Automobile, truck, tractor, jet ski, and lawn mower manufacturers face tighter environmental regulations while price pressures put harsh cost demands on makers of jets, subway cars, and trains in the current tough economic environment. These market drivers show no signs of abating and, indeed, promise to grow even tougher in the coming decades.
These requirements for more-for-less and faster put a significant burden on test and measurement engineers to provide high quality field-test data to their design engineering teams. In-field testing is used primarily to validate the models created by CAD/CAM programs and fine-tune the calculations.
In general, these programs demand more data points collected at higher resolution to validate and improve their models. As a result, test engineers must acquire data with higher channel counts, sensitivity, and sampling rates, often from tests conducted over longer durations.
Instead of testing 20 channels of sensor data collected at 500 Hz, test requirements with hundreds of channels collected at
10 kHz no longer are unusual. Indeed, for some applications such as wind tunnel or military survivability, test requirements can easily run into thousands of channels. Having more channels available also allows test engineers to provide the required data in a shorter time with fewer test runs.
Hardware Considerations
To meet the demands faced by test engineers, modern in-vehicle data acquisition systems now offer several features:
Tough Design to Handle Shock, Vibration, Water, and Dust
Data acquisition systems must be small and rugged enough to survive 120° conditions in the Arizona desert as well as -40° in Sweden. It’s impossible to bolt a PC-based recorder to the back of a bulldozer and have it survive.
Low Power Draw With Intelligent Grounding and Wiring Schemes
The weight of extra batteries required to run some tests can exceed the gross vehicle weight. Older data acquisition systems often draw hundreds of watts in contrast to newer battery-powered systems that use circuitry borrowed from laptop PCs, PDAs, and cell-phone technology. Intelligent wiring and grounding schemes offer practical solutions to everyday problems such as chasing ground loops and tracing broken lead wires.
Signal Conditioning for Dozens or Hundreds of Mixed Channels
A wide variety of gages and sensors converts physical measurements to electrical signals. Strain gages, accelerometers, load cells, and thermocouples are among the most commonly used.
Sampling Rates up to 100 kHz per Channel
While most physical vibrations in vehicles occur at relatively low frequencies, measurement of acoustic noise, vibration, and harshness requires higher data sampling rates. Many tests are run at 2N sample rates (1,024 or 2,048 Hz) for easy FFT analysis and playback to simulators and shaker controllers. Most data acquisition systems support multiple sampling rates to measure strain gages at 20 kHz together with thermocouples at 10 Hz to minimize file size.
Synchronous Sampling Across Multiple Recorders
Samples must be time aligned to determine cause-and-effect relationships, which requires independent digitizers per channel. Precise phase alignment is not trivial and can be tricky to achieve, particularly for multi-unit acquisition networks, such as multiple data acquisition systems tied together with common clock and trigger signals to collect data in parallel. Differences in filtering also contribute to large discrepancies in recorded data (see sidebar).
Distributed Signal Conditioning
Stringing dozens of sensitive signal cables throughout a vehicle is inconvenient and time-consuming, leading to short circuits, crimped wires, broken leads, and noise pickups. Using multiple distributed digitizer units allows engineers to run test leads a short distance to a local module that digitizes the signals and passes them serially along a single cable back to the main acquisition unit.
Mixture of Analog, Digital, and Serial Bus-Based Inputs
In addition to analog sensor outputs, test engineers acquire digital signals for frequency/counter inputs and switch closures. Use of GPS parameters for position, altitude, and velocity has become widespread. Additionally, monitoring parameters from vehicle controllers provides key test data without setting up extra sensors.
In the automotive world, use of the CAN bus is commonplace; military applications often acquire parameters using the 1553 bus. Many other serial buses are used worldwide with widely varying transmission protocols.
Data Storage to PC-Compatible Devices
When collecting higher channel counts at higher sampling rates, stored data files can easily grow to hundreds of megabytes or gigabytes. While many data acquisition systems use internal PC hard drives to collect data, they are susceptible to shock and vibration and can be power hogs.
Although hard drives can hold much more data than other memory devices, the small size and low power of Compact Flash or PC Cards are attractive for in-vehicle applications. Today, 2-GB Compact Flash cards that handle 1,000 g’s shock are available, and larger sizes are on the way.
