Vibration testing of today’s electronic products is becoming more complex, calling for sophisticated profiles such as sine on random, random on random, and reproduced road tests. Vibration equipment designers are responding by leveraging customers’ intuitive understanding of Windows-like graphical user interfaces (GUIs) to make the equipment easier to use. Many vibration test equipment manufacturers say they are actively developing GUI-based application-specific solutions to help non-expert users.
“MB Dynamics has developed GUIs on PC platforms that can be customized to particular test applications,” said MB Dynamics president, Richard McCormick. “Special software has been delivered that uses current PC controls to handle test-specific functions, allowing the non-expert to remain productive. Overarching this trend is one guiding theme: Customers want suppliers to deliver whole, bundled solutions, not just products.
“Customers have less and less time available to decide how to adapt standard products to their test applications, build all the nonstandard fixtures and instrumentation, and make all the multiproduct interfaces,” he said. “They value suppliers who help them develop test methods and processes to make the best use of their equipment. They want suppliers to take turnkey responsibility for what is delivered and be knowledgeable in working with them to decide which tests should be run.”
Dr. Sri Welaratna, president and CEO of Data Physics, described the user interface on one of his company’s products. “The Windows 95/NT-based interface offers a look and feel in common with all modern PC applications. The screen is arranged with pull-down menus, straightforward dialog boxes, and a toolbar for quick access to tables and editors. Setting up and running tests and viewing data are simple point-and-click mouse operations.”
Test Waveforms Grow in Complexity
Automotive product quality is attracting continued attention, according to Robert Mercado, U.S. sales and marketing manager at Trig-Tek. “We are seeing a drive toward zero defects in automotive electronics. This requires more in-house screening by the manufacturer.
“To keep costs and screening time down, manufacturers are looking at performing quick vibration screens on all their electronics before shipping them to the customer. For this application, wideband random vibration must be used to excite all the important vibration modes of the product,” he continued.
There is an accompanying trend toward tests that more closely represent the actual vibration seen by a product under normal use. This point was made by Mr. McCormick of MB Dynamics. “Measuring the real-world conditions that act on the product in its service environment will reduce the likelihood of overstressing or understressing. This includes analysis of the power spectral density (PSD) shape, bandwidths, crest factors, transient time histories, and accelerations.
“Our business strategy is to deliver affordable solutions to help our customers find and fix the causes of squeaks and rattles in vehicles,” he said. “Problems exist in instrument panels, doors, seats, roof assemblies, radios, airbag modules, and seat-belt retractors. We want to provide solutions that encapsulate our application know-how and training along with the equipment.”
Vibration testing is an important design verification tool for many types of vehicles, not just cars. The tests come close to being a true simulation of the field environment to which the test items will be subjected. Vince Murray, sales application engineer with Unholtz-Dickie, explained the relevance of sine on random testing.
“The random component of the test represents the vibration that the device under test (DUT) realizes from the engine, and the swept sine components represent the actual sine vibration realized from belt-driven components attached to the engine,” he said. “Random on random tests used by the military are generated from helicopter and tank vibration analyzed in the field. Again, the broadband random vibration is a simulation of engine noise, and the swept random components of the test are generated by helicopter blades, tank tracks, and gunfire.”
Laboratory vibration testing representing real-life service conditions can be reproduced from recorded test-track data. Software editors are available that allow you to select those parts of a field test recording that best suit your test requirements and to add or modify the data. Alternatively, tests can be designed to simulate service conditions by combining random and sine components.
Controllers Become Comprehensive
Swept sine vibration testing has been used for decades. The quick screening Mr. Mercado of Trig-Tek referred to could be accomplished more slowly by the swept sine approach. However, applying a wideband random vibration signal excites all the product’s vibration modes simultaneously, regardless of their specific resonant frequencies. As a result, the wideband test is faster to run.
Analog swept sine testing is straight forward. You sweep an oscillator across the frequency range of interest, apply this drive signal to the test article, and measure the results. Of course, it’s not quite that simple because a control loop is involved which forces the shaker table to follow the reference input sine. The output from an accelerometer attached to the DUT then can be monitored for resonant peaks as the sine input is swept.1 In the traditional swept sine method, only one frequency at a time was being controlled.
Modern digital controllers control random, wideband excitation by comparing the spectrum of the feedback control accelerometer signal to a reference spectrum. These controllers also compensate for resonances within both the basic shaker and the fixturing required to mount the DUT.
In addition, several channels of vibration data from different parts of the DUT may be recorded and monitored simultaneously. In a multichannel controller, you can determine which of the channels are to be used for control of the shaker and which are to be monitored continuously to determine if preset limits have been exceeded. If they are exceeded, the test will be aborted. Before the test is run, it’s very important to set appropriate limits that will protect the DUT and the shaker system.
There’s a Lot Going on Under the Hood
The first area to consider is the closed-loop control of the basic shaker and test fixture. In the past, resonances in this assembly were difficult to deal with, so fixturing had to be made very rigid and often was expensive.
