The early days of personal computers. Stephen B. Gray.
Twenty years ago, while I was the computers editor on Electronics magazine at McGraw-Hill, I realized there was much I could learn from building a computer. It didn't take long to find out how difficult it was just to get started. There were no kits, no "cookbooks." Computer textbooks usually contained partial schematics, but none told how to connect the various sections.
After several years of trying to build a digital computer in my spare time, I began to realize how difficult it must be for other hobbyists. So, to solicit information to help me build a machine and to share what little information I had been able to learn on my own, I sent a letter to seven electronics and computer trade magazines and three hobby publications on May 5, 1966:
This is an invitation to those readers who are building their own computers to join the Amateur Computer Society, a nonprofit group open to anyone interested in building and operating a digital computer that will at least perform automatic multiplication and division, or is of a comparable complexity.
The society publishes a bimonthly newsletter containing problems and answers, information about where to get parts and schematics and cheap ICs, and articles on subjects such as Teletype equipment and checking out magnetic cores.
Will interested readers please write to me, giving details on their proposed or in-the-works computers, such as word length, number of instructions, sources of parts and schematics, clever solutions to previous problems, etc.? Response to the Letter
Five of the magazines printed some or all of the letter, and responses began to arrive. The original idea of the Amateur Computer Society, or ACS, was a membership organization with chapters and a newsletter or two. But the people who wrote in were so widely scattered that local chapters never got beyond the idea stage.
Initially, more than 160 men (but not one woman) wrote from five countries and 27 states, and 110 eventually became early "members" of the ACS, although the most they got was the newsletter--$3 for the 11 issues in the first volume, from August 1966 to December 1968; $3 for the 12 issues of Volume II, from April 1969 to March 1972; and $5 for the 15 issues of Volume III, from June 1972 to June 1976. Only two issues of Volume IV were published: August and December 1976; the ACS Newsletter was then discontinued, with these words:
"Times have changed, and now that kits are so prevalent, there are other publications that serve the readers' purpose better than the ACS Newsletter. Also, the ACS Newsletter always depended heavily upon reader input, and this input has dwindled. . . . Thank you for your support over the last 10-1/2 years. It was fun while it lasted."
ACS membership never totalled more than a few hundred. Nor did I try actively to increase the number, because of the work involved in producing even a few hundred copies of each issue in my spare time. I was doing all the work, including typing, collating, folding, stuffing, and stamping. Had there been enough potential advertisers, the newsletter might have been turned into a magazine, but up until 1974 (and even later) there weren't enough to permit starting up a magazine devoted to computer-building.
Each of the first half dozen issues of the ACS Newsletter was devoted mostly to an individual topic such as sources of schematics, input/output (mainly Teletype), logic circuits, memory, designing a computer kit for the ACS, mounting and interconnections, reference sources (where to find articles and books about computers), etc.
Responses from prospective members ranged all the way from "I've been thinking about building a computer for some time" (two dozen of these) through "I have the shift registers completed" (a dozen of these) to "I've build a computer and am now programming it" (two of these). Building a Computer in 1966
Back in the mid-sixties, to build a simple computer accumulator, which could do no more than add successive inputs, using toggle switches for input and lamps for output, cost several dollars per bit. To build an extremely simple "computer" with four-bit words and without memory, and which divided the easy way (by repeated subtraction without shifting), could cost two or three hundred dollars.
Used vacuum tube computers were occasionally available, but such machines brought with them problems of size, power requirements, air-conditioning, and tube replacement costs.
Used transistor computers were seldom available at a price a hobbyist could afford; a Recomp III, even at five percent of its original cost, was still $4750. The cheapest third generation computer was still expensive; a PDP-8/E, made by Digital Equipment Corp., cost $5000 without a Teletype.
Building one's own computer was such a complicated undertaking that very few were ever completed, and nearly all of those were built by electronics engineers working in the data processing industry.
The main problem in building a computer was (and still is) the many technologies involved. Computer companies had specialists in logic, input/output, core memory, mass memory, peripherals, and other areas. To build one's own computer required learning a great deal about each one.
If the computer hobbyist was an electronics engineer working for a computer manufacturer, he could drop in on a friend down the hall or in the next building and ask what kind of drivers might be needed for a core memory with such-and-such specs. Most hobbyists had no such resources.
