[cctalk] Re: Intel 4004

The F14 flight control (CADC) computer was a chipset, with different functional aspects built into each chip. The design was done by Garrett AirResearch. The requirements of the system were quite arduous, and thus the computer was reasonably powerful for its time, especially considering its size and power supply requirements. Once the logic was all tried and true via bread boarding the system, The logic was given to American Micro-systems, Inc., (AMI) who laid out and fabricated the chips. At the time, AMI was one of the few companies that could make large scale MOS ICs. AMI did a lot of secret work for the US Government which is how it got its start in MOS LSI, and a lot of AMI's early history is somewhat shrouded in mystery because most of the work they did was secret. It appears that AMI's first MOS LSI calculator chipset was for Smith Corona Marchant (SCM), with an eight-chip set that was partitioned into two ROMs, a control chip that decoded the microcode in the ROMs into control signals, a digit parallel, serial in/out ALU, an input processing IC for scanning the keyboard, de-bouncing, and generating more signals going to the microcode control chip, an output chip that took in serial data representing a digit, decoded the BCD into 1-of-10 signals, sent that out to the common Nixie bus, and also strobed the appropriate digit, as well as keeping track of decimal point information, a register chip that contained three 68-bit serial-in/serial out (with perhaps one position 4-bit parallel out) shift registers that represented the storage for the working registers of the calculator, and lastly, a data routing chip that took care of gating serial data streams to/from the register chip, ALU chip, and output chip. Technically, this chipset was kind of a 4-bit micro-coded engine that was microprogrammed to operate as a calculator, but with different I/O chips and microcode, it could have been micro-coded to be a small, general-purpose four-bit processor. The resulting calculator(s), the SCM Cogito 414 (introduced first on 23-April-1969), and it's little brother, the SCM Cogito 412 (identical chipset including ROMs, but has a jumper on the main board that limits the machine to 12 digits versus the 414's 14 digits - and introduced a bit later to allow sales of the 14-digit version to ramp up before introducing a lower-cost model with two fewer digits). Was that chipset developed for SCM a microprocessor chipset? That's really tough to say one way or the other. It could have fairly easily been turned into a small general purpose (probably decimal based rather than binary) computer with some different I/O chips and microcode, but does that count as a microprocessor, either as-is, or with modifications? There was also a chipset that was developed by an individual entrepreneur that was intended to function as the compute engine for a small portable computer. At the moment, I can't recall the name of the person. He claimed his design was truly the first "CPU on a chip". It had all of the requisite bits (excuse pun) in the design to make it a full processor. At some point fairly recently, after arguing his case for many years, it went to court, with his claim being that he beat Texas Instruments to implementation. TI had their single-chip microprogrammed "calculator" processor that only needed display drive electronics, an external clock generator, and a keyboard. The claim was that the chip that this guy had developed was a complete CPU, whereas the TI chip, in order to do anything other than serve as a calculator (with different mask-programmed microcode) would require additional support ICs to do anything really useful as a computer. A major point of the decision was that the engineer had some of the core CPU chips and determined that one of them was still working, and built a small demonstration computer using it. It was slow, but had a full keyboard and a LCD display and could do simple application-like functions. The court sided on the independent guy, although it was a very contentious decision. I'm sorry I don't have the details of this at hand at the moment, but there was quite a splash in the technical media regarding this decision. At least for now, as far as patent law is concerned, this was the first single chip microprocessor. So, technically, the CADC chipset for the F-14 was very likely the first general-purpose processor implemented in MOS on a reasonably small number of chips. The machine was a 20-bit machine, and had to do a lot of math, very quickly, so its math functions were heavily optimized for the types of calculators necessary for positioning the control surfaces of the F-14 in a "fly-by-wire" environment. There were dedicated multiplier and divider chips to do these calculations in hardware as quickly as possible. So, in some ways, the CADC, while it could likely be micro-coded to operate as a more conventional computer, a lot of the details of its implementation are still secret as far as I know. The chips were never made available to anyone but the integration developer for the F-14 project, and were all kept under