matrix like Datamaster’s, though configured differently. The controller is unified with the pad card once again, though of course still implementing the IBM PC Mode 1 interface.

The IBM 5291 Model 1 Display Station and 5292 Color Display Station were both 5250-family terminals announced on 6^th September 1982. They are plug-compatible with the ’70s IBM 5251 Display Station (model 11), ^[27][28] though the 5291 was intended as a low-cost replacement and the 5292 as a colour-enhanced supplement. 5291 Model 2 joined them on 2^nd October 1984,^[29] which was functionally identical to 5291-1 but smaller, intended to be more ergonomic, and with a relocatable screen, whereas 5291’s was tightly integrated into the base unit.

The 529X keyboard is a fun one. It’s like the F/XT but dialled up to eleven. It has considerable bezels just like the 5324’s, though with a slightly smaller ‘forehead’. But it also has absurd three-setting riser feet, the longest of an IBM keyboard I have seen. Perhaps both of these facts lend themselves to the 529X keyboard’s common nickname – the “bigfoot” Model F. The key nomenclature is back to being 5250-style given 5291’s and 5292’s application. The 529X keyboard CSA is the same as the Type 1 F/XT’s but without the controller daughtercard. Instead, the terminal itself would be its controller, making the keyboards effectively “brainless” on their own. This is where I can see why IBM designed the Type 1 F/XT differently from Datamaster, as alluded to before – more component commonality between higher-volume products. Both PC and (in my observations) at least 5291 were more so than Datamaster. Ultimately, the simpler, more elegant Type 2 solution won out for PCs at least.

The 5291-2 and 5292 shared the same completely detachable keyboard with a DA-15 plug at the end of it, dubbed the Type 2 529X keyboard. 5291-1’s Type 1 was instead always tethered, as its cable connected to a 14-pin IDC-style header inside the terminal, meaning the terminal has to be opened to detach it. The Type 2 is typically the far more common of the two today.

Of all the keyboards discussed today, the 529X keyboard was produced the longest. Lexmark, which was founded from IBM Information Products Corporation’s divestiture on 27^th March 1991,^[32] produced 529X keyboards into at least mid-1994. This makes it one of the last Model Fs in production. Lexmark also refurbished existing 529X keyboards into 1996, their final year in the keyboard business. Lexmark-produced and Lexmark-refurbished 529X keyboards can usually be distinguished by the latter being mislabelled “Model M” on their rear labels instead of “F”.

On 16^th Feburary 1984, IBM announced the 5155 Portable Personal Computer. It is essentially a PC/XT assembled into a luggable form and given a cute 9″ amber CRT display and one or two now-half-height 5¼” floppy diskette drives.^[34] It was IBM’s answer to the Compaq Portable, one of the first IBM PC clones that launched just under a year earlier and happened to be in the same form factor.

Compared to the original F/XT, the 5155’s version is housed in an all-plastic and two-tone cover set with a compartment to stuff its cable into, its base plate is a shiny, silvery aluminium instead of chromated steel, and it has a 6-pin modular plug instead of DIN. This results in the keyboard being considerably lighter and sounding somewhat higher-pitched than the original. The cover set can also be opened toolessly as the dark plastic bezel surrounding the keys can be lifted at any time. Electrically, the keyboard remains unchanged, and it can be passively adapted to 5-pin DIN if so desired.

When not in use, the keyboard should be mounted upright to the front of the 5155, covering its entire front. The keyboard has two clips towards the top that are used to release it and allow it to pivot down. You can additionally deploy its riser feet to completely detach the keyboard from the host PC.

The IBM Instruments Computer System 9000 (CS/9000, also known simply as the System 9000) was a Motorola 68000-powered laboratory computer introduced in May 1982. It was notable for its vertical stack including an integrated printer/plotter and a 57-key membrane touch panel. On 21^st February 1984, IBM introduced the System 9002 Desk-Top Computer and renamed the CS/9000 to the System 9001 Bench-Top Computer. In Q2 1985, the System 9003 Industrial-Floor Computer also joined them.^[36]

The 9002 was designed to be a space-saving alternative to the now-9001, shedding the printer/plotter and integrating the membrane touch panel into the keyboard itself (discussed more later).^[37] The 9003 was intended for use in manufacturing and process control applications, being able to operate as standalone or link with an IBM host processor via Systems Network Architecture (SNA). The entire 9000 family was withdrawn effective 2^nd December 1986,^[38] as ultimately all three members were unsuccessful in their respective markets.