In the June 1997 issue of EE-Evaluation Engineering, dynamically poor payloads were discussed along with adaptive filters that could compensate for them. Adaptive filtering is used to determine the inverse transfer function of the shaker and its load at the point where a control feedback accelerometer is mounted. This process is termed inverse modeling.
“Most adaptive-filter algorithms operating in the time domain require the implementation of…convolution. In this case, convolution integrates the products of two time-varying functions, [the output from the shaker/fixture assembly and a delayed version of the reference waveform].” 2
Frequency-domain multiplication is equivalent to time-domain convolution, and it’s faster. Consequently, adaptive filters usually operate in the frequency domain.
This digression into adaptive filtering is important because such a filter can account for the resonances of the combined shaker and fixture mechanics. These are major benefits that the vibration test engineer did not enjoy before the advent of digital controllers. Adaptive filtering may afford the opportunity to reduce fixturing costs, and it also may help to produce more accurate test results more easily.
Preconditioning the reference signal by passing it through the filter ensures that the reference signal is reproduced faithfully at the DUT. Because the adaptive filter coefficients are continuously re-evaluated, temperature-induced changes in resonant frequencies also are compensated for.
The PSD curve describing a random vibration signal is continuous, and the corresponding time-domain waveform is constantly changing. For random testing, the average of many fast Fourier transforms (FFTs) of the vibration signal is used as an estimate of the actual PSD at the DUT. The controller adjusts power within frequency bands to ensure that the estimate closely matches the desired PSD. Because the average of the FFTs consists of discrete frequency lines, it can only approximate the continuous PSD curve.
In the case of a controller reproducing data from a road test, both the phase and magnitude of each frequency component will match those originally recorded. Randomness in this context refers to the uncontrolled, naturally occurring noise sources that were recorded, such as tires on gravel and engine sounds.
Digital filtering, which requires extensive multiply and accumulate operations, and FFTs, also requiring many so-called butterfly computations, form the heart of modern controllers. Rather than attempting to solve time-domain convolution integrals, controllers multiply FFTs quickly in the frequency domain and then perform an inverse FFT to return to the time domain. The digital output word is applied to a digital-to-analog converter (DAC), and the DAC output forms the input to the power amplifier which eventually drives the shaker.
To cope with large dynamic range signals such as those encountered close to and far away from resonances, 18-bit resolution analog-to-digital converters change at least one manufacturer’s controller inputs to digital form. If rounding, overflow, and accuracy are considered, resolution greater than 20 bits must be maintained throughout the math operations.
How Fast They Go and Why It Matters
Dynamic performance of digital controllers depends upon throughput which, in the case of Data Physics’ Model SignalCalc 550, is provided by three 40-MHz DSP ICs per four channels. They perform the required FFTs, adaptive filtering, and inverse FFTs in real time.
This means that all the required computations are performed on successive frames of sampled data points without missing any data. In the case of the SignalCalc 550, 800 data points are sampled at 4 kS/s, taking 200 ms total. The FFT of this data has 400 lines, a 2-kHz bandwidth, and 5-Hz resolution. The controller’s loop time is 200 ms.
The shorter the loop time can be made, the more responsive the controller can be to changes either within the reference PSD or the DUT and shaker system. However, the lower limit is the time taken to acquire the necessary number of data points. Faster DSPs can operate simultaneously upon more channels and allow higher sampling rates while still completing all required computations within the minimum loop time.
Dr. Welaratna of Data Physics commented about future improvements: “The trend is to continue using higher-powered DSPs to make the product smaller and less expensive and at the same time a better performing product. This trend already has produced the market expansion of vibration controllers and will continue.”
Sine on random testing highlights an important limitation of random controllers. The sine frequency can only be changed after each data frame unless a separate swept sine facility is provided. Some manufacturers do this, summing the smoothly changing sine with a random signal that is only updated after each loop time interval. Other controllers step the sine by controlling it along with the random component.
Testing Portable Products Requires High Frequencies
There are other trends in product design that have affected vibration testing. In some cases, vibration frequencies are as high as 4 to 8 kHz. Also, the very small, portable telephones and calculators being made today require shock testing at high frequencies because they are so lightweight.
Neill Doertenbach, product manager at QualMark, added, “To more effectively test small items, we have designed a new actuator system. It allows you to tune the table characteristics, shifting the energy toward the high end of the spectrum. This capability is a significant feature in a random shock vibration system.”
Test Engineers Are Vanishing
Test engineers must be aware of the dynamic effects they could encounter when using a digital controller. But at the same time, they shouldn’t have to become DSP experts to understand how to apply one.
Adding emphasis to the test engineer’s dilemma are comments from Mr. McCormick of MB Dynamics. “Skilled test people continue to be a vanishing breed in today’s test lab. Labs are being staffed more by engineers and technicians who are generalists. They have to wear many hats to get their jobs done.”
