Basic Intent
This
lecture will provide some general examples of accelerometers. After the general
overview, a careful attention will be given to a particular commercial product,
the ADXL50
accelerometer by Analog Devices.
This exploration will include commentary on the design, fabrication, packaging,
and intended application of this device. The student is expected to gather some
familiarity with the operation of this device, and for some of the tradeoffs
which constrained its design.
Accelerometer
examples
Fig. 1: Mass on a Cantilever
Suppose an accelerometer is made by suspending a mass at the end of a cantilever, and embedding a doped silicon strain gauge in the cantilever to measure deflection of the cantilever. A drawing of the configuration is shown in Fig. 1.
The mass is 1 mg, and it is located 1 mm from the support of the cantilever. The cantilever is 100 microns long and 10 microns thick. Suppose the resonant frequency of the cantilever with the mass is 1 kHz. Then a 1 milli-g acceleration would cause a deflection of
The Angle of deflection of the cantilever is given by the trigonometric relation:
Therefore, the radius of curvature, r, of the 100 micron length of cantilever is given by
Because of this radius of curvature, the length of the upper surface of the cantilever is greater than the length of the lower surface of the cantilever. The strain is given by:
So, the strain in the upper surface of the cantilever during the 1 milli-g acceleration is approximately 100 ppm. How much of a resistance change does this correspond to? Remember that the gauge factor for a typical silicon strain gauge is of order 100.
Therefore, we would expect a 1 ppm change in the resistance of the silicon strain gauge. This is pretty small, but can be measured with a decent resistance bridge circuit.
Fig. 2: Silicon Accelerometer
Fig. 2 shows a drawing of a silicon accelerometer with a doped silicon strain gauge of the type we've been considering. Several small companies are marketing such silicon strain gauge accelerometers with minimum resolution of order a few milli-gs, and at cost of 10-50 dollars per device. These cost levels are accessible only because the technology for making these devices lends itself to large scale parallel fabrication - meaning that the parts are made in hundreds or in thousands at a time.
A somewhat more sophisticated accelerometer technology relies on the use of capacitive displacement transducers within a micromachined silicon structure. A good example of such a device is the 7754-1000 Accelerometer by Endevco. This accelerometer is capable of detecting signals down to 0.3micro-g/sqrt(Hz) with a resonant frequency of 9 kHz. This means that the capacitive displacement transducer in this device is capable of measuring deflections as small as 10^-15 m/sqrt(Hz), which is extremely good performance for any miniature transducer. This level of performance does not come cheaply - these devices are a few thousand dollars each.
ADXL50
Accelerometer
A large development in the automotive industry has been the Air Bag safety system. Initially made available by Chrysler in the mid 1980s, the Air Bag has become a standard item in most passenger vehicles sold in the US and throughout the world.
About 10 million automobiles are sold in the US every year. Thus, this is a large enough market to drive technology development.
Early air-bag systems featured a crash sensor which consisted of a ball bearing in a tilted tube with a contact switch at the upper end. If a sufficiently large acceleration occurred, the ball bearing rolled up the tube, closed the switch, and the airbag was inflated. This rolling ball sensor was manufactured by TRW among other companies, and it was offered to the automotive industry as a $5 part. Its performance was not spectacular, but it was rugged and reliable.
Remember that the operation of such a safety system on a vehicle represents a legal liability situation of potentially catastrophic dimensions. Systematic failures of the system could lead to hundreds of deaths, which could lead to billions of dollars in legal costs. Profits of large automakers can (sometimes) be a couple of billion dollars on sales of a couple of million cars. The airbag safety system represents a small fraction of this profit (10-100 million), and so companies are unwilling to be exposed to billions of dollars in legal claims unless they can be assured that the system performance is very good.
Now, the ball-in-a-tube design is a little clumsy, but with some packaging, can be expected to perform reliably throughout an automotive lifetime which might be more than 10 years, and include exposure to extreme cold and heat for extended periods of time.
In spite of these concerns, the ball-in-a tube has not been perfect. One possible problem has been the tendency to deploy upon the encounter of a long, deep pothole. Unnecessary deployment is somewhat dangerous, because it briefly incapacitates the driver. In addition, it causes some expensive damage to the vehicle dashboard, and occasionally injures the driver. There have been numerous reports of hearing loss due to the abrupt change in pressure caused by deployment in a sealed car. Some injuries to the face (eyes in particular) have also been reported.
