Verifying a Novel Approach to Acoustic Testing: Acoustic Devices & Control Systems
As published at the 32″d Aerospace Testing Seminar
Dale Schick, Thomas Hoffmann
mtp international, Verona,
NJ Eric Friedlander, Ed Kinsella
Acoustic Research Systems, Shepherdstown, WV
When conducting an acoustic test utilizing a direct field approach, the optimal pairing of speaker and control systems is of the utmost importance. With that in mind, this paper discusses novel approaches with these interconnected systems to improve acoustic test quality, performance, and reliability. First, to expand the capabilities of a loudspeaker-based acoustic testing system, a “blank-slate” approach is discussed for creating an acoustic device exclusively for the unique and exceptional demands of direct field testing. The existing method of utilizing COTS music-reinforcement systems is compared with a new method of acoustic testing utilizing purpose-built acoustic devices, and this new method is reviewed regarding existing benchmarks for field diffusivity, uniformity, and previously published limitations on test time and output levels.
During launch, the payload on a rocket (e.g., a satellite) is exposed to a high energy sound pressure field. Typically, the field has characteristics that are without any distinct harmonics leading to a uniform field inside the rocket’s fairing . Depending on the structural resonances of the payload, the acoustic excitation can possibly cause unacceptable damage. Therefore, the expected sound pressure levels from launch are recreated in a lab to safely test the effects of the field, first on mass-mockups, then on the payload itself. Historically, Reverberant Field Acoustic Test (RFAT) chambers were the only possible solution to recreate the needed sound pressure fields due to their low acoustic energy dissipation and predictable sound pressure fields. Multiple nitrogen-powered horns with bandpass frequency characteristics provided sufficient acoustic energy to recreate the launch environment, but with some restrictions, such as limited frequencies which can be controlled. The reverberant chamber itself has a long reverberation time, for example 20s for low frequencies and 3s for high frequencies [2, 3]. Reverberant chambers are expensive to operate, maintain and refurbish and there are personnel safety issues to consider, especially if the test is done under Nitrogen atmosphere [4, 5, 6].
An alternative for acoustic testing in reverberation chambers is Direct Field Acoustic Noise (DFAN) testing. There, multiple acoustic transducers, typically built for music-reinforcement of concerts, are placed around a device under test (DUT) in a circle. The speakers are powered through amps that are driven by an acoustic control software, which shapes the launch environment by processing microphones within the field to close the feedback-loop.
The last few decades have seen vast improvements in non-reverberant acoustic testing for space tech applications as it strives closer toward the uniformity and achievable sound pressure levels that reverberant chambers provide . Initially, control systems provided a single (coherent) drive signal which was then processed by traditional live-sound equipment as used for music-reinforcement, which applied crossover filters to deliver the band passed signals to each individual component of the acoustic transducer system. Innovations were made to deliver multiple (semi coherent) drives controlled by multiple discreet control microphone locations, which reduced the amount of standing waves, increasing uniformity in the field. Though greatly reduced, standing waves in the sound pressure field cannot be eliminated unless all sources are completely incoherent [8, 9].
The longstanding problem with traditional sound equipment has been twofold. Firstly, the speakers lacked the power efficiency that would otherwise allow for all positions to be completely decoupled because they require coherence from adjacent speakers to produce enough energy to meet the demands of most aerospace acoustic requirements . Secondly, the separate drivers require that ‘sections’ of the speakers produce energy isophasically in the same bandpass region. Notably, small drivers designed to be lightweight and hung from a support structure to project high-frequencies far into large audiences were made separate from their heavier low-frequency woofers that provided the bass energy, which travels long distances more easily, and are therefore typically placed on the ground.
In this study an innovative type of acoustic excitation device, called NEUTRON, is investigated. It is specially designed and manufactured by Acoustic Research Systems for the sole purpose of DFAN testing. Where the previously discussed solution of traditional sound reinforcement equipment requires separate ‘sections’ of speakers to reproduce the entire test spectrum, each device provides the full test range of 20 – 10,000 Hz in a fully developed wavefront at the front of the cabinet and were designed with adequate efficiency to deliver launch specification levels from just a single device.
With these functional developments, a fully incoherent operation mode can be achieved by individually controlling each device with a single drive without the need to apply any mixing technologies. To show the advantages of fully incoherent testing, comparisons of the effects of operating completely incoherent devices versus semi-coherent (the current standard) as well as completely coherent were conducted.
