Behind the Improvement of Automobile "IQ"--Application and Challenges of Sensors

Sensors for measuring quantities such as pressure, temperature and acceleration have been a staple of automotive electronics for many years . But the vast array of systems that require such functionality, such as fuel injection control, fuel economy and safety systems (including "smart" air bags and tire pressure monitoring), have moved beyond simply placing a sensor where it is needed and sending the sensing signal back to a control unit. The advent of various digital bus systems has made it easy to centralize processing, simplify wiring harnesses, and take advantage of the processing intelligence built into the sensors responsible for data acquisition. But such an architecture also brings reliability concerns; in addition, typical automotive applications face challenges in the continued drive to reduce costs. High-speed buses such as the Controller Area Network (CAN) and the more powerful FlexRay bus have traditionally been used for computationally intensive, fast processing applications such as engine and chassis control. The low-cost, single-wire Local Interconnect Network (LIN) was developed for body electronics applications such as seat positioning and temperature control that are not speed-critical but focus on simplicity and low cost. Single-wire LIN also means lighter weight, which can lead to better fuel economy. Matthias Poppel, global advanced embedded control marketing manager at Texas Instruments (TI), said that being able to place the control IC and sensor on the mechanical part being monitored can save space and simplify the processing work of the system's central processor. But he went on to say that the reliability of such mechatronic sensors that integrate mechanics and electronics is an issue. "It is also not flexible enough (for design engineers) because there may be only one supplier for such integrated sensors, while there are generally several suppliers for components combined from discrete devices," he added. In Poppel's view, "the transition to 32-bit MCUs and satellite processor/sensor structures connected to mechanical parts has yet to be tested." For example, he said that a PCB equipped with a mechatronic sensor attached to a measuring device must be tested and verified to ensure stability and reliability under any expected motion, load, temperature and vibration conditions. Steve Henry, senior marketing manager for programmable controller power at Freescale Semiconductor's sensor business unit, also mentioned concerns about sensor reliability and the impact of packaging on it. "For example, the challenge with sensor-accelerometers is that users want smaller packages for the integrated sensor and controller IC." Henry said Freescale offers a 6mm×6mm QFN surface-mount microelectromechanical system (MEMS) device, but the dies must be stacked to meet the smaller overall package requirements so that they can be mounted in increasingly compact spaces. When stacking accelerometers (which can monitor mass and resonance) on processor chips, you need to pay attention to the impact of stress sensitivity from external environments such as load and vibration, Henry said. "Encapsulating the accelerometer with silicone, RTV or other materials will insulate it from the package," he pointed out. But then you need to develop a different die bonding technology because "it may try to be connected to a pillow using wire bonding technology," Henry said. In addition, this will also affect reliability. "Because you can't put all the functionality on one piece of silicon, you need to stack the die to optimize the process," said Mark Shaw, marketing, applications and systems manager for Freescale's sensor products. You can take advantage of the high chip logic density and high voltage resistance (of the sensor processor) without being limited by the MEMS process, he said. Henry, on the other hand, believes that sensors should be viewed from a packaging perspective. The conclusion is: not put it on one piece of silicon, but put it on two chips, the processor and the sensor. Shaw pointed out that the larger MCU chip itself has a larger die size. "MEMS has a higher failure rate and a lower yield," he said, noting that putting the processor and MEMS sensor on the same chip will pay a greater price. If the MEMS in the sensor area fails, even the best processor part will be scrapped. While agreeing that sensor yields must be improved, Frank Cooper, president of mixed-signal chip supplier ZMD America, sees advantages in the simplified packaging provided by moving to what he calls a "single-die solution." This approach is better than wire-bonding the ASIC to the sensor and connectors to wire-bond them together and then packaging them together. “A true single-chip solution places the G sensor (accelerometer), temperature sensor or flow sensor on the same chip as the signal processing, which minimizes wire bonds,” Cooper said. This leaves fewer places in the sensor installation for popping, shorting, fatigue and contamination.

