Usagi Electric’s Paper Tape Reader Is Ready To Hop With The Tube Computer

September 16, 2024 by 23 Comments

After previously working out a suitable approach to create a period-correct paper tape reader for his tube-based, MC14500B processor-inspired computer, [David Lovett] over at the Usagi Electric farm is back with a video on how he made a working tape reader.

The assembled paper tape reader as seen from the front with tape inserted. (Credit: David Lovett, Usage Electric, YouTube)
The assembled paper tape reader as seen from the front with tape inserted. (Credit: David Lovett, Usage Electric, YouTube)

The tape reader’s purpose is to feed data into the tube-based computer, which for this computer system with its lack of storage memory means that the instructions are fed into the system directly, with the tape also providing the clock signal with a constant row of holes in the tape.

Starting the tape reader build, [David] opted to mill the structural part out of aluminum, which is where a lot of machining relearning takes place. Ultimately he got the parts machined to the paper design specs, with v-grooves for the photodiodes to fit into and a piece to clamp them down. On top of this is placed a part with holes that line up with the photodiodes.

Another alignment piece is added to hold the tape down on the reader while letting light through onto the tape via a slot. After a test assembly [David] was dismayed that due to tolerance issues he cracked two photodiodes within the v-groove clamp, which was a hard lesson with these expensive (and rare) photodiodes.

Although tolerances were somewhat off, [David] is confident that this aluminum machined reader will work once he has it mounted up. Feeding the tape is a problem that is still to be solved.  [David] is looking for ideas and suggestions for a good approach within the limitations that he’s working with. At the video’s end, he mentions learning FreeCAD and 3D printing parts in the future.  That would probably not be period-correct in this situation, but might be something he could get away with for some applications within the retrocomputing space.

We covered the first video and the thought process behind picking small (1.8 mm diameter) photodiodes as a period-correct tape hole sensor for a 1950s-era computing system, like the 1950s Bendix G-15 that [David] is currently restoring.

Continue reading “Usagi Electric’s Paper Tape Reader Is Ready To Hop With The Tube Computer” →

Large Language Models On Small Computers

September 7, 2024 by 6 Comments

As technology progresses, we generally expect processing capabilities to scale up. Every year, we get more processor power, faster speeds, greater memory, and lower cost. However, we can also use improvements in software to get things running on what might otherwise be considered inadequate hardware. Taking this to the extreme, while large language models (LLMs) like GPT are running out of data to train on and having difficulty scaling up, [DaveBben] is experimenting with scaling down instead, running an LLM on the smallest computer that could reasonably run one.

Of course, some concessions have to be made to get an LLM running on underpowered hardware. In this case, the computer of choice is an ESP32, so the dataset was reduced from the trillions of parameters of something like GPT-4 or even hundreds of billions for GPT-3 down to only 260,000. The dataset comes from the tinyllamas checkpoint, and llama.2c is the implementation that [DaveBben] chose for this setup, as it can be streamlined to run a bit better on something like the ESP32. The specific model is the ESP32-S3FH4R2, which was chosen for its large amount of RAM compared to other versions since even this small model needs a minimum of 1 MB to run. It also has two cores, which will both work as hard as possible under (relatively) heavy loads like these, and the clock speed of the CPU can be maxed out at around 240 MHz.

Admittedly, [DaveBben] is mostly doing this just to see if it can be done since even the most powerful of ESP32 processors won’t be able to do much useful work with a large language model. It does turn out to be possible, though, and somewhat impressive, considering the ESP32 has about as much processing capability as a 486 or maybe an early Pentium chip, to put things in perspective. If you’re willing to devote a few more resources to an LLM, though, you can self-host it and use it in much the same way as an online model such as ChatGPT.

CL-32: The Minimum Possible For A Useful Handheld Computer

September 1, 2024 by 19 Comments

For almost as long as there have been microcomputers, there have been attempts with varying success to make tiny handheld microcomputers. Sometimes these have been very good, and other times they’ve missed the mark in some way. Latest to find its way to us is the CL-32 from [Moosepr], it’s a handheld computer with an ESP32 as brains, an electronic paper display, and a QWERTY keyboard in its smart printed case.

The hardware is relatively standard, save for the keyboard which is a dome-switch design in which the membrane carrying the domes is hand-made. We like this, and don’t think we’ve seen anyone else doing that. Expansion is taken care of by a novel socket arrangement in which boards nestle in a recess in the surface. Some experimentation was required as always to drive the display, but the result is a functional computer.

Sadly there’s little detail in terms of what the software will be, and no hardware files as yet. But what we can see is promising enough to make this one to watch, so we’ll look forward to what they come up with. If an ESP32 OS is a problem, there’s always badge.team, who have been continuously improving theirs since 2017.

The Computer We All Wish We’d Had In The 8-Bit Era

August 31, 2024 by 38 Comments

The 8-bit home computers of yore that we all know and love, without exception as far as we are aware, had an off the shelf microprocessor at heart. In 1983 you were either in the Z80 camp or the 6502 camp, with only a relatively few outliers using processors with other architectures.

But what if you could have both at once, without resorting to a machine such as the Commodore 128 with both on board? How about a machine with retargetable microcode? No, not the DEC Alpha, but the Isetta from [RoelH]— a novel and extremely clever machine based upon 74-series logic, than can not only be a 6502 or a Z80, but can also run both ZX Spectrum games, and Apple 1 BASIC. We would have done anything to own one of these back in 1983.

