Authored by: |
More and more products depend on electronic smarts embedded inside. The growth of embedded systems has been spurred by computation power that lets engineers work at the module level, rather than design systems from scratch. Early embedded systems had limited memory, CPU power, and I/O. Operating systems did little more than boot-load a program that performed one specific, noncomplicated task. The basic rule of thumb became, “If it needs to do something else, add another processor to handle it.”
If size and power were no object, than that was indeed the path to take. Today, however, the reality of batterypowered operation and energy efficiency has intervened.
Battery life comes at a premium in embedded-system applications found in consumer electronics, home appliances, aviation, automotive, medical electronics, and industrial automation and controls, to name just a few. Many times over it was a case of those who knew the product being charged with adding the embedded control, of which they knew very little.
Designers of these systems need a host of skill-sets related to computer hardware, embedded software, and electronics, as well as a background in the area of application. To let these individuals focus on the application details and not the electronics, embedded-system makers have crafted more-powerful systems with features that also simplified the design process. Recent devices demonstrate the power and sophistication found today in the embedded world.
The power of sight
The say a picture is worth a thousand words. But it takes millions of bits to make a picture. Processing that many bits of data takes time, so practical vision systems need a lot of processing power. It was quite common to find early vision systems housed in enclosures that dwarfed a refrigerator. Today, thanks to embedded electronics, it’s quite possible to obtain a powerful vision system that fits in your hand. Processors able to handle the torrent of data in these situations include the Blackfin line from Analog Devices, Norwood, Mass.
The Blackfin consists of both 16 and 32-bit designs that target the demands and power constraints of embedded audio, video, and communications applications. It uses what’s called Micro Signal Architecture (MSA) jointly developed with Intel Corp., Santa Clara, Calif., to better handle data streams. Its instruction set resembles those of reduced instruction set computers (Risc). This, combined with dual 16-bit multiply accumulate (MAC) signal processing registers, lets Blackfin processors perform equally well in both signal-processing and control-processing applications, eliminating the need to specify separate processors for each task.
The four latest additions to this line add dual-core capabilities at clock speeds up to 1 GHz. The ADSP-BF606 and ADSP-BF607 target general-purpose digital signal-processing (DSP) applications such as wireless communications, industrial process control, and electric power-grid monitoring/ protection. The ADSP-BF608 and ADSP-BF609 include a high-performance video analytics accelerator called the Pipelined Vision Processor (PVP) — a set of configurable processing blocks that accommodate up to five concurrent image algorithms for high-level analytics. These processors, along with the PVP, are touted for applications such as automotive advanced-driver-assistance systems (ADAS), industrial machine vision, and security/surveillance systems.
All four processors have two 500-MHz cores. Features optimized for low-power operation include a high-bandwidth switched-fabric datamovement infrastructure that lets data route more quickly for calculations or analysis. Typical power consumption is 400 mW or less. On the development side, the Blackfin processors are supported by the CrossCore Embedded Studio, a software-development tool specifically for the Blackfin processor families. Based on the Eclipse development platform, this highly visual, intuitive IDE supports both proprietary and open-source tools and technologies through pluggable modules.
This lets the system handle codes from widely used programming languages such as Ada, C, C++, Cobol, Perl, PHP, Python, R, and Ruby (including Ruby on Rails) with the proper module. Additional programming support comes in the form of the Analog Devices C/C++ compiler and the Micrium μC/ OS-IIITM operating system. Blackfin processors also include on-chip peripheral options such as timers, counters, pulse-width modulators (PWM), serial peripheral interfaces (SPI), and controller-area network (CAN) interfaces. Development accessories include high-speed in-circuit emulators, EZ-Kit development boards, and application-specific EZ-Extender cards.
The embedded PC
In the world of embedded systems, there are singlechip computers, but no such thing as a single-chip personal computer. Thus, the need for system-on-chips (SoC) and computer-on-module (COM) approach.
