Sensors to measure such quantities as pressure, temperature and acceleration have been a staple of automotive electronics for many years. But the myriad systems required for such functions as emissions control, fuel economy and safety (including "smart" airbag systems and tire pressure monitoring) go far beyond sticking a simple transducer where needed and wiring it back to a control box.

The advent of various digital bus systems allows for centralized processing and simplified wiring, and makes it possible to take advantage of processing intelligence that's packaged with the sensor right where the data is acquired. But such an architecture brings with it concerns about reliability and the classic automotive gremlin of ever-lower cost requirements.

A high-data-rate bus, such as the widely used controller-area network and the more powerful, deterministic FlexRay bus, are traditionally used for such applications as engine and chassis control where intensive, fast processing is required. The lower-cost, single-wire local-interconnect network (LIN) was developed for such body electronics applications as seat positioning and climate control, where speed is not of the essence, so simplicity and low cost dominate. Single-wire LINs also mean less weight for greater fuel economy.

Being able to put a control IC along with a sensor on the mechanical part being monitored saves space and processing duty back on a system's central processor, said Matthias Poppel, worldwide advanced embedded-control marketing manager for Texas Instruments Inc. (Dallas). But he added that the reliability of such a combined mechanical and electronic sensor is a concern. "There is also less flexibility [for a design engineer], in that there may be only one supplier for such integrated sensors rather than several for each discrete component of a device," he added.

In Poppel's view, "The move to 32-bit MCUs and the construction of satellite processors/sensors attached to mechanical parts will have to be proven." As an example, he noted that a pc board with a mechatronic sensor attached to a component where the measurement takes place has to be tested and proven for any expected movement, loads, temperature and vibration, in addition to the integrity of the attachment mechanism itself.

Also touching on sensor reliability concerns and the influence of packaging on them was Steve Henry, power market manager, programmable controllers, for Freescale Semiconductor Inc.'s Sensor Operations (Tempe, Ariz.). "The challenge with sensors-accelerometers, for instance-is that customers want smaller packages with a sensor and control IC."

Henry said that Freescale offers such a microelectromechanical-system (MEMS) device in a 6 x 6-mm QFN surface-mount package, but the die has to be stacked to meet the footprint requirements for a smaller overall package that can be mounted in tighter and tighter places.

Stacking the accelerometer-which brings concerns about mass and resonance-onto a processor chip demands attention to stress sensitivity from external influences (loads and vibrations), said Henry. "Coating the accelerometer with a silicon gel, RTV or other material can isolate it from the package," he noted. But then it becomes necessary to develop different die-attachment techniques because "it is like trying to wire-bond to a pillow," Henry said. And that affects reliability.

"Stacking dice to optimize the process is needed, because you can't integrate all functions on one piece of silicon," said Mark Shaw, manager for marketing, applications and systems for Freescale sensor products. You can take advantage of high chip-logic density and high standoff voltage (for the sensor processor) and not be limited by MEMS processes, he said.

Henry said the sensor should be looked at from a packaging point of view. The result is that instead of putting it on one piece of silicon, it is put on two chip elements-a processor and a sensor.

Shaw pointed out that a big microcontroller chip is a large die in itself. "Because MEMS have a higher defect rate with lower yield," he noted, having both processor and MEMS sensor on the same chip would result in greater expense. Good processor sectors would end up being rejected due to MEMS defects in the sensor portion.

While also acknowledging that sensor yield has to come up, Frank Cooper, president of mixed-signal silicon supplier ZMD America, nevertheless sees an advantage in migrating to the simplified packaging offered by what he terms a "single-silicon solution." This methodology goes beyond having to wire-bond an ASIC chip to a sensor and a connector, and then overmolding a package.

"The true single-silicon solution puts the G-sensor [accelerometer], temperature sensor or flow sensor on the same chip with the signal processing for the fewest wire bonds," said Cooper. Thus, there are fewer sites for cracking, shorting, fatigue and contamination of the fabricated sensor assembly.

Rubber meets road
Reliability is also an issue, because sensors are being subjected to harsh environments they never previously encountered. One such prominent application is tire pressure sensing (TPS). The National Highway Traffic Safety Administration mandates that TPS sensors be available for 20 percent of the new-car fleets in the 2006 model year.

John McGowan, head of Infineon Technologies' Sense and Control Group, said that sensors for TPS are in a "tight, hot place" and have to be rugged and long-lived-but still come in at a reasonable cost. Company engineers developed such a sensor by mounting a CMOS ASIC, for data processing and signal conditioning, and a piezo pressure-measuring element on a common lead frame. A "triple-stack silicon sandwich," consisting of the ASIC between two layers of glass, provides robustness, McGowan said.