If retargetable microcode is new to you, imagine the instruction set of a microprocessor. If you take a look at the die you’ll find what is in effect a ROM on board, a look-up table defining what each instruction does. A machine with said capability can change this ROM, and not merely emulate a different instruction set, but be that instruction set. This is the Isetta’s trick, it’s not a machine with a novel RISC architecture like the Gigatron, but a fairy conventional one for the day with the ability to select different microcode ROMs.

It’s a beautifully designed circuit if you’re a lover of 74 logic, and it’s implemented in all surface mount on a surprisingly compact PCB. The interfaces are relatively modern too, with VGA and a PS/2 keyboard. The write-up is comprehensive and easy to understand, and we certainly enjoyed digging through it to understand this remarkable machine. We were lucky enough to see an Isetta prototype in the flesh over the summer, and we really hope he thinks about making a product from it, we know a lot of you would be interested.

IBM’s Latest Quantum Supercomputer Idea: The Hybrid Classical-Quantum System

August 28, 2024 by 15 Comments

Although quantum processors exist today, they are still a long way off from becoming practical replacements for classical computers. This is due to many practical considerations, not the least of which are factors such as the need for cryogenic cooling and external noise affecting the system necessitating a level of error-correction which does not exist yet. To somewhat work around these limitations, IBM has now pitched the idea of a hybrid quantum-classical computer (marketed as ‘quantum-centric supercomputing’), which as the name suggests combines the strengths of both to create a classical system with what is effectively a quantum co-processor.

IBM readily admits that nobody has yet demonstrated quantum advantage, i.e. that a quantum computer is actually better at tasks than a classical computer, but they figure that by aiming for quantum utility (i.e. co-processor level), it could conceivably accelerate certain tasks for a classical computer much like how a graphics processing unit (GPU) is used to offload everything from rendering graphics to massively parallel computing tasks courtesy of its beefy vector processing capacity. IBM’s System Two is purported to demonstrate this when it releases.

What the outcome here will be is hard to say, as the referenced 2023 quantum utility demonstration paper involving an Ising model was repeatedly destroyed by classical computers and even trolled by a Commodore 64-based version. Thus, at the very least IBM’s new quantum utility focus ought to keep providing us with more popcorn moments like those, and maybe a usable quantum system will roll out by the 2030s if IBM’s projected timeline holds up.

A man in glasses and a black sweatshirt sits in front of an orange and black computer screen just below eye level at the table in front of him. His keyboard sits on the table below. He appears to be in a park as there are trees and grass in the background.

Flying Lotus Is A Framework-Powered Portable All-in-One Computer

August 24, 2024 by 12 Comments

One of the things that we love about the modern era of computing is the increasing ease by which you can roll your own custom computer, as seen with the cyberdeck phenomenon. The Flying Lotus is another awesome build in this vein.

Built around the Framework ecosystem, this device was built to suit the very specific use case of its designer, [Carlos Aldana]. He found himself traveling a lot and that the ergonomics of a laptop left a lot to be desired, especially when in the air. Add to it the fact that he has trouble typing on typical laptop keyboards for any length of time, and you can see how an ergonomic keyboard plus a laptop just doesn’t really work on a tray table.

The Flying Lotus takes the screen, modular ports, and mainboard of a Framework laptop and puts them into a single computing block that can be hung from the clever tabs at the top or mounted on a stand that puts the screen at a more ergonomically ideal height from the work surface. [Aldana] describes it as an “iMac that’s portable.” Since it doesn’t have an integrated keyboard, you can run it with whatever keyboard you like from super duper ergo to a teeny game controller sized unit.

We’ve talked about why we like Framework so much before, and if you’d like another take on a modern portable computer, how about this portable Mac mini?

Continue reading “Flying Lotus Is A Framework-Powered Portable All-in-One Computer” →

A Game Of Snake On A LEGO Mechanical Computer

August 23, 2024 by 6 Comments

Really, [OzzieGerff] had us at “LEGO.” But then he took it to another place entirely and built a completely mechanical, nearly 100% LEGO version of Snake. And it’s just as cool as it sounds.

Mind you, it’s a little hard to grok how this whole contraption works, which has been in the works for a while, but we’ll try to summarize as best we can. The most important aspect of this build is that there are no electronics at all — everything is done with mechanical linkages, with some Technics pneumatic components and a couple of electric motors to provide the oomph. The three main components are the input section, which consists of a two-axis joystick, a tail buffer, which keeps track of the length of the snake’s tail as gameplay progresses, and the largest component, the 16×16 display.

The joystick translates user inputs into pneumatic signals which pass through a mechanical filtering unit that prevents the head of the snake from doubling back on itself. The filtered inputs then pass into the screen reader, a complex device that probes the status of a given pixel on the display and determines the status of the snake’s head. If it touches a snake pixel, the game’s over. Hitting a blank pixel moves the head of the snake by one and takes one pixel off the end, while a food pixel extends the snake’s length.

Keeping track of the length of the snake is the job of the buffer, which uses Technics tank tracks and levers. Setting a one is done by flipping the lever to one side as it passes under the write head; a read head further down the track senses which way the lever is flipped and translates it into a pneumatic signal. The buffer has four channels, one for each possible direction the snake’s head could be moving. The signals drive a screen writer, which moves a pyramidal follower across a series of push-rods that flip the corresponding pixel on the display to show the proper icon. Simplicity itself? No, but the video below will make things a lot clearer.

It doesn’t look like [Ozzie] is quite done with this game, as he doesn’t show any actual gameplay yet. We’d love to see and hear that — we suspect it’ll make quite a racket. We’ll be keeping an eye out for this one, but while we wait, check out this rope braiding machine or watch Lego break steel.

Continue reading “A Game Of Snake On A LEGO Mechanical Computer” →