Qualification of DFAN Field & Improved Acoustic Control Methods

Eric Friedlander – Acoustic Research Systems
Ed Kinsella – Acoustic Research Systems
Dale Schick – m+p International Inc.
Thomas Hoffman – m+p International GmbH

Abstract

There exist several approaches to control and qualification of Direct Field Acoustic Noise (DFAN) environments, with various degrees of stability, safety, and convenience. This paper aims to compare the methods available today and to expound upon previous claims from the authors. The next generation of DFAN testing is further examined to verify the quality of field for larger configurations, while a method for vertically decoupling acoustic devices is vetted via acoustical and structural response measurements, in pursuit of an even more diffuse and uniform acoustic environment.

Introduction

A previously published effort by the authors from summer 2021 was about the development of a novel simulation and test workflow for Direct Field Acoustic Testing. While the primary focus of that study was to validate newly developed test and simulation methods for DFAN testing with a smaller-scale test configuration, several secondary studies were conducted which have not yet been published. One of these secondary studies focused on how various DFAN drive methods might affect the structural excitation of a simple panel. It was noted at the conclusion of this secondary study that a test panel showed varying levels of structural excitation when exposed to different types of DFAN system drive methods, and possible under- or over-test conditions were seen with more coherent drive signals. (Shorter, et al., 2023) This finding was noted by the authors for additional study and was one motivating factor behind the larger-scale study described below. In Spring 2023, the authors were granted access to a full-size mass model representative of a typical medium-sized satellite. This model was instrumented with accelerometers in locations which were consistent with previous acoustic tests, providing a logical platform for furthering the previously discussed panel study to the next step, both in terms of test article complexity as well as DFAN system size. Then the authors conducted a similar study of drive methods, field conditions, and structural excitation.

Test Methodology

There exist several approaches to control and qualification of Direct Field Acoustic Noise (DFAN) environments, with various degrees of stability, safety, and convenience. This paper aims to compare the methods available today and to expound upon previous claims from the authors. The next generation of DFAN testing is further examined to verify the quality of field for larger configurations, while a method for vertically decoupling acoustic devices is vetted via acoustical and structural response measurements, in pursuit of an even more diffuse and uniform acoustic environment.

  1. Conduct Modal Survey of Mass Model – Intended to provide predictive excitation data for comparison with subsequent operations.
  2. Conduct Fully Incoherent DFAN Test – To validate previous studies’ extrapolations of diffusivity with a larger test configuration.
  3. Conduct DFAN Test in Semi-Coherent and Stack-Coherent Configurations – Intended to further explore the ramifications of a DFAN test which is only coherent on a 2D plane, such as one conducted with traditional concert speakers.
  4. Compare Results Through Data Analysis – Compare data from various excitation methods with the acceleration response of the test article.

Phase 1 – Modal Survey

There exist several approaches to control and qualification of Direct Field Acoustic Noise (DFAN) environments, with various degrees of stability, safety, and convenience. This paper aims to compare the methods available today and to expound upon previous claims from the authors. The next generation of DFAN testing is further examined to verify the quality of field for larger configurations, while a method for vertically decoupling acoustic devices is vetted via acoustical and structural response measurements, in pursuit of an even more diffuse and uniform acoustic environment.

In keeping with the precedent set forth in the earlier panel study, the modal survey was conducted on the mass model utilizing a calibrated impact hammer and response accelerometers. The locations of the response accelerometers were chosen to correlate to a Finite Element Model (FEM) for later studies as well as to compare directly with the recorded acoustic response. The accelerometer locations were then applied to a reduced geometric model to animate mode shapes. Though the full map of response accelerometers is proprietary, the mode shapes and Frequency Response Functions (FRF’s) are shared throughout this paper, with an example provided in Figure 1 below.

Figure 1 – NRL Chamber Horn Configuration

Five distinct locations on the mass model were excited with an impact hammer fitted with a tip which provided broadband excitation with good coherence from 50 to 1600 Hz, where the bulk of the modal response was expected. An example of the coherence from one of these locations is in Figure 2 below.

