Full of sound and fury, signifying everything
Don Morris was fuming—mad enough to leave the Calculator Products Division (CPD)—because he’d just overheard his boss, Rex James, discussing opportunities in the calculator R&D lab with another engineer. Both James and the unnamed engineer belonged to the Church of Jesus Christ of Latter Day Saints (also known as the LDS church or the Mormons). The engineer had complained to James that more managerial promotions should go to LDS members. James agreed, and counseled patience. Morris, who had worked with James on earlier calculator projects, immediately planned his exit from the calculator lab. He walked down the hall to HP’s Loveland Instrument Division (LID) to check into transfer opportunities.
CPD lab manager Bob Watson, who was a member of the LDS church himself, found out that Don Morris wanted to leave and why. He met with Morris and convinced him that James’ overheard comments were wholly inappropriate and that HP management in general and Watson in particular didn’t share James’ stated views. Watson promised to talk to James and deal with this issue.
Later, Morris himself confronted James, in language both loud and quite profane. James put his head down on his desk for a few moments, then raised it and admitted being totally out of line. James promised to amend his ways and asked for Morris’ forgiveness. He also noted that Morris’ loud use of profanity in the very public lab environment (HP had no offices, only cubes) was unprofessional. Morris briefly argued the observation, but in the end, both men apologized to and forgave each other. Both were apparently changed by the confrontation.
These events occurred just before work on HP’s third-generation desktop computers was about to commence. They would ultimately thrust Don Morris into the role of project manager for the HP 9825A desktop calculator, and would make him it’s Godfather.
In 1973, James and Morris were coming off a string of three related calculator projects: the HP 9805, HP 46, and HP 81. All three of these desktop machines shared a plastic case made of structural foam. They were the first calculators to employ structural foam, which would also be used in the HP 9825A. Earlier calculators employed metal cases made of cast or extruded aluminum.
The HP 9805 was a statistical calculator, based on Loveland’s own architectural design. It was not a user-programmable machine; it shared no electronics with earlier HP desktop calculators; and it was a short-lived, technological dead end for CPD. The HP 46 used the same Mostek-produced LSI chips developed for the HP 45 handheld calculator but new ROMs with code written specifically for the HP 46 added hard-copy printing. Similarly, the HP 81 was a desktop version of the HP 80 handheld business calculator and used the LSI chips from that machine. None of these three desktop machines was especially successful.
After creating this trio of machines, James became a section manager and Morris became one of his project managers for a new desktop calculator, the HP 9815, which would become the third-generation version of HP’s original desktop calculator, Tom Osborne’s brainchild, the HP 9100A. The second-generation version of this machine, the HP 9810A, was significantly slower than the HP 9100 so there was plenty of room for improvement.
Morris and James flew from Colorado to Phoenix, Arizona to meet with Motorola’s Semiconductor Products Division to hear about a proposal for a new microprocessor. Intel had only introduced the first commercial microprocessor in late 1971 and Motorola needed to catch up. The microprocessor Motorola would propose to HP would be the 8-bit, NMOS MC6800. It was destined to become the heart of the HP 9815. On the flight to Phoenix, the two men agreed that they wanted and needed a solid win after the less-than-satisfactory results of the HP 9805/46/81 projects. They were a team and the past, if not forgotten, was the past.
CJ and out
After they returned from Phoenix, James was scheduled to go into a CPD R&D lab staffing meeting. CPD’s third-generation calculator projects were about to ramp up. As part of their effort to achieve that desired solid win, James asked Morris to create a ranked list of the top CPD lab engineers to work on the HP 9815 project. Morris created a list of engineers that he most desired on the project. He made the list twice as long as needed to ensure that he got good team members, because he expected the “draft” to alternately give the top engineers to James and to the other section manager in the lab, Fred Wenninger, who was charged with developing the successor to the HP 9820A desktop calculator.
When James returned from the staffing meeting, he had gotten every engineer on his HP 9815 design team that Morris had wanted. However, he no longer had Morris as project manager. He’d traded Morris to Wenninger, with Watson’s full approval, in exchange for Morris’ “A team.” Watson had approved Morris’ transfer to Wenninger’s section because he was aware of the earlier religion-induced friction between James and Morris, even though that was old news and the two men had reconciled.
