Mouser examines a human-centric approach to robotics. Soft Robotic designs are less metal and mechanical and utilize gentle, compliant mechanisms and actuators built using fluids and flexible materials. These enable a wide range of motion and are suitable for exoskeletons or wearables.
IKALOGIC is giving away an IkaScope! (retail value $379)
The IkaScope WS200 is a pen-shaped battery-powered wireless oscilloscope that streams captured signals to almost any Wi-Fi-connected screen.
The IkaScope WS200 offers a 30M Hz bandwidth with its 200 Msamples/s sampling rate and maximum input of +/-40 Vpp. It provides galvanically-isolated measurements even when a USB connection is charging the internal battery. The IkaScope WS200 will work on desktop computers (Windows, Mac and Linux) as well as on mobile devices like tablets or smartphones. The free application software can be downloaded for whichever platform is needed.
The IkaScope WS200 has no power switch. It detects pressure on the probe tip and turns on automatically. Patented ProbeClick technology saves battery life: all power-consuming circuitry is only turned on when the probe tip is pressed, and the IkaScope WS200 automatically shuts down completely after a short period of non-use. The internal 450 mAh battery lasts about one week with daily regular use before recharging is necessary. An isolated USB connection allows for recharging the internal battery: two LEDs in the unit indicate battery charge and Wi-Fi status.
Clicking the Autoset button on the IkaScope software automatically adjusts gain and time-base to quickly view the signal optimally. The IkaScope WS200 also knows when to measure and when to hold the signal display without the need for a Run/Stop button. The IkaScope’s innovative Automatic History feature saves a capture of the signal when releasing pressure on the ProbeClick tip. The History Database is divided into Current Session and Favorites, where signal captures are permanently saved, even after the application is closed. Previously measured signals can quickly be recalled.
Most desktop oscilloscopes have a static reference grid with a fixed number of divisions, but the IkaScope allows pinch and zoom on touch screens (or zoom in/out with a mouse wheel), stretching the grid and allowing an operator to move and zoom through a signal capture for detailed review. The associated software even has a share button on the screen: simply click on it to share screenshot measurements.
IKALOGIC | www,ikalogic.com
Are embedded vision solutions complex? Expensive? Strictly about software? Get answers to your top questions about developing embedded vision solutions, right from Avnet & Xilinx.
But isn’t embedded vision complex? Lacking scalability? Rigid in its design capability?
Truth be told, most of those ideas are myths. From the development of the first commercially viable FPGA in the 1980s to now, the amount of progress that’s been made has revolutionized the space.
So while it can be complex to decide how you’ll enter an ever-changing embedded vision market, it’s simpler than it used to be. It’s true: Real-time object detection used to be a strictly research enterprise and image processing a solely software play. Today, though, All Programmable devices enable system architects to create embedded vision solutions in record time.
As far as flexibility goes, you’ll find something quite similar. In the past, programming happened on the software side because hardware was preformatted. But FPGAs are more customizable. They contain logic blocks, the programmable components and reconfigurable interconnects that allow the chip to be programmed which allows for more efficiency of power, temperature and design—all without the need of an additional OS.
Ready to bust some more myths around embedded vision? Watch our video breaking down the five biggest myths around embedded vision development.
Filtering pulsed signals can be a tricky prospect. Using a recent customer problem as an example, Robert highlights various alternative approaches and describes the key concepts involved. Simulation results are provided to help readers understand what’s going on.
By Robert Lacoste
Welcome back to the Darker Side. A couple of months ago, one of our customers was having trouble with its project and called us for help. As is often the case, the problem was more a misunderstanding of the underlying concepts than any kind of hardware or software issues. We helped him, but because the same issue could jeopardize your own projects I thought it would be a nice topic for this column.
What is it about? Of course, I won’t be able share the details of our customer’s project, but I will describe a close example. Let’s imagine you need to build an ultrasonic ranging system. Just as bats do, you want to transmit short bursts of ultrasound, then listen for echoes. As you probably know, the time between transmission and reception divided by twice the speed of sound will give you the distance of the obstacle.