The CS/9000 and 9001 used the IBM System 9000 Standard Keyboard (part number 4780898), which was designed to sit on top of a protrusion from the host computer. The Standard Keyboard notably sports a rectangular “IBM System 9000” branding, no flip-out riser feet, a cable that sprouts from the left, and a choke just before the cable’s DIN plug. Besides that, it is otherwise the same and electrically compatible with the standard F/XT.

The 9002’s keyboard -the IBM System 9002 Hybrid Keyboard – is a lot more interesting, though. As aforementioned, this keyboard notably integrates the 57-key membrane touch panel that was originally on the front of the 9001 into the keyboard. Presumably, “Hybrid” refers to its nature of having both a buckling-spring keyboard and the touch panel. The resultant design is very reminiscent of the IBM 104-key Model F Converged Keyboard used for the IBM 3290 Information Panel and 5085 Graphics Processor, with the panel taking the place of the 24-function-key bank but retaining its noticable bezels and two-stage riser feet. Despite this, the keyboard itself remains the F/XT-style assembly.

Whilst a 9002 computer has turned up in recent times, the keyboard has not, nor are there any photos of the keyboard from outside the 1980s. No one that I know owns one to test it or dig inside, thus its electronics remain a mystery to me. In my ‘professional’ opinion, the 9002 Hybrid Keyboard is the rarest Model F keyboard known, perhaps at most tied with another keyboard I plan to talk about soon. At least all the other known variants have been contemporaneously pictured…

At present, the 9003 or its keyboard also remain unseen to me. Computerworld’s reporting suggests the keyboard and 57-key touch panel were separate, just like the 9001.^[41]

The IBM Tempest PC (TPC) family are IBM PC-related products that were designed or modified to qualify under the TEMPEST program. TEMPEST is a NATO-recognised, United States National Security Agency specification regarding protecting against spying with electronic equipment. In particular, qualifying systems are designed not to radiate electromagnetic emanations to counter Van Eck phreaking. IBM referred to such systems as being ‘TEMPESTed’, usually adding special covers and filters to meet the requirements. The TPC family ultimately included the IBM 4450 TPC1 (TEMPESTed PC), 4455 TPC2 (PC/XT), 4456 TPC3 (3270 PC), 4459 TPC4 (PC/AT), and 4460 TPC5 (3270 PC/AT).^[42]

TPC1 and TPC2 indeed come with an F/XT variant named TPC Keyboard I and TPC Keyboard II, respectively, which despite their name and part numbers, do not differ from each other. Additionally, an IBM EMR Keyboard and EMR Keyboard II are known, although it was unclear to me what exactly those two were used for (ie., are they also TPC family products or are they intended for TEMPESTed IBM systems outside of the TPC family).

All four keyboards have clay covering the controller, hardier, non-coiled cables, and unique plugs compared to their ‘civilian’ counterparts. EMR I uniquely used a 180° 5-pin DIN plug not unlike the Type 1 F/XT’s, whereas the other three (and TPCs 3 through 5) used a DE-9 connector instead. This means EMR at least is not physically compatible with the known TPCs. EMR II, however, seems to be identical to TPC1 and TPC2.

Regarding the PCs themselves, at least based on Cathode Ray Dude’s analysis on the TPC4 and comparison with a standard PC/AT, they appear to be mostly as you would expect from the standard models inside. The motherboard, power supply and ISA card complement were typical, and the differences lay with the reinforced outer casing, the power switch location and function, and some buffer space between the rear I/O and the actual back of the TPC.^[49]

ASK. Admiral Shark’s Keyboards original content. License/note: CC BY-NC-SA 4.0.