Old news or not, Morris became the HP 9825A project manager under section manager Fred Wenninger. James went off with his “A team” to develop the HP 9815—code named CJ. The moniker CJ stood for Caesar Julius because the abbreviation for Julius Caesar, JC, could be “misinterpreted” as a religious reference to Jesus Christ and religious issues had already occupied too much of the division’s time and energy.
The summer of ‘73
At first, Morris had no project engineers because the ones Rex James didn’t get for the CJ project were all still working on other projects. That gave Morris the luxury of the summer of 1973 to conduct some market research. Historically, HP specialized in next-bench marketing, which meant asking the engineer at the next bench for opinions on new product ideas. Morris teamed with Dick Barney, who was spending a tour in CPD marketing as the product manager for the HP 9810/20/30 interface cards after working in the lab on the HP 9866A thermal page printer. In the summer of ‘73, Morris asked Barney to set up customer visits with the best HP field salesman in each sales region. Morris asked to visit three types of customers:
- A lost sale (lost to Wang, Digital Equipment, or some other competitor)
- A happy HP 9820A or HP 9830A customer
- An unhappy customer for the HP 9820A or HP 9830A
Morris and Barney then took an extremely significant road trip to visit these customers and prospects. During this trip, they saw many different systems with HP 9820A and HP 9830A desktop calculators at the center. Some of these systems were traditional computer systems with mass storage and printers attached. These systems were performing computational tasks such as accounting and billing.
Other systems surrounded the desktop calculator with more exotic peripherals such as voltmeters, frequency counters, and other sorts of instrumentation. In these systems, the desktop calculator’s role was that of controller. The controller directed the operation of various data-gathering devices; accumulated and processed the data; and then either stored the processed data or presented it on a display, a printer, or a plotter.
True market segmentation
It’s at this point that Morris feels he made his first major contribution to the development of HP’s desktop machines with Dick Barney’s help. Each evening, they discussed the events of the day. Eventually, a coherent picture started to form in Morris’s mind that split the desktop calculator market into two segments: computational machines and controllers. This segmentation implicitly included price segmentation, which is the way the HP 9810/9820/9830 series had been developed. With their limited key-per-function keyboards and abbreviated displays, the lower-priced, second-generation HP 9810A and HP 9820A desktop calculators readily fit the controller role. The HP 9830A—with its full keyboard, wide alphanumeric display, advanced BASIC language, HP 9866A thermal page printer, and large hard disk drive—better fit the computational role.
All of the second-generation machines had similar hardware I/O abilities, because they were all based on the same processor hardware. However, they had very different software I/O abilities because the I/O features in the HP 9810A, HP 9820A, and HP 9830A desktop calculators were implemented in option ROMs that varied widely in ability from machine to machine. The HP 9830A had more native I/O abilities than the two lower-cost machines, because of its BASIC programming language, and it required more advanced I/O option ROMs to control advanced peripherals such as its HP 9880 hard-disk controller.
The substantive differences in the HP 9830’s I/O abilities further reinforced Morris’ new segmentation model. This segmentation was instrumental in helping to define the next-generation machines. The planned third-generation HP 9815A and HP 9825A desktop calculators would fall into the controller category. An even larger machine—with more mass storage, a full QWERTY keyboard, and a much better display—would be sold as a high-end computational machine. That machine, code named QWERT for its QWERTY keyboard, would be sold as the HP 9845.
Based on this insight, lab manager Bob Watson created three sections within the division’s R&D lab to develop mutually supportive products. A computational section would develop the high-end desktop calculators. A controller section would develop the low- and middle-segment desktop calculators. Finally, a peripherals section would develop new peripheral devices and interface cards to support all of the division’s computational desktop machines and controllers.
Saved by software
Software veteran Jack Walden was slated to become the HP 9825’s software project manager, but he was tied up with an unfinished project. Walden had been an associate professor at Oklahoma State and was on Morris’ doctoral committee, but he left the university in 1969 to join a classmate of his, Fred Wenninger, who happened to be working at HP in Loveland. Walden’s first assignment was coding software for the HP 9810A.
