from EETimes: http://ift.tt/1tn6EAU
via Yuichun
Mailbag Monday.
Dave opens his mail
Forum HERE
Spoilers:
Keithley 177 Multimeter teardown schmeatic and Manual
Commodore GL987R Calculator Teardown and VFD probing.
Sharpie Hanger Kickstarter
Some interviews from the 2014 Sydney Mini Maker Faire at the powerhouse museum.
Winstar has introduced a 3.5in version of its M Series colour TFT LCDs.
WF35M measures 100x66mm and under 6mm thick (plus connector) has a built in RS-232 interface. The actual display is 70x53mm, and 320xRGBx240.
There is a resistive touch panel option, and an SPI interface version is on the way. The LED backlight is dimmable.
The built-in controller is a PIC 24, which can receive in-situ software up-grades.
At the same time, Winstar released a roadmap for the rest of the TFT M Series.
In Q3 it will be introducing a 4.3inch version, with RS-232 or SPI interfaces and both resistive and capacitive touch options.
Before the end of the year, and with the same features, there will also be 5.7inch M Series display.
The firm has demo software for user-interface development.
UMC has paid $48 million for a 9.3% stake in a newly formed foundry company which is a subsidiary of Fujitsu Semiconductor including its 300mm wafer manufacturing facility located in Kuwana, Mie, Japan.
UMC’s 40nm technology will be licensed to Fujitsu.
“Partnering with a leading Japanese semiconductor company like Fujitsu Semiconductor for a local joint venture will not only offset the time, risk and cost of building a new fab, but also provide access to another 300mm manufacturing source in addition to UMC’s own 300mm operations in Taiwan and Singapore,” says UMC CEO Po Wen Yen, “with three 300mm operations in different regions throughout Asia, both parties will hold unique positions to be able to serve customers looking to mitigate manufacturing risk, such as Japan’s automotive chip makers, who vigorously seek suppliers that implement robust business continuity plans. UMC will also be able to leverage this strategic partnership to gain new foundry business within the Japanese market.”
“Fujitsu has been in search of a partner with whom to jointly operate a new foundry company based on its Mie 300mm wafer fab,” says Haruki Okada, President of Fujitsu Semiconductor, “Fujitsu Semiconductor and UMC have built up a good relationship while Fujitsu Semiconductor has outsourced its products to UMC for years. The joint venture will provide foundry services to customers, based on a wide variety of CMOS technologies, the 40nm process technology licensed by UMC and high quality manufacturing system certified by automotive customers.”
Anite has added carrier aggregation and Category 6 device support to its Nemo Outdoor laptop-based drive wireless network tester.
Phone network operators worldwide are testing and trialling carrier aggregation, and are launching LTE-Advanced (LTE-A) networks with the first phase of Cat 6 devices, using two carrier components with down-link bandwidths of 20MHz.
“Demonstrating Cat 6 peak data rates shows Anite’s commitment to ensure that our products support the latest wireless broadband technologies,” said Anite CTO Kai Ojala. “We have achieved application data throughput of 293Mbit/s with a 40MHz bandwidth using two-component carrier aggregation and the latest test terminals with Anite’s Nemo Outdoor test solution in a laboratory environment.”
Nemo supports over 270 terminals and scanning receivers from a range of vendors, and many network technologies.
Carrier aggregation allows peak data rate to be increased by sharing a single data stream between more than one LTE-A frequency band.
CSR has appointed a bank to advise it on offers from other semiconductor companies to take it over, reports the FT.
It is thought by the FT that the company could attract a bid of up to $3 billion.
After news of the sale was leaked, the shares shot up 23% to over £7, valuing the company at around $2 billion.
Two years ago, CSR sold its WiFi, Bluetooth and GPS chip development operations to Samsung for $310 million and Samsung paid £2.23 a share for a 5% stake in the company which it sold earlier this year, in January, for £6.60 a share.
PC unit shipments are expected to fall by 3.7% in 2014, says IDC.
PC shipments in mature regions are now projected to grow by 5.6% in 2014 – the highest since 2010 – with both consumer and commercial segments showing positive growth.
On the other hand, the outlook for emerging market has been lowered slightly to reflect reduced stability and economic conditions in Asia/Pacific, Latin America, and Central Europe, the Middle East and Africa (CEMA).