tight secrecy for many, many years. So, while it might qualify as a microprocessor distributed over a number of large scale ICs, it was all top secret and definitely not an open commercial product. I actually spent some time on the telephone with the primary engineer on the CADC (Ray Holt) many years ago, and while I learned a lot, there were many of my questions that he had to politely decline to answer because he was still bound by duty to the US Government not to disclose information that would be considered in the interest of the national security of the United States. I believe that some more information has come out concerning the chipset in the interest of the historical record. Here's an interesting article about Ray Holt and the CADC. https://www.eejournal.com/article/ray-holt-and-the-cadc-the-worlds-first-mi… Here is a website maintained by Ray: https://firstmicroprocessor.com/ The opinions expressed on Ray's site are his own. Another interesting, but often overlooked multi-chip microprocessor set was designed by Computer Design Corp., which marketed higher-end calculators under the Compucorp brand, as well as making OEM versions for Monroe and a few others (Sumlock, Deitzgen among them). The chipset was called the "HTL" chipset, and consisted of bit-serial ROM and RAM with mask-programmed addressing, a core micro-coded engine that implemented an eight-bit processor, with an instruction set that was fairly general-purpose, but had a lot of stuff specific to calculator operations, such as instructions for operating on half-word (4-bit) quantities in binary and BCD. The core "CPU" chips used the data in the microcode ROMs as the interpreter for the instruction set, with the rest of the microcode ROM containing the specific "macrocode" implementation of the calculator. There were I/O chips that were designed to run a multiplexed Nixie tube display, as well as a Shinsu-Seiki (Seiko/Epson) drum impact printer. A generalized Input chip took care of the keyboard, but its operation was quite programmable through control instructions sent to it that would configure it for various different keyboard designs. There were some optional chips that together combined to be called the "LEMP", for Learn Mode Programmer" that had specialized serial RAM for storing program steps (stored as key codes learned from the keyboard), and a sequencer that took care of doing things like programmed looping, branching, comparisons, etc., as well as taking keyboard input and stuffing the key codes into the program RAM. The LEMP RAM could hold 256 steps, or optionally with an additional two chips, 512 steps. Lastly, there was a CLEMP chip, another I/O chip that interfaced an optional optical card reader that could read punched cards into LEMP RAM, for quick loading of applications, since the RAM technology used was volatile, and register and program step storage was lost when power was removed (and it was initialized to null on power-up as part of the RAM sizing microcode). There were un-used key codes that could encode actual instructions of the CPU instruction set which could be used by those with the appropriate documentation to write code directly for the microcode-implemented instruction set of the chipset. The first operable HTL chipsets appear to have been running (the chipsets were laid out and fabricated by AMI) sometime in the mid-to-latter part of 1969, putting the chipset in the timeframe for consideration. Computer Design Corp. from its inception was involved in designing electronic calculators under contract to calculator manufacturers. Initially, they used early DTL bipolar ICs made by Signetics. Computer Design Corp. could design and build the calculators under contract, or just provide all of the documentation for the customer to put together a complete calculator from the design and sell it as their own. Interestingly enough, Nippon Calculating Machine Co., (NCM) which had marketed two early transistorized calculators (basically a copy of the design of Italy's IME 24 Calculator, without any royalties or rights paid to IME) called the Busicom 161 and 141. Since their first two calculators were essentially engineered for them in Italy, NCM, while it had some digital electronics competency, did not have much in the way of the skills needed to design a machine as complex as a calculator. Since the 161 and 141 had caused some ruckus, with Industria Macchine Elettroniche (IME)complaining publicly that the Busicom machines were unlicensed copies, there never was any legal process involved. NCM decided to look outside for calculator design engineering, and right away became one of Computer Design Corporation's early customers. Computer Design Corporation developed the Busicom 202, 207 and 2017, with CRT display and optional printer attachment on the 207 and 2017), which were built with DTL logic and magnetostrictive delay line memory with punched card programming. These machines were targeted at higher-end environments. Computer Design Corp. also designed the Busicom 162 and 162C, smaller desktop calculators designed more for bookkeeping/accounting and general math. The 