Figure 2 – Impact Coherence Example

For the intent of this study, the accelerometer positions provided adequate predictive capabilities, but the system was under-instrumented for a full survey. Shaker-driven excitation would have provided more accurate FRF’s, but time constraints forced a simpler test. With this in mind, the authors discovered primary modes of interest both in the structural component and the “solar arrays” of the mass model, which are assumed to be large flexural absorbers with low-frequency excitation. The solar array models exhibited modes at 52 Hz, 66 Hz, 85 Hz, 128 Hz, and 179 Hz, confirming this assumption. The structure exhibited modes were discovered at 146 Hz, 521 Hz, 1055 Hz and 1383 Hz, the first of which is shown in the simplified geometric representations below with the solar arrays included.

Figure 3 – 52 Hz Solar Array Mode (Dotted lines are undeformed)

Phase 2 – Fully Incoherent DFAN Test

Following the initial modal study of the mass model, the next step in the study was to better qualify the acoustic field and structural excitation generated by a large scale DFAN system using purpose-built acoustic devices and uncorrelated drive signals intended to produce a uniform and diffuse acoustic field.

The DFAN system comprised of (32) ARS NEUTRON acoustic devices, deployed in an 8×4 configuration (eight columns, each four high), utilizing the standard cabling and amplification infrastructure that supports the NEUTRON devices. An m+p International acoustic control system operated the system with (32) acoustic outputs and (60) inputs, while an additional system of m+p VibRunner inputs served as a data acquisition system for the (24) accelerometers mounted on the mass model. While these two m+p International systems could have been combined into a single acoustic test control and data acquisition system, for scheduling reasons outside of the scope of the study it was necessary to keep two systems separate.

Figure 4 – 8×4 Configuration with Approximate Microphone Placement

To adequately control and define the acoustic field, (51) microphones were arranged in the test circle around the mass model. (32) microphones were placed in front of each NEUTRON device to function as a control, as shown in Figure 4 in red, and (19) microphones were placed in monitor positions around the test article, as shown in Figure 4 in green. The authors used (32) microphones to capture a more detailed image of the acoustic field as well as enable the study of alternative drive methods with more control loops. Of the (19) monitoring microphones which were not part of the control loops, (12) were placed in the exact positions that would have been utilized if the test were conducted with a narrowband Multiple Input, Multiple Output (MIMO) acoustic controller, and the additional (7) microphones in positions outside the target area of excitation to study the behavior of the acoustic field away from the unit under test.

This placement strategy proved to be an important step towards validating the uniformity of the acoustic field, as the fully incoherent DFAN system could be controlled by the microphones intended for the octave-band controller (in front of each NEUTRON device), while monitor microphones captured the acoustic field in positions typically used for MIMO narrowband acoustic control, allowing for the uniformity of the two approaches to be compared. The controller utilizes ANSI Standard 1.11 (American National Standard, 2004)

Figure 5 – (32) Acoustic Devices Arranged Around the Mass Model, Amplification Racks in the Foreground

Over the course of three days, the mass model was exposed to several acoustic profiles. The majority of which were envelopes of multiple launcher environments reaching levels of 145dB at one- or two-minute durations. Multiple control strategies were tested as well at varying levels of coherence, with varying device coupling arrangements, and several microphone control strategies. Because the DFAN system did not require “cool-down” time between high level runs and could complete entire two-minute runs without stopping to confirm the integrity of the test hardware, it was possible to run multiple 145dB profiles back-to-back, increasing the amount of data that could be collected in a short amount of time.

The additional drive methods were designed to introduce various levels of coherence into the DFAN system. While an idealized diffuse field would have an infinite number of incoherent sources impacting the test article from all directions, reality dictates that a finite number of maximally incoherent drivers are feasibly deployed. This approach, whereby the acoustic devices are maximized by how many can be arranged in a physical footprint around the device under test and driven with as many uncorrelated drives as possible, has thus far produced predictable and repeatable results. For this larger study, three distinct drive methods were utilized featuring 32, 16, and 8 drives, the arrangement of which is shown below in Figure 6.