Freescale's Henry also cited the "media compatibility" issues to which TPS sensors can be exposed-"interesting chemicals" and fluids can splash on a tire in a garage, including battery acid, mounting lubricant, dust, chemical residues from the manufacturing process as well as moist air inside an inflated tire.

Infineon's McGowan said that placing processing at the sensor ensures accuracy from functions that include temperature compensation, self-calibration and detection of failure modes. Controlling cost comes from the integration of functions and features on the single chip (as opposed to the discrete passive components used in the past) as well as volume production. Finally, such intelligent satellite sensors allow for a smaller central processor that can be freed from number crunching to allow faster decision processing.

Future sensors will be packaged for embedding in the tire structure.

Current tire pressure monitoring sensors are either inflation stem mounted or strapped to the interior of the wheel rim. Because these devices are powered by coin-cell batteries, McGowan said, tier-one suppliers are pushing for a 10-year battery life. "To achieve that goal, we use vehicle information within the processing algorithm to reduce the sample and transmission rates if the car is not moving," he added.

Future pressure monitoring may be done with sensors embedded directly in the tire structure. These would have to be powered by what McGowan termed "energy scavenging," using the tire flexing to drive a piezo for sensor energy. The concept could be extended, say, to using engine vibrations to power knock sensors. Alternatively, embedded tire pressure sensors could be powered inductively from outside the tire. Concerns here include the effects of any metal antenna loops in the tire walls on the physical characteristics of the tire.

Preliminary work on an "intelligent tire" done by a group headed by Magneti Marelli goes beyond merely measuring pressure. The group's results were reported at the SAE 2005 World Congress last month (paper 2005-01-1481) by project leader Andrea Neponte and strategic innovation manager Piero De La Pierre. The tire tested would not only measure pressure; a triaxial accelerometer on the inner liner would also provide data on tire forces along three axes, tire contact patch size and road surface conditions (via vibration data).

Although the tests used a battery to ensure a communication link, at the sensor power levels needed (300 milliwatts) the team believed the piezo method would not have enough power for the application. Such a tire data system could be used to provide information for a vehicle chassis control system or determine if tire or suspension service was required.

Down the road
Additional applications of sensors within, say, the next five years will probably include more gyro-based components, according to Shaw at Freescale. These will supply angle-rate data for roll stability control and other axes' closed-loop control. These gyros will be based on MEMS, whose machining costs should come down with increased production volume.

Tire-pressure-sensing ICs will go into 20 percent of 2006-model cars.

Peter Knittl, marketing manager for pressure and Hall-effect sensors at Infineon, sees added performance for airbag-triggering impact sensors by going to pressure-based devices rather than the current G-sensors. "This changeover to 'active' sensors is being driven by new government mandates [FMVSS-201] for side impact," he said. "A G-sensor triggers after the structure is deforming. But a pressure sensor in a door will detect a sound wave sooner-in roughly 5 to 6 milliseconds, compared with up to 10 ms for a G-sensor." A future airbag system could use both types of sensors to provide redundancy.

Automotive sensor systems trends revolve not just around where sensors might be used but "how the various bus systems will have to work together and which domains each will respectively conquer," said TI's Poppel.

Two or three wires?
ZMD's Cooper said he is surprised the single-wire LIN bus is not yet "more dominant in cars. Typical applications still employ three-wire radiometric sensor interfaces. Perhaps with the advent of new digital protocols emerging, nobody in automotive wants [to risk] a recall or to be on the bleeding edge." Thus, developers are comfortable with proven legacy components and want to be able to dip into a readily stocked parts bin, he added. "While the industry direction may be to a digital interface, they are still looking for an analog [three-wire] output signal."

In five years, Cooper expects smaller, lighter, more process-capable sensors, but fewer electrical control unit (ECU) modules overseeing them. And, he said, the drive toward digital-interface standard infrastructure will be aided by its reduced wiring (lower weight) for greater fuel economy, lower emissions and elimination of another layer of data interpolation.

"With more than 100 ECUs in some cars today, the goal is to reduce the number of boxes in the car," said TI's Poppel. "Code reusability will also contribute to lowering costs. And while the leading edge of technology today can be found in handheld wireless applications, [because of their short design cycles] these products aren't around long enough to wear out. Automobiles, on the other hand, are longer-term systems that [need electronics] to deliver quality and reliability for years."

Rick DeMeis (jldmld @comcast.net), editor of CMP's AutomotiveDesignLine.com and a veteran journalist covering automotive engineering http://www.eet.com