Hackaday Podcast Episode 263: Better DMCA, AI Spreadsheet Play, And Home Assistants Your Way

No need to wonder what stories Hackaday Editors Elliot Williams and Al Williams were reading this week. They’ll tell you about them in this week’s podcast. The guys revisit the McDonald’s ice cream machine issue to start.   This week, DIY voice assistants and home automation took center stage. But you’ll also hear about AI chat models implemented as a spreadsheet, an old-school RC controller, and more.

How many parts does it take to make a radio? Not a crystal radio, a software-defined one. Less than you might think. Of course, you’ll also need an antenna, and you can make one from lawn chair webbing.

In the can’t miss articles, you’ll hear about the problems with the x86 architecture and how they tried to find Martian radio broadcasts in the 1920s.

Miss any this week? Check out the links below if you want to follow along, and as always, leave your comments!

Direct download in DRM-free MP3.

Continue reading “Hackaday Podcast Episode 263: Better DMCA, AI Spreadsheet Play, And Home Assistants Your Way”

Weird Things To Do With FPGAs

There’s an old joke about how can you find the height of a building using a barometer. One of the punchlines is to drop the barometer from the roof and time how long it takes to hit the ground. We wonder if [Alexlao512] had that in mind when he wrote a post about unconventional uses of FPGAs. Granted, he isn’t dropping any of them off a roof, but still. The list takes advantage of things we usually try to avoid such as temperature variation, metastability, and the effects of propagation delays.

For example, you probably know that hooking up an odd number of inverters into a loop forms an oscillator—the so-called ring oscillator. The post discusses how you can use an oscillator like that to measure propagation delay or even as a strain gauge. If you put pressure on the FPGA chip, the frequency of the ring oscillator will subtly vary.

Continue reading “Weird Things To Do With FPGAs”

The Long Strange Trip To US Color TV

We are always fascinated when someone can take something and extend it in a clever way without changing the original thing. In the computer world, that’s old hat. New computers improve, but can usually run old software. In the real world, the addition of stereo to phonograph records and color to photography come to mind.

But there are few stories as strange or wide-ranging as the path to provide color TV. And it had to be done in a way that a color set could still get a black and white picture and black and white sets could still watch a color signal without color. You’d think there would be a “big bang” moment where color TV burst on the scene — no pun involving color burst intended. But there wasn’t. Instead, there was a long, twisted path with many competing interests and ideas to go from a world in black and white to one tinted with color phosphor.

Background

In 1928, Science and Invention magazine had plans for building a mechanical TV (although not color)

It is hard to imagine, but John Logie Baird transmitted color images as early as 1928 using a mechanical scanner. Bell Labs had a demonstration system, also mechanical, in 1929. Baird broadcast using his system in 1938. Even earlier, around 1900, there were attempts to create mechanical color image systems. Those systems were fickle or impractical, though.

Electronic scanning was the answer, but World War II froze most consumer electronics development. Baird showed an electronic color system in late 1944. However, it would be 1953 before NTSC (the National Television System Committee) adopted the standard color TV signal for the United States. It would be almost 20 years later before SECAM and PAL were standardized in other parts of the world.

Of course, these are all analog standards. The world’s gone digital now, but for nearly 50 years, analog color TV was the way people consumed TV in their homes. By 1941, NTSC produced a standard in the United States, but not for color TV. TV adoption didn’t really take off until after the war. But by 1950, the US had some 6 million TV sets.

This was both a plus — a large market — and a negative. No one wanted to obsolete those 6 million sets. Well, at least, the government regulators and consumers didn’t. But most color systems would be incompatible with those existing black and white sets. Continue reading “The Long Strange Trip To US Color TV”

Repairing A Gear With A Candle (and Some Epoxy)

You have a broken gear you need to fix, but there’s no equivalent part available. That’s the issue [Well Done Tips] faced with a plastic gear from a lawnmower. While we’d be tempted to scan the gear, repair the damage in CAD and then 3D print a new one, we enjoyed hearing about his low-tech solution. In addition to the write up, there’s a video showing the process you can watch below.

The idea is pretty simple. Using a piece of pipe and melted candle wax, he prepared a mold of an undamaged section of the gear. Then he cast epoxy resin in place to recreate the missing pieces. There are a few tricks, like putting holes in the remaining part of the gear so the epoxy flows into the existing part. Depending on the gear’s purpose and original material, you might be able to just use it as-is. However, you could also use the repaired gear as a template to create another mold and then cast an entire gear from resin or even metal if you can cast metal.

You can argue whether resin is better or worse than PLA, but of course, it depends on the kind of resin—photopolymers are different from epoxy resins you’d use for this sort of thing. If you think you might like to make your new gear out of aluminum, you might find some inspiration in a previous post.

Continue reading “Repairing A Gear With A Candle (and Some Epoxy)”

Breadboard SDR Doesn’t Need Much

[Grug Huhler] built a simple Tayloe mixer and detector on a breadboard. He decided to extend it a bit to be a full-blown software defined radio (SDR). He then used WSJT-X to monitor FT8 signals and found that he could pick up signals from all over the world with the little breadboard system.

A Raspberry Pi Pico generates a quadrature clock that acts as the local oscillator for the radio. All the processing of the input signal to a quadrature signal is done with a 74LV4052A, which is nothing more than an analog multiplexer. In principle, the device takes a binary number from zero to three and uses it to connect a common signal to one of four channels. There are two common lines and two sets of four channels. In this case, only half of the chip is in use.

An antenna network (two resistors and a capacitor) couples the antenna to one of the common pins, and the Pi generates two square waves, 90 degrees out of phase with each other. This produces select signals in binary of 00, 01, 11, and 10. An op amp and a handful of passive components couple the resulting signals to a PC soundcard, where the software processes the data. The Pi can create clocks up to about 15 or 20 MHz easily using the PIO.

The antenna is a 20-meter-long wire outside, and that accounts for some of the radio’s success. There are several programs than can work with soundcard input like this and [Grug] shows Quisk as a general-purpose receiver. If you missed the first video explaining the Tayloe mixer design, you can catch it below the first video.

This isn’t the first breadboard SDR we’ve seen, but they all use different parts. We’ve even seen a one-bit SDR with three components total (not including the microcontroller). Seriously.

Continue reading “Breadboard SDR Doesn’t Need Much”

Learn AI Via Spreadsheet

While we’ve been known to use and abuse spreadsheets in the past, we haven’t taken it to the level of [Spreadsheets Are All You Need]. The site provides a spreadsheet version of an “AI” system much like ChatGPT 2. Sure, that’s old tech, but the fundamentals are the same as the current crop of AI programs. There are several “lesson” videos that explain it all, with the promise of more to come. You can also, of course, grab the actual spreadsheet.

The spreadsheet is big, and there are certain compromises. For one thing, you have to enter tokens separately. There are 768 numbers representing each token in the input. That’s a lot for a spreadsheet, but a modern GPT uses many more.

Continue reading “Learn AI Via Spreadsheet”