Fortunately, many solutions exist for the problems of test engineers. The following associations offer a variety of courses, literature, and reference material specifically relating to vibration testing. The companies mentioned in the article also are good sources of information.
The Institute of Environmental Sciences and Technology (IEST), (847) 255-1561, www.iest.org.
The Society for Experimental Mechanics, (203) 790-6373.
The Society of Automotive Engineers (SAE), (724) 776-4841, www.sae.org.
References
1. Prosuk, A. and Bossaert, G., “Vibration Control Testing,” Realtime Update, Hewlett-Packard, Fall 1994, pp. 7-9.
2. Nobari, S., “Vibration Controllers Handle Complex Profiles and Payloads,” EE-Evaluation Engineering, June 1997, pp. 79-81.
NOTE: This article can be accessed on EE’s TestSite at www.evaluationengineering.com. Select EE Article Archives and use the key word search.
Vibration Test Products
Controller Can Be Enhanced
By Adding Software Modules
The Model VCP_PC Vibration Control System runs under Windows NT and interfaces to the HP E1432A VXI Digitizer and the HP 3565S Dynamic Signal Analyzer. Vibration modes include broadband random with notching, classical shock, automatic notching for random and sine testing, sine on random, and random on random. The software is modular, so you can build your system using only those elements relevant to your tests. Windows NT supports exporting data, charts, and graphs to word-processing programs and spreadsheets. m+p International, (973) 239-3005.
Module Converts PC
Into Vibration Controller
The DVC-4 Controller converts a PC into an instrument providing operating screen displays and electronic control-panel indications. The system features four-channel monitoring and control and operates in the swept-sine, random, or classical shock control modes. Software and a small interface card that plugs into the PC are included. The DVC-4 control package measures 7″ × 15″ × 1 ½” and connects to the control card via a cable. The controller provides current sources for accelerometers. A demo disk is available. $9,995. Vibration Test Systems, (330) 562-5729.
Software Package Organizes
HALT and HASS Testing
The Test Manager program handles all HALT and HASS test setup, control, monitoring, and reporting. A wizard-based system speeds up profile generation and chamber setup. Temperature and vibration profiles can be generated if you require them to be time or temperature driven. You can change test flow dynamically based on test results, chamber status, or product responses. Collected data can be stored for incorporation into reports. Because Test Manager uses National Instruments’ HiQ™ data presentation language, custom report packages can be created. QualMark, (303) 254-8800.
Control Software Recreates
Actual Time History Waveforms
Time Replication Acceleration Control (TRAC) is control software that allows the closed-loop reproduction of stored acceleration data in the time domain on electrodynamic or electrohydraulic shakers. Targeted applications include buzz, squeak and rattle testing; accelerated life testing; endurance testing; failure-mode generation; and environmental conditioning. A time history waveform up to 10 minutes long can be stored and recreated. Waveform editing and looping are supported. Lower (1.5 Hz minimum) and higher (500 Hz maximum) frequency limits and control resolution (0.5 Hz maximum) are selectable. Unholtz-Dickie, (203) 265-3929.
Accelerometers Are Calibrated
Accurately by PC-Based System
Matched signal-conditioning channels in the Model 2000A Transducer Calibrator ensure that inputs from the reference and test accelerometers are treated equally. Sensitivities from 0.1 to 999 pC/g and 0.1 to 999 mV/g are handled. The calibrator provides from 2- to 10-mA accelerometer drive, controls excitation and measures response from 2 Hz to 20 kHz, and has an accuracy of 0.3% @ 100 Hz, 0.5% from 10 Hz to 10 kHz, and 1% from 2 Hz to 20 kHz. A Pentium base computer is recommended with 1,024 × 768-pixel monitor resolution. Trig-Tek, (714) 956-3593.
Vibration Controller Uses
ActiveX to Automate Routines
The SignalCalc 550 Vibration Controller simulates various production and field environments with random, sine, shock, sine-on-random, and resonant search and dwell. ActiveX is supported in SignalCalc 550 and the company’s SignalCalc ACE dynamic signal analyzer to automate test sequences and share information with other test reports. Custom control panels can be created, and the test process can be controlled from an external user interface. SignalCalc 550 provides reproductions of vibration environments on electrodynamic and hydraulic shakers. Data Physics, (408) 371-7100.
Vibration Control Software
Also Schedules Tests
WinVCS Control Software is a Windows 95-based package that runs on the manufacturer’s existing VCS-800 Vibration Control System platform. Sine, random, and shock control and automatic test scheduling and profile sequencing are included. Displays of the drive signal, control response, monitor response, and transmissibility can be viewed simultaneously. Over- and under-stress limits feature scalable tolerance and abort thresholds. The VCS-800 provides over 80-dB dynamic range, from 100- to 800-line resolution, and a frequency range up to 4,500 Hz. Four channels are standard, but the system can be expanded to eight. Thermotron, (616) 393-4580.
Copyright 1999 Nelson Publishing Inc.
January 1999