Moreover, the shift in frequency between transmitted and received bursts will give you the relative speed of this obstacle, thanks to the so-called Doppler shift. Ok, but how will you design such a ranging device? First, you’ll need to generate and transmit bursts of sine waves—also called tone bursts—with the proper ultrasonic frequency, say 40 kHz. That’s easy to do even with a pair of trusty NE555 chips or NAND gates, or maybe with a microcontroller if you prefer dealing code rather than a soldering iron. These bursts will need to be as short as possible—maybe 1 ms or so—because this will improve the distance resolution.
The transmit side is easy, but the receiver will be a little more complex. In real life, the received signal will have a very low amplitude and probably plenty of added noise. This is especially true if you consider that the Doppler shift could be significant, meaning with fast-moving objects. In that case you will not know the exact frequency of the burst you should detect.
One possible architecture to avoid this problem, while minimizing noise, could be the one illustrated on Figure 1. First, do a spectrum analysis of the received signal. Because this signal contains noise plus the received ultrasonic echo, its frequency spectrum will show a peak at the frequency of the received ultrasonic carrier. Therefore, you can measure this actual reception frequency. Assume it is 40.5 kHz due to Doppler shift. You can use this information to tune a very selective band-pass filter, which will isolate the received ultrasonic burst from any other noise. Why not a 40.5 kHz +/-100 Hz filter? You will then recover a clean version of the received pulse and measure the time difference between transmission and reception with a detector and a time counter. Brilliant idea, isn’t it? If you agree, then please read on. This was the concept used by our customer, and unfortunately it doesn’t work! At least not as described. In this article I will explain why, using some easy to understand simulations and as little math as possible. So, don’t’ be afraid. Come with me to the Darker Side of pulsed signals.
Before going into the explanation, I need to present you an alternative version of this intended receiver. Because you are a reader of Circuit Cellar, you know that developing such a design would be far easier using digital signal processing than trying to build analog spectrum analyzers and precisely tuned filters. The digital equivalent of this receiver is illustrated on Figure 2. Just compare it with the former, you will find the same concepts.
Here the received signal is preamplified and directly digitized with a properly selected analog-to-digital converter (ADC). Its frequency spectrum can then be calculated with a Fourier Transform, using the well-known Fast Fourier Transform (FFT) algorithm, for example. The frequency peak can then be searched into this spectrum. Then a narrow band-pass filter can be created and tuned to this frequency and the filtered signal can be calculated. …
Read the full article in the August 337 issue of Circuit Cellar
Don’t miss out on upcoming issues of Circuit Cellar. Subscribe today!
Coming to your inbox tomorrow: Circuit Cellar’s Analog & Power newsletter. Tomorrow’s newsletter content zeros in on the latest developments in analog and power technologies including ADCs, DACs, DC-DC converters, AD-DC converters, power supplies, op amps, batteries and more.
Bonus: We’ve added Drawings for Free Stuff to our weekly newsletters. Make sure you’ve subscribed to the newsletter so you can participate.
Not a Circuit Cellar Newsletter subscriber?
Don’t be left out! Sign up now:
Our weekly Circuit Cellar Newsletter will switch its theme each week, so look for these in upcoming weeks:
Microcontroller Watch. (8/14) This newsletter keeps you up-to-date on latest microcontroller news. In this section, we examine the microcontrollers along with their associated tools and support products.
IoT Technology Focus. (8/21) Covers what’s happening with Internet-of-Things (IoT) technology–-from devices to gateway networks to cloud architectures. This newsletter tackles news and trends about the products and technologies needed to build IoT implementations and devices.
Embedded Boards.(8/28) The focus here is on both standard and non-standard embedded computer boards that ease prototyping efforts and let you smoothly scale up to production volumes.
Murata Power Solutions has introduced its DCM20 series of multifunction panel meters. For DC systems, these meters measure DC voltage and current, calculate power up to 96 kW, and display values either manually selected or continuously cycling. The miniature panel-mount product provides an input voltage range of 0.5 VDC to 72 VDC, with 10 mV of resolution. The meter also supports current measurement ranges from 5 A to 1,200 A when used with an external user-supplied resistive shunt. Targeted for use in 12 V, 24 V or 48 V systems, out-of-the-box accuracy of the product is +/-1 % for voltage and +/-2 % for current.
