Relay Computer Starts With An Adder That Makes A Racket

Computers built using discrete logic chips? Seen it. Computers from individual transistors? Impressive, but it’s been done. A computer built out of electromechanical relays? Bring on the ozone!

The aptly named [Clickity Clack]’s new YouTube channel promises to be very interesting if he can actually pull off a working computer using nothing but relays. But even if he doesn’t get beyond the three videos in the playlist already, the channel is definitely worth checking out. We’ve never seen a simpler, clearer explanation of binary logic, and [Clickity Clack]’s relay version of the basic logic gates is a great introduction to the concepts.

Using custom PCBs hosting banks of DPDT relays, he progresses from the basic AND and XOR gates to half adders and full adders, explaining how carry in and carry out works. Everything is modular, so four of his 4-bit adder cards eventually get together to form a 16-bit adder, which we assume will be used to build out a very noisy yet entertaining ALU. We’re looking forward to that and relay implementations of the flip-flops and other elements he’ll need for a full computer.

And pay no mind to our earlier dismissal of non-traditional computer projects. It’s worth checking out this discrete 7400 logic computer and this all-transistor build. They’re impressive too in their own way, if a bit quieter than [Clickety Clack]’s project.

49 thoughts on “Relay Computer Starts With An Adder That Makes A Racket

      1. The pre-WWII machine (V1/Z1*) was actually fully mechanical. The V2/Z2 used a combination of relay logic and mechanical memory however that was build after the WWII outbreak.

        (* the machines were named V1/V2 etc. after Versuchsmodel* but was retroactively renamed to Z* after Zuse*)

  1. He uses way too many relays.

    In the bad old days when we did have relay based computers we used grey code and not traditional binary. Also the relays had 3 stacks of up to 18 contacts.

    Diodes are the answer here.

    (a)-|>|-(o)-|<|-(b) active high o = a OR b
    (a)-||-(b) active low o = a AND b
    Active high can go through a relay coil to ground
    Active low can go through a relay coil to positive supply

    1. Another trick is to use pass logic. You can make an AND gate by attaching A to the armature, and B to the coil.

      Or if you have active high and active low signals, you can attach them to both sides of the coil.

    2. Grey code? Don’t see how that would work (though I haven’t actually tried it out).
      Many people that make relay logic don’t accept diodes as early diodes were inefficient and expensive – so not used for (most?) real relay computers.

      1. Grey code is like binary except that only ever one bit changes between consecutive number representations. Binary tends to put huge spikes on power rails because so may relays change together and in really large computers that could cause other relays to falsely synchronise or even cause system wide oscillation.

        Diodes (previously called rectifiers) were poor in early day computers but later systems used them extensively. Old telephone exchanges (from the Crossbar era) were essentially relay computers and were often stories high.

        1. Not really. A computer uses input passed through a process to output something. A telephone exchange is just a very large multi-contact switch. It is not processing anything. it is not outputting anything. It has about as much input as a ceiling light.

  2. What I’d like to see is a home-built fluidic computer, based on coanda-effect fluidic amplifiers…

    In an alternate history novel series called “1632” this kind of computer comes up as an “aqualator”.

      1. Yes. But so far I haven’t found evidence somebody has actually built one. There are discussions about it, even some objects on thingiverse, but even those basic blocks don’t seem to have been built and shown to work. I’ve seen the most basic versions work in two examples, but its hard to see how that would scale up.

        I have that problem sitting on a shelf of my mind, I have researched the topic of pure fluidic amplifiers, and my next step would be trying to simulate them in a 2d or 3d fluid simulation. Unfortunately I know nothing about fluid simulation, so I have to learn that first. Ultimately the goal would be to design 3D printable gates…. But that’s very far offf.

        1. “…But that’s very far offf [sic].”

          Very far off from what? Very far off from WHEN? In the mid-1960s I led a design team charged with designing an integrated comm-nav-ident (ICNI) aircraft. The only beyond-the-state-of-the-art technology which did NOT make it into the final design was the use of fluidic logic…not dense enough, don’cha know. I do, however, still have some fluidic gates and fluidic flip flops rattling around here somewhere…
          Tell you what: simply expend the effort to look up fluidic logic, and you’ll find that, once again, there’s nothing new under the sun…by at least fifty years, in this scenario.

          1. “Far off” for me. Yes, I did research the topic. I have found references to “fluidic logic” aplenty. Digging deeper though, it’s not that simple. Certainly I have not come across successful DIY implementations. And the papers and texts I have found so far don’t tell a lot of details like dimensions, pressures, flow rates etc.. Then there is the problem of “power supply” and distributing it across a more complex device. That’s why I want to simulate this stuff first, to get the design parameters even into the ballpark of something that would work.

            I’m talking specifically about pure fluidic amplifiers, not about hydraulics, mechanical valves, “bubble logic” etc.