Figure 1: Front and side view of the NEUTRON device
To overcome the performance issues of directly radiating concert speakers for high sound pressure levels (SPL), the NEUTRON, which will be referred to as ‘acoustic device’ in this study, Figure 1, makes use of a proprietary high-efficiency horn loading technology like traditional reverberation chamber acoustic transducers. It is notable that, in some sense, music-reinforcement speakers are designed to maximize SPL per unit volume for handling reasons, but for the purpose of DFAN testing, the SPL per unit area with respect to the cylinder area around the test article is more important. Thus, the device was designed deep enough to employ full horn-loading throughout the frequency range. As shown later in this document, this design provides the ability to run high-level tests on smaller DUT’s, as well as eliminating complications that arise from trying to phase align different passbands that are spatially (and therefore temporally) separated. The physics of acoustical horn-loading is well understood, and outside the scope of this paper. But to illustrate the points above, Figure 2 shows a measured efficiency plot of the acoustic device’s low frequency section vs a conventional direct radiating LF unit. Here, radiating into a steradian space of 2m was assumed.
Figure 2: Comparison between the acoustic efficiency between the low frequency section of the NEUTRON device and a traditional low frequency concert speaker
Figure 3: Experimental test setup with 6 NEUTRON acoustic transducers, six control microphones and a measurement microphone array
For the following three tests, six acoustic devices were arranged in a 120 inch diameter circle around a mock test article made of wood, metal, and covered in fabric, Figure 3a. Six microphones were used to close the feedback loop, with one microphone per acoustic device. The feedback loop was controlled by mtp international’s VibControl DF-ACS controller software and accompanying data acquisition hardware.
For all control and monitor locations, the free-field PCB 378A12 microphone was chosen due to its dynamic range (182 dB ref. 20 uPa), relatively flat response (‡2 dB from 5 Hz to 20000 Hz), and the convenience of Internal Charge Piezo Electronics (IEPE); which provide preamp-free operation with the acquisition hardware. An independently calibrated PCB model CAL 200 microphone calibrator was used prior to testing for each configuration to account for any changes in pressure and humidity.
A twelve-microphone array, Figure 3b, was placed above the test article. This location was judged to have the least uniformity of field and would therefore be a conservative representation of the UUT exposure. It was constructed from PVC pipe to which four wooden dowels were affixed. Plastic clips were used to attach the microphones to the wooden rails. The microphones on the array were randomized in all three axes within about three inches (76 mm), to avoid accidentally ignoring standing waves that happen to share a wavelength with the distance between each location.
The number of acoustic devices were chosen as an optimal balance of acoustic coverage and headroom, while still leaving adequate gaps between them for ease of access to the DUT. Each was powered by a single eight channel audio amplifier operating at 415VAC. Drive signals were patched from the controller outputs via a Direct Out Prodigy.MP audio processor and delivered to the amplifiers digitally via the Dante™ Audio-Over-IP protocol. Measurements showed that each acoustic device generated a broadband acoustic wavefront starting at the front grill, significantly reducing the distance required between it and the DUT. As such, a 120-inch diameter was chosen to ensure optimal acoustic coverage and proximity to the DUT.
Figure 4: Photograph of the test setup for consistency study without test specimen
The 1540E Workmanship’ specification on test requirements for space vehicles was used for comparison testing because its broadband acoustic target is at a relatively moderate OASPL level, which makes it good for evaluation purposes of various drive methods. Since the specification did not particularly challenge the capabilities of the acoustic system, it allowed the study to focus strictly on variances in uniformity.
Due to the portable nature of Direct Field Acoustic Noise (DFAN) systems, it is ideal to create a test configuration that comes as close as possible to reaching the diffusivity expectations of an RFAT chamber but without the difficulty of operation that accompanies such a facility. Furthermore, if this ideal configuration can be reached, field uniformity would be sufficient so that the microphones, acoustic devices, and even the Device Under Test (DUT) need not be expertly placed.
Figure 5: Sound pressure simulation of the direct field at f = 1000 Hz. Top view
Since a typical DFAN configuration doesn’t have the benefit of using reflections to create a sufficiently diffuse field, in theory, the most efficient way to create this type of field is to have as many devices as possible project completely incoherent signals from as many locations as possible. Though broadband simulation is not widely available, it can be helpful to test the theoretical waveform interactions at discreet frequencies with an accurate computer simulation.
Wave computer modeling software was used in conjunction with real-world recordings to help analyze discreet frequency waveform interaction at 1 kHz, Figure 5. Though simulation was important for the research and development of incoherent testing techniques, the nature of a truly incoherent field is such that simulation for purposes of control microphone or speaker placement does not seem to provide any added benefit, unless traditional semi-coherent techniques are applied.