Packaged in a rugged, one-piece polycarbonate housing, with dimensions of 2.1″ x 1.43″ or 53.3 mm x 36.3 mm, the DCM20 fits in ‘0U’ and ‘1U’ racks making it well-suited for laboratory instrumentation as well as industrial and telecom equipment. Threaded mounting studs and caged terminal blocks for application wiring ensure reliable operation in harsh environments.
Applications for the product include, but are not limited to, real-time monitoring and display of DC power in telecom power distribution systems, battery management/backup systems, laboratory instrumentation and alternative energy and marine installations.
The DCM20 features a large (0.36″ /9.2 mm) bright red display easily readable at 15 feet (5 m), with green or blue displays a future option. A front-panel capacitive touch sensor is incorporated for selection of operating mode, avoiding wear-out issues possible with a membrane of other mechanical switches. Using the touch sensor control, the user may configure the unit to display voltage, current or power, or set the unit to continually cycle between the three measurements.
The unit can be self-powered from the measured voltage or powered separately from an external power supply, which can range from 9 VDC to 72 VDC. When self-powered, the input voltage range that can be measured is 9 VDC to 72 VDC and when externally powered the lowest measurable input voltage extends down to 0.5V. Current consumption of the DCM20 is generally negligible compared with the measured current being typically 6 mA at 12 V and only 2 m A at 72 V input.
A DIP switch on the DCM20 allows selection of 16 different full-scale current readings from 5 A to 1,200 A providing compatibility with a wide range of external shunt resistors, available both from Murata and other manufacturers. A fine adjustment potentiometer is also provided to calibrate the unit to compensate for shunt resistor tolerance for improved system measurement accuracy. The external shunt resistor may be placed in either the ‘high’ or the ‘low’ side of the power system, as the DCM20 has a common-mode voltage range of 72 V. A jumper is available to set where the voltage is actually measured, either remotely or at the shunt resistor. In this way, high or low side current sensing is practical and power measurement can exclude losses in wiring and the shunt resistor itself.
Murata Power Solutions | www.murata-ps.com
Vicor has added 25 new products to its family of DC-DC converter modules (DCMs) with tighter output voltage regulation of ±1%. With unrivaled power densities of 1,032 W/inches-squared, the new series allows engineers to drive loads requiring tighter regulation with minimal additional circuitry or downstream components.
The DCM ChiP (Converter housed in Package) is a DC-DC converter module that operates from an unregulated, wide range input to generate an isolated, regulated DC output. With its high frequency zero-voltage switching (ZVS) topology, the DCM converter consistently delivers high efficiency across its entire input voltage range.
The new DCMs are used broadly across defense and industrial applications that require tighter output voltage regulation. These applications include UAV, ground vehicle, radar, transportation and industrial controls. The DCM ChiPs are available in M-grade, which can perform at temperatures as low as -55°C.
Vicor | www.vicorpower.com
Maxim Integrated Products has announced a series of power-management ICs (PMICs) that enable designers to optimize power for automotive advanced driver-assistance systems (ADAS) functions to achieve high performance, small size, efficiency and electrical protection.
ADAS functions, many of which are now mandatory or will be soon, increase vehicle safety and enhance the driving experience. These features include smart braking for collision avoidance, GPS/navigation, adaptive cruise control, lane centering, lane-departure warning, and back-up/surround video. Although these functions receive considerable design attention, managing DC power in electrically harsh vehicle environments is a less-publicized yet critical challenge which involves significant issues of functions, features, performance, efficiency and footprint.
Maxim’s array of application-optimized ICs, which manage DC power, solve the top-level designer pain points for various ADAS functions involving a combination of package size, operating efficiency, quiescent current, electrical protection, and EMI generation.
