No self-respecting audiophile—or weekend CD warrior, for that matter—needs an introduction to the Bose name or the array of speaker systems the company produces. But those endless hours of listening pleasure are no accident: Each Bose speaker is 100% tested before it leaves the factory.
So, how does the company translate “sounds great” into a consistent, reliable testing process that doesn’t bog down the assembly line? This was the concern when Bose geared up for production of the Bose Lifestyle® 20 audio system featuring the new Jewel Cube® speakers.
The company had used VXI to test other products, but realized that the earlier implementation was not appropriate for the new line. Company officials knew VXI could do the job, but some important features of VXI were not supported by the previous vendor. Testing the Jewel Cube transducer (the actual cone/magnet/coil assembly itself) would be more complex, and the testing cycle must be much faster.
Figure 1 illustrates the architecture for the Jewel Cube VXI test system. The UUT is placed in a specially constructed acoustic environment and monitored by special microphones at many locations, the details of which represent both untold years of optimization research and a competitive advantage.
Audio waveforms from an arbitrary waveform generator (arb) drive the UUT. The waveforms—different for each subtest—may comprise up to 4 MB of 16-bit sample points, representing anything from simple sine waves to multiperiod, evolving complex waveforms.
The output of the UUT is measured by the microphones and makes its way back through an interconnection assembly to a VXI signal-conditioning module for gain and anti-alias filtering. The conditioned audio signals are then applied to an analog-to-digital converter (ADC) and the digitized data is transferred through a Slot 0 controller to a SUN Sparc5™ workstation for digital analysis. In addition, a few TTL control signals pass between the VXI hardware and the buttons, LEDs and custom electronics of the test booth.
More Challenges
Several potential difficulties must be overcome for this system configuration to meet all the testing criteria:
Some subtests must run at alternate sampling rates. For these tests, a different set of microphone channels must be automatically selected for routing to the ADC. No wiring changes are allowed from one subtest to another.
Transfer of ADC-sampled data to the computer must not interfere with the provision of stimulus sample points to the arb.
Relative timing information between the excitation waveform and the digitized signals from the microphones is important in some cases. Clocking output points at the arb must be “locked” in step with the ADC sampling.
Allowance must be made for possible expansion to include real-time digital signal processing of sampled data.
The tests must run faster than on previous production lines. The development cycle for the test stand—hardware and software—also must be much shorter.
The Solution
Selection of the channels to be automatically routed to the ADC module is handled without any wires. The KineticSystems V207 ADC and V252 Gain/Filter Modules use MUX-bus™ communication, an analog multiplexing/communication mechanism over the VXI Local Bus. The choice and order of channel selection are programmed into Scan Tables before each subtest. The analog signals are then transferred over environmentally isolated Local Bus backplane lines.
Potential data-transfer conflicts are avoided by using the buffering capabilities within the arb and the ADC. Each has enough on-board memory to avoid data conflicts over the VXIbus while the tests are running.
Of most concern was the need to “lock step” the arb output and ADC sampling from an asynchronous start—like a button push. The VXI implementation provided trigger bus lines to communicate timing. For instance, a start digital control signal was passed to the ADC over one trigger line and, in response, the ADC sample clock was passed across a second trigger line to the arb to be used as its output clock.
The use of register-based VXI instruments was a big factor in reducing both test times and program development. Direct access to registers and dual-port waveform memory meant fast data transfers and much simpler programming that shortened the development cycle.
Bose Jewel Cube testing is representative of most acquisition and control applications in that its success ultimately depends on the many detailed interactions of its component parts. The choice of platform—in this case, VXI—can only promise the potential for this success. The key is how well these modules work together as a family.
About the Author
Michael Skubic is the Product Marketing Specialist at KineticSystems. He joined the company in 1981 with a B.S.E.E. degree from the University of Indiana-Urbana. He moved to Marketing in 1995. KineticSystems, 900 N. State St., Lockport, IL 60441, (815) 838-0005.
Copyright 1996 Nelson Publishing Inc.
April 1996