Shrinking Hardware, Increasing Functions
There's no question that board- and chip-level control products are shrinking in physical size, while continuing to add embedded functions. It's part of a natural progression to smaller, portable, more functional devices and gadgets desirable in the commercial and consumer world. The trend extends into industry, as well, where original equipment manufacturers (OEMs) face continual demands to re...
There's no question that board- and chip-level control products are shrinking in physical size, while continuing to add embedded functions. It's part of a natural progression to smaller, portable, more functional devices and gadgets desirable in the commercial and consumer world. The trend extends into industry, as well, where original equipment manufacturers (OEMs) face continual demands to reduce product size and cost.
Producing tiny components and systems is still another aspect of shrinking hardware. This is where tradeoffs among design efficiency, manufacturing cost, thermal management, power consumption—and something as basic as the handling of many miniature parts—must be worked out. Key to making it all happen is the increasing level of integration in circuit boards and microprocessors.
Programmable or off-the-shelf
'The biggest factor driving the 'shrinking hardware, increasing functions' trend is availability of programmable technologies that allows high integration devices to be custom tailored to the need of particular applications,' notes Nathan John, director of strategic marketing at Cypress MicroSystems (a subsidiary of Cypress Semiconductor Corp.). Application-specific integrated circuits (ASICs) offer a way to obtain high integration, though only with sufficiently high end-product volumes. More recently, other programmable technologies have come along, allowing the average engineer to produce an integrated solution, explains John. He cites CPLDs (complex programmable logic devices) and FPGAs (field-programmable gate arrays) as examples for purely digital designs, and products such as Cypress MicroSystems' PSoC tool for mixed signal designs (see CE , April 2004, p. 39).
In Kontron America's experience, designers of embedded system products want more performance, in a smaller package that requires less power. Meanwhile, consumers continually compare embedded computing technology with that available in a desktop PC, according to Ingo Kupper, senior electronics engineer at Kontron.
Chip designers and manufacturers, such as Intel, support the trend of less hardware-more functions. Kupper mentions Intel's XScale PXA255 device as an example of recent processors designed specifically for embedded computing applications. Because of that processor's high integration (plus speeds up to 400 MHz), Kontron used it as the basis for its X-Board&PXA> CPU module to develop a very compact embedded format that delivers a full feature set, yet requires relatively low power (typically 1.5 W at 3.3 V).
In the device networking and communication sector, the main triggers for miniaturization are smaller installation spaces and complex applications, explains Charles Chen, product manager at Moxa Technologies Inc. This includes transferring acquired data from devices into Web-based data management or databases, without a front-end processor. 'In today's IT industry, smaller size means lower cost, whereas more functions always shorten the development time to market,' he says.
Responding to the market trend, Moxa recently launched its NE-4000 Series Network Enabler (NE) embedded modules for industrial serial devices. Particularly the smallest NE board in the series (cover photo) eases integration with a user's PCB or small device form-factor. 'This not only saves space, but also means that the board can be incorporated into the user's existing serial devices,' continues Chen.
Faster to market
Moreover, added embedded functions help designers integrate their legacy devices quicker with Ethernet capability. For instance, Moxa NE-4000 Series provides for easy configuration via Web browser, serial console, or Microsoft Windows utility. Its application-ready TCP/IP communication firmware reportedly allows users to make their legacy serial devices Ethernet-ready without circuit modification—speeding time to market, hence lowering R&D or production costs. 'Using a full-function embedded module instead of chip-level solution can also help designers save development time from three to six months,' remarks Chen.
MEN Micro Inc. has two diverse perspectives about the trend toward smaller, more integrated embedded solutions—new control applications continually create marketplace demand, and smaller, more functional electronics create new types of control applications. 'While marketplace demand fuels the development of more integrated electronic solutions, highly integrated solutions unleash the creativity and innovation that results in new applications,' says MEN Micro President, Ernest Godsey, P.E. 'In the end, both forces drive toward smaller, yet more functional electronics.'
In a larger sense, natural forces always push smaller hardware with greater functionality. Godsey cites the force of competition and its effects on the cost of electronics, among reasons. Smaller hardware takes less material to produce, which becomes inherently more cost-effective once volume production is reached. 'A powerful way for an electronics company to win market share from its competitors is to provide more functionality at less cost...[via] smaller hardware with greater functionality,' he remarks.
PC technology has driven the migration of microprocessors from the office into embedded industrial control applications, in the view of Robert Burckle, vice president of WinSystems. Readily available microelectronics and a vast software infrastructure supporting PCs has been leveraged into products that survive in harsh industrial environments. 'The result of this union is shrinking hardware with increasing functions, while offering faster time-to-market with less risk and reasonable cost,' he says.
'[However], inheriting technology from the PC industry is a two-edged sword,' Burckle continues. While great advances have produced extraordinary price/performance for components essential to industrial control, parts also are quickly abandoned in the quest for technology with even greater functionality and complexity. 'A critical issue is product obsolescence,' he remarks.
