Along with the obvious contribution of saving lives, the automobile airbag has created some interesting–but manageable–challenges for testing high-volume, low-cost accelerometers. The techniques we used to test these accelerometers may be of some help the next time you have to test high-volume devices.
Functionally, the airbag accelerometer is a fairly simple device which outputs an electrical signal, usually an analog voltage, in proportion to the acceleration or force imparted. Unlike fully electronic ICs, the DUT is not static during testing.
For this reason, standard ATE cannot to test these accelerometers, and custom test equipment must be designed. A high throughput rate with thorough test coverage must be achieved to produce a cost-effective, high-quality component.
The airbag application requires a relatively tight, single-axis sensitivity specification over a wide temperature range. This, in turn, makes testing a significant challenge.
The dominant device design used for airbag accelerometers uses capacitive sensing on a silicon element or chip. Usually a device will be comprised of a sensor element and a signal-conditioning IC. These two chips are then assembled and packaged using standard IC techniques.
Because the geometries of the sensor element cannot generally be controlled precisely enough to achieve a raw device which meets specifications, a means of electronic compensation via trimming or calibration must be built into the signal-conditioning IC. In-package electronic programming techniques are typically used, since this allows the device to be vibrated, measured and trimmed at the same time.
A viable test system for fully automatic testing consists of these major components:
(o) Vibration Subsystem.
(o) Handler Subsystem.
(o) Test/Instrumentation Subsystem.
The vibration subsystem consists of one or more power amplifiers and shakers. The fixture or nest in which the DUT resides is affixed to the shaker table. A high-stability piezoelectric reference accelerometer is threaded into the fixture. The output signal from the reference accelerometer is measured simultaneously with that of the DUT during vibration tests.
Optimization of test throughput and coverage generally favors an open-loop system; that is, servo-loop controllers are not used. It is crucial–and all the more so without the averaging effects of a low-bandwidth servo loop–that the fixture be designed rigidly with no resonances inside the DUT bandwidth so that the reference signal accurately reflects the force applied to the DUT. Other error sources, such as the axis-of-orientation for both DUT and reference device, must also be minimized.
The handler subsystem must move devices in and out of the nest or test site. It also must move the devices through forced-temperature testing chambers and soak chambers. In combination with the test subsystem, it must keep track of the position and test status of the devices within the overall in-line system, and, finally, bin the devices according to final test status.
The instrument subsystem must provide signal switching, sourcing and measuring required to test the DUT. These tests include:
(o) Contact Within the Nest–Verify test-probe contact with the DUT independent from DUT faults, such as bad bond-wire connections.
(o) At-Rest Sensor–Verify the DC parameters of the DUT, power-supply current and output voltage, along with output noise. The output noise serves as the 0 g level output of the device.
(o) Vibration Sensor–Verify the dynamic performance of the DUT by measuring sensitivity vs frequency and total harmonic distortion (THD). All of these tests are performed while the DUT is being vibrated.
(o) IC Functional–Device-specific and pin parametric measurements.
The number of terminals for an airbag accelerometer is small (about 10 or fewer). This minimizes the cost of packaging and final assembly within the airbag system.
Generally, the test site is designed with two contacts per DUT terminal. This allows testing of the contact, independent of internal DUT faults such as open bond-wire connections. In addition, Kelvin (remote) sensing can be performed on any terminal.
Functional sensor/IC testing generally requires the major instruments shown in Figure 1:
(o) Digitally synthesized waveform generator–drives power amp.
(o) Waveform digitizer(s)–measures DUT/reference output.
(o) Programmable voltage supplies–supplies DUT power and other inputs.
(o) DC voltage meter–measures at-rest parameters.
(o) Reed-relay switch(es)/multiplexer(s)–connects instrument resources.
(o) Digital I/O–tests digital ports.
(o) Charge amp(s)–converts piezoelectric output to voltage.
These functions can be packaged quite efficiently using primarily VXI instrument cards. The VXI backplane trigger bus is useful for various vibration-measurement synchronization functions and software/CPU-triggered measurements. A C-sized VXI cage provides an adequate platform for most of the instrument requirements within a system.
A typical system may include between 400 and 1,000 signals/conductors cabled to VXI instruments. The VXI data bus provides a high-speed link between the system CPU and all VXI instruments with one inexpensive interface.
Custom front-/back-end signal-processing functions are required to interface the DUT to the off-the-shelf instrumentation and CPU. This active circuitry resides in the instrument rack, leaving only passive components in the nest. With the addition of parametric test/measurement software, the overall instrument subsystem can be used for both manufacturing test and product engineering; that is, characterization.
The waveform generator and digitizers, together with the vibration-measurement software, form a subsystem with the capabilities of an FFT analyzer. Unlike a benchtop instrument, this subsystem directly measures DUT sensitivity (V/g) as opposed to voltage by incorporating the information in the calibrated-reference signal. As with the FFT analyzer, it is possible to program various parameters such as resolution bandwidth or full-scale range, and measure both amplitude and phase.
The frequency selectivity of the FFT measurements is crucial to accuracy, since both the DUT and the vibrator will exhibit a few percent THD at high levels of acceleration. To correctly measure DUT nonlinearity vs signal level, harmonic distortion must be rejected. Similarly, to measure DUT THD, the harmonics must be separately detected.
To reject the effect of harmonic distortion in the vibration signal, the measured DUT sensitivity and reference signal can be used to calculate and subtract the input distortion reflected to the DUT output. Alternatively, if the vibrator is stable enough in response, the signal produced by the waveform generator can be precalibrated to cancel out the distortion in the power amp and shaker, resulting in a clean vibration signal.
The recent introduction of side-impact airbags has led to a requirement for accelerometers which must be tested at levels up to several hundred g’s. Sustained vibration at these levels is not practical, so testing is typically performed using shock pulses.
Shock can be applied using a pneumatic shock machine. Mechanical filtering techniques are used to achieve the desired pulse shape. A low-sensitivity device migh…
March 1995