Close
0%
0%

ONAN 1111

A Retrobrew TTL CPU - Classic yet Baroque

Public Chat
Similar projects worth following
A Retrobrew TTL CPU. 20 bit word-length. 74181-style ALU re-implemented with gate chips. Internal connections through a crossbar switch. Software in microprogram form only. Soviet-era 74x series chip clones.

Project sponsored by PCBWay (many thanks to them).

All right. One more retrobrew TTL CPU.

What did make me want to do this kind of thing? Really, I always wanted to do it. As a kid (12-13), I played with microprocessors in the 1980s (6809 mostly). Had a C64 too. But even back then I wondered about these little 14 pin cockroaches and asked myself what it would take to build an entire computer out of them. Now I am over 50 and have quite a lot of time on my hands so I told myself 3 months ago or so that it was time to look seriously into starting my own TTL CPU retrobrew project (as they are called now)

What am I trying to achieve?

A Retrobrew project is a work of art. The goal is aesthetic, not practical. It has to look right and feel right. Here are the aesthetic canons I intend to follow:

1. Classicism

I aim for a classic late 70s look and feel, like a VAX-11/780 or a Xerox Alto, In particular:

  • PCBs are green (not red or blue)
  • Components are through-hole, not SMD
  • LEDs are round 5mm, red, green or amber. No bar-graph modules. No rectangular shapes. Not even 3mm round ones. Of course, no RGB LEDs or other monstrosities.
  • Each LED has its own current-limiting resistor. No resistor arrays.
  • C language is the basic system programming language. No Forth or other weird stuff.

Finally, a retrobrew project needs to be seen. So no enclosures. Just PCBs on a mounting support.

2. Baroque

Within these classical guidelines, I feel an urge for the baroque, the outlandish, the mildly outrageous. This urge lead me to the following options:

  • Use of old soviet TTL chips. Early on in the 70s, the USSR started cloning the 74 series chips from TI and other US manufacturers. Some of these clones are still produced and much new old stock is still floating around. I have found a number of reliable sources on E-bay and will try to source all the chips I need from them. Soviet clone chips add a touch of John Le Carré flavor to the project and even (to me at least) a hint of post-apocalyptic atmosphere.
  • Only microprogramming. The ONAN 1111 CPU will have no ISA. All software tasks will be handled at the microprogram level. That way, all instructions are by construction 1 clock cycle long. The C compiler will directly output microprogram binary. ONAN 1111 will be a single-task machine, in the spirit of Apollo's AGC, for example. Of course, this task will itself include a number of features (to be disclosed more fully later).
  • 20 bit length. I have heard of 36, 48, 12 or 24 bit lengths, but not 20. Why? 20 bits has a number of (aesthetic) advantages:
    • 2^20 is in the neighborhood of 1 million, a nice round number. Both address and data buses will be 20 bit wide: a total memory of 1 million 20 bit words.
    • 20 bits is enough to hold all UNICODE characters within a word. A character string on ONAN 1111 will be a string of 20 bit words. In C terms: char = short = int = 20 bit.
    • 20 is 4*5. Since almost all online PCB manufacturer require a minimum quantity of 5 per PCB run, this will make it possible to build ONAN 1111 in 4 bit wide slices, each component having 5 slices. The PCB in the picture above shows a 4-bit ALU slice (a reimplementation of the 74x181)
  • Crossbar switch. Most CPUs have 1 to 3 internal buses to carry data between the different component. ONAN 1111 will have an 8-way crossbar switch instead. As a result all possible transfers between the 3 registers, the SP, the uPC, the ALU and the Ram address/data buses will be possible within a clock cycle. The crossbar switch alone will require 180 К555КП7 (74x151 clone) chips. Luxury has its uses in art.

74181 Board PCB.pdf

PCB Layout of the 74181 re-implementation board

Adobe Portable Document Format - 306.03 kB - 09/04/2022 at 12:00

Preview
Download

74181 Board Schematic.pdf

Kicad Schematic of the 74181 re-implementation board

Adobe Portable Document Format - 333.99 kB - 09/04/2022 at 11:56

Preview
Download

  • Multiplexer PCBs just arrived (great stuff)

    Nicolas REIMEN6 days ago 0 comments

    The multiplexer PCBs for the crossbar switch have just arrived. Here they are:

    These were sent free of charge by PCBWay as part of the sponsorship deal.

