New Part Day: The RISC-V Chip With Built-In Neural Networks

October 8, 2018 by Brian Benchoff 87 Comments

After exploring a few random online shops one day, [David] (thanks for sending this in, by the way) ran across a very interesting chip. It’s a dual-core, RISC-V chip running at 400MHz. There’s 6 MB of SRAM on the CPU, and there’s 2MB for convolutional neural network acceleration. There is, apparently, WiFi on some versions. There are already SDKs available on GitHub, and a bare chip costs a dollar or two. Interested? Log in to Taobao, realize Taobao does pre-orders, and all this can be yours.

This is a preorder — because apparently you can do that as a seller on TaoBao, but the Sipeed M1 K210 is available as a ‘core’ board with 72 pins in a one-inch square package, a version with WiFi, or as a complete development board with an OV2640 camera, 2.4 inch LCD, microphone, and onboard USB. There are videos of this chip running a face detection routine. It found Obama.

A bit of googling tells us this chip comes from a company named Kendryte, and here the specs are repeated: this is a dual-core RISC-V with an FPU, a bunch of RAM, and can run TensorFlow. Documentation is available, although the datasheet will need to be translated, and as of this writing there’s a GitHub filled with SDKs and examples, with some of the repos updated in the last hour.

Over the years we’ve seen a few RISC-V chips given development boards, and you can buy them right now. The HiFive 1 is an exceptionally powerful microcontroller with processing power that puts it right up against the Teensy (which is built around a Freescale chip), but it’s also fairly expensive. We’re not sure the Arduino Cinque (also RISC-V) ever made it to production, but again, expensive. The idea that a RISC-V microcontroller could be available for just a few dollars is very interesting, it even comes with SDKs and utilities to make the chip useful.

Perfecting The Solar Powered Web Server

October 8, 2018 by Tom Nardi 35 Comments

Running a server completely off solar power seems like it would be a relatively easy thing to do: throw up a couple of panels, tack on a charge controller and a beefy battery, and away you go. But the reality is somewhat different. Most of us hackers are operating on a relatively limited budget and probably don’t have access to the kind of property you need to put out big panels; both pretty crippling limitations. Doing solar on a small-scale is hard, and unless you really plan ahead your setup will probably be knocked out on its first cloudy day.

So when [Kris de Decker] wanted to create a solar-powered version of his site “Low-tech Magazine”, he went all in. Every element of the site and the hardware it runs on was investigated for potential power savings, and luckily for us, the entire process was written up in meticulous detail (non-solar version here). The server still does go down from time to time if the weather is particularly poor, but in general it maintains about 90% uptime in Barcelona, Spain.

The solar side of the equation is fairly simple. There’s a 50 watt photovoltaic panel charging a 12V 7Ah lead-acid battery though a 20A charge controller. With an average of 4 to 6 hours of sunlight a day, the panel generates 300 Wh of electricity in the best case scenario; which needs to be split between charging the battery and running the server itself.

As for the server, [Kris] chose the Olimex Olinuxino A20 Lime 2 in part because of it being open source hardware, but also because it’s very energy-efficient and includes a AXP209 power management chip. Depending on processor load, the Olimex board draws between 1 and 2.5 watts of power, which combined with charging losses and such means the system can run through two days of cloudy weather before giving up the ghost. A second battery might be added in the future to help improve the run time during low-light conditions, but for now its been working pretty well.

Perhaps the most interesting part of the whole project are the lengths to which the website itself was optimized to keep resource utilization as low as possible. Images are compressed using dithering to greatly reduce their file sizes, and the site eschews modern design in favor of a much less processor intensive static layout. There’s even a battery capacity display integrated into the page through some clever use of CSS. Even if you aren’t looking to set up your own sun worshiping website, there are tips here for building efficient web pages that could absolutely be put to use in other projects.

If you’re interested in solar projects, we’ve got you covered. From an open source charge controller to building DIY photovoltaic panels, there’s plenty of prior art you should find very…illuminating. Please clap.