While the HP 9825’s software was on hold waiting for Walden to free up, an engineer named Wayne Covington, who had worked with Fred Wenninger and Ed Olander on the early simulation studies of HPL for the HP 9820 and on BASIC language firmware for the HP 9830 showed up to discuss a new language technology with Don Morris.
Covington had become excited about a new table-driven language parser that had been developed in 1969 by Franklin DeRemer for his PhD dissertation at MIT. This parser is now called the LALR (look-ahead, left-to-right) parser. It is particularly well known by computer language specialists today because of the YACC (yet another compiler-compiler) and Bison parser generators, which appeared in 1985 and 1988 respectively. But DeRemer’s work was barely known in 1973. He didn’t start to publish articles about his technique until October, 1970. Covington wanted to use DeRemer’s parser in the HP 9825. The parser would allow more separation between the machine’s high-level language structure and its software bowels.
Up to this point, that was absolutely not the way HP Loveland was developing its machines. The preferred software structure, which had been defined by Tom Osborne’s original HP 9100 design, was to very tightly knit the software together, all the way from the low-level hardware drivers to the high-level operations initiated by the keyboard. Tightly interwoven code reduced the need for costly ROM memory but it locked the software up tight. Once jelled, software changes were either very difficult or impossible.
Covington convinced Morris that if the HP 9825 software ever needed to be changed, that a more modular and layered design approach would become vital, so it was best to start out that way. Morris consulted with Jack Walden, who agreed that Covington’s recommendation had merit. However, Walden was still not available to work on the implementation and Morris didn’t feel that he could hand the entire project to Covington, who had worked on Fred Wenninger’s ill-fated Synchro-Rectro-Flash analog computer, so he went looking for someone else to honcho the HP 9825 software project.
He found Chris Christopher working in production engineering, HP’s version of purgatory (because real engineers design). Christopher may have been on probation in production engineering or he may just have fallen from grace in some engineering manager’s eye, but “he impressed me immensely,” said Morris.
Christopher had proven himself as someone who could get software into production and out the door. He had developed the software for the HP 9830’s tape-cassette mechanism and for the HP9830’s disk-based HP 9880 Mass Memory storage system. Morris made Christopher the software project leader for the HP 9825 and added Covington to his team. (Eventually, Chris Christopher would become a senior vice president and general manager of HP’s Business PC Global Business Unit and its Workstation Global Business.)
Morris also added another Synchro-Rectro-Flash alum on the HP 9825 team. He made Geoff Chance the project manager for the HP 9825’s I/O subsystem. Chance had developed “The Link,” an early I/O subsystem for HP’s original desktop calculator, the HP 9100. During his talks with Don Morris on the summer of ‘73 tour, Dick Barney became so excited about the potential I/O capabilities of the third-generation machines that he decided to leave the marketing department to return to the R&D lab so that he could design some of those new I/O products. Barney and Mike Kolesar, who would design the hardware for the HP 98032A parallel interface card and the HP 98034A HPIB interface card, worked with John Nairn to define just what would turn the HP 9825A into Don Morris’ vision of a controller. Nairn would then go on to create the I/O ROMs for the HP 9825A that would transform the vision into reality.
Going for baroque and then throwing out Barney’s baby
It turned out that Wayne Covington’s idea to use DeRemer’s table-driven parser for HPL in the HP 9825 was indeed fortuitous. Originally, the HP 9825 was being designed to follow in its predecessor’s footsteps. It was supposed to map functions directly to individual keyboard keys like the HP 9820. In this sort of machine, no parser is needed. Each pressed key directly represents a language token so the entered program is self parsing. However, the HP 9825 development team began to discover the tremendous and unexpected capability and flexibility provided by the microprocessor-powered hardware. Consequently, the definition and complexity of the HP 9825’s programming language began to swell with the result that each keyboard key acquired a second and then a third function overlay, which required the addition of shift keys and operating modes. The entire concept of a key-per-function was quickly becoming overly baroque and unworkable. Morris eventually decided to scrap the key-per-function design and adopt a QWERTY typewriter-like keyboard similar to the one used on the HP 9830. That decision pointed the team towards using BASIC instead of HPL.
