By Eric Chapman
Principal R&D Engineer
Edited by Kenneth Korane
It's no secret that design engineering is a matter of tradeoffs. For instance, when marketing wants a new product that offers better performance in a smaller package that costs less, engineering typically tells them to pick two out of three.
But, that was before the advent of FPGA (field-programmable gate arrays) and SoC (system-on-a-chip) technologies. Now, with these rapidly evolving silicon technologies, designers can sometimes have their cake and eat it too.
Parker Hannifin, for instance, is a motion and control company with the majority of its engineering staff having roots in mechanical engineering. Most of our engineers are aware of the revolution in electronics by virtue of having their desktop computers upgraded on a regular basis due to improvements in the speed, size, and cost of technology. But most of us don't have a real feel for the magnitude of the change over the past 10 years, or how it has affected the industrial and aerospace communities.
To illustrate the change, consider a Parker electronic design from 1992 still used today that has twice been adapted to newer electronic technologies. The circuit provides timing signals for operating a charge coupled device (CCD). Found in digital cameras, scanners, and fax machines, a CCD is the electronic sensor array that acquires light.
One application used a CCD to read a section of a bar-codelike printed-metal strip and determine absolute linear position of the reader relative to the bar code. The original circuit used 21 VLSI (very-large-scale integration) chips, 35 capacitors, 7 resistors, a transistor, and 3 diodes. Just to wire the CCD components together required a 20-in.2 board. It worked well, but line effects on the conductors between pins of the chips limited transmission speed.
The circuit was for an optical position sensor on the trailing-edge flap actuator for a NASA Systems Research F-18 aircraft tested at California's Edwards Air Force base. For this application, the previous design's 20 in.2 of components were packed into one EPLD (erasable programmable-logic device) chip. The circuit had all the resources of the EPLD in an area only about 0.25 in.2 By integrating components into a single chip, circuit speed increased and EMI (electromagnetic interference) significantly declined. This seemed to be the ultimate design — until FPGA technology came along.
FPGA is similar to EPLD except that it packs a much higher density of logic gates. An FPGA can be described as a "sea of logic gates." With a variety of development tools — such as VHDL (very-high-density logic) compiler, simulator, place, and router — designers can connect the gates to form most any logic device. Designers can also purchase premade, modular units to plug into their own designs.
Microprocessors, UARTs (universal asynchronous receiver-transmitters), counters, timers, motor commutators, CAN-bus modules, or any other digital component can all be packed into a single device. Now, in our most current application, the CCD controller circuit has become a small modular portion of what is nearly a system on a chip. It uses only about 1⁄20th of the resources of an FPGA, in this case an Altera 10K10. The full chip has a 0.62-in.2 area, so the controller circuit — which once required 20 in.2 — effectively uses 0.03 in.2 The cost of the original circuit was approximately $11.00, whereas the latest circuit, 1⁄20th of a $15.00 chip, costs 75 cents. The balance of the Altera FPGA provides multiple functions previously performed by discrete components, such as RAM, timing, and UART communications, again greatly reducing component count and board space.
Even greater improvements are in the offing, thanks to a new technology called System-on-a-Chip (SoC). For complete industrial and aerospace systems, FPGA technology lacks the ability to integrate analog, light, and micromachine circuits. But SoC is often referred to as "mixed-signal" because it incorporates analog and digital circuits in the same silicon die area. This opens the possibility of a computer, analog interfaces, radio-frequency circuits, accelerometers, and ultrasonic arrays all on one chip. Optoisolators, LEDs, or lasers can be embedded into the same one-chip package. In fact, much of this is already being done.
Producing proprietary SoC devices requires large volumes and can be fairly pricey, but it is becoming possible for companies to form partnerships with thirdparty start-up companies and share in their electronic research and development successes.
Xemics (http://www.xemics.ch) is a Swiss company producing SoC, low-cost, low-power, ultrasmall, onechip Risc (reduced instruction-set computer) systems with analog data-acquisition circuits on-board.
Microtune (http://www.microtune.com) has a Bluetooth 2.4-GHz radio, with complete software stack, 8051 IP Processor core for user applications along with digital parallel and serial I/O.
For microelectromechanical systems (MEMS) solutions, companies like Sensant (http://www.sensant.com) produce arrays of ultrasonic transducers integrated with common silicon circuits. Another outfit, MotionSense (http://www.motionsense.com) integrates accelerometers, signal conditioning, memory, clocks, and a processor together on a single die.
And then, of course, there is Altera (http://www.altera.com) a major supplier of FPGA components and tools. They recently introduced the Excalibur development system for integrating SoC technology.
Light-emitting components may not be on board yet — but they're coming. Motorola has announced a materials breakthrough (http://www.infoworld.com/articles/hn/ xml/01/09/04/010904hnmotorola.xml) that will facilitate fast, cheap GaAs/silicon components to be developed, and allow the integration of silicon chips into laser or fiber-optic applications.
These days, electronic designs are getting smaller, better, and cheaper without sacrificing any parameters. Moreover these improvements are tending towards exponential. With FPGA and SoC technologies, engineers can have it all.