Movidius Deep Learning USB Stick by Intel

Last week, Intel launched the Movidius Neural Compute Stick, which is a deep learning processor on a USB stick.

This USB stick was not an Intel invention. In fact, Intel had acquired Movidius company that had produced last year the world’s first deep learning processor on a USB stick based around their Myriad 2 Vision Processor.

Neural Compute Stick is based around the Movidius MA2150, the entry level chip in the Movidius Myriad 2 family of vision processing units (VPUs). Using this stick will allow you to add some artificial visual intelligence to your applications like drones and security cameras. 

Movidius Neural Compute Stick form factor device enables you prototype and tune your deep neural network. Moreover, the USB form factor connects to existing hosts and other prototyping platforms. At the same time, the VPU provides machine learning on a low-power inference engine.

Actually, the stick role comes after training your algorithm where it is ready to try real data. All you have to do is to translate your trained neural network from the desktop using the Movidius toolkit into an embedded application inside the stick. Later on, the toolkit will optimize this input to run on the Myriad 2 VPU. Note that your trained network should be compatible with Caffe deep learning framework.

It is a simple process

  1. Enter a trained Caffe
  2. Feed-forward Convolutional Neural Network (CNN) into the toolkit
  3. Profile it
  4. Compile a tuned version ready for embedded deployment using the Neural Compute Platform API.

An outstanding feature is that the stick can work without any connection to cloud or network connection, allowing to add smart features to really small devices with lower consumption. This feature may be on of the revolutionary ideas to start combining IoT and machine learning devices.

Neural Compute Stick Features

  • Supports CNN profiling, prototyping, and tuning workflow
  • All data and power provided over a single USB Type A port
  • Real-time, on device inference – cloud connectivity not required
  • Run multiple devices on the same platform to scale performance
  • Quickly deploy existing CNN models or uniquely trained networks
  • Features the Movidius VPU with energy-efficient CNN processing

“The Myriad 2 VPU housed inside the Movidius Neural Compute Stick provides powerful, yet efficient performance — more than 100 gigaflops of performance within a 1W power envelope — to run real-time deep neural networks directly from the device. This enables a wide range of AI applications to be deployed offline.” — Remi El-Ouazzane, VP and General Manager of Movidius.

At the moment, the stick SDK in only availble for x86, and there are some hints to expand platforms support. Meanwhile, developers are hoping to have ARM processor support since many of IoT applications rely on ARM processor. However, this may be not possible since the stick is an Intel product.

This stick is available for sale now, and costs $79. More information about how to get started with the stick is available on the Movidius developer site. Also check this video by Movidius:

 

Pulse Generator For Stepper Controller Using AD654

This stepper pulse generator project is an easy solution for stepper controller drive. It’s a very important tool and can be used to drive stepper in standalone mode. It generates square wave pulses in frequency range 0-50Khz. This project has multiple features which are a must for stepper controller. It has on board frequency generator with wide span of frequency, Slide switch for direction control and jumper for enable or disables the stepper controller. AD654 is heart of the project and its generate the pulse for stepper controller, output frequency 0-50Khz, higher frequency output is possible by changing CT capacitor value connected between pin 6 and 7. Refer to data sheet of AD654 for alteration.

Pulse Generator For Stepper Controller Using AD654 – [Link]

Next Generation Solar Cell That Can Capture Nearly All Energy of Solar Spectrum

Researchers developed a multijunction solar cell on a GaSb substrate that can efficiently convert the long-wavelength photons typically lost in a multijunction solar cell into electricity. This prototype cell has an efficiency of 44.5% which is higher than conventional solar cells.

Next Generation Solar Cell To Absorb Nearly All Solar Spectrum
Next Generation Solar Cell To Absorb Nearly All Solar Spectrum

A GaAs-based cell is stacked mechanically with the GaSb-based materials to create a four-terminal, five junction cell with a spectral response range covering the region containing greater than 99% of the available direct-beam power from the Sun reaching the surface of the Earth. By comparison, the most typical solar cell can convert only one fourth of the available energy into electricity.

The working principle of this new solar cell is slightly different than the commonly available one. The cell is assembled in a mini-module that has a lens with a geometric concentration ratio of 744 suns. The lenses to concentrate sunlight onto tiny, microscale solar cells. As the solar cells have a very tiny form factor of  1 mm², solar cells using more complicated materials can be developed cost effectively.