This industrial-design mockup of an HP 9825 shows the original non-QWERTY, key-per-function keyboard that gave the machine its code name: Keeper (key per function). Eventually, the HP 9825’s planned abilities grew so much that even three functions mapped to each key could not accommodate all of the planned capabilities.
Photo courtesy of Fred Wenninger
That problem and Morris’ decision gave rise to an even bigger problem—a political problem. Barney Oliver, the head of HP Labs, absolutely loved the HP 9820 and its HPL language. Oliver had made several substantial contributions to the HP 9820’s programming language during the machine’s development, so he considered HPL to be, in some sense, “his” language. One of the things Oliver really liked about HPL was its ability to infer multiplication operations. So, for example, you could multiply the three variables A, B, and C simply by typing “ABC.” That’s something that’s easy to manage in software if each variable (A, B, and C in this case) has its own key (HPL had 26 simple variables, one for each letter of the alphabet), but it became unworkable with a QWERTY keyboard because there is no way of telling if “PRINT” is a command to print something or if it’s the implied product of five simple variables: P, R, I, N, and T.
Consequently, Morris’ decision to scrap the key-per-function keyboard and use a QWERTY keyboard instead meant that HPL had to be scrapped as well, in favor of a more conventional programming language like BASIC, which had already been implemented in the HP 9830. Barney Oliver wasn’t buying the decision to terminate his beloved HPL and he had the clout to reverse Morris’ decision if he really wanted to do so. Morris writes, “My bosses knew what I didn’t. An angry Barney was very dangerous.” Sure enough, Bill and Dave’s “hatchet man” Ralph Lee soon made the trip from Palo Alto to Loveland for a special meeting to review Morris’ decision. Barney Oliver accompanied Lee.
The meeting took an entire day. Morris presented his arguments and his reasons for the decision. Oliver apparently said nothing all day, but he did listen. At the end of the meeting, Lee agreed with the decision to drop the HP 9825’s key-per-function straightjacket, even if that meant driving a stake through the heart of HPL as well. Oliver would not easily give up his baby however. As the meeting ended, he asked Morris if HPL might be saved if Oliver found a way to solve the problem with the implied-multiply syntax. “Of course,” said Morris, who was more concerned with the keyboard practicalities than with the religious aspects of programming languages.
By that evening, Oliver telephoned Morris with the fix. “Variables are capitals and functions are written in lower case,” he said. With that simple syntactic change, Oliver saved HPL and made it compatible with a conventional computer keyboard. Thanks to Covington’s parser technology, the entire language was running on a simulator two days after Oliver’s brainstorm.
NMOS II runs off the rails, almost
Meanwhile, things weren’t going well for the NMOS II effort, which meant that the HP 9825’s hybrid microprocessor was in trouble. Larry Lopp, who had been a central figure in getting HP’s original NMOS process up and running had just left Loveland and moved to HP’s Cupertino division to work on the silicon-on-sapphire process for the Cupertino division’s minicomputer-in-a-desk, code named Amigo. Cupertino was one of several HP divisions that had noted HP Loveland’s success with small computers and Amigo was one of several projects designed to go after much the same market.
The potential for disaster grew as more HP divisions decided to compete against each other in the desktop computer market. Dave Packard, whose vision had gotten HP into computers in the first place, stepped in. Packard knew that the desktop market was too important to HP for the company to make a mistake. At the same time, the Amigo and Keeper teams were certain that the other team would fail. In a Solomon-like move, Packard allowed both projects to continue, with the proviso that the products might compete in the market, but the product-development teams had to cooperate and help each other out. Thus HP would get the technology development it needed for future projects and the market could decide which team had the better product concept.
As the NMOS II work continued to founder, the pressure grew to abandon it for Larry Lopp’s SOS process technology. Lab manager Bob Watson demanded a decision. Section manager Fred Wenninger deferred the decision to Morris because it was a crucial decision for the HP 9825 and that was Morris’ baby. Morris picked NMOS II, which eventually turned out to be a winner, but only after the NMOS II team overcame several very large problems.