    They are of excellent quality. Before that, I had ordered my PCBs from "another supplier" (blink-blink, hint-hint, everyone knows who I am talking about, right?). The PCBs I just received are much better. I think anyone can tell just by looking at the picture and comparing it to the previous ones in the project logs or on the main page. Everything looks better, the silk screen lettering, the solder mask, the tinning of the pads, ...

    As I said, a retrobrew machine is a work of art and it makes a lot of sense, therefore, to have gleaming PCBs like these.

    PCBWay's website is also quite nice and convenient to use for specifying the order and following its progress. They even have a feature that allows you to follow the PCBs as they go through each step of the fabrication process. Practically, I think it is totally useless of course, but it is super cool. Nice touch!

  • Crossbar-Switch PCBs on the way

    Nicolas REIMEN09/20/2022 at 08:17 0 comments

    The 50 PCBs for the crossbar switch are on the way! (thanks to PCBWay) Wait a minute?! Did you say fifty?!

    Yes, fifty.

    The crossbar switch, the heart of ONAN 1111's design, will allow data to be exchanged between (nearly all) internal components of the processor on each clock cycle. It consists of ten 20-bit wide 8-to-1 multiplexers. Each multiplexer is split into five 4-bit wide slices. 10 x 5 equals 50. QED.

    Each board will contain four 74x151 clones (К155КП7), for a total of 200. Each chip is a 1-bit wide 8-to-1 multiplexer. The boards will also contain 8 LEDs and a 74x138 clone (К555ИД7) to indicate which data-path is selected, when the processor is running in single-step mode or at a low clock speed. This alone will contribute 400 blinkenlights to the overall design. What is a retrobrew project without the blinkenlights ? ...

    Below is a Kicad-generated 3D picture of one of the boards:

    This one is the multiplexer that will feed a loading value to the microprogram counter (uPC) whenever necessary (jumps, interrupts, ...) The dip-switches will be used to enter the address of the interrupt-handling routine.

  • Support by PCBWay

    Nicolas REIMEN09/18/2022 at 05:42 0 comments

    The online PCB company PCBWay has generously offered to support this project. I feel honored by this offer as PCBWay has supported a number of very cool projects in the past (notably, CuriousMarc's YouTube Channel). Many thanks to them.

  • People I look up to in the Retrobrew Community

    Nicolas REIMEN09/06/2022 at 08:51 0 comments

    As you may have guessed, my all time favorite is Bill Buzbee and his Magic 1 TTL 16-bit 70s style homebrew minicomputer. He is no doubt my greatest inspiration and his example is the chief reason I decided to go ahead with my own project.

    However, Bill is not alone. Many other people have inspired me over the years. Namely:

    • The CuriousMarc YouTube channel. These guys are not strictly Homebrew CPU builders but their resurrection of an AGC (Apollo Guidance Computer) is certainly the coolest vintage computing project I have ever seen. CuriousMarc is sponsored by PCBWay and so are we!
    • Ken Shirriff aka "Master Ken". A member of the CuriousMarc restoration team, whose excellent blog I avidly read and strongly recommend.
    • All the guys from the Homebrew Computers Web-ring. The list seems mostly dormant nowadays but it once was the Mecca of Retrobrew building. Among the members of the ring, I found the following particularly worthy of note:
      • The already mentioned Bill Buzbee ...
      • Jeroen Brinkman, whose 100kg relay-based beast is in a category all its own.
      • James Newman, whose discrete transistor-based Megaprocessor is the main contender to Jeroen Brinkman's Mercia for the title of most unwieldy retrobrew machine.
      • fritzler-avr.de's Spaceage 2 32-bit MIPS I ISA TTL implementation. Another heavyweight, although not in the size department but in terms of computing power.
      • Gianluca G's Apollo 181 4-bit educational TTL computer. Simple, elegant, Italian.
    • Two YouTubers, Ben Eater and James Sharman, whose numerous and highly educational video series I followed regularly over the years.

    I am sure I am forgetting many interesting projects. These are just the ones that came to my mind today.