The series of PMICs which Maxim has released include:
MAX20019 Dual Synchronous Step-Down Converter—Provides the industry’s smallest 3.2MHz dual step-down power supply in a 2mm × 3mm package size (compared to the closest competing solutions that offer single channel parts in either a 2mm x 2mm or 3mm x 3mm package size)
MAX20087 Quad Camera Power Protector—ASIL-grade camera module protector IC includes an I2C interface to report on over/undervoltage/fault conditions; monitors up to four 600 mA coax channels and isolates faults from individual camera modules
MAX20075 and MAX20076 Synchronous Step-Down Converters—Offer the industry’s lowest quiescent current with peak and valley mode options; provide a high peak efficiency of 91% for always-on applications compared to competing solutions, while featuring a 40 V load-dump tolerance
MAX20014 Triple-Output Converter—Features one synchronous boost and two synchronous step-down converters for smaller, simpler, and lower cost designs (competing approaches require two ICs plus discrete components); features 2.2 MHz switching frequency and spread-spectrum capability for reduced EMI and comes in a small 4 mm x 4 mm package size
Maxim Integrated | www.maximintegrated.com
Analog Devices has announced the Power by Linear LTC3372, an integrated power management solution for systems that require multiple low voltage outputs generated from an input voltage as high as 60 V. The LTC3372 features a 60 V synchronous buck switching regulator controller followed by four configurable synchronous monolithic buck regulators. This combination provides up to five high efficiency low quiescent current outputs in a single IC, well-suited for automotive, industrial and medical applications.
The LTC3372’s buck controller operates over a 4.5 V to 60 V input voltage range and drives an all N-channel MOSFET power stage. Its output can be programmed to either 3.3 V or 5 V and can generate an output current up to 20 A. The controller output is typically used to feed the four monolithic buck regulators. Each monolithic buck channel can be programmed to regulate an output voltage as low as 0.8 V with a configurable output current up to 4 A. Eight 1 A integrated power stages are programmed by the C1-C3 pins into one of eight unique configurations, from a quad 2 A buck to a dual 4 A buck. This allows only one inductor per channel.
The LTC3372 offers a low IQ solution ideal for battery-powered or automotive applications in which one or more power supply rails are always on. With just the high voltage controller enabled, the device draws 15 µA from a 12 V input supply while regulating the output to 5 V at no load. Each monolithic buck regulator adds only 8 µA of additional IQ per channel enabled. The LTC3372’s monolithic buck switching frequency can be programmed from 1 MHz to 3 MHz and can be synchronized to an external clock while the buck controller switches at 1/6 of this frequency. Additional features include foldback current limiting, soft-start, short-circuit protection and output overvoltage protection.
The LTC3372 is available from stock in a thermally enhanced 48-pin 7 mm × 7 mm QFN package. E and I grades are specified over an operating junction temperature range of –40°C to 125°C, and the H grade features operation from –40°C to 150°C.
Analog Devices | www.analog.com
TDK has announced a new series of EPCOS current-compensated ring core power line double chokes. The RoHS-compatible B82724J8*N040 series features a high rated voltage of 800 VDC or 250 VAC in continuous operation and was developed specifically for use as DC link chokes in frequency converters. A further advantage of the new series is its good thermal behavior in comparison with conventional types. That means that chokes with high inductance values can be operated with high currents, even at a high operating temperature.
The new chokes offer a range of inductance values extending from 0.5 mH to 47 mH and rated currents of between 1.6 A and 10 A, depending on the inductance. Their rated ambient temperature without derating is +70°C. For the additional attenuation of symmetrical interferences the power line chokes feature a stray inductance of around 0.5 % of the rated value. The dimensions for all types are 18.5 mm x 31.3 mm x 33.2 mm, just like the standard B82724J series. This means that the new chokes can be designed into existing layouts.
The new ring core chokes are manufactured using a flame retardant plastic that is compliant with UL 94 V-0 and certified in accordance with IEC 60335-1, Clause 30 (glow-wire and ball pressure test). In addition, the winding is completely potted, which allows its use in highly polluted environments. Typical applications include frequency converters and power supplies.
EPCOS, a TDK Group Company | www.epcos.com
Eurotech has announced that AVR, a potato harvester manufacturer based in Belgium, has chosen the ReliaGATE family of intelligent edge computers running Eurotech’s Everyware Software Framework and Everyware Cloud to manage the edge devices for its smart agriculture project to connect its harvesting machinery. These IoT building blocks are integrated by AVR partner Delaware Consulting with a Microsoft MS Azure-based IoT platform that gathers, analyzes and visualizes data from sensors on tractors and other farming vehicles.