Software Considerations
Hardware specifications top the list of requirements when selecting a field-based recorder. All test engineers would agree that data integrity is the most important feature in an in-field data acquisition system. Having said that, the compromises in software that field-based engineers historically have been forced to make for solid, reliable recorder systems are enormous compared to recorders designed for laboratory-based applications.
Engineers no longer have the luxury of time to learn the ins and outs of poorly designed software. Those who only use a recorder occasionally cannot go through a long, slow learning curve before starting the test. The constraints of time and pressure to acquire data quickly are forcing equipment manufacturers to upgrade their software to modern standards.
The ever-increasing software requirements for an in-field acquisition system include the following characteristics:
Ease of Use
Users of all levels of training and experience, from design engineers through test technicians, may be involved in a test. Balancing the demands to provide hundreds of settings and features with simple, one-button operation puts a burden on the software developers.
Features such as auto-identification of channels, numerical meter displays, trigger and alarm warnings, and flexible trace displays are essential to performing a test. A common expectation is support for a sensor database, where engineers can load calibration values from a standard spreadsheet or smart sensors containing transducer electronic data sheets.
Experienced engineers typically develop sophisticated test settings off-line prior to the test. At run time, operators open a test file, double check for expected sensor values, then press the go button.
Diagnostics
Test diagnostics are essential prior to, during, and after a test run to make sure data is properly collected for post-run analysis. In addition to self-tests and on-board calibration sources, modern systems can verify sensor wiring and excitation. Ideally, the channel provides more information about out-of-boundary readings than only good/bad indications. For example, supplying voltage values for the four legs of a Wheatstone bridge circuit is very helpful to determine if a strain gage suffers from a short circuit or a clipped excitation voltage lead.
PC Compatibility
Some systems use the hard drive of a PC to record test data, with some restrictions due to susceptibility to shock, vibration, and temperature. Other units save data to RAM or flash memory, which are more rugged but offer less storage capacity. In either case, all in-vehicle recorders use PCs for setting up tests, real-time displays, and data analysis.
Data can be transferred over Ethernet or USB to networked PCs, file servers, memory sticks, or CDs. Engineers will analyze the collected data, select the most important events and channels, and transfer the data to advanced analysis software such as MATLAB, nCode, or RPC Pro.
Scalability/Parallel Acquisition
Projects to acquire data on large platforms such as aircraft frames, launch pads, or wind tunnels often use multiple data acquisition systems to acquire sufficient channels in parallel. Data from different units must be synchronized by a common clock signal. Ideally, all data may be accessed by multiple networked PCs during acquisition, enabling each engineer to view data of particular interest.
Customization
Each engineering organization has different procedures and protocols to acquire and analyze data. Use of industry-standard software interfaces such as DCOM, .NET, and XML allows open access for convenient transfer of data to engineering databases and design programs.
Data acquisition software should provide different levels of access to specific features and settings. For example, an operator may be allowed to recalibrate a sensor but not modify the time base or filter settings.
Conclusion
A test engineer faces ever-increasing demands for more data at higher resolution in shorter periods of time. Appropriate selection of data acquisition systems is essential to help meet these demands.
About the Author
Jay Roberts has served as a product manager for Nicolet Technologies for eight years. He holds a B.A. from Kalamazoo College and attended graduate school at Indiana University. LDS-Nicolet, 8551 Research Way, M/S 140, Middleton, WI 53562, 608-276-5600, e-mail: [email protected]
Signal-Conditioner Filtering
The type of analog signal filtering used depends on whether the key interest is time-domain or frequency-domain analysis. Filters show either good transient (step) response or good frequency response but not both.
Bessel filters most often are used for analyzing step response, such as measuring the amplitude of a signal spike, due to the linear phase response and flat bandpass. In contrast, commonly used filters, such as Butterworth or elliptical, demonstrate excellent frequency response but distort step response by as much as 20% (Figure 1).
If Bessel filters are so useful for transient response, why are they not used more widely for frequency response? As shown in Figure 2, Bessel filters display very slow 3-dB roll-off, so they lower the frequency bandwidth that can be accurately measured. In contrast, Butterworth or elliptical filters allow much higher frequencies to be monitored, almost up to the Nyquist frequency or one-half of the sample rate.
Bessel Filter—Black; Butterworth Filter—red; Input Step—brown
Bessel Filter—Black; Butterworth Filter—red
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2003 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
December 2003