Safety and engine efficiency will continue to be the focus of current and future automotive sensor applications

When the tires touch the road, reliability is also an issue because sensors are exposed to harsh environments they have never encountered before. A representative application is tire pressure sensing (TPS). The U.S. National Highway Traffic Safety Administration has mandated that 20% of new cars be equipped with TPS sensors for the 2006 model year. John McGowan, director of sensing and control at Infineon Technologies, said TPS sensors are used in "tight, hot places" and must be rugged and have a long life, as well as reasonable cost. Infineon engineers developed such a sensor by placing a CMOS ASIC for data processing and signal modulation and a piezoelectric pressure measurement element on a common lead frame. The "three-layer silicon sandwich" structure that sandwiches the ASIC between two layers of glass is rugged, McGowan said. Freescale's Henry also touched on the issue of "media compatibility," where TPS sensors can be exposed to "interesting chemicals" and fluids that can splash onto tires in the garage, including acid from battery spills, assembly lubricants, dust, chemical residues from the manufacturing process and humid air from an inflated tire. Infineon's McGowan said integrating processing with the sensor ensures accuracy for functions such as temperature compensation, self-calibration and failure mode detection. Cost control can be achieved by integrating multiple functions and features on a single chip (as opposed to discrete passive devices in the past) and mass production. Finally, this smart auxiliary sensor allows smaller central processing units to be freed from data crunching for faster decision processing. Current tire pressure monitoring sensors are either mounted as a raised piece on the outside of the tire or fixed to the inside of the wheel rim. Because these devices are powered by button batteries, McGowan said that first-tier suppliers are looking for a battery life of 10 years. "To do this, we use vehicle information in our processing algorithms and reduce the sampling and transmission rates when the car is stationary," he added. Future pressure monitoring may be done by sensors embedded directly in the tire structure. These sensors must be powered by a technique McGowan calls "energy scavenging," which uses the flexing of the tire to drive a strain device (piezo) that provides energy to the sensor. The concept can be extended, for example, to use engine vibration as operating energy for crash sensors. Another approach is to drive the embedded tire pressure sensor by induction from the outside of the tire. Issues to be concerned about here include the effect of any metal antenna ring in the tire hub on the physical properties of the tire.

The latest dual-chip accelerometer in a QFN package has a die stack size of 6mm × 6mm × 1.45mm (right).
The accelerometer core is on the top of the die, and the CMOS control IC is on the bottom. (Top) The QFN sensor stacking process is also shown

A group led by Magneti Marelli has done early work on a "smart tire" that goes beyond simple pressure sensing. The group's work was presented at the SAE 2005 World Congress by project leader Andrea Neponte and strategic innovation manager Piero De La Pierre (paper 2005-01-1481). The tire will test more than just pressure; a three-axis accelerometer on the inner pad will also provide tire force data along three axes, the size of the tire's contact patch, and road conditions (via vibration data). Although

the test used a battery to ensure a reliable communication link, the power level required by the sensor (300 milliwatts) is such that the team believes that the strain gauge approach will not provide enough power for this application. Such a tire data system could be used to provide information to the vehicle's chassis control system or determine whether tire or suspension system service is needed.

Automotive Sensor Outlook

Over the next five years, other applications for sensors will likely include more gyroscope-based devices, Freescale's Shaw said. These devices will provide angular rate data for roll stability control and other axis closed-loop control. The gyroscopes will be based on MEMS, whose processing costs will come down as volumes increase.

Peter Knittl, marketing manager for pressure and Hall-effect sensors at Infineon Technologies, believes that enhanced performance for crash sensors triggered by airbags will require pressure-based devices rather than the G-sensors currently used. "This move to 'active' sensors is driven by new (U.S.) government regulations (FMVSS-201) for side impact protection," he said. "G-sensors will trigger when the structure deforms. But pressure sensors in the door will detect a sound wave sooner (about 5 to 6 milliseconds), while G-sensors have a detection time of 10 milliseconds." Future airbag systems may use both sensors at the same time for increased redundancy.

Trends in automotive sensor systems are not only a bellwether for where sensors will be used, but also reflect "how various bus systems must work together and which application areas each bus is dedicated to," said TI's Poppel.

Two-wire or three-wire?

ZMD's Cooper said he was surprised that the single-wire LIN bus has not yet taken a greater lead in cars. Typical applications still use three-wire radiometric sensor interfaces. Perhaps with the advent of new digital protocols, no one in the automotive industry wants to risk recalls or bloodshed. So, proven legacy devices are handy for developers who want the convenience of using stock parts, he continued. “While the industry may be moving toward a digital interface, they are still looking at analog (three-wire) output signals.”

Cooper foresees smaller, lighter sensors with more processing power in the next five years, but with fewer electronic control unit (ECU) modules to handle them. In addition, he believes that the improved fuel economy and lower emissions from fewer wires (less weight) and the elimination of another layer of data interpolation will help drive the shift to a digital interface standard infrastructure.

"Some cars today have more than 100 ECUs, and the goal is to reduce the number of ECUs in the car," said TI's Poppel. "Code reusability will also help reduce costs. In addition, while you can see today's leading-edge technology in handheld wireless applications, these products do not have a long life cycle (due to their short design cycles). On the other hand, cars are long-term systems that require high quality and reliability from the electronics."