PC market drivers for the short term still mainly rest on business spending in mature markets. However, other variables have also appeared. In addition to the stimulus from the end of support for Windows XP, competition from tablets has declined as tablet penetration rises and volume has shifted to smaller sizes that are less competitive with PCs.
At the same time, PCs have made some progress addressing tablet competition with more slim, touch, and low-cost models available.
Growth of Chromebook demand, particularly in education in mature markets, is also contributing to the PC outlook. In emerging regions, the competition from other devices remains a larger factor, and the boost from mature market drivers less pronounced.
Replacement of Windows XP-based systems should also decline over time as the installed base of XP systems shrinks.
“Programs to reduce PC prices, such as Windows 8.1 with Bing, have helped to improve PC shipments in some segments,” said IDC’s Jay Chou. “Coupled with a shift toward more mobile PCs, the market has seen a quickened pace of innovation and a focus on price points. Nevertheless, the prospects for significant PC growth in the long term remain tenuous, as users increasingly see PCs as only one of several computing devices.”
“Direct competition from tablets seems to be waning,” said Loren Loverde, vice-president of worldwide PC trackers. “However, PC replacement cycles have expanded as users have alternative computing platforms and places to spend money. The launch of Windows 9 Threshold in 2015 could potentially boost demand, although it will be hard to gauge the actual impact until the advantages to consumers and commercial users in functionality and integration with specific devices is more apparent. For the moment, we continue to see PC demand coming primarily from replacements with overall shipments declining slightly through the end of the forecast.”
Researchers at Harvard University’s school of engineering have been puzzling over autonomous robot collaboration – in particular, how do you get robots that can only sense a short distance to make something big without an all-seeing eye or central control to direct them.
And recently they have had some success – getting 1,024 ‘Kilobots’ to huddle into pre-defined patterns, and three construction robots to make simple buildings.
Kilobots are extraordinarily simple, designed at Harvard to be ant-like and mass-produced cheaply – they are now available from a Swiss company called K-Team.
Each Kilobot is essentially a 33mm diameter PCB standing on three 20mm long rigid wire legs.
Intended for use on a smooth surface – Harvard has a 2.4×2.4m table – Kilobots are shaken forward by vibration from two phone-style vibrator motors on the PCB.
By spinning the motors at different speeds, Kilobots can creep forward, turn left, or turn right.
Power for three hours of travel comes from a Li-ion battery carried on top. Charge control is included on the PCB, with charge power coming from voltage applied between any leg and a spring on top of the battery.
On-board, an 8bit Atmel microcontroller (2kbyte RAM, 32kbyte ROM) does the thinking, and a top-mounted tri-colour LED displays status to human observers.
Perhaps the most crucial piece of equipment is a wide-angle infra-red transceiver mounted underneath in the centre of each Kilobot. Infra-red from this bounces off the white table top to be picked up by other nearby Kilobots.
Kilobots swarm following the map on the left.
Communication range is up to 10cm, and the intensity of received IR gives the robot some idea of the distance between it and its neighbours. Distance estimation varies somewhat from robot to robot, and is not identical in all directions, so control algorithms have to be tolerant of these vagaries.
This limited communication range gives swarm members their short-sightedness. If a neighbour is more than 10cm away, a robot cannot have any direct knowledge of it, only knowledge that has been passed via others.
For convenience, all members of the swarm can be programmed simultaneously by an over-head optical transmitter. All members of the swarm, except four in the pattern-forming experiment, get the same programme.
The universal programme includes a map of the required pattern in the form of a coarsely-pixelated grid (see photo), with pixels set to 0 (leave empty) or 1 (fill with robots). Pixels only indicate the pattern to be filled, not exact final robot positions – there is no intention to achieve one-robot-per-pixel in a square grid. Robots fill the required shape in a largely random arrangement with average inter-robot spacing determined by a pre-programmed value – more of this later.
All the Kilobots are closely packed into one arbitrarily-shaped swarm, adjacent to where the final pattern is required (see photo).
The four unique robots are seeds, placed within range of each other on the edge of the starting swarm.
The seeds will not be moving, and will eventually be just inside the edge of the finished pattern. One of them is programmed to be the origin of the finished pattern: x=0, y=0.
All robots transmit and receive frequently. When a message is received, as well as decoding it, the recipient estimates how far away the transmitting robot was.