Clock Monitors Deep Space Network, Keeps Vigil Over Lost Mars Rover

October 7, 2018 by Dan Maloney 7 Comments

It’s been a long, long time since we heard from Opportunity, the remarkable Mars rover that has shattered all expectations on endurance and productivity but has been silent since a planet-wide dust storm blotted out the Sun and left it starved for power. Right now, it’s perched on the edge of a crater on Mars, waiting for enough sunlight to charge its batteries so it can call home. All we can do is sit, and wait.

To pass the time until Opportunity stirs again, [G4lile0] built this Deep Space Network clock. Built around an ESP32 and a TFT display, the clock monitors the Deep Space Network (DSN) website to see if mission control is using any of the huge antennas at its disposal to listen for signals from the marooned rover. If the DSN is listening, it displays a special animation exhorting the rover to phone home; otherwise, it shows which of the many far-flung probes the network is communicating with, along with a slideshow of Mars mission photos to keep the spirits up. When the day finally comes that Opportunity checks in, an alarm will sound so [G4lile0] can pop the champagne and celebrate with the rest of us.

We realize that the odds that Opportunity will survive this ordeal are decreasing by the Sol. It’s an uphill battle; after all, the machine was 55 times its original 90-day design life when it went dark, so it’s an uphill battle. Then again, it has beaten the odds before, so there’s still hope.

Continue reading “Clock Monitors Deep Space Network, Keeps Vigil Over Lost Mars Rover” →

The Bells! The Bells! One Battery Since 1840

October 7, 2018 by Al Williams 18 Comments

It is good advice to change batteries in your fire alarms at least once a year. Even our low-power LCD calculators need new batteries from time to time. But at the University of Oxford, they have an electric bell that has been ringing essentially non-stop on one set of batteries for about 178 years! Is the energy crisis solved then? Perhaps not. The bells require a high voltage but very little current and the pair of batteries — piles in the parlance of 1840 — have kept the charge flowing for about 10 billion rings. As you can see in the video below, though, the ringing isn’t very vigorous.

How does it work? When you think of converting electrical power to mechanical motion you probably think of a motor, even though there are plenty of other transducers like speakers, muscle wires, and solenoids. Arguably the first device was electrostatic bells that were invented by a Scot named [Andrew Gordon] around 1742. [Ben Franklin] made them famous, though, so they are often called Franklin bells.

Continue reading “The Bells! The Bells! One Battery Since 1840” →

Hackaday Links: October 7, 2018

October 7, 2018 by Brian Benchoff 14 Comments

Ah, crap. We lost a good one, people. [Samm Sheperd] passed away last month. We’ve seen his stuff before, from a plane with a squirrel cage fan, to completely owning a bunch of engineering students by auditing a class. The obit is available as a Google Doc, and there’s a Samm Sheperd Memorial Fund for the Big Lake Youth Camp in Gladstone, Oregon.

FranLab is closing down! Fran is one of the hardware greats, and she’s being evicted. If you’ve got 2000sqft of workshop space in Philly you’d like to spare, you know who to talk to. There will, probably, be a crowdfunding thing going up shortly, and we’ll post a link when it’s up.

The Parallax Propeller is probably one of the most architecturally interesting microcontrollers out there. It’s somewhat famous for being a multi-core chip, and is commonly used in VGA generation, reading keyboards, and other tasks where you need to do multiple real-time operations simultaneously. The Parallax Propeller 2, the next version of this chip, is in the works, and now there’s real silicon. Everything is working as expected, and we might see this out in the wild real soon.

Thought artistic PCBs were just a con thing? Not anymore, I guess. There has been a lot of activity on Tindie with the Shitty Add-Ons with [TwinkleTwinkie] and [Potato Nightmare] releasing a host of very cool badges for your badges. Most of these are Shitty Add-Ons, and there will be an update to the Shitty Add-On spec shortly. It’s going to be backwards-comparable, so don’t worry.