As a result, manufacturers and designers face difficulties in trying to project long-term availability of boards and components. 'For commercial PC technology, two years is a lifetime. Yet for industrial markets, the timeline stretches from seven to 10 years and beyond. Guaranteeing availability of parts has become difficult, if not impossible,' Burckle says.
Some, though not all, semiconductor companies offer selected products with ex- tended availability. Major CPUs and chipsets are more obtainable, but components, such as frequency synthesizers, video controllers, and network controllers 'may be difficult or impossible to buy,' according to Burckle. Designers often must resort to multiple surveys of various vendors for projections about critical parts availability. This can lead to sooner-than-expected redesigns or advance parts purchases for products to be built years in the future. 'Both options are costly,' he adds.
Along the same lines, Robert Oglesby, president of Host Engineering—a specialist in Ethernet interface cards and technology provider for AutomationDirect ( www.automationdirect.com ) —notes that Host has historically looked to microcontrollers containing the functions needed in the quest to put more functionality into smaller packages at lower prices. 'This can be very rewarding or very frustrating, depending on the availability of the right part,' he says.
Recent advances in FPGA and soft processor technology have given board-level developers the benefits of having exactly the right part, without concerns for part obsolescence, availability, or vendor dependency, explains Oglesby. FPGAs enable Host Engineering to 'build' the exact processor it needs, very quickly and at low cost. '[It] gives us the equivalent of a custom ASIC for every application. We expect the majority of our future designs to be FPGA-based.'
In the past, FPGAs were too costly for low-end embedded applications. 'However, like other chip technologies, FPGAs have followed Moore's Law and reduced pricing while increasing density,' Oglesby continues. 'In the last couple of years, the price vs. density relationship has made them viable for our products.' Host sees even more advantage ahead for FPGAs, coming from continuing semiconductor process improvements. 'The logical lower limit for price is related more to the chip package than the chip contents,' he concludes.
Todd Dobberstein, DAQ product manager at National Instruments (NI), notes that board-level designers continue to find new ways to add functionality within a smaller size while maintaining efficiency—despite issues like thermal management, reliability, and cost. '[This] trend will progress until technology takes us to a point of diminishing return. Until we find a point where efficiency preempts size, industry will continue to shrink size and increase functionality,' he says.
He gives an example of this trend in National Instruments' reconfigurable I/O (RIO) technology, where LabView graphical programming language does hardware design, while a software tool automatically designs an optimized, reconfigurable hardware circuit to implement the application in silicon. Digital loop rate for LabView can be as high as 40 MHz. NI's method takes a step beyond an off-the-shelf hardware approach popular with control system OEMs that use ASIC chips. With RIO technology, users can configure and reconfigure the FPGA for each control application in LabView. It results in 'the ultimate level of hardware customization,' allowing control system OEMs to deliver a solution with more customizable functionality in a compact size, explains Dobberstein.
NI sees reconfigurable I/O technology evolving and expanding into new and different types of devices/systems, from plug-in computer modules with integrated I/O circuitry to embedded machine vision systems.
Power, I/O structure limits ahead?
Most experts agree that still smaller and more functional embedded products will continue to evolve for some time.
While it's dangerous to forecast limits on technology innovation, embedded control system designers face some real physical limits, according to MEN Micro. As integrated circuits continue to shrink and add denser functionality, a basic problem of input/output (I/O) channels remains for control systems. 'As processors and compute modules get smaller, it becomes more and more difficult to connect to 'real world' I/O,' explains Godsey.
'Much of the problem is basically mechanical,' he says. 'Connectors for high point count control system take up space.' (In fact, some tiny boards coming on the market sport a connector as large as, if not larger than, the basic product.) Minimum overall system size is further constrained if power switching or electrical isolation is required. Godsey mentions that some developers claim the problem will disappear when all I/O channels become wireless. He believes that '[wireless] just 'spreads' I/O over a larger footprint, hiding the wiring/connector issues.'
Providing for signal access is likewise noted as a potential size limit at Kontron. 'Routing of signals and connectivity also presents a challenge as the industry moves forward with miniaturization,' says Kupper. 'Buses and pins take up space.' Kontron X-board design addresses this issue, in part, by eliminating legacy buses.
Other limits may also be on the horizon. For one, Kupper mentions the minimum power requirements of materials used at the core of today's CPUs and chips. For example, silicon requires a minimum of 0.7 V to execute functions. 'Chipmakers are beginning to work with some alternative materials that allow for even lower power requirements, but as long as we are working with electrons, we can only go just so far,' he says.
Of course, higher function density leads to more power consumption. Mitigating the associated heat generated becomes a big issue. '[Often], customers who want small board-level solutions for their embedded applications demand no moving parts and placement in fully enclosed environments,' adds Kupper. Kontron applies various thermal management solutions in its 'Computer-on-Module' products. For ETX boards, solutions include thermal pads, a heat-spreader plate, and heatsink/fan as needed. For E2Brain—a RISC-based embedded computer—a so-called 'Brain Cap' helps to dissipate heat and makes the module more rugged.