One COM device is the conga-TM77 from congatec inc., San Diego. Emphasizing energy efficiency, it uses third-generation Intel Core processors based on Intel’s new 22-nm Ivy Bridge technology with 3D trigate transistors. The Ivy Bridge chips are die-shrunk versions of the second- generation 32-nm Sandy Bridge designs that use dual-gate planar transistors. While fully compatible with Sandy Bridge, the Ivy Bridge chips reduce electrical power needs but also sport a 20% boost in processor power.
The COM Express module uses either the Intel Core i7-3612QE, a quad-core processor running at 2.1 GHz and total power dissipation of 35 W, or the Intel Core i7- 3615QE quad-core at 2.3ŽGHz and total dissipated power of 45ŽW. Both processors possess a 6-Mbyte L2 cache memory and are used with the Mobile Intel HM76 Express chipset. This chipset offers native USB 3.0 support and accesses up to 16ŽGbytes of dual-channel DDR3-1600 memory. It’s literally a full PC-type system on a 3.74 × 4.92-in. board.
Of course, a system like this needs a display. For graphic work, the integrated Intel HD4000 graphics core performs up to 50% better than its predecessors with the ability to control up to three independent displays. It also uses the Intel Flexible Display Interface (FDI) that runs DirectX 11 OpenGL 3.1 OpenCL 1.1. The OpenCL programming environment distributes computing tasks over a variety of processor systems involving a number of hardware units. This way multiple tasks can execute in parallel using a single step process known as Single Instruction Multiple Data (SIMD.) This approach supports classic parallel computer architecture as well to better handle analytical problems well suited to parallel processing.
A high-performance MPEG-2 hardware decoding unit decodes multiple full-HD videos in parallel. In addition, Blu-ray support is offered through the WMV9 (VC-1) and H.264 (AVC) codecs at rates up to 40 Mbps.
Eight USB ports connect a variety of peripherals with four of the ports capable of USB 3.0 Superspeed operation at 5,000 Mbps. With USB 3.0, data transfer is considerably faster, energy consumption lower, and it becomes possible to simultaneously send and receive data through the port.
The COM module offers seven PCI Express 2.0 lanes, a PCI Express graphics 3.0 (PEG) that provides 16 lanes for high-performance external graphics cards, four SATA interfaces handling up to 6 Gbytes/sec and RAID support, an EIDE and a gigabit Ethernet interface. Fan control, a low-pin-count (LPC) bus for simple connection of legacy I/O interfaces, and high-definition audio capabilities round off this embedded system’s range of features.
Suitable operating systems include Windows 7, Windows XP, Windows Embedded Standard, Embedded POS Ready (WEPOS), and Linux 3.0. A matching evaluation carrier board for the Type 6 COM Express modules is also available.
Analog front ends for the real world
In the field of energy management, smart-grid technology uses electronics embedded in smart meters to monitor energy use and, eventually, give home owners realtime feedback about how to cut energy bills. Smart meters that monitor energy consumption use analog front ends (AFEs) to convert the power measurement into digital signals for the embedded-control system, helping designers create smart meters that maximize performance while minimizing costs.
One such AFE used in energy management is the MCP3911 from Microchip Technology Inc., Chandler, Ariz. The MCP3911 features two 24-bit, delta-sigma analog- to-digital converters (ADCs) that operate at 3„V with an accuracy of 94.5-dB SINAD and 106.5-dB THD. The low noise and low-distortion reading lets the AFE provide better energy-meter and power-monitoring accuracy from start-up to maximum current then that possible with standard ADCs. Four different power modes let the chip handle low-power designs, at approximately 0.8 mA/channel, up to designs that capture higher-speed signals, like those necessary for analyzing harmonic content.
The MCP3911 can help keep down circuit costs. For example, the unit runs on 2.7 to 3.6 V for both analog and digital operation, so it can be powered from the same power rail as the microcontroller. This eliminates the need for voltage-interface circuitry. An internal, low-temperature- coefficient voltage reference, along with programmable- grid arrays (PGAs) on each channel, further enhance metering and monitor designs.