There are some limitations with respect to computer simulation that reduce the accuracy of the simulated spatial variance. Namely, the acoustic devices are modeled as a point source at the front/center of the device and, as such, assume similar horizontal/vertical dissemination of the energy. This can be improved by measuring an individual acoustic device at various distances and offset angles from the front-facing portion as well as characterizing rear-facing rejection. Such a study is planned and will help improve the simulation’s ability to pre-determine the number of devices needed for a specific test article size.
In the absence of more precise source properties, the point-source method was used in the simulation depicted in Figure 5. While there are still some variations in the field, standing waves are greatly reduced when driving the speakers incoherently. In a real test, when the sources are incoherently averaged over many excitations the result is expected to be smoother and more diffuse.
By virtue of the acoustic device’s ability to project the entire test-frequency range from each location, this study was able to avoid any additional variations, especially in the lower frequencies. For broadband results, however, a real-world measurement as described in the next section with independent controls and monitors helps to reveal the presence of standing wave interactions that cannot be modeled simultaneously.
Figure 6: Measurement results of 1540E workmanship specification using a fully coherent drive signal on all acoustic transducers
The 1540E specification was used for this test (137.84 dB OASPL) with a single drive output controlled by the average of the six control microphones using the octave control method. Due to the difficulties described previously with coherence and standing waves this test, predictably, suffered from low field uniformity even from one control microphone to the next, Figure 6. It is worth noting that no known modern system operates with perfectly coherent drive signals. This test, however, served as a good way to imagine a ‘worst-case’ scenario with respect to control of the field.
The microphone array, placed above the test article, was over the abort limit specifications (+6 dB) from roughly 100 to 500 Hz and significantly under specification (-3 dB) from 200 to 7000 Hz. The overall sound pressure for the microphones in the array ranged from 137.4 dB to 144.4 dB, the highest of which was 4 dB over the 3 dB limit that the specification requires. Narrowband analysis further reveals the presence of standing-waves, visible as protruding peaks in the spectrum, and a general lack of diffusivity.
Figure 7: Measurement results of 1540E workmanship specification using semi coherent drive signals on all acoustic transducers
The 1540E specification was also used for this test (137.84 dB OASPL) with a single drive output controlled by the average of the six control microphones using the octave control method. Unlike the previous MISO test, a proprietary ‘smoothing’ method was applied which artificially decouples each device from the next. The purpose of artificially decoupling is to simulate the partial mixing methods that are used in existing technologies without infringing on existing intellectual property of any proprietary in-process methods.
The microphone array response was more controlled across the entire spectrum with respect to upper-frequency tolerances but did suffer from undertest from 40 Hz to 200 Hz due to a comb-filter-like artifact, which seems to have been a side effect of using the smoothing algorithm, Figure 7. Though diffusivity still suffered across the spectrum, decoupling somewhat still seemed to have a positive effect on the field in general, as revealed by the linear average of the narrowband. Especially any standing waves observable in Figure 6 could be reduced. The scattering in the narrowband spectrum of the array microphones above 100 Hz is caused by reflections at the DUT and interference caused by the semi coherent drive signals.
Figure 8: Measurement results of 1540E workmanship specification using fully incoherent drive signals on all acoustic transducers
Again, the 1540E specification was used for this test (137.84 dB OASPL) but this time with six completely incoherent drives to each of the six acoustic devices, each one controlled by its corresponding control microphone using the octave control method, Figure 8. No additional mixing, decoupling, or filtering were used in the amplifier network. The array, placed above the test article, was slightly below the -3 dB OASPL but very closely resembled the general shape of the control microphone plots. Diffusivity and uniformity were greatly improved in this configuration, as revealed by the linear average PSD plots in Figure 8. The measurements support the prediction that completely incoherent signals produce minimal standing waves and provide an environment that is more like a reverberant chamber.
Again, the microphone array used in this configuration was placed in an area where the field was not expected to meet specifications across the entire spectrum, so additional testing needed to be done to prove that these results scale to a larger test article and a more realistic monitor microphone network.
To further examine the effects of completely uncorrelated signals, twelve acoustic devices were arranged in six stacks of two configured radially around the test article. For a configuration example, refer to Figure 4. Seven individual runs were made, two at 10′, two at 12′, and three at 15′ diameters, with control microphone placement completely randomized. Four monitor microphones were placed at random distance and height around the test article and remained at the same locations for all seven runs. In this way, the monitors were representative of the general quality of field even as the configuration around them was drastically changed. A test article based on the Low Earth Orbit OCO-2 satellite dimensions was constructed of internally braced wood with asymmetrical aluminum and steel features as it most accurately represented the type of object that a system of this size would be able to test.