In addition to having to learn a great deal about computer electronics, the hobbyist also had to get into mechanical areas such as packaging, back-plane wiring, metal working, plastics, and many others. Memory
Magnetic drum memories were sometimes available, but usually from equipment that had been sledgehammered before being discarded, and thus were often damaged. Read/write heads only a few thousandths of an inch out of alignment can scratch a drum surface beyond repair.
A variety of core memories was available, but, as one computer memory engineer put it at the time, "The used and surplus memory planes I have seen on the market are real antiques. There are several possible reasons for core plans being in the reject bin. One is that too many cores in the matrix needed to be replaced. Another is that too many were replaced to pass the quality control requirements of a given project." In seeking core memories, the buyer thus had to be extremely knowledgeable.
As for tape drives, one surplus company advertised a Potter model without electronics or even a rack for $150. The Average ACS Computer
In the seventh issue of the ACS Newsletter, dated November 1967, a survey form was included asking for details of each member's computer, whether in the works or only in the planning stages. The next newsletter gave the survey results.
Most of those who returned the survey form planned on using core memory, the hardest part of the computer to get working; most wanted 4K or 8K words, but few got core up and running.
Teletype was the most common input/output device. Some members also used paper tape, Nixie readout tubes, magnetic tape, and electromechanical typewriters.
Clock speeds of the amateur computers averaged 0.5 MHz.
Generally speaking, beginning amateurs hoped to use a large number of instructions, between 50 and 100. Those who had gotten fairly well into the construction used no more than 11 to 34.
The average length of data words and instruction words was 12 bits for each. The speed required for addition ranged from eight microseconds down to ten milliseconds.
The number of registers ranged from two to 11, with three the most popular. One member projected two registers for memory, two for data, one for operation code, and five for address.
As to "cost so far," the range was from zero to $1500, with an average (among those reporting a cost) of $650. For "estimated cost when complete," the range was from $300 to "over $10,000," with an average of $2100. Education
One of the most significant areas on the survey was education. Most of those responding had at least one technical degree. After noting this high level of education among the membership, the ACS Newsletter commented:
"Because the greatest majority of those sending in the survey have technical degrees, and because those who sent it in are among those who have advanced the most with their computers, it seems that lack of a technical education is holding back many ACS members from pushing ahead with their machines, or perhaps from just getting started. Unlike amateur radio, there just isn't enough circuit level information available on how to build computers.c
Several members gave progress reports on the survey form: "Teletype controller and memory operational. Can presently transfer data from Teletype to register to memory and back. Delay line memory stability problems solved--successfully retrieved data after eight hours." Later, he had to drop the delay line memory, because of poor long term stability. It would work fine for a while, but later would shift by one or two bits, throwing it out of synchronization with the external clock. He tried core memory, then bought a used magnetic drum memory taken from an airborne computer for $100, giving him 8K words. Innovate or Copy?
Many non-engineer ACS members, unable to design their own computers, tried copying existing designs. Several patterned their instruction set after that of the IBM 1401 or IBM 1620 computer. One Long Island member had software similar to that of the 1620 and hoped that his "IBM 1620 Model III" would be about 25 percent faster than IBM's 1620 Mod II, and would have all of its 60-plus instructions.
Most members who borrowed an instruction set already in use were copying that of the PDP-8 family, manufactured by Digital Equipment Corp. By that time, DEC had sold more than 10,000 of the PDP-8, which was attractive because of its comparatively low price, variety of programs available, and a simple yet powerful set of instructions. Completed Computers
Only two of those surveyed reported being anywhere near completion of their computers.
Jim Sutherland, an engineer with Westinghouse in Pittsburgh, noted that his Echo IV took a year to build and would need ten years to program. Echo IV was seven feet long, one and a half feet deep, and six feet high. The central processor was complete but, as with all amateur computers, the input/output system was still growing.
Echo IV used 2N404 transistors and NOR logic elements; the NOR gates were originally used in process control systems built by Westinghouse a dozen years before, and had been declared scrap. The gates were mounted on etched circuit boards with 35-pin connectors. A total of 120 boards of 16 types was used in the entire system.
The memory unit, an Ampex 4096-RQ-30A, came from an obsolete process control computer. Memory cycle time was six microseconds, but since the NOR gates required from one to three microseconds to switch, the add time was pushed up to 216 microseconds.
Echo IV had four flip-flop registers, and three registers in core memory. There was 8K words of 15-bit core memory; clock speed was 160 KHz; and there were 18 instructions.
Input was by six alphanumeric control keyboards, eight-channel paper tape reader, 15 interrupts, and 75 relay-contact closures. Output included two printers, 60 relay-contact closures, eight-channel paper tape punch, and four digital clocks. Interconnections were wire wrapped. The ACS Newsletter
The first volume of the ACS Newsletter (1966-68) provided information about computer trainers, Teletype equipment, circuit boards, ICs, kits, and details of computers built by members. The second volume (1969-72) included information about Nixie readout tubes, core memory, buying reject ICs, memory drums, and the MITS desk calculator kit. It also described the first commercial computer kit, the National Radio Institute NRI 832 (1971). this kit, designed by Lou Frenzel who later moved to Heathkit, had 52 TTL Ics, 17 storage locations, and 15 instructions. the memory consisted of slide switches to simplify the teaching of bit storage.
Volume III (1972-76) looked into Don Tarbell's computer (which multiplied a 140-digit number by itself in 40 seconds), Intel's 4004, and 8008 chips, the Scelbi-8H kit, Radio-Electronics Mark-8 kit, Hal Chamberlain's HAL-4096 computer, and several of the early commercial kits and boards. Only two issues were published of Volume IV in 1976; these dealt exclusively with commercial kits and peripherals as well as several products shown at the first personal computing show in Atlantic City. A Flattering Accolade
An article by Sol Libes on "The First Ten Years of Amateur Computing"
(Byte, July 1978, pp. 64-71) was taken largely from items in the ACS Newsletter. Written "to set the record straight," because many people thought personal computing "started only two or three years ago, with the introduction of the Altair 8800 by MITS," it continued:
"If one could find a specific date for the birth of personal computing, it would be May 5, 1966. For it was on that date that Stephen B. Gray founded the Amateur Computer Society and began publishing a quarterly called the ACS Newsletter."
Very flattering, but not true; it was like saying Henry Ford was the father of the automobile. If anything, the ACS Newsletter was the first publication in the world about personal computers. Apparently it is also the only detailed source about the early days; the Smithsonian Institution has asked for a set of the newsletters.
In 1966-67, hoping to get financial backing for full time operation of the Amateur Computer Society and for a projected lower level Amateur Digital Society, I wrote to several foundations and large computer manufacturers, but to no avail. The same thing happened several years later when I sought financing for several full-time years to write the detailed history of the early days of personal computing.
The following is taken from the only chapter written for that unpublished history, in which the Scelbi-8H was to have been an important milestone. First Advertised Personal Computer Using a Microprocessor
The first advertisement for a personal computer based on a microprocessor appeared in the March 1974 issue (p. 154) of QST, an amateur radio magazine. The ad was for the Scelbi-8H, manufactured by Scelbi Computer Consulting, Inc., of Milford, CT.
Scelbi's founding father, Nat Wadsworth, was a design engineer with General DataComm Industries in Danbury, CT, when Intel gave a seminar nearby on the 8008 microprocessor. But when he and several other young engineers tried to talk management into simplifying products with the 8008, they got nowhere.
Wadsworth, intrigued by the capabilities of the 8008, cornered several other company engineers and asked, "Why don't we design a nice little computer and each build our own to use at home?" Two of them agreed, and Wadsworth and Robert Findley designed most of the system.
"We had planned on building three computers," Wadsworth said later, "because there were three of us in the initial group. But the work on the first prototype was so exhausting that we decided to lay out actual PC boards and have a small quantity made up. Thus, for about the same amount of time involved, we would have a better quality unit. Also, I think the idea of going commercial had always been in the back of my mind." Incorporating Scelbi
Wadsworth quit his job in the summer of 1973 to work full time on the computer. Scelbi was incorprated in August; the name stood for Scientific, Electronic, and Biological, "because we saw this computer as capable of going into any of those areas." The full name was never used; legally the company was Scelbi Computer Consulting, Inc. "Most people called it Skell-bee, but we pronounced it Sell-bee."