  • Reimplementing the 74x181

    Nicolas REIMEN09/04/2022 at 13:12 1 comment

    The 74181 is the classic-est of classic 4-bit slice ALU chips. Initially, I thought I might use these chips in the design. However, I decided against it when I came across a batch of old soviet chips that made it reasonably simple to re-implement the venerable ALU with gate chips. The soviet chips I found are КР531ЛР9s (KP531LR9 = 74x64 clones) which contain a combination of AND and NOR gates making it relatively straightforward to implement the binary functions of the 74181. The schematic can be found in the "Files" section, as well as the PCB layout.

    The first stage of the 74181 contains 4 identical elements (1 for each bit). Each of these elements are implemented using 2 КР531ЛР9s (for a total of 8). The second stage contains functions of increasing complexity which perform the equivalent of a carry-lookahead, plus a few other things. In total, I used 4 КР531ЛР9s, 12 КМ155ЛА3 (4 NAND gates, = 74x00) and 4 КМ155ЛА4 (3 triple-input NAND gates = 74x10) to implement the 2nd stage.

    Here is a picture of the finished board:

    The first stage is the set of 10 vertical chips in the upper half of the board. The second is the set of 20 horizontal chips below. All other components are only for display purposes and will not be included in the final version of the ALU board.

    The Green LEDs on top are the two 4-bit inputs (A and B). The 16 red LEDs on the right indicate what function the ALU is currently executing, based on the S input. Only one of these LEDs is lit at a time. They are driven by a pair of К555ИД7s (=74x138) located right of the green input-display LEDs. At bottom, from left to right are located: the output display (4 green LEDs), various indicator signals and finally, 8 red LEDs showing the output of the first stage and input of the second. These signals play a central role in the operation of the ALU, as explained in Ken Schirriff's excellent blog post on the 74181.

    The ONAN 1111 CPU will contain 5 boards equivalent to this one (5 slices of 4 bits = 20 bits wide). These final ALU boards will have a little more logic to handle shifting operations (which the 74181 does not do).

  • Testing the Soviet Chips

    Nicolas REIMEN09/04/2022 at 12:15 1 comment

    Because most soviet clone chips are new old stock (Over 30 years old in most cases), a little testing was in order. Therefore, I made the chip tester module below:

    The Arduino Mega is a "generic" Chinese clone but it works all right. The last Zif socket (32 pins) is not yet installed because it is still in the mail. It should arrive shortly. I will have to test 32 pin chips when I start dealing with static RAM. I know this is not exactly "classic", according to my own definition, but I decided that using DRAM from the period (4164, 41256) would have really been too much of a hassle.

    I am happy I made this tester because some batches of chips contained up to 25% duds.

    If anyone is interested, I can publish the documentation for this tester here (Schematic, PCB and Arduino test programs).

  • Sourcing Soviet TTL Chips

    Nicolas REIMEN09/04/2022 at 11:45 0 comments

    There are a number of good suppliers on ebay. Particular thanks to:

    The Soviets copied the TI catalog of TTL 74xx chips from the 70s onwards. The oldest chips I have come across so far have date-codes from the early 80s. The following page has a very useful correspondence table between the Soviet model numbers and those of the 74xx series. With their Cyrillic markings and inimitably soviet-style plastic containers, these chips have an undeniable coolness factor to them. Here is a pictures of several of them:

View all 7 project logs

Enjoy this project?

Share

Discussions

zpekic wrote a day ago point

Amazing project! Looking forward to dig into all the project logs and learn a lot I am sure. Btw, check out my microcoded designs (including a compiler), and a serial-CPU using 16*16 crossbar switch :-)

  Are you sure? yes | no

Nicolas REIMEN wrote 17 hours ago point

Yeah, I have already looked into it. It looks pretty neat

  Are you sure? yes | no

Nicolas REIMEN wrote 09/06/2022 at 16:37 point

  Are you sure? yes | no

Nicolas REIMEN wrote 09/05/2022 at 17:48 point

Thanks a lot. The information about this machine seems to be quite scarce on the Internet. A quick search just turned up a few mentions in articles about the early Arpanet. Do you know if there is more detailed info available somewhere?

  Are you sure? yes | no

Steve Toner wrote 09/06/2022 at 16:02 point

There's not much detailed info out there, but  a bit of an overview here: [ih] BBN C-series computers (isoc.org) and here: Design of a user-microprogrammable building block (walden-family.com)

  Are you sure? yes | no

Steve Toner wrote 09/05/2022 at 15:52 point

FYI: The BBN C/70 was a 20-bit machine.

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates