Hackaday Prize judge [Ben Krasnow] has been busy lately. He’s put his scanning electron microscope (SEM) to work creating an animation of a phonograph needle playing a record. (YouTube link) This is the same 80’s SEM [Ben] hacked back in November. Unfortunately, [Ben’s] JSM-T200 isn’t quite large enough to hold an entire 12″ LP, so he had to cut a small section of a record out. The vinyl mods weren’t done there though. SEMs need a conductive surface for imaging. Vinyl is an insulator. [Ben] dealt with this by using his vacuum chamber to evaporate a thin layer of silver on the vinyl.
Just imaging the record wouldn’t be enough; [Ben] wanted an animation of a needle traveling through the record grove. He tore apart an old phonograph needle and installed it in on a copper wire in the SEM. Thanks to the dual stage setup of the JSM-T200, [Ben] was able to move the record-chip and needle independently. He could then move the record underneath the needle as if it were actually playing. [Ben] used his oscilloscope to record 60 frames, each spaced 50 microns apart. He used octave to process the data, and wound up with the awesome GIF animation you see on the left.
[Ben] wasn’t done though. He checked out a few other recording formats, including CD and DVD optical media, and capacitance electronic disc, an obscure format from RCA which failed miserably in the market. The toughest challenge [Ben] faced was imaging the CD media. The familiar pits of a CD are stored on a thin aluminum layer sandwiched between the lacquer label and the plastic disc. He tried dissolving the plastic with chemicals, but enough plastic was left behind to distort the image. The solution turned out to be double-sided tape. Sticking some tape down on the CD and peeling it off cleanly removed the aluminum, and provided a sturdy substrate with which to mount the sample in the SEM.
We’re curious if stereo audio data can be extracted from the SEM images. [Oona] managed to do this with a mono recording from a toy robot. Who’s going to be the first one to break out the image analysis software and capture some audio from [Ben’s] images?
Continue reading “Phonographs Through The Eye Of An Electron Microscope”
Way before you kids had touch screens and mice, we had to walk uphill both ways to tell a computer where we were pointing at on the screen. I speak, of course, of light pens. When these photodiodes in a pen were pointed at a CRT, the display driver would tell the computer where the pen was pointing. It’s a pretty incredible video hack today, and these things were around in the 1970s. You could, of course, use a light pen with most of the old 8-bit home computers, including the Commodore 64.
[Jan] has a soft spot for the light pen on the C64. So much so he made a new input device using this tech. It’s great, and if this existed in 1985, all the cool kids would have known about it.
The build is called the LightHammer. It’s a light pen, inside the head of a plastic hammer, with a few springs, nuts, and washers to tell the computer to read the light pen input. The light pen itself is just a photodiode with a few transistors; it was a simple circuit in the 80s, and it’s a simple circuit today.
A new input device isn’t worth anything without an app to show off the tech, and [Jan] is about three steps ahead of us here. He wrote a game for this LightHammer – a digital version of Whac-A-Mole and Simon. They’re exactly what you think they are: the classic ‘repeat the computer’ and ‘murder rodents’ games.
If that’s not enough, [Jan] also built an arcade cabinet for his C64 setup, with the monitor, joysticks, a 1541, and a TV mounted in a cabinet that would look great in a bar. You can check out a video of that and the games using the LightHammer below.
Continue reading “A Revolutionary Input Device, 30 Years Too Late”
Minecraft wizard, and record holder for the Super Mario World speedrun [SethBling] is experimenting with machine learning. He built a program that will get Mario through an entire level of Super Mario World – Donut Plains 1 – using neural networks and genetic algorithms.
A neural network simply takes an input, in this case a small graphic representing the sprites in the game it’s playing, sends that input through a series of artificial neurons, and turns that into commands for the controller. It’s an exceedingly simple neural network – the network that can get Mario through an entire level is less than a dozen neurons – but with enough training, even simple networks can accomplish very complex tasks.
To train the network, or weighting the connections between inputs, neurons, and outputs, [SethBling] is using an evolutionary algorithm. This algorithm first generates a few random neural networks, watches Mario’s progress across Donut Plains 1, and assigns a fitness value to each net. The best networks of each generation are combined, and the process continues for the next generation. It took 34 generations before MarI/O could finish the level without dying.
A few members of the Internet’s peanut gallery have pointed to a paper/YouTube video by [Tom Murphy] that generalized a completely different technique to play a whole bunch of different NES games. While both [SethBling]’s and [Tom Murphy]’s algorithms use certain variables to determine its own success, [Tom Murphy]’s technique works nearly automatically; it will play about as well as the training data it is given. [SethBling]’s algorithm requires no training data – that’s the entire point of using a genetic algorithm.
Continue reading “Neural Networks And MarI/O”
Ethernet has been around since the mid-70s, but if you think it was always Cat5 and 10BaseT, you’d be sorely mistaken. The first ethernet was built with coaxial cable, vampire taps, AUI adapters, and a whole bunch of other network hardware that will make wizened networking veterans cringe. [Matt] had heard about these weird physical layers back when he started building networks in 1997, but he had never seen one. Now it’s an ancient and forgotten footnote in the history of computer networking. Is it possible to build a Thicknet in this modern era? It turns out, yes, it’s possible. It’s not easy, though.