Surprisingly, regardless of control microphone location or distance of the acoustic devices to the DUT, the static monitor microphones generally remained in-tolerance, even when examined at 1 Hz resolution in Power Spectral Density, as shown in Figure 9. Additionally, the standard deviation for all four static DUT microphone locations for all seven runs was less than 1.5 dB in the individual octave bands and less than 0.23 dB OASPL. With respect to distance, it is known that the quality of directivity improves when horn-loading devices are used, as opposed to devices that are directly radiating. More data is needed to prove, but it is assumed that this improved directivity without the need of adjacent device reinforcement is somewhat responsible for the ability to resize the test circle diameter without greatly reducing OASPL or quality of field.
When restricted to a single diameter (15′, in this example) the deviation across the three runs that shared this configuration was less than 0.6 dB in the individual octave bands and less than 0.16 dB OASPL.
Figure 9: Measurement results of consistency study. PSD of device-under-test microphones of four separate runs with random control microphone placement.
State of technology advice with respect to control and device placement suggests that there are specific rules to follow to ensure the control microphones are properly controlling the field, but do not accurately measure the energy away from these control points, creating a ‘control field’ that is representative of only those specific locations and an ‘actual field’, which is all the places in the field that are not being controlled. The results of this study indicate that when using DFAN, it is important to make a distinction between the ‘control field’ and the actual field’ and measure both, as the two are quite different when standing waves are present.
The conclusion of this measurements arguably suggests that selecting the correct number of acoustic devices for a specific test article and setting a consistent distance are more important factors than choosing accurate microphone placement when all sources are completely incoherent. For this reason, it may be worthwhile to use computer simulation of the field so that accurate power capability is chosen for the test article size and mass. Though not plotted in this paper, it also seems that using a subset of control microphones to be averaged for each zone provides better overall control of the field when uniformity is already achieved by decoupling the sources.
To further understand the effect that exponentially averaged narrowband (FFT) based control has on test-article exposure, two tests were performed to test the reaction speed, level-step methodology (or how quickly the system jumps from one level to the next), and the system’s ability to protect the UUT in an over/under test situation. The results were so extensive, that a separate, more focused paper will be written on the topic. Some brief conclusions of this study are as follows:
A: Exponential averaging significantly slowed the system’s ability to make timely adjustments when controlling in narrowband. Ten to thirty-six seconds was typical, depending on the frequency resolution. Whereas octave control was only constrained by the duration of the time constant, typically one second. B: When stepping-up in level from one segment to the next, narrowband control methods assumed linearity and “jumped’ directly to the next level. Octave control methods made second-to-second adjustments that gently increased the level, considering the nonlinear nature of the acoustic field. C: Due to the nature of exponential averaging, the narrowband control systems did not reveal over testing conditions and failed to abort the test. Octave control methods aborted the test at precisely the speed they were programmed to.
Maximum Output Study
Understanding the results of the previous study with regards to field uniformity and acoustic device placement, it was decided to push the limits of the system to study whether the uniform behavior of the sound field at lower OASPL levels would translate to higher OASPL tests. To do so, the EELV Secondary Payload Adapter (ESPA) launch specification was chosen and 6 dB OASPL added to the reference. This specification is particularly difficult to achieve with traditional DFAN configurations due to the low frequency content requiring more low frequency and subwoofer components than are practical for the relatively small test-article size.
Again, twelve devices were used in six stacks of two each and placed around the test article in a 7.5 foot (2300 mm) diameter circle and controlled by an average of two microphones per stack in uncorrelated octave control.
The results were nominal at 150.5 dB OASPL with all control and monitor microphones showing in-tolerance values when standard tolerances were applied to 1/3 octave and PSD linear averages. Software limiters for the acoustic devices and amplifiers indicated headroom and the test was able to run for a full minute (as required by the specification) without any noticeable degradation in device impedance. It’s notable that the system was able to operate a similar 148 dB test afterward with only a few minutes of ‘cooldown’ time between runs.
Figure 10: Experimental result of maximum achieved overall sound pressure level. Test setup was changed so that the acoustic transducers touch at their corners (i.e. without gaps)
The capabilities documented in these experiments call into focus existing, published guidance for conducting DFAN tests (such as NASA-HDBK-7010) that are based on the abilities of music-reinforcement systems. The goal of DFAN testing, to produce a field that more exactly matches a reverberant field acoustic testing (RFAT) environment, can be even more fully realized with advances in the technologies and control methods used to perform DFAN tests.