The stacked cell acts like a filter with a particular material in each layer to absorb a specific range of wavelength of sunlight. The stacking procedure uses the transfer-printing technique which enables three dimensional modeling of these super-tiny devices with a high degree of precision.

Around 99 percent of the power contained in direct sunlight reaching the surface of Earth falls between wavelengths of 250 nm and 2500 nm. The entire range is not accessible by conventional solar panels as they are made from abundant, cheaper materials, such as silicon. Matthew Lumb, the lead author of the study and a research scientist at the GW School of Engineering and Applied Science, said,

Our new device is able to unlock the energy stored in the long-wavelength photons, which are lost in conventional solar cells, and therefore provides a pathway to realizing the ultimate multi-junction solar cell.

The cost of this specific solar cell is pretty high due to the high-end materials used and complex technologies implemented. However, the researchers achieved the upper limit of possibility in terms of efficiency. The new solar cell shows much promise in spite of being highly expensive. perhaps in future, the production cost can be reduced and the similar solar cell will be available commercially in the market.

290Hz Narrowband Laser On Chip For Numeros Photonic Applications

Researchers from the MESA+ research institute at University of Twente have collaborated together with the provider company of the customized microsystem solutions “LioniX International” to achieve the lowest bandwidth tunable diode laser on a chip.

The newly-developed laser operates in the IR region at 1550 nm with an 81 nm tuning range, which means that users can choose the color of the laser themselves, within a broad range. The laser is an integrated InP-Si3N4 hybrid laser consists of two different photonic chips, optically connected to each other.

Photonics is a key technology that makes numerous other innovations possible. So that, scientists and researchers are making big efforts at this field including deployment of photons for transporting and processing data.

To make photonic chips function as efficient as possible, we need to be able to control the light signals. Which means that all transmitted light particles should have the same frequency and wavelength as possible. The university researchers have succeeded developing a tiny laser on a chip with a maximum bandwidth of just 290 Hertz.

Our signal is more than ten times more coherent – or clean – than any other laser on a chip.
~ Professor Klaus Boller, the research leader

This record laser will have countless applications especially in fiber optic communications that require high data rate. This applications includes 5G mobile networks, accurate GPS systems and sensors for monitoring the structural integrity of buildings and bridges.

You can find out more details here.

The Aurora Boxealis – A Color Sensing and Mirroring Project

Besides looking damned good on an otherwise bland and ordinary desk, this project is about more than just being attention grabbing eye candy.  It’s about demonstrating a small portion of our single board computer capabilities by hooking up a color sensor, RGB light strip, and enclosing it in a nice looking wooden enclosure.  We’re dubbing it the “aurora boxealis”, and it’s made to stand out from the crowd at trade shows and provide a fun, interactive way to professionally demonstrate an interesting sensor, in this case a color sensor.  Grabbing a color swatch from the table and placing it on the top of the box will trigger the lights to mirror that color.

The Aurora Boxealis – A Color Sensing and Mirroring Project – [Link]

Arduino PICO, The Tiny Arduino-Compatible Board

MellBell, the Canadian-based hardware and electronics company, has launched their first product: Arduino PICO!

At first, the company says that Arduino PICO is the smallest Arduino compatible board ever, since it is 0.6″ x 0.6″ inch sized (~15mm squared). This tiny fully-fledged arduino-compatible board has a Leonardo-compatible 16MHz ATMEGA32U4 chip and a micro-USB port. The main cause of building PICO was to have a really small brain to use in many application with worrying about size or allocated space.

PICO’s Technical Specifications

The 16MHz ATMEGA32U4 integrates 2.5KB SRAM and 32KB flash, 4KB of which the bootloader uses. The 1.1-gram PICO has 8x digital I/O pins, 3x analog inputs, a PWM channel, and a reset button. In addition, the board has a 7-12V power with 5V operating voltage, where each I/O pin uses 40mA. It is worth to mention that PICO is competing with 12 x 12mm, $18 µduino, which similarly offers an Arduino Leonardo compatible ATMEGA32U4 MCU and which is smaller in size.

Moreover, MellBell provides an aluminum version that comes with the same ATMEGA32u4 core processor. With an Aluminum not regular fiber-glass, this makes PICO more reliable for overheated applications and environments.

Arduino PICO is now live on a Kickstarter campaign that two days ago had achieved its goal! Fortunately, there is still a chance until 17 Aug 2017 to pre-order one of PICO’s packages. You can get your early bird PICO for CA$18 ($14) and Aluminum edition for CA$32($25). Also, there is a special edition that includes  Aluminium PICO, four colored PICOs,  PICO joystick shield, micro drone kit, PICO solar station,  dual PICO board,  micro li-ion battery, PICO starter kit,  MiniMega board and finally a special “THANK YOU” video for CA$ 960 ($765).

