2025-07-12 19:00:17
Sometimes, all you need to make something work is to come at it from a different angle from anyone else — flip the problem on its head, so to speak. That’s what [Keizo Ishibashi] did to create his Cantareel, a modified guitar that actually sounds like a hurdy-gurdy.
We wrote recently about a maker’s quest to create just such a hybrid instrument, and why it ended in failure: pressing strings onto the fretboard also pushed them tighter to the wheel, ruining the all-important tension. To recap, the spinning wheel of a hurdy-gurdy excites the strings exactly like a violin bow, and like a violin bow, the pressure has to be just right. There’s no evidence [Keizo Ishibashi] was aware of that work, but he solved the problem regardless, simply by thinking outside the box — the soundbox, that is.
Unlike a hurdy-gurdy, the Cantareel keeps its wheel outside the soundbox. The wheel also does not rub directly upon the strings: instead, it turns what appears to be a pair of o-rings. Each rosined o-ring bows 2 of the guitar’s strings, giving four strings a’ singing. (Five golden rings can only be assumed.) The outer two strings of this ex-six-string are used to hold the wheel assembly in place by feeding through holes on the mounting arms. The guitar is otherwise unmodified, making this hack reversible.
It differs from the classic hurdy-gurdy in one particular: on the Cantareel, every string is a drone string. There’s no way to keep the rubber rings from rubbing against the strings, so all four are always singing. This may just be the price you pay to get that smooth gurdy sound out of a guitar form factor. We’re not even sure it’s right to call it a price when it sounds this good.
Thanks to [Petitefromage] for the tip. If you run into any wild and wonderful instruments, don’t forget to let us know.
2025-07-12 16:00:34
Few electronic ICs can claim to be as famous as the 555 timer. Maybe part of the reason is that the IC doesn’t have a specific function. It has a lot of building blocks that you can use to create timers and many other kinds of circuits. Now [Stoppi] has decided to make a 555 out of discrete components. The resulting IC, as you can see in the video below, won’t win any prizes for diminutive size. But it is fun to see all the circuitry laid bare at the macro level.
The reality is that the chip doesn’t have much inside. There’s a transistor to discharge the external capacitor, a current source, two comparators, and an RS flip flop. All the hundreds of circuits you can build with those rely on how they are wired together along with a few external components.
Even on [stoppi]’s page, you can find how to wire the device to be monostable, stable, or generate tones. You can also find circuits to do several time delays. A versatile chip now blown up as big as you are likely to ever need it.
Practical? Probably not, unless you need a 555 with some kind of custom modification. But for understanding the 555, there’s not much like it.
We’ve seen macro 555s before. It is amazing how many things you can do with a 555. Seriously.
2025-07-12 13:00:15
Magnetic Core memory was the RAM at the heart of many computer systems through the 1970s, and is undergoing something of a resurgence today since it is easiest form of memory for an enterprising hacker to DIY. [Han] has an excellent writeup that goes deep in the best-practices of how to wire up core memory, that pairs with his 512-bit MagneticCoreMemoryController on GitHub.
Magnetic core memory works by storing data inside the magnetic flux of a ferrite ‘core’. Magnetize it in one direction, you have a 1; the other is a 0. Sensing is current-based, and erases the existing value, requiring a read-rewrite circuit. You want the gory details? Check out [Han]’s writeup; he explains it better than we can, complete with how to wire the ferrites and oscilloscope traces to explain why you want to wiring them that way. It may be the most complete design brief to be written about magnetic core memory to be written this decade.
This little memory pack [Han] built with this information is rock-solid: it ran for 24 hours straight, undergoing multiple continuous memory tests — a total of several gigabytes of information, with zero errors. That was always the strength of ferrite memory, though, along with the fact you can lose power and keep your data. In in the retrocomputer world, 512 bits doesn’t seem like much, but it’s enough to play with. We’ve even featured smaller magnetic core modules, like the Core 64. (No prize if you guess how many bits that is.) One could be excused for considering them toys; in the old days, you’d have had cabinets full of these sorts of hand-wound memory cards.
Magnetic core memory should not be confused with core-rope memory, which was a ROM solution of similar vintage. The legendary Apollo Guidance Computer used both.
We’d love to see a hack that makes real use of these pre-modern memory modality– if you know of one, send in a tip.
2025-07-12 10:00:10
Recently AI risk and benefit evaluation company METR ran a randomized control test (RCT) on a gaggle of experienced open source developers to gain objective data on how the use of LLMs affects their productivity. Their findings were that using LLM-based tools like Cursor Pro with Claude 3.5/3.7 Sonnet reduced productivity by about 19%, with the full study by [Joel Becker] et al. available as PDF.