One number passed between robots is ‘gradient’. All robots listen for gradient values coming from neighbours. They assume a gradient value one more then the lowest value they hear, and transmit this new value.
At the start, only the origin robot knows its gradient value, which is set to ’0′. This it transmits, and a neighbour receiving this 0 gradient value assumes its own gradient value ’1′, and transmits this. Depending on how they are positioned, neighbours of these two hear 0 or 1 as the lowest value in earshot, and assume 1 or 2 respectively.
The overall effect is that gradient numbers, on average, increment along straight lines away from the origin robot like spokes in a cartwheel. Lines drawn between robots with equal gradient values will form loops around the origin.
These loops are actually near to circular as there is a further restriction in gradient dissemination: robots have to ignore values sent from greater than a pre-programmed range. When this is set closer to 33mm, gradient value contour loops are closer to circular.
Any robot that can only hear robots of lower or equal gradient value, know they are on the edge of the swarm, and know roughly the size of the swarm, even though the centre is way beyond their 10cm communication range.
Until now, all robots have also been transmitting an ‘I am stationary’ message.
Robots that know they are on the edge of the swarm start to move forwards and right, while staying a pre-programmed fixed distance from any neighbour they approach – it is this distance which will eventually govern final image inter-robot spacing.
This simple rule causes edge robots to orbit the coastline of the swarm, each performing a part-circle around every individual on the coast before passes on to its neighbour.
The four seeds have the beginnings of an x-y grid programmed into them and, by each measuring their distance to the other three (many times to reduce noise), they can calculate an estimate of their x-y coordinates on the required image pixel grid. Once they have estimated their position, they are ‘localised’ and transmit their x-y positions.
Triangulating on three neighbours gives three different positions on the plane. A simple mathematical algorithm gets best positional accuracy by ‘minimising the spread’ of the three positions, said Harvard.
Kilobots swarm following a different map.
Once the seeds have localised, their neighbours can also localise and the x-y grid spreads through the swarm as soon as any non-localised robot can triangulate on three localised neighbours.
Moving edge robots continually recalculate their position (and gradient value) to remain localised, and continue to transmit their position for the use of others.
Eventually, an edge-following robot, comparing its estimated position to its internal pixel map, will know it has entered the required pattern. It will continue to edge-follow inside the image.
It will stop when it: is about to leave the required image, or it bumps into a robot of equal gradient value that has already stopped – which the moving robot knows because the stationary one will be sending out an ‘I have arrived’ message.
These two simple stopping rules cause the image to fill-out with approximately concentric arcs of stationary robots centred on the origin robot.
All sorts of mix-ups are possible – for example: too many robots moving off the starting coastline at the same time, sticking vibration motors, faster robots catching up with slower ones, or clumsy robots knocking others from their final resting place.
A few simple rules have been developed to sort these situations.
Where rules cause stalemate between two robots, the robot with the highest ID number wins.
These ID numbers are randomly self-allocated. If two robots with the same ID come into range, they both randomly recalculate – which can result in a ripple of re-calculation should a new number match another local. ID numbers are locally unique, but may not be globally unique.
Algorithms cause and cope with odd situations
In experiments, 1,000 Kilobot images (see K photo) have taken up to 11 hours to form.
Multiple attempts at forming identical shape and size rectangles with a smaller number of Kilobots have shown the algorithms are robust – the rectangle is always formed, even though the pattern of robots within each rectangle is different in each case.
Why not just simulate the whole thing?
“We can simulate the behaviour of large swarms of robots, but a simulation can only go so far,” says Professor Radhika Nagpal. “The real-world dynamics, the physical interactions and variability, make a difference, and having the Kilobots to test artificial intelligence algorithms on real robots has helped us better understand how to recognise and prevent the failures that occur at these large scales.”
In a separate project, called Termes, Harvard has taken inspiration from termites.
This involves 3-d rather than 2-d construction, and robots carrying and placing square tiles to make a structure rather than making a pattern out of their own bodies.
The tiles have the proportions of Scrabble tiles, although they are considerably larger.
Termes robots have four types of sensor and three actuators. They can travel horizontally, climb one tile, or descend one tile, while carrying another tile, which they can deposit in front of themselves – so they can make their own staircase of tiles.