The network [Matt] is building is a true 10Base5, or Thicknet, network. The backbone of this network is a coaxial cable 9.5mm in diameter. [Matt] discovered that while the common belief that Thicknet used RG-8/U cable. This appears to be incorrect, as the connectors for this cable – vampire taps that pierced the insulation and shield of the cable – are designed for cable manufactured by Belden, part number 9880.
[Matt] assembled the cable, vampire taps, AUI cables, and even found a few ISA NICs that would still work with a reasonably modern computer. He even went so far as to build a USB Ethernet adapter with an AUI interface. This impossibly retro device uses a standard USB to 10BaseT Ethernet adapter, with a chip designed to convert 10BaseT to AUI hacked onto a circuit board. That in itself is an incredible piece of engineering, with a handful of power supplies to get the correct 2.5, 3.3, 5, and 12 Volts to the right places.
As far as exercises in computing history go, [Matt] is at the top of his game. In the process of building it, he also figured out why no one uses Thicknet anymore; once it’s in place, you can’t change it, the cable is big, bulky, and the connectors are terrible. Still, it’s an amazing example of how far we’ve come.
A few years ago, Roombas — everyone’s favorite robotic trash can — graced the pages of Hackaday with reverence. There was nothing this little robot couldn’t do, save for going up stairs. Roomba hacks have died off since then, and these little trash cans have been swallowed up by dumpsters. It’s all very sad, really.
[Mike] has had one of these Roombas around for a while, sitting in a closet, waiting for someone to make use of it. He recently dug it out, looked it over, and watched the LEDs light up after troubleshooting a problem with the batteries. Then the problem was how to control it.
He had wanted to connect it to a VIC-20, but the handy serial port on the Roomba only accepted baud rates between 19.2k and 57.6k. The VIC-20, with the ancient 6522 VIA, could only bitbang a serial port up to 2400bps. Then the idea hit him. In his closet of ancient technology, [Mike] had a Tandy 102, a slightly upgraded TRS-80 Model 100 that could easily drive a serial port at 19.2k.
When it comes to a mobile retro robotics platform, [Mike] couldn’t have found a better computer. The Tandy 102 has a display, a BASIC interpreter, enough RAM to run a Roomba, and is powered by a few AA batteries. He did need a little bit of level conversion for the serial port, but a MAX232 took care of that easily.
With everything put together, [Mike] had a robot and a computer that is at least as good as the old Heathkit HERO robot. You can check out a video of the Tandy bot below.
Continue reading “Hackaday Retro Edition: The RadioShack Roomba”
[TheBackyardScientist] has been living up to his name, this time by casting a pewter sword in his yard. Pewter is a soft alloy of mostly (85–99%) tin along with copper, antimony and bismuth. Older pewter castings often used lead as well. The great thing about pewter is its low melting point of 170–230 °C. At such low temperatures, pewter can be melted down on a common hot plate. Think of it as an easy way to get into the world of metal casting – no forge required. Of course, anyone who has been splashed with solder will tell you that hot molten metal always deserves a lot of respect.
[BackyardScientist] obtained his metal by hunting the local thrift stores. He used the “lost foam” method of casting, by carving a sword out of styrofoam. The sword was embedded in a 5 gallon bucket of sand with a riser to allow the mold to be filled. The pewter was melted on a cheap hot plate, and poured into the mold. The hot metal melts the foam on contact, simultaneously filling up the cavity left over in the sand mold. [BackyardScientist] was left with a solid pewter sword. It won’t hold an edge, but it is a great illustration of the technique.
Click past the break to see [TheBackyardScientist’s] video.
Continue reading “From Scrap To Sword: Casting Pewter”
It only takes one mistake to realize electrolytic capacitors have a polarity, but if you’re working with old tube gear, tube amps, or any old equipment with those old orange dip, brown dip, or green dip foil capacitors you also have to watch your polarity. These old caps were constructed with a foil shielding, and there’s always one side of these caps that should always be connected to the chassis ground. If you don’t, you’re going to get interference – not something you want in an amplifier circuit.
Old caps that have long since given up the ghost usually have a black band designating whatever side of the cap the ‘foil ground’ is. This is the side that should be connected to ground. If you look at modern foil caps, you might also see a black band on one side of the cap, which should – if we lived in a just world – also designate the foil ground. This is not always the case.
To properly test foil caps and determine which side should be closer to ground, you can construct a small tester box that’s more or less an h-bridge with a single switch and a pair of alligator clips in the middle. Connect the cap to the clips, put the output of the circuit in your scope, and flick the switch: the direction that has the least amount of interference is the denotes the foil ground of the cap. Replace those old caps in your vintage equipment with a new, correctly oriented cap, and you’re well on your way to having a great sounding amplifier.
Continue reading “How To Tell If You’re Installing Foil Capacitors Backwards”