With the benefits and challenges of producing an incoherent field at least initially investigated, further research must be done to build on the results of the methods demonstrated in this paper to ensure that this approach meets the scalability and reliability goals of DFAN testing.
Control-wise, when comparing the computer simulation of theoretical decoupling to measurements of completely incoherent acoustic devices, the evidence arguably indicates that current industry standards of partial coupling are restrictive in pushing the technology forward because standing waves cannot be completely eliminated. When using DFAN, it is important to make a distinction between the control field’ and the ‘actual field’ with respect to the DUT and to take measurements of both, as the two are quite different when standing waves are present. Furthermore, exponentially averaged narrowband acoustic control, though useful in increasing diffusivity for semi-coherent devices, adds additional restrictions to the safety and accuracy of acoustic testing. It was shown that no preliminary simulations on the control microphone locations are needed when using fully incoherent source signals as the sound pressure field is sufficiently distributed at all locations of interest. Also, even at very high sound pressure levels over 150dB, field uniformity persisted.
- W. O. Hughes and A. M. McNelis, “Investigation of Acoustic Fields for the Cassini Spacecraft: Reverberant Versus Launch Environments,” in 5th Aerocoustics Conference and Exhibit, Bellevue, WA, 1999.
- Q. Z. S. W. J. Z. Xiaojun QIU, “A case study on the new reverberation room built in University of Technology Sydney,” Proceedings of the 23rd International Congress on Acoustic, 2019.
- F. W. Grosveld, “Characterization of the Reverberation Chamber at the NASA Langley Structural Acoustics Loads and Transmission (SALT) Facility,” NASA/CR-2013-217968, 2013.
- J. S. Peterson and R. C. Bartholomae, “Design and Instrumentation of a Large Reverberation,” in Proceedings of the 2003 National Conference on Noise Control Engineering, Cleveland, Ohio, 2003.
- A. A. Tan, N. Salim and S. A. I. M. I. Noor Hidayah Tauhid Ahmad, “Relative Characteristic Analysis on Reverberation Acoustic Facility,” WSEAS TRANSACTIONS on ACOUSTICS and MUSIC, 2019.
- L. Freyber and A. Grillenbeck, “New Acoustic Test Facility,” in Proceedings of the 4th International Symposium on Environmental Testing for Space Programmes, Liege, Belgium, 2001.
- A. R. Kolaini, M. R. O’Connell and W. B. Tsoi, “ACOUSTICALLY INDUCED VIBRATION OF STRUCUTRES: REVERBERANT VS. DIRECT ACOUSTIC TESTING,” in 25th Aerospace Testing Conference, Manhattan Beach, CA, 2009.
- P. Larkin, “Developments in Direct-Field,” Sound and Vibration, November 2014.
- A. R. Kolaini, B. Doty and Z. Chang, “Impact of Acoustic Standing Waves on Structural Responses: Reverberant Acoustic,” Jet Propulsion Laboratory, California Institute of Technology, Pasadena, 2012.
- E. C. Stasiunas, T. J. Skousen, V. Babuska and D. Gurule, “Designing a Direct-Field Acoustic Test of a Flight System;” Sandia National Laboratories, Albuquerque, NM,
Dale Schick is a Support Engineer for mtp International Inc. He has extensive experience in environmental testing as applied to Aerospace and over 14 years of practical experience in the industry.
Thomas Hoffmann is product manager of m+p international and coordinator of the Direct Field Acoustic Noise Testing activities. Before he joined m+p international he worked as a researcher at Institute of Dynamics and Vibration Research in Hannover, Germany. There he worked on his PhD thesis on the nonlinear dynamics of friction damped turbine blades. He holds a Bachelor’s and a Master’s degree in mechanical engineering from Darmstadt Institute of Technology in Germany
Eric Friedlander is the CTO and Co-Founder of Acoustic Research Systems. He has a BS in Audio Technology from American University in Washington, DC. Before founding ARS, Eric was a product manager at Harman Professional for the JBL Professional Tour Sound portfolio, and prior to that he worked as an acoustic testing engineer at MSI-DFAT.
Ed Kinsella is the director of acoustic engineering for Acoustic Research Systems. He has over two decades experience designing high-output acoustic devices for a variety of industrial applications. Ed was educated originally at Charterhouse in Surrey, and then at St John’s College, Cambridge, where he graduated with honors.