Also, as they can turn around on the spot, they can also construct a staircase with right-angled bends.
The desired structure is described as an x-y-z grid of whole tiles.
Rather than provide each robot with a 3-d map of the finished structure and have it decide its own method of construction – with the inherent risk of one robot stranding another – an off-line compiler turns the structure definition into a construction plan, for any number of robots, which is a series of paths, consisting of flat sections and staircases. All directions are left, right or straight. There are no diagonal paths.
To avoid congestion, this plan has one or more routes up and one or more paths down.
Where ever a robot is, which it knows by always starting at tile 0,0,0 and counting whole tiles in x, y and z, its stored path information tells it which direction, or directions, it can move away from its current position.
When it come across a place where a tile should be, it deposits its load and moves down for another tile.
The exact set of rules followed by Termes robots are created by the compiler for the particular job. The compiler also rejects structures that are impossible to build.
Like Kilobots, once Termes robots leave the construction site they edge-follow the incomplete structure to the point that come across spare tiles, then edge follpw around to tile 0,0,0 whee the construction path always begins.
Harvard has proved that locally-collaborating robots can construct large pre-defined structures, rather than needing central control and long-distance communication.
And while centralised systems can have good group efficiency and fast recovery from problems, Harvard points out that central control fails when the central controller breaks, and it also proposes that central control might not be the best solution with large numbers of robots, over wide areas, or in remote locations.
“We’ve proven the extreme end of the scale: that it could be just like the termites. And from the termites’ point of view, it’s working out great,” says Nagpal. “It may be that, in the end, you want something in between the centralised and the decentralised system.”
The Kilobot experiment is described in the 15th August issue of Science, and Termes in the 14th February issue and at the AAAS 2014 Annual Meeting. Research was supported by the Wyss Institute for Biologically Inspired Engineering at Harvard University.
There is a Harvard Kilobot swarm video. Scroll down.
FTDI Chip and MCCI Corporation have developed TrueTask USB – an embedded USB host stack designed for use with FTDI Chip’s FT900 product family.
TrueTask USB is an embedded USB host stack supporting the FT900 series of MCUs. With today’s announcement, embedded product developers who select FT900 MCUs will have access to a USB platform capable of supporting a variety of USB peripherals – such as USB storage, portable media players, video/audio conferencing system peripherals, mobile phones, Wi-Fi dongles, etc.
“The FT900 is a powerful, truly embedded MCU with an elegant architecture and rich connectivity offerings,” said MCCI CEO Terry Moore,adding that it is field proven and comes with an extensive class driver portfolio. “By tightly integrating TrueTask USB with the FT900 MCU we were able to optimise performance and memory without compromising on quality and features,” said Moore.
“TrueTask USB from MCCI is known throughout the embedded industry for high quality USB,” said Fred Dart, CEO and founder of FTDI Chip. “Having a pre-integrated, best-in-class USB stack is aligned with our product strategy for rapid go-to-market of our customers. We have worked extensively to ensure both a technical and commercial alignment between TrueTask USB and the FT900 series, in order to ensure that our customers experience the best value for money.”
Infineon Technologies’ decision to pay $3bn for International Rectifier may have surprised the semiconductor industry, but the deal is all about acquiring a process technology.
That process technology is gallium nitride-on-silicon.
GaN-on-silicon, as it is called, is a hot commodity in the power electronics industry and IR has one of the few working GaN-on-silicon processes in production.
Why is the technology so hot at the moment?
For one thing it runs cool.
The recent NeuLandproject, which was funded by the German Federal Ministry of Education and Research (BMBF), concluded that SiC and GaN-on-silicon can halve the energy loss in switched mode power supplies for PCs and can increase the efficiency of power electronics by one third.
The project was directed by Infineon, which uses SiC in its Jfets and diodes for the 600V to 1700V voltage class primarily used in switched-mode power supplies for PCs or televisions and in motor drives.
With the IR acquisition, Infineon now has production ready GaN-on-silicon, the other piece in the jigsaw.
GaN-on-silicon technology is also a motivation behind Google’s $1m power inverter design challenge.
“We’re looking for someone to build a kW-scale inverter with a power density greater than 50W/inch3,” said the firm, which is suggesting wide bandgap semiconductors (GaN or SiC, for example) could be part of the answer.