          2. Strange that you (that obviously can’t f**king read simple text) led anything. Perhaps you were an intern and now making up shit of your glorious past? Or perhaps you are a teenager trying to troll? You seem to have the manner of a teenager however some people never grow up… But i’ll be helpful and make a simple list that you may understand:

            . Nobody claimed that fluidic logic is a new innovation. NOBODY.
            . Many of us KNOW about fluidic logic (and some of us realize why there are better alternatives even for hobbyists).
            . Your ranting is hilarious. Do you really claim that you (while “leading” the team mentioned) did 3D printable gates in the 60ies?

  3. Every relay computer project I see posted only reminds me that useful computers could have been built long before they were. Unlike a lot of technology that had to wait on advancements in materials or developments in theory, all of the elements for a practical relay computer were available no later than the last third of the 19th century. George Boole had established the foundations in the theory of logic, these were well known, and widely known and understood. The relay had been a standard piece of apparatus used in telegraph and other signaling systems in fact it was included in the original 1840 telegraph patent of Samuel Morse. So everything was in place about a hundred years before anyone thought to build one.

    Make you wonder what might be right in front of our noses now that we are not seeing.

    1. Well there’s the Antikythera mechanism so no telling what has been lost to history. With that being said even if we had primitive computers back in the day, there’s no telling that they would have put them to as spectacular uses as we do these days.

      1. The Antikythera mechanism is a state machine, and there were other examples of those sorts of calculating devices down through the ages. A relay computer can implement a Turing machine – a true computer in every sense of the term.

        1. It isn’t possible to implement a Turing machine using any technology as we can’t get infinite storage. What we can build are finite state machines which can solve a subset of the problems a Turing machine can handle.

          1. No we cannot implement the abstract machine with current hardware but we can make those that can implement Turing complete programing that would if there was infinite storage so these stand above finite-state machines and other logic automatons.

          2. The reference above to a “Turing machine” was intended to be a reference to a Turing complete computer.

            Most people are unaware that there is a completely separate abstraction model called a “Turing machine” so they use these terms interchangeably.

            There is still a very important distinction between a Turing complete computer and finite state machine.

            If you code in Verilog of VHDL then the distinction is obvious. You can double the number of states of a pure state machine by adding just one single bit register or flip flop.

            You can also create a classic state machine by emulation with a Turing complete computer but this doesn’t scale in the same way as a pure state machine.

            The term ‘infinite storage’ is misleading because only one small part needs to be accessed at any one time. Given that time is also a factor, you could argue that the memory only needs to have the capacity to be expanded at a higher rate than the progress of the computer to fulfil this requirement. Is the internet then a ‘Turing machine’?

      1. I’m glad to see that he will be honoured thusly. However, he was not in his own day an obscure figure, he was the first professor of mathematics at Queen’s College, Cork (now University College Cork) in Ireland, he was a Fellow of the Royal Society and won the Keith Medal. There was a lot of awareness of his ideas and how important they were.

  4. The ‘done before’ comments are missing the point completely. Mechanical clocks have been ‘done before’, but watching Chris make one, step by step, on the Clickspring channel is magical. Same thing here – we all know it can be done and has been done – but watching it done and explained is a different thing.

  5. You blockhead millenials ever get tired of absolutely not having a CLUE? About ANYTHING which didn’t happen before you were born?
    Meditate on the following for three or four days, and then keep the results to yourself: Charles Babbage; Difference Engine; Analytical Engine.

    1. completely besides the point were talking about relay bases systems here Babbage’s designs were pure mechanical and of the two only the Analytical engine is a true computer in the sense that its (at least in theory) Turing complete

      1. First of all, define what is meant by the exact definition–the canonical form–of a ‘Turing Complete’ machine.

        Having defined, precisely, a Turing Complete machine, please enlighten us all as to why one of Babbage’s creations was, and why one was not, ‘Turing Complete’.

      2. A computer doesn’t have to be Turing complete to be a computer. A computer is a device (or individual) that computes, there isn’t a need to be general purpose to earn the title…

      1. Translation: you have NO idea what you’re even talking about, what IS being talked about, or how to contribute to the conversation. Two CLUELESS STRIKES for the bonehead millennial.

        “I would like to take you seriously, but to do so would be an affront to your intelligence.”–William F. Buckley, Jr.

        1. When you started your post with “You blockhead millenials”, were you expecting a warm response?

          I am an older person and I understand that people learn different things now. I find some of what they learn interesting and they find some of my knowledge interesting.

          Every generation is the same, we pick up the tools that the last generation made and use them to make better tools for us and the next generation.

        2. I see that you think I’m millennial… Very wrong.

          When someone starts a post like an asshole (like you did) then expect to be treated like one. Why should I weigh my words when you obviously have no interest in a civilized discourse?

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.