Wadsworth worked up to 18 hours a day turning the prototype board into commercial PC boards, and private investors started lining up. But several months later, in November 1973, at the age of 30, Wadsworth had a heart attack. This stopped everything for a while, and the investors disappeared. But Wadsworth recovered, finished the design, and the company started advertising. "We chose QST because we knew that many hams were dyed-in-the-wool electronic enthusiasts." All that came with the computer were the assembly instructions; the user was expected to know--or learn--something about programming.
Response to the ad was large, so Findley joined the company, others were hired, and then in May 1974 Wadsworth had a second heart attack. This "pretty well removed Scelbi from having a chance at making it big in those days, and it never became a major factor. But we continued to limp along. It became--for me, anyway--a different thing. I started doing it more for love, for the joy of still being alive, than for any ideal of a commercial enterprise."
While in the hospital, Wadsworth started writing a book, Machine Language Programming for the 8008. "We published it ourselves, on an offset press, as it came off a Teletype. It was absolutely horrendous esthetically, but to our utter amazement, when we casually advertised it, it sold something like 1500 copies within a month or two at $20 apiece.
"Soon we were taking in more from the manual than we were on computers! I figured out that if Scelbi was to keep going, it wasn't to be as a computer manufacturer. We stopped advertising the computer to concentrate on books and software.
"Altogether we sold about 200 computers--half assembled, half kits." Half were Scelbi-8H hobby computers with up to 4K of memory; the rest, differing mainly because of more memory (up to 16K), were Scelbi-8B business computers. The Scelbi-8H first went on sale March 1974, the 8B in April 1975.
"For that time, we had a very sophisticated system, a complete system. We had a tape cassette interface that actually worked a lot better than the ones MITS started selling. We had a CRT based on an oscilloscope, and Teletype interface, and we developed a combination monitor, editor, and assembler in ROM."
Scelbi began to "develop software products to support the 8H and the 8B, but which were put in book form: our editors, monitors, and assembler. We sold thousands of copies. After the success of those books, we went on to modify them for the 8080, for the Altair, the Imsai, whatever was out at that time." Cookbooks
"During this evolutionary period, we got the idea for the cookbooks. The first two were written by Bob Findley. The first was for the 8080, patterned after my machine language programming book. What we had was an engineer's handbook that presented the instruction set and utility routines. We had a full floating point package in there, which I had developed for the 8008. That was an extremely successful book. Tens of thousands of the cookbooks have been sold in virtually every edition we put out. It became a classic.
"We lost money the first several years of operation at an average of $500 per computer. We did not start making money in this company until we were in publishing, and then we made up for the losses. Otherwise we wouldn't have been able to stay in business." The Scelbi Influence
Asked what he thought was his influence on personal computers, Wadsworth replied in 1981, "Well, I think I had a lot of influence I never knew about. I know lots of instances of companies that were started by individuals who were initially dealing with me. For example, one of our first systems was sold to a fellow from the Midwest. We then had conversations with them which we later determined were nothing more than snooping on our marketing and everything. And they started a company producing microcomputers.
"One thing I've always found amusing is that when MITS began advertising, they claimed they had the first computer for personal use. And then in their newsletter they offered to trade our computers for theirs."
Wadsworth sold the rights to publish Scelbi books to other publishers to concentrate on a newsletter for pocket computers.
Did he get any recognition for being one of the pioneers? "Not really. It has always escaped people's attention. Certainly MITS wasn't going to say, 'Well, really, Scelbi was up there.' Jonathan Titus and those boys, they say, had the first mass produced computer, the Mark-8, which was not really a commercial product. It was a hobby thing.
"I don't think people really know today that we were the first commercial producer of computers. And I don't stay up nights worrying about that, because I never cared about being first. It was what I was interested in doing, and we went ahead and did it, and we would have done it a lot more successfully had I not had the health problem. It could have been a whole different story." The MITS Altair
The $179 MITS desktop calculator kit was soon overwhelmed by fully assembled Japanese calculators that were cheaper than the kit. To stay alive, MITS brought out a computer kit for which there was no time for the further development it could have benefitted from. Although the Altair was not easy to assemble or use, MITS knew how to market it with four-color ads in professional and hobby magazines.
However, that's another story, to be found elsewhere in these pages.