Recent years have seen the evolution of the air bag systems in response to the need for more accurate detection of threatening situations. For some automakers, this has led to a desire for a sensor which measures all acceleration signals, instead of just a threshold, and allow signal processing to more accurately determine the need for deployment.
The size of this market has stimulated several small sensor companies to develop and introduce micromachined piezoresistive accelerometers of the type studied in the last lecture. Accelerometers from Lucas Novasensor and EG&G IC Sensors are presently onboard several imported vehicles, and are being considered by Ford, GM and Chrysler for domestic automobiles. (The big 3 are always cautious about becoming dependent on a small company as a single source for a critical automobile component, since an earthquake, fire, or other disaster can interrupt the supply of needed parts)
In addition, Analog Devices , Freescale, and Bosch have developed accelerometers for this application, which are now nearly ubiquitous in automobiles.
Fig. 3: ADXL50 Sensor and Signal-conditioning Circuits
The effort at Analog is particularly interesting, partially because it has been well-publicized, and partially because it has been based on a large departure from the techniques and approaches of the competition.
The approach taken by Analog relies upon the commercial development of a silicon micromachining. This approach relies on the deposition and patterning of a series of polycrystalline silicon and silicon dioxide layers on the surface of a wafer. After the patterning is complete, hydrofluoric acid is used to etch away all of the oxide layers, leaving behind the polysilicon structural layers. This fabrication technology emerged from academia in the last two decades, and has been looked at for anything from miniaturized motors to medical sensors.
Fig. 4: Bulk-micromachined Accelerometer
In the sensor research group at Berkeley, (SeeMicrolab at Berkeley ) this technology was adapted to the development of supported resonant structures which can be pushed in directions parallel to the wafer surface by electrostatic forces between interdigitated comb-drives. A simple structure of this sort is shown in Fig 4. In this figure, a floating element is suspended by a long, thin beam whose opposite end is anchored to the substrate. The floating element has some fingers which are interdigitated with the fingers of a fixed element. When a voltage is applied between these elements, the generated force pulls the interdigitated elements closer together. Since the motion of the floating element is constrained by the support system, it is 1-dimensional in nature. Using this technology, the Berkeley group has made a comprehensive family of resonant devices for many applications.
These comb-drive elements can also be used as capacitive displacement transducers. As the floating element moves closer to the fixed element, the capacitance due to the overlap of the fingers increases linearly with displacement.
One very important feature of this style of structural fabrication is that the sequence of steps required to accomplish these mechanisms is very close to being compatible with the steps necessary to build integrated circuitry. Because of this compatibility, it is feasible to build a mechanical sensor element on the same substrate as the signal conditioning circuitry. For capacitance measurement, this is a very big advantage, since it minimizes the length of leads from the sensor to the circuit, and therefore reduces the errors from stray capacitances.
A bit more than a decade ago, researchers at Analog decided to make an integrated accelerometer based on this process. Success required development of a real sensor/circuit process with independently optimized electrical and mechanical characteristics. The target application for this development was the automotive air-bag accelerometer market, which was expected to require resolution down to a few milli-gs, bandwidth of 10 kHz, excellent linearity, very little sensitivity to temperature and other noise sources, and very low cost. The automotive industry placed the acceptable part cost at $5/part.
Analog, Bosch, Freescale, ST Microelectronics and others have since sold millions of accelerometers for airbag systems. These miniaturized accelerometers have found their way out of the car and into consumer devices, such as the Nintendo Wii and iPhone, as well.
Fig. 5: ADXL50 Accelerometer Sensor Element and Circuit
Some illustrations of this device are shown in Fig 5. The sensor element is a suspended slab of polysilicon with comb-drive and sense electrodes on two sides. The detection scheme works by detecting the difference between two capacitances. This difference signal is amplified, and compared with a threshold detector, which sends positive or negative pulses to the control electrodes in an effort to minimize the difference signal. By counting the difference between positive and negative pulses, it is possible to measure the acceleration of the system.
This electronic detection scheme would be complicated to build off-chip, but is easily designed within the electronics process which runs around the sensor element in the Analog process. As a result, a fully-integrated device is possible.
After all the deposition and etching steps are complete, the mass elements are released by a brief dip in HF, and the wafers are diced, mounted, and packaged in ordinary transistor headers. The details of this dicing and packaging process are also very complicated, but will not be described here...
Fig. 6: External Signal Conditioning Circuit