162/162C Nixie-display calculators used DTL bipolar logic, and a small magnetic core array for memory storage. These two machines were essentially replacements for the Busicom 161/141 that had limited lifetimes because of their discrete transistor construction. Along with designing calculators for others, primarily Busicom as is known at this point, Computer Design Corp. was also working on the design of its own MOS chipset that could be programmed to make up a sophisticated calculator. When that chipset was operational and tested to be working in a calculator prototype, the company thought they would just be a supplier of boards stuffed with the MOS chipset to anyone who wanted them and they could integrate them into their own calculator. Computer Design Corp. had created its Compucorp division to market a line of machines using the chipsets, and was working on putting ramping up its production facilities. There were actually small numbers of Compucorp-branded Nixie-display calculators coming off the line that were being sold through a number of independent business machine distributors that they had lined up. Sales weren't particularly brisk, though, because the company was a newcomer to a very well-established calculator marketplace. Even though their calculators were more capable than most of the competitors on the higher-end of the electronic calculator business, their name was not well known when compared with the giants of the industry: Wang Laboratories, Hewlett Packard, and others like Monroe, SCM, and Friden. Monroe happened to learn of these new calculators and that Computer Design Corp. was eager to find OEMs to increase sales, and jumped on the bandwagon big-time, making a deal with Computer Design Corp. to be the exclusive retailer of calculators made by Computer Design Corporation using the HTL chipset. Computer Design Corp. was forced to stop making Compucorp-branded calculators, and badge their calculators as Monroe products. Monroe would buy them and distribute them to all of their business machine retail outlets. This was seemingly wonderful for Computer Design Corp., as it immediately had a gigantic and very well-established sales network for its calculators. As it turned Out, Monroe was making a killing on the machines, and Compucorp wasn't getting much out of their part of the deal. Compucorp still wanted to sell their own machines through independent retailers, with different features than the Monroe models. In time, Compucorp asked Monroe for release from the exclusivity agreement, but Monroe would have nothing of it. That did not go over well with Computer Design Corp. management. At one point, Computer Design Corp. announced it was going to buy the Monroe division of Litton Industries from Litton, as a way to get out of the contract that so limited them, but that fell on its face, as Litton wanted way more than Computer Design Corp. could muster. Then, Litton announced that they intended to buy Computer Design Corporation. All of this craziness was going on just as the production lines at Computer Design Corporation were churning out Monroe-badged machines at a frenzied rate. Computer Design Corporation was able to somehow (I don’t know how) thwart the threat of buyout by Litton (who easily could have bought up all of Computer Design. Corp.'s stock at an extremely attractive price), and as part of it, they were able somehow able to get rid of the exclusivity arrangement with Monroe/Litton, and they immediately began selling Compucorp-branded calculators that were identical to the Monroe machines other than cabinet styling and color scheme, as well as some subtle functional differences to differentiate Compucorp machines from Monroe's versions. At some point it was announced that the contract with Monroe had been severed, giving free-reign to Compucorp to sell its own calculators, as well as to OEM to other makers, and surprisingly, Monroe signed up as an OEM customer. This arrangement suited Computer Design Corp. much more favorably. I'll close by mentioning that the relationship between Nippon Calculating Machine Co. and Computer Design Corp. might have seemed to end with the design of the 162/162C, but there was continued business between the two companies. Nippon Calculating Machine engineers had designed the logic for their own complex MOS LSI calculator chipset that could allow the chips to be combined in different ways to make varying types of calculators. At that time, Japan did not have anything but university and corporate labs doing very early work on LSI MOS, and there was nowhere near any kind of production capability in the country. Nippon Calculating Machine turned to the US to try to find someone to make their chips for them. Initially, all of the chipmakers they went to rejected them. They also visited Computer Design Corp. since they were long-time customers, and there was discussion about seeing if Computer Design Corp. might be able to serve as an intermediary to US chipmakers to get their chips made. This wasn't really something that Computer Design Corp. was interested in doing, but given the long history of working