With a showcase version up and running, AVR plans to release the platform for end users later in 2018, gathering market feedback to drive the development of new capabilities. No financial information has been disclosed. AVR has a decades-long history in the field of potato agriculture, designing and manufacturing harvesters, planters and cultivators. It’s a niche market, but they are one of the world’s biggest players, exporting equipment to every continent. However, even a traditional industry like agriculture is being impacted by emerging IoT innovations.
According to AVR, Agriculture adopts new tech relatively slowly compared to other sectors. But the key words “smart farming: and :precision agriculture” are cropping up more and more often. In the past, AVR focused much more on the mechanical side of agriculture. Now, its goal is to develop smarter machines with many more sensors and use the data its collects to bring value and transparency to stakeholders along the entire value chain.
Eurotech | www.eurotech.com
IBASE Technology has announced two SBCs, both powered by an NXP i.MX 6Dual Cortex-A9 1.0GHz high performance processor. The IBR115 2.5-inch SBC and the IBR117 3.5-inch SBC are designed for use in applications in the automation, smart building, transportation and medical markets.
IBR115 and IBR117 are highly scalable SBCs with extended operating temperature support of -40°C to 85°C and an optional heatsink. Supporting 1 GB DDR3 memory on board, the boards provide a number of interfaces for HDMI and single LVDS display interface, 4 GB eMMC, Micro SD, COM, GPIO, USB, USB-OTG, Gbit Ethernet and a M.2 Key-E interface. These embedded I/Os provide connection to peripherals such as WiFi, Bluetooth, GPS, storage, displays, and camera sensors for use in a variety of application environment while consuming low levels of power.
Both models ship with BSPs for Yocto Project 2.0 Linux and Android 6.0. They both run on dual-core, 1 GHz i.MX6 SoCs, but the IBR115 uses the DualLite while the IBR117 has a Dual with a slightly more advanced Vivante GPU.
- With NXP Cortex-A9, i.MX 6Dual-Lite (IBR115) / i.MX 6Dual (IBR117) 1GHz processor
- Supports HDMI and Dual-channel LVDS interface
- Supports 1 GB DDR3, 4 GB eMMC and Micro SD (IBR115) / SD (IBR117) socket for expansion
- Embedded I/O as COM, GPIO, USB, USB-OTG, audio and Ethernet
- 2 Key-E (2230) and Mini PCI-E w/ SIM socket (IBR117) for wireless connectivity
- OpenGL ES 2.0 for 3D BitBlt for 2D and OpenVG 1.1
- Wide-range operating temperature from -40°C to 85°C
IBASE Technology | www.ibase.com.tw
Nordic Semiconductor has launched “nRF Connect for Cloud”, a free service for Cloud-based evaluation, test, and verification of Bluetooth Low Energy (Bluetooth LE) designs employing Nordic’s nRF51 and nRF52 Series multiprotocol Bluetooth LE SoCs. nRF Connect for Cloud features an intuitive workflow and offers much of the functionality of Nordic’s “nRF Connect for Desktop” and “nRF Connect for Mobile” which are popular applications used for building and developing Bluetooth LE products. nRF Connect for Cloud also supports an extensive range of standard Bluetooth services together with proprietary services such as nRF UART.
Operating with all popular browsers, nRF Connect for Cloud uses web Bluetooth application programming interfaces (APIs) to push and extract data to and from the Cloud, enabling the developer to test and modify the behavior and performance of prototypes. By using the front-end and visualization features of nRF Connect for Cloud, historical data can be extracted from databases and analyzed in a browser. The product also allows engineers to monitor and interact with remote wireless IoT designs enabling the collaboration of geographically separate development teams on a single project.
nRF Connect for Cloud is supported by the nRF Gateway App available for iOS and Android-powered mobile devices. The nRF Gateway App enables Nordic Bluetooth LE devices to use a smartphone-enabled Internet gateway to convert Bluetooth LE messages to ReST/MQTT/IP protocols for Cloud interoperability.