This study was also intended to establish a methodology to assess the impact from introducing LLM-based tools in software development. In the RCT, 16 experienced open source software developers were given 246 tasks, after which their effective performance was evaluated.
A large focus of the methodology was on creating realistic scenarios instead of using canned benchmarks. This included adding features to code, bug fixes and refactoring, much as they would do in the work on their respective open source projects. The observed increase in the time it took to complete tasks with the LLM’s assistance was found to be likely due to a range of factors, including over-optimism about the LLM tool capabilities, LLMs interfering with existing knowledge on the codebase, poor LLM performance on large codebases, low reliability of the generated code and the LLM doing very poorly on using tacit knowledge and context.
Although METR suggests that this poor showing may improve over time, it seems fair to argue whether LLM coding tools are at all a useful coding partner.
2025-07-12 07:00:48
Who doesn’t want an X-ray machine? But you need a special tube and super high voltage, right? [Project 326] says no, and produces a USB-powered device that uses a tube you can pick up two for a dollar. You might guess the machine doesn’t generate X-rays with a lot of energy, and you’d be right. But you can make up for it with long exposure times. Check out the video below, with host [Posh Arthur].
The video admits there are limitations, of course. We were somewhat sad that [Project 326] elected not to share the exact parts list and 3D printed files because in the unlikely event someone managed to hurt themselves with it, there could be a hysterical reaction. We agreed, though, that if you are smart enough to handle this, you’ll be smart enough to figure out how to duplicate it — it doesn’t look that hard, and there are plenty of not-so-subtle clues in the video.
The video points out that you can buy used X-ray tube for about $100, but then you need a 70kV power supply. A 1Z11 tube diode has the same basic internal structure, but isn’t optimized for the purpose. But it does emit X-rays as a natural byproduct of its operation, especially with filament voltage.
The high voltage supply needs to supply at least 1mA at about 20 kV. Part of the problem is that with low X-ray emission, you’ll need long exposure times and, thus, a power supply needs to be able to operate for an extended period. We wondered if you could reduce the duty cycle, which might make the exposure time even longer, but should be easier on the power supply.
The device features a wired remote, allowing for a slight distance between the user and the hot tube. USB power is supplied through a USB-C PD device, which provides a higher voltage. In this case, the project utilizes 20V, which is distributed to two DC-DC converters: one to supply the high-voltage anode and another to drive the filament.
To get the image, he’s using self-developing X-ray film made for dental use. It is relatively sensitive and inexpensive (about a dollar a shot). There are also some lead blocks to reduce stray X-ray emission. Many commercial machines are completely enclosed and we think you could do that with this one, if you wanted to.
You need something that will lie flat on the film. How long did it take? A leaf image needed a 50-minute exposure. Some small ICs took 16 hours! Good thing the film is cheap because you have to experiment to get the exposure correct.
This really makes us want to puzzle out the design and build one, too. If you do, please be careful. This project has a lot to not recommend it: high voltage, X-rays, and lead. If you laugh at danger and want a proper machine, you can build one of those, too.
2025-07-12 04:00:27
Building a simple 8-bit computer is a great way to understand computing fundamentals, but there’s only so much you can learn by building a system around an existing processor. If you want to learn more, you’ll have to go further and build the CPU yourself, as [MINT] demonstrated with his EPROMINT project (video in Polish, but with English subtitles).
The CPU began when [MINT] began experimenting with uses for his collection of old memory chips, and quickly realized that they could do quite a bit more than store data. After building a development board for single-chip based programmable logic, he decided to build a full CPU out of (E)EPROMs. The resulting circuit spans four large pieces of perfboard, weighs in at over half a kilogram, and took several weeks of soldering to create.
The star of the system is the ALU, which runs an instruction set inspired by the Z80, but with some optimizations and added features. In particular, it has new operations for multiplication, division, bitstream operations, more advanced bit shifting, and a wide range of mathematical functions, including exponents, roots, and trigonometric functions. [MINT] documented all of this in a nicely-formatted offline booklet, available under the project’s GitHub repository. It’s currently only possible to program for the CPU using opcodes or a custom flavor of assembly, but there are plans to write a C compiler for it.
Even without being able to write in a higher-level language than assembly, [MINT] was able to drive a VFD screen with the EPROMINT, which he used to display some clips from The Matrix. This provided an opportunity to demonstrate basic debugging methods, which involved dumping and analyzing the memory contents after a failed program execution.
Using memory chips as programmable logic gates is an interesting hack, and we’ve seen Lisp programs written to make this easier. Of course, this isn’t the first CPU we’ve seen built without any chips intended for logic operations.
Thanks to [Piotr] for the tip!