Figure 6 – Drive Patching Methods

Following the completion of the study, the first topic investigated was field uniformity. For this, linear averages of the entire exposure at the test run 0dB level were generated to qualify that the long-term exposure of the mass model to the acoustic field was within the target parameters.

While quick-print 1/3Oct band plots with a linear average of the entire time at 0dB were immediately available following each test run, further post-processing of the microphone responses was undertaken at the end of the study to produce linearly averaged Power Spectral Density (PSD) plots, coherence plots, and waterfall plots.

One area of particular interest was the DFAN system’s ability to produce a uniform acoustic field around the entire mass model free from standing waves or anomalies. Previously published studies, (Underwood, 2021) posit that a uniform acoustic field is only achievable using phase targeting by a narrowband controller to produce spectrum at targeted microphone locations. Furthermore, they argue that concert line arrays should be used to approximate Reverberant Acoustic Test Facilities (RATF) field properties. In this type of control method, the rest of the field, or all points outside of the targeted controls, are not as well addressed. While the NASA-HDBK-7010 (Roe, 2016) does discuss some of the anomalies that can arise in a phase-controlled acoustic field, colloquially referred to as “hot-spots” or “cold-spots”, the potential damage to a unit under test that these field anomalies can cause is often mitigated in line array and subwoofer based systems by rotating the unit under test so as to minimize the exposure of sensitive surfaces, usually flexural absorbers like solar arrays, to intensely focused areas of energy.

Because of the test program’s interest in comparing the acoustic field uniformity of octave-band vs. narrowband control systems, as discussed previously, microphones were placed simultaneously at control locations typical of octave-band controlled systems and of narrowband controlled systems. This way the field uniformity could be studied both at the acoustic devices, as well as in the gaps between the acoustic devices and at the microphone positions typically associated with narrowband control.

Having run the entire test program with an octave-band controlled system, the results below show the control system microphones referred to as “controls”, and the microphones in positions typically used to control narrowband controlled systems as “monitors”. The monitor positions, in this case, are not controlled in any way, but are instead intended to validate the control field as they are more tightly placed around the test article as seen in Figure 4.

Figure 7 – 32 Drive Controls vs Monitors (1/3 Octave)

Figure 8 – 32 Drive Controls vs Monitors (PSD)

While the agreement between the control microphones for the octave-band test system and the monitor microphones in the narrowband control positions is strong, an unexpected feature of these responses was that in most areas the monitor microphones – arguably placed in less “ideal” positions, both in gaps between the acoustic device columns and in positions correlating to legacy control strategies – held a tighter grouping with more uniform response than the control microphones did. While this might seem counterintuitive, if one considers that each control microphone is in the direct field of the device it is in controlling, while the monitor microphones are at points where they capture the combining of multiple uncorrelated acoustic devices, it makes sense that these monitor positions exhibit the diffusivity and uniformity characteristics of the broader acoustic field. This enhanced diffusivity and uniformity was achieved without the need for complex control strategies attempting to “enforce” field quality requirements at those locations.

By using monitor microphones, the field is more adequately qualified via points not in the control loop. Additionally, this large test with a full-scale mass model appears to disprove the decade-old theory that a uniform field is only achievable by using phase-targeted control microphones.

Phase 3 – Alternative Drive Methods

Following the validation of the acoustic field’s uniformity and diffusivity characteristics over several test runs with various test spectra; the next study was to explore whether drive methods with different levels of coherence affect the level or quality of excitation that the mass model sees during the acoustic test.

To validate the fully incoherent operation method, the same test levels were produced with 16 and then 8 drive outputs, with the latter producing a single drive per column of acoustic devices like a more traditional DFAN system using concert line arrays and subwoofers. Unlike traditional concert line array and subwoofer systems utilizing mix matrix processors, the 8-drive output system still produced 8 stochastic outputs for the low-frequency range between 20-200 Hz, an important distinction as any additional coherence at low frequencies produces vastly different acoustic coverage in the excitation field. The same acoustic target profile was used for each run, and the same microphone locations were utilized as from the previously discussed uniformity study.