The Gateway App communicates with the nRF Connect for Cloud back-end hosted on Amazon Web Services (AWS) and is based on Software as a Service (SaaS) components. By leveraging AWS industry-grade components, the app implements end-to-end data and device connectivity, guarantees reliability, and scales from a few to hundreds of Bluetooth LE devices.
nRF Connect for Cloud currently supports Bluetooth LE solutions but future versions will also support Nordic’s nRF91 Series low power, global multimode LTE-M/NB-IoT System-in-Package (SiP) for cellular IoT.
nRF Connect for Cloud works out-of-the-box with the Nordic Thingy:52 IoT Sensor Kit, Nordic nRF5 development kit (DK), and software development kit (SDK) examples. A quick-start guide is available from www.nrfcloud.com.
Nordic Semiconductor | www.nordicsemi.com
Renesas Electronics has announced an update to its Embedded Target for RH850 Multicore model-based development environment for multicore MCUs for automotive control applications. The update supports development of systems with multirate control (multiple control periods), which is now common in systems such as engine and body control systems. This model-based development environment has become practical even in software development scenarios for multicore MCUs, and can reduce the increasingly complex software development burdens especially in control system development of self-driving cars.
Renesas’ earlier RH850 multicore model-based development environment automatically allocated software to the multiple cores and although verifying performance was possible, in complex systems that included multirate control, it was necessary to implement everything manually, including the RTOS and device drivers. Now there’s ever-increasing requirements to boost engine and vehicle performance, and at the same time shorten product development time. By making this development environment support multirate control, it is possible to directly generate the multicore software code from the multirate control model. This has made it possible to evaluate the execution performance in simulation.
Not only does this allow execution performance to be estimated from the earliest stages of software development, this also makes it easy to feed back the verification results into the model itself. This enables the completeness of the system development to be improved early on in the process, and the burden of developing the ever-larger scale, and increasingly complex, software systems can be significantly reduced. Renesas is accelerating the practical utility of model-based development environments in software development for multicore processors and is leading the evolution of green electric vehicles as proposed in the Renesas autonomy concept.
Control functions development requires multirate control, such as intake/exhaust period in engine control, the period of fuel injection and ignition, and the period with which the car’s status is verified. These are all different periods. By applying the technology that generates RH850 multicore code from the Simulink control mode to multirate control, it has become possible to directly generate multicore code, even from models that include multiple periods, such as engine control.
Renesas also provides as an option for the Integrated Development Environment CS+ for the RH850, a cycle precision simulator that can measure time with a precision on par with that of actual systems. By using this option, it is possible to estimate the execution performance of a model of the multicore MCU at the early stages of software development. This can significantly reduce the software development period.
The JMAAB (Japan MBD Automotive Advisory Board), an organization that promotes model-based development for automotive control systems, recommends several control models from the JMAAB Control Modeling Guidelines. Of those, Renesas is providing in this update the Simulink® Scheduler Block, which conforms to type (alpha) which provides a scheduler layer in the upper layer. This makes it possible to follow the multirate single-task method without an OS, express the core specifications and synchronization in the Simulink model, and automatically generate multicore code for the RH850 to implement deterministic operations.
Along with advances in the degree of electronic control in today’s cars, integration is also progressing in the ECUs (electronic control units), which are comparatively small-scale systems. By supporting multirate control, making it easier to operate small-scale systems with different control periods with a multicore microcontroller, it is now possible to verify the operation of a whole ECU that integrates multiple systems.
The updated model-based development environment is planned to support Renesas’ RH850/P1H-C MCU that includes two cores by this fall, and also support for the RH850/E2x Series of MCUs that include up to six cores is in the planning. In addition, Renesas plans to deploy this development environment to the entire Renesas autonomy Platform, including the “R-Car” Family of SoCs.
Renesas is also continuing to work to further improve the efficiency of model-based software development, including model-based parallelization tools from partner companies and strengthening of related multirate control support execution performance estimation including the operating system. Moving forward, Renesas plans to apply the model-based design expertise fostered in its automotive development efforts in the continually growing RX Family in the industrial area which is seeing continued increases in both complexity and scale.
Renesas Electronics | www.renesas.com
Brainy System ICs
Long gone now are the days when FPGAs were thought of as simple programmable circuitry for interfacing and glue logic. Today, FPGAs are powerful system chips with on-chip processors, DSP functionality and high-speed connectivity.