'Denser functions can have a dramatic effect on the overall system power consumption,' says Cypress MicroSystems' John. 'One of the biggest sources of power usage in integrated circuits is the I/O structure that communicates information to other devices on the board.' More system functions incorporated in fewer on-board devices, or ideally in a single system-on-chip, will lower overall system power consumption, he adds.
'Miniaturized design always burdens the main processors with full-load processing,' states Moxa Technologies' Chen. It results in higher operating temperatures, which traditional solutions attempt to solve by adding a fan or a cooling system on the control board. But, this is not acceptable in many cases. 'However, to downsize the board design, the best way of balancing thermal management is to use low power consumption processors, such as Intel 8051 or a RISC-CPU,' he says
Moxa addresses thermal management by holding power consumption to under 1.5 W in its NE-4000 Series embedded device networking boards. These include NE-4000T Serial (TTL) to 10M Ethernet drop-in module; NE-4100T Serial (TTL) to 10/100M auto-sensing Ethernet 26-pin drop-in module; and NE-4120S/A, an RS-232/422/485 to 10/100M auto-sensing Ethernet networking board. These boards ease users' concerns for overheating, reliability, and product lifetime.
Connecting human operators to the control system presents another critical I/O function. Invariably this restricts just how small the total system can be, regardless of how small the electronics shrink, notes MEN Micro's Godsey. For example, displays must be readable by the operator; also switches, buttons, touchscreens, etc., needed for operator interaction with the system, expand the hardware footprint. 'Ultimately, these mundane yet necessary factors will place a floor on the extent to which a total system can be miniaturized,' he concludes.
Integrated system modules vs. open-bus systems.
In the move to higher levels of integration, experts have noted a further control industry trend of replacing open-bus systems with highly integrated system modules, systems-on-a-chip (SoCs), and increasingly smaller printed circuit board assemblies. MEN Micro Inc. differs with this prediction of the demise of open-bus systems.
“We believe that open bus and discrete solutions will continue to have a place, but that place will be pushed to more high-end applications, where the compute bandwidth and high number of I/O channels can only be achieved with an open-bus system,” says MEN Micro President, Ernest Godsey, P.E. He believes that both SOC-based and open-bus solutions will become more powerful (and less expensive) over time.
“Over time, SoCs will become viable solutions for more existing applications,” adds Godsey. “Open bus systems won’t disappear, but they will be forced up the food chain.” They will serve newer, high-end applications that may not be viable with any of today’s technology.
Differentiating‘system component’ and standard electronic component
Robert Burckle, vice president of WinSystems , observes that the shortage of skilled engineers is forcing companies to move from proprietary in-house designs to embedded PCs and PC/104 modules used as system components. It allows more focus on core competency. He stresses that there’s a difference between standard electronic components and system components. Developers increasingly rely on the latter as key to making this transition.
Standard components are those familiar active devices, such as integrated circuits, microprocessors, memory, diodes, transistors, etc., along with “passives,” such as resistors, capacitors, and inductors, explains Burckle. “They’re the basic elements needed to mount on a circuit board for a customized, application-specific design.” In contrast, system components have active and passive parts mounted on circuit boards configured for a specific task. System components can be either single- or multi-function modules that serve as highly integrated building blocks of a system. “A system component can be as simple as a digital I/O board or as complex as a computer with video, memory, networking, and I/O points all on a single board,” he says. “The net result is that a control engineer can select a standard platform to solve problems in a non-standard world.”
Miniaturization across the board
More capability in ever smaller packages is spreading to wider embedded control components.
Analog Devices Inc. (ADI) recently introduced the first members of its nano DAC family of digital-to-analog (D/A) converters that offer very compact size and low power consumption for space-constrained applications. The smallest of the devices, AD5641, takes 70% less board area and 80% less power than comparable devices, says ADI. It delivers 14-bit resolution and 'guaranteed monotonic behavior' with maximum power consumption of 100
Size and performance result from ADI's design and packaging methods—both patented. 'Segmented string' architecture provides accuracy with significant die-size savings, while advanced chip-on-lead die assembly technology maximizes package cavity size for a minimal board footprint.
Other lower resolution devices of the family—AD5621 (12 bits), AD5611 (10 bits), and AD5601 (8 bits)—provide alternatives for applications not needing the highest performance. These devices are sampling now, with production quantities slated for December 2004.
Besides its small footprint (3.5 x 3.5 mm) and low profile (1.20-mm max. mounted height), Fairchild Semiconductor's new FODB100 Microcoupler boasts wide temperature range of -40 to 125 °C, suitable for industrial as well as consumer applications. The surface-mount device is called the first single-channel optocoupler with a transistor output housed in a ball-grid array (BGA) package.
Microcoupler is designed for voltage isolation in feedback circuits of power systems, also for signal isolation in high-voltage, electrically noisy environments, according to Fairchild. It features 2,500 V rms isolation for one second and 'high current transfer ratio at low input current levels.' FODB100 is available now with eight-week delivery ARO.