Programming without writing code
Embedded systems are indeed computers, and as such they must be programmed. However, programs written for real-time control of machinery typically need specialized development systems geared more toward control.
One such programming suite for automation technology comes from ISaGRAF, Brossard, Quebec, Canada. Introduced in 1990, it was originally designed to bridge the gap between microcomputer systems and programmable logic controllers (PLCs.) Today, ISaGRAF covers open automation, traditional automation applications, embedded control, and soft logic.
The ISaGRAF 6.1 Workbench is now based on Microsoft Visual Studio 2010 for speed, a modern user interface, and better windows docking and package management. The new workbench design is complemented by V5.3 of the C5 firmware. To keep the software agile and flexible, ISaGRAF features are supported using plug-ins. New features may be added to the environment simply by adding another plug-in.
The ISaGRAF 6.1 Workbench adds an IEC 61499 editor for creating software in terms of function blocks. Applications are built by networks of function blocks that generally provide an interface for event I/Os and data I/Os. Programmers, thus, build their automation applications by stringing function blocks together rather than writing direct code. They simply drag and drop any function block from the block library into a program as needed. Functions and function blocks can be chosen by project, device, or resource. The blocks are also grouped by scope and category, or programmers can simply search for the block name.
Also new in this edition is a Version Control System that lets multiple users work on the same elements without conflicts via a check-in/check-out process. It also lets developers manage multiple versions of a project, backup and restore entire projects or only certain elements, and compare files between different versions.
Because different types of applications need different types of interrupts (e.g., time, pulse, I/O…), ISaGRAF provides a toolkit that lets OEMs define and map their interrupts to an ISaGRAF application. It also includes a plug-in for end users to configure and program the interrupts.
Finally, the Failover mode is a backup operating mode where the functions of the primary-control system are assumed by a secondary-control system should the primary system become unavailable, either due to equipment failure or scheduled downtime. Its purpose is to make control systems more fault-tolerant. The Failover feature lets the user modify control decisions and change the conditions under which a controller gains or loses control.
Real-time motor control
A large percentage of the energy used in the U.ŸS. goes into powering electric motors. No surprise, then, that embedded systems are increasingly used in motor controls. The Piccolo line of microcontrollers from Texas Instruments Inc. (TI), Dallas, is one such embedded system designed from the ground up to control motor operation. Latest additions to that line include the TMS320F2803x and TMS320F2806x.
The new models add a C-programmable, integrated control law accelerator (CLA) coprocessor. The CLA is a 32-bit floating-point math accelerator, designed to work independently of the TMS320C28x CPU core to off-load complex, high-speed control algorithms. This off-loading frees the CPU to handle input/output and feedback loop metrics, resulting in up to a 5× boost in performance for closed-loop applications. The CLA also provides direct access to on-chip peripherals for parallel execution of algorithms to accelerate system response time and improve efficiency.
Programming the Piccolo system is handled through TI’s ControlSuite software tools. It provides libraries, examples, and support with devicespecific drivers and software. The ControlSuite software platform includes the CLA C-compiler software libraries and system examples available for motor-control applications that let developers create customized designs with typically a 15 to 20% performance improvement over prior versions.
Metalanguage tools give designers direct access to onchip peripherals for parallel execution of algorithms to improve flexibility and compatibility. The CLA also enables faster system response, high-speed control loops, and improved triggering and fault detection.
As part of their motor-control functions, the F2803x microcontrollers include pulse-width modulators that have a 150 psec resolution; a 3-MSPS, 12-bit analog-todigital converter (ADC) with 16 input channels; and two 10-MHz oscillators on chip.
A complex math and CRC unit (VCU) provides 75 tailored math instructions to accelerate processing of communications algorithms. Also included on-chip are USB 2.0 and CAN interfaces for improved communications throughput and three analog comparators with 10-bit reference, that help eliminate external parts and simplify the external circuitry.
The embedded field is growing as makers of embedded systems continue to expand their abilities. While we’re not at the point of a pure drag-and-drop embedded device, the growing set of features and interfaces could make that a reality.