By Jeff Child, Editor-in-Chief
Today’s FPGAs have now evolved to the point that calling them “systems-on-chips” is redundant. It’s now simply a given that the high-end lines of the major FPGA vendors have general-purpose CPU cores on them. Moreover, the flavors of signal processing functionality on today’s FPGA chips are ideally suited to the kind of system-oriented DSP functions used in high-end computing. And even better, they’ve enabled AI (Artificial Intelligence) and Machine Learning kinds of functionalities to be implemented into much smaller, embedded systems.
In fact, over the past 12 months, most of the leading FPGA vendors have been rolling out solutions specifically aimed at using FPGA technology to enable AI and machine learning in embedded systems. The two main FPGA market leaders Xilinx and Intel’s Programmable Solutions Group (formerly Altera) have certainly embraced this trend, as have many of their smaller competitors like Lattice Semiconductor and QuickLogic. Meanwhile, specialists in so-called e-FPGA technology like Archonix and Flex Logix have their own compelling twist on FPGA system computing.
Exemplifying the trend toward FPGAs facilitating AI processing, Intel’s high-performance line of FPGAs is its Stratix 10 family. According to Intel, the Stratix 10 FPGAs are capable of 10 TFLOPS, or 10 trillion floating point operations per second (Figure 1). In May Microsoft announced its Microsoft debuted its Azure Machine Learning Hardware Accelerated Models powered by Project Brainwave integrated with the Microsoft Azure Machine Learning SDK. Azure’s architecture is developed with Intel FPGAs and Intel Xeon processors.
Intel says its FPGA-powered AI is able to achieve extremely high throughput that can run ResNet-50, an industry-standard deep neural network requiring almost 8 billion calculations without batching. This is possible using FPGAs because the programmable hardware—including logic, DSP and embedded memory—enable any desired logic function to be easily programmed and optimized for area, performance or power. And because this fabric is implemented in hardware, it can be customized and can perform parallel processing. This makes it possible to achieve orders of magnitudes of performance improvements over traditional software or GPU design methodologies.
In one application example, Intel cites an effort where Canada’s National Research Council (NRC) is helping to build the next-generation Square Kilometer Array (SKA) radio telescope to be deployed in remote regions of South Africa and Australia, where viewing conditions are most ideal for astronomical research. The SKA radio telescope will be the world’s largest radio telescope that is 10,000 times faster with image resolution 50 times greater than the best radio telescopes we have today. This increased resolution and speed results in an enormous amount of image data that is generated by these telescopes, processing the equivalent of a year’s data on the Internet every few months.
NRC’s design embeds Intel Stratix 10 SX FPGAs at the Central Processing Facility located at the SKA telescope site in South Africa to perform real-time processing and analysis of collected data at the edge. High-speed analog transceivers allow signal data to be ingested in real time into the core FPGA fabric. After that, the programmable logic can be parallelized to execute any custom algorithm optimized for power efficiency, performance or both, making FPGAs the ideal choice for processing massive amounts of real-time data at the edge.
ACAP for Next Gen
For its part, Xilinx’s high-performance product line is its Virtex UltraScale+ device family (Figure 2). According to the company, these provide the highest performance and integration capabilities in a FinFET node, including the highest signal processing bandwidth at 21.2 TeraMACs of DSP compute performance. They deliver on-chip memory density with up to 500 Mb of total on-chip integrated memory, plus up to 8 GB of HBM Gen2 integrated in-package for 460 GB/s of memory bandwidth. Virtex UltraScale+ devices provide capabilities with integrated IP for PCI Express, Interlaken, 100G Ethernet with FEC and Cache Coherent Interconnect for Accelerators (CCIX).
Looking to the next phase of system performance, Xilinx in March announced its strategy toward a new FPGA product category it calls its adaptive compute acceleration platform (ACAP). Touted as going beyond the capabilities of an FPGA, an ACAP is a highly integrated multi-core heterogeneous compute platform that can be changed at the hardware level to adapt to the needs of a wide range of applications and workloads. An ACAP’s adaptability, which can be done dynamically during operation, delivers levels of performance and performance per-watt that is unmatched by CPUs or GPUs, says Xilinx… …
Read the full article in the August 337 issue of Circuit Cellar
Don’t miss out on upcoming issues of Circuit Cellar. Subscribe today!