with Nippon Calculating Machine Co., they agreed that they would take a crack at taking the design that NCM had developed and work on partitioning it in a way that would work with chip complexity constraints and packaging, and see if they could come up with a chipset that they could get AMI to fabricate. Computer Design Corp. already had the knowledge to do this from their own internal project to develop their own chipset. It isn't known exactly what kind of agreement was forged, and if up-front money was put up by NCM, but it is known that NCM put a firm deadline in place stating that a calculator using the chipset would have to be delivered in person to NCM headquarters and a presentation and demo of the machine made to NCM executives and engineers. Work began immediately on taking the NCM logic design and turning it into a batch of chips, which, by initial estimate was going to require 19 chips, and that was assuming higher levels of integration that were only being perfected by AMI at the time. As the NCM delegation was preparing to return to Japan with their hopes dashed of making an agreement directly with a chipmaker, a call was received from a fledgling chip manufacturing company that they had briefly visited when they were in the Santa Clara, CA. area. The company specialized in making memory chips, though, and it was initially thought that the huge amount of random logic that was involved in the NCM calculator chipset design was beyond the capabilities of this company. Memory devices are just repetitions of the same memory cell pattern over and over on the chip to make a memory array, with additional logic to do the address decoding and read/write (for RAM) or read(for ROM) circuitry and I/O buffers. This type of chip design has some aspects that allow a degree of automation to be used in laying out the chips, and also had very little in the way of random logic to complicate the chip design. The NCM calculator design was a whole bunch of random logic...gate upon gate, flip-flop upon flip-flop wired together in a rats nest of logic for each chip. After NCM left this fledgling company named Intel, higher-ups at Intel had learned of the visit by NCM, and told their underlings to get back in touch with NCM and let them know they would see what they could do for NCM. Intel was cash-hungry at the time, and it was thought that perhaps this project could serve double-duty...to bring in some needed cash for Intel's production capacity expansion, as well as to perhaps push Intel's IC fabrication technology into areas other than RAM and ROM. Intel was doing well with their memory ICs, but they were limited by the number of chips they could produce. Chip manufacturing technology is very expensive, so Intel was investing everything they made in enhancing their production capabilities. So, maybe working with NCM might help bring in some funds to help that effort. A deal was forged between Intel and NCM. A firm deadline was part of the contractual language where Intel would have to build a prototype calculator based on the chips, and bring it to NCM's headquarters on a specific date and present/demonstrate it to NCM executives and engineers. In the end, the NCM chip design was set aside as it was simply beyond the capability to be fabricated by Intel. Had Intel not had an idea of how to effectively do what the massively complex chipset did in a much simpler fashion, who knows what the future would have held. The solution to Intel's problem of abandoning NCM's chipset is what became the Intel 4004. Intel developed and fabricated the 4004 and some support chips to make it into a useful controller for a calculator, including some shift register ICs for I/O, some combination RAM and I/O chips, as well as ROM and I/O chips. They built a very simple prototype calculator that could do integer-only math as their demo, and presented it to NCM. The Busicom 141-PF was the resulting production calculator, arguably the first consumer device to use a microprocessor to create its functionality. The rest is history. There's a piece of the story left hanging, though. That's a very, very interesting story that will have to wait until I get all of my ducks in a row to tell, which I will do on the Old Calculator Museum website at some point in the not too distant future. I'll certainly announce it here when it's finalized on up on the website. Rick Bensene The Old Calculator Museum https://oldcalculatormuseum.com -----Original Message----- From: ED SHARPE via cctalk [mailto:cctalk@classiccmp.org] Sent: Tuesday, November 21, 2023 2:00 PM To: General Discussion: On-Topic and Off-Topic Posts <cctalk(a)classiccmp.org> Cc: ED SHARPE <couryhouse(a)aol.com> Subject: [cctalk] Re: Intel 4004 I had heard something about a f14 chip pehS being first but not avail. To general  public???Ed# Sent from AOL on Android On Tue, Nov 21, 2023 at 2:41 PM, Joshua Rice via cctalk<cctalk(a)classiccmp.org> wrote: On 21/11/2023 09:03, ED SHARPE via cctalk wrote: The 4004 was definitely the first commercially available single-chip CPU on the market, but if you include multi-chip LSI designs, the lines get blurry.