Check out the campaign video:

Romeo BLE – An Arduino Based Powerful Robot Control Board With Bluetooth 4.0

Romeo BLE is an all-in-one Arduino based control board specially designed for robotics applications from DFRobot. This platform is open source and it’s powered by thousands of publicly available open-sourced codes. Romeo BLE can easily be expanded using Arduino shields. The most important feature—Bluetooth 4.0 wireless communication, allows the board to receive commands via Bluetooth. So, users can now use their smartphone, tablet, or computer to interact with the control board.

Control Robot From Smartphones by Bluetooth 4.0
Control Robot From Smartphones by Bluetooth 4.0

Even the codes can be uploaded over Bluetooth a USB Bluno Link adapter, without requiring any wired USB connection between the board and a PC. This is a great advantage for mobile applications where codes are debugged and uploaded frequently.

The Romeo BLE also includes two integrated two-channel DC motor drivers and wireless sockets, which makes project development more hassle-free. One can start the project immediately without needing an additional motor driver circuitry. The motor driving section also supports extra servos which need more current.

There are two ways to power the Romeo BLE board. But, the polarity must be correct. Otherwise, the board may get permanently damaged as there exists no reverse polarity protection. The two powering methods are:

  • Power from USB: Plug in the USB cable to the Romeo controller from a power source (i.e. wall jack or computer). If the input voltage and current are sufficient, the Romeo BLE board should turn on and a LED should light up. While powered from USB, do NOT connect anything else like motor, servo etc. except LED. Because the USB can only provide 500mA current which is certainly not enough for driving loads like motors.
  • Power from External Power Supply: Connect the ground wire from your supply to the screw terminal labeled “GND” on Romeo board, and then connect the positive wire from your supply to the screw terminal labeled “VIN”. The maximum acceptable input voltage is 23 volts. Do not exceed this value anyway.
Romeo BLE Board Pin Diagram
Romeo BLE Board Pin Diagram

Specifications:

  • Microcontroller: ATmega328P
  • Bootloader: Arduino UNO
  • Onboard BLE chip: TI CC2540
  • 14 Digital I/O ports
  • 6 PWM Outputs (Pin11, Pin10, Pin9, Pin6, Pin5, Pin3)
  • 8 10-bit analog input ports
  • 3 I2Cs
  • 5 Buttons
  • Power Supply Port: USB or DC2.1
  • External Power Supply Range: 5-23V
  • DC output: 5V/3.3V
  • Size: 94mm x 80mm

Features:

  • Auto sensing/switching external power input
  • Transmission range: 70m in free space
  • Support Bluetooth remote update the Arduino program
  • Support Bluetooth HID
  • Support iBeacons
  • Support AT command to config the BLE
  • Support Transparent communication through Serial
  • Support the master-slave machine switch
  • Support USB update BLE chip program
  • Support Male and Female Pin Header
  • Two-way H-bridge motor Driver with 2A maximum current
  • Integrated sockets for APC220 RF Module

You can program Romeo BLE using Arduino IDE version 1.8.1 or above. Select Arduino UNO from Tools –> Boards in the IDE. Go to arduino.en.cc to download the latest version of Arduino IDE. Read the Romeo BLE wiki to learn more.

Harvesting Sound Energy From Passing Cars

by Mechanical Attraction @ instructables.com:

There is energy everywhere around us and in many different forms. Many devices have been developed to harvest light, wind, waves, and more. One unusual place of energy harvesting is from passing cars. As cars pass by some of their energy is released in form of sound. Even though the overall energy maybe small it can be harvested. In this Instructable I will show how to apply the solution of Euler–Bernoulli beam theory to design a cantilever beam to oscillate at such a frequency to adsorb sound waves as well as converting its mechanical motion into electricity.

Harvesting Sound Energy From Passing Cars – [Link]

Should you build or buy your next single-board computer?

by Markku Riihonen @ EDN Europe:

Does it make sense to design and build your own single-board computers? It used to be the sensible option for anyone concerned about matching features to production cost.
Traditionally, with your own board design, you have the freedom to add only the components that are absolutely vital to achieving the right level of functionality for the target application. But the relentless rise of the system-on-chip (SoC) device has changed that equation in a number of ways.

Should you build or buy your next single-board computer?- [Link]