The next big target application for GaN-on-silicon is to increase energy efficiency and reduce cost in the design of solar inverters.
However, the market is not ready for GaN-on-silicon yet. The cost of components will have to fall even more for wide-scale application in solar inverters, and that means further research is required on reliability, service lifetime and costs.
With the acquisition of IR, Infineon is making a statement that it believes in the future of GaN-on-silicon and sees an opportunity to become a market leader, but that will need a two-way fight with the other GaN powerhouse, Cree.
What’s inside a dial-in pacemaker monitor system?
Datasheets:
LMV824
MAX4330
CNY64 Optocoupler
ADSP-2185N DSP
CX88168 Smart Modem with Voice CODEC
MAX994 10bit ADC
Meder Reed Sensor
Probe shock video showing the piezoelectric effect
Forum HERE
TROY, MI—Contract manufacturer Unified Business Technologies could add 171 jobs over the next three years after winning an Army contract to make components for surveillance and communications.
Fairchild is laying off about 15% of its workforce as it closes plants in Utah, Malaysia and Korea.
Fairchild is to stop five-inch silicon production in Bucheon, Korea and reduce its six-inch wafer capacity.
The closures are expected to take effect from Q2 2015 to Q4 2015.
“An adaptive supply chain must be the foundation of any global manufacturer’s operations in the increasingly dynamic semiconductor solutions market,” says Fairchild CEO Mark Thompson.
“The realignment we are announcing today will maximise the utilisation of eight-inch factories and reduce the complexity of our manufacturing footprint, while creating the flexibility to support ongoing customer demand through a greater use of external manufacturing sources. Fairchild will continue operating eight-inch wafer fabrication lines in South Portland, Maine and Mountain Top, Pennsylvania, as well as the Bucheon six- and eight-inch fabrication lines. Fairchild will also continue operating assembly and test facilities in Cebu, Philippines and Suzhou, China.”
Through the combined actions, Fairchild expects to incur approximately $36 million in cash restructuring and other costs. The company also plans to record during the closure process non-cash charges of approximately $25 million for accelerated depreciation.
Once completed, the company expects to realise annual savings of approximately $45 to $55 million from a second quarter of 2014 financial baseline. Of these estimated savings, approximately 75 percent are expected to be cash savings, with the balance attributable to lower depreciation costs. after the closures have been completed, and about 75% of that is expected to be cash savings.
Cypress and foundry Shanghai Huali Microelectronics Corporation (HLMC), have developed functioning silicon cells using Cypress’s Sonos (silicon oxide nitride oxide silicon) 55nm embedded flash.
The silicon cells are designed for smartcards and the Internet of Things (IoT).
The technology and design IP will be available for high-volume manufacturing by HLMC customers in the second half of 2015.
HLMC licensed Cypress’s 55nm Sonos embedded nonvolatile memory (NVM) process in January 2014.
Sonos only requires three mask layers to insert it into a standard CMOS process compared with the nine to 12 additional masks generally needed for other embedded Flash technologies. This mask reduction results in lower manufacturing costs.
Sonos does not alter standard device characteristics or models when it is added to baseline CMOS process, preserving existing design IP.
It promises high yields and reliability, 10 years of data retention, 100,000 program/erase endurance cycles, and resistance to soft errors.
Cypress has demonstrated the ability to scale Sonos to 40nm and 28nm nodes.
FAIRFIELD, CT—Scientists at GE are experimenting with a technology, called Direct Write, that uses special “inks” to print miniature sensors directly inside jet engines, gas turbines and other hot, harsh and hard-to-reach places.
The SEMI book-to-bill ratio for July was 1:07, down from the 1.10 of a June but an improvement on the 1.00 of May, the1.03 of April, the 1.06 of March and the 1.01 of February.
July bookings were $1.41 billion which is 2.8% lower than June’s $1.46 billion, and is 17.1% higher than the July order level of $1.21 billion.
July billings were $1.32 billion which is 0.7% lower than June’s $1.33 billion and 9.4%?higher than the July 2013 billings level of $1.20 billion.
“Order activity for semiconductor equipment has held at a steady level so far for 2014,” says Denny McGuirk, president and CEO of SEMI, “this trend, along with improvements in semiconductor device sales and unit shipments, is consistent with our outlook for strong equipment sales growth this year.”