Unnecessary drama!?! In my 3D printing community?!? Yes, it’s true, there was a small tiff over the Midwest RepRap Festival this week. Here’s what went down. You got three guys. John, Sonny, and Steve. Steve owns SeeMeCNC, based in Goshen, Indiana. John worked for SeeMeCNC until this year, and has been the ‘community manager’ for MRRF along with Sonny. Seeing as how the RepRap Festival is the only thing that ever happens in Goshen, Steve wanted to get the ball rolling for next year’s MRRF, so he sent out an email, sending the community into chaos. No, there’s not some gigantic fracture in the 3D printing community, John and Sonny, ‘were just slacking’ (it’s five months out, dudes. plenty of time.), and Steve wanted to get everything rolling. No problem here, just a bunch of unnecessary drama in the 3D printing community. As usual.

Towards Open Biomedical Imaging

October 7, 2018 by Brian Benchoff 9 Comments

We live in a world where anyone can build a CT machine. Yes, anyone. It’s made of laser-cut plywood and it looks like a Stargate. Anyone can build an MRI machine. Of course, these machines aren’t really good enough for medical diagnosis, or good enough to image anything that’s alive for that matter. This project for the Hackaday Prize is something else, though. It’s biomedical imaging put into a package that is just good enough to image your lungs while they’re still in your body.

The idea behind Spectra is to attach two electrodes to the body (a chest cavity, your gut, or a simulator that’s basically a towel wrapped around the inside of a beaker). One of these electrodes emits an AC signal, and the second electrode measures the impedance and phase. Next, move the electrodes and measure again. Do this a few times, and you’ll be able to perform a tomographic reconstruction of the inside of a chest cavity (or beaker simulator).

Hardware-wise, Spectra uses more than two electrodes, thirty-two on the biggest version built so far. All of these electrodes are hooked up to a PCB that’s just under 2″ square, and everything is measured with 16-bit resolution at a 160 kSPS sample rate. To image something, each electrode sends out an AC current. Different tissues have different resistances, and the path taken through the body will have different outputs. After doing this through many electrodes, you can use the usual tomographic techniques to ‘see’ inside the body.

This is a remarkably inexpensive way to image the interior of the human body. No, it doesn’t have the same resolution as an MRI, but then again you don’t need superconducting electromagnets either. We’re really excited to see where this project will go, and we’re looking forward to the inevitable project updates.

Reading Old Data Tapes, The Hard Way

October 7, 2018 by Dan Maloney 29 Comments

Those who were around for the pre-floppy days of computer mass storage were likely to have made the mistake of slipping a cassette tape containing data into a stereo tape deck. Instead of hearing the expected Awesome Mix, the speakers gave off an annoying bleat, warbling between two discordant tones and no doubt spoiling the mood.

What you likely heard was the Kansas City standard, an early attempt to provide the budding microcomputer industry with a mass-storage standard. It was successful enough that you can still find KCS tapes in need of decoding to this day. That job would be a snap with a microcontroller, which is exactly why [matseng] chose to do it the hard way and built a KCS decoder with nothing but discrete components.

The goal was to decode the frequency-shift-keyed (FSK) signal into an 8-bit parallel output, and maybe drive a seven-segment display as the characters came off the tape at a screaming 300 baud. Not an IC is in sight in the schematics; as [matseng] says, it’s nothing but “Qs, Rs, and Cs.” All the amps, flip-flops, and counters needed are built from a forest of transistors, and even the seven-segment display is a DIY affair of LEDs in a 3D-printed and hot-glue frame. The video below shows the display doing its best to show the alphanumeric characters encoded on the audio tape. And for who absolutely need a dose of Arduino, [matseng] used one along with a dead-bug low-pass filter to emulate KCS signals, for easier development.

We always appreciate hackers who take the road less traveled to arrive at a solution, but if you’re pressed for time to decode some KCS tapes, fear not – all you need is a PC and Audacity.

Continue reading “Reading Old Data Tapes, The Hard Way” →