HP 9825 Evolution
When production HP 9825As first shipped in mid 1976, they did so with only 8 kbytes of RAM, of which the user got only 6844 bytes to hold programs and data. The original HP 9825A also had a clicky “Cricket” keyboard derived from design of HP’s handheld calculators. The first customer shipments of the HP 9825A had been delayed for a month so that the memory boards could be upgraded to use the new 4-kbit dynamic RAMs that were just starting to ship in volume. This delay significantly enhanced the early machine’s usability by doubling the RAM capacity of the low-end model. Early HP 9825A programmers would probably have found machines equipped with only 2748 bytes of RAM to be far too limited. HP also introduced memory-upgrade boards that increased the HP 9825A’s RAM capacity to 16, 24, or 32 kbytes, which was the maximum amount of RAM the machine could hold. (The other half of the HP 9825’s address space was consumed by language and option ROMs.) However, the 32-kbyte option was not especially popular. To get that last 8 kbytes of RAM, you had to give up the String, General I/O, and Advanced I/O ROMs due to address conflicts.
The “Cricket” keyboard was an attempt to make reduced-cost keyboard for the HP 9815A and 9825A. Work on the Cricket keyboard technology had started before Don Morris made the fateful decision to deviate from the “key-per-function” orientation and to put a full QWERTY keyboard on the HP 9825A. The Cricket keyboard would probably have worked well on a key-per-function machine, but it was a touch-typist’s nightmare because key travel was minimal. In addition, the keyboards had reliability problems. Within two years, HP redesigned the keyboard to use high-quality, full-travel keyboards manufactured by Cherry, a well-respected switch manufacturer. Many existing HP 9825A owners ordered keyboard retrofit kits for their machines when HP announced the kit’s availability.
It wasn’t long before customers were soon bumping up against the HP 9825A’s RAM limits. As customers often do, many HP 9825A owners were demanding more. HP initiated a project, code named “Skoal,” to find a way to cram more RAM into the HP 9825A’s already filled-to-capacity memory space. The result was the HP 9825T, which incorporated a special state machine that looked at each memory-access operation and determined if the operation was directed to RAM or ROM. Obviously, store operations were always directed at RAM (with one strange exception caused by a latent bug in the Systems Programming ROM). Memory reads had to be analyzed using knowledge about the operating-system code itself. That knowledge was coded into a ROM in the state machine that guided the choice between RAM and ROM for each memory access. The state machine implemented a simple bank-switching scheme, with ROM in one 64-kbyte bank and RAM in the other. Addition of the memory-access state machine boosted the HP 9825T’s maximum RAM capacity to an amazing 63 kbytes.
Over the years, HP developed a growing number of peripheral products for the HP 9825 desktop computers including printers, plotters, hard- and floppy-disk drives, and digitizers. In addition, the HP 9825 more than realized Don Morris’ vision of an instrument controller. HP’s Loveland Instrument Division incorporated the HP 9825A, B, and T models into many automated test systems. The HP 9825’s excellent control capabilities over the HP 98034A HPIB interface made it the premiere system controller for rack-and-stack instrument systems of the late 1970s and 1980s.
HP finally retired the HP 9825 by 1983. The machine was hugely successful by any standard. It firmly placed HP in the top spot as the vendor of choice for instrument controllers and automated test systems. By some reports, HP sold around 28,000 HP 9825 desktop computers. That’s a huge number of machines for the times, although it represents much less than a day’s worth of worldwide PC sales today.
The HP 9825 was supplanted by the HP 9826, introduced in 1982. The HP 9826 was based on Motorola’s MC68000 microprocessor. Originally, the HP 9826 was not supposed to run HPL; it was scheduled to run HP’s Rocky Mountain Basic. However, the BASIC programming project for the HP 9826 ran into problems and a tiger team was assembled to add HPL to the HP 9826. The project only took four people half a year to complete, although it was canceled four times over the course of that six months. However, that’s a story for another Web page.
Information on this page came from personal discussions with Don Morris, Geoff Chance, John Nairn, Dyke Shaffer, and Joe Beyers.