Figure 9 – 16 Drive Controls vs Monitors (1/3 Octave, 120s Linear Average)

Figure 10 – 16 Drive Controls vs Monitors (PSD, 120s Linear Average)

Figure 11 – 8 Drive Controls vs Monitors (1/3 Octave, 60s Linear Average)

Figure 12 – 8 Drive Controls vs Monitors (PSD, 60s Linear Average)

Until the recent advent of DFAN testing with octave-band control (Friedlander, et al., 2022) direct field acoustic testing was conducted with idealized coherence targets enforced onto the acoustic field by a narrowband controller, for years this method had provided a better connection between the novel test method and heritage reverberant chamber data. However, it is also well known (Kolaini, et al., 2012) that acoustic fields generated by reverberant chambers have limitations, most notably their inability to maintain diffuse characteristics into the lower frequencies where larger test articles are excited.

As shown in Figure 9 through Figure 12, acceptable field uniformity can be achieved by a DFAN system with multiple different drive methods, so long as a minimum number of stochastic, uncorrelated drives are utilized. However, as mentioned earlier in this paper, the authors wanted interested to further investigate the effects of these multiple types of drive methods with varying amounts of coherence on structural excitation.

Phase 4 – Data Analysis

In addition to linear averages of 1/3 octave and PSD plots of the microphones, the test team discussed best practices for generating periodic or “rolling” averages, as it is typical with narrowband MIMO acoustic control systems to provide a single exponential average of the acoustic field to review the uniformity of the field over time. Since pulling rolling-averages from exponentially averaged spectral information carries the risk of obfuscating potential over- or under-test conditions, it was decided to utilize several linear averages taken from discreet points in the time history to qualify the performance of the acoustic field. These linear rolling-averages were produced every (5) seconds to demonstrate the stability of the system over the course of the test, as shown in Figure 13. Note that the 16-drive configuration contains twice the number of averages due to its 120 second run time.

Figure 13 – Rolling Averages for 32, 16, and 8 Drive Configurations

Given the opportunity to deploy a large DFAN system to test a full-size mass model with a fully incoherent field capable of producing adequate excitation at lower frequencies without the “hot/cold spots” inherent to traditional concert array systems, the authors felt it was important to validate whether the modern fully incoherent approach still produced a high quality of structural excitation.

Acoustic excitation, though not mechanically coupled to the test article, influences the overall acceleration response. To quantify the excitation of the various methods, linear averages at 1Hz resolution were gathered from the time-domain recordings of the data acquisition system and compared to the impact excitation as shown in Figure 14. Note the coupled solar array mode at 180 Hz.

It is worth noting that due to the nonlinear nature of the panels, the higher overall excitation of the acoustic test causes many modes to shift in both frequency and amplitude when compared to the impact test. As mentioned before, exciting these modes with a random shaker test at various levels would have helped to predict these nonlinearities, but time did not permit.

Figure 14 – Solar Array vs. Acoustic Excitation Modes

In both the 32 drive and 16 drive configurations, where columns of acoustic devices were fully or partially uncoupled, respectively, the acceleration response was mostly unchanged with less than 1g RMS variation between the two. In Figure 15, there is 0.87g RMS difference between the 0dB test with fully incoherent operation and partially incoherent operation.

Figure 15 – Acceleration Response Average, All locations (32 vs. 16 Drive)

With the assumption that columns of devices that are at least partially incoherent produce an acoustic field that is more sufficiently diffuse, (Larkin, 2014) a comparison between fully incoherent (32 drives, 1 per device) and fully coherent (8 drives, 1 per column) columns of acoustic devices was conducted. It was found that fully coherent columns produced responses that were 11% higher overall when compared to incoherent columns, especially in frequencies below 200 Hz.

Figure 16 – Acceleration Response Average, All locations (32 vs. 8 Drive)

Especially susceptible to lower frequency excitation fluctuations are flexural absorbers like solar array structures. Further isolating the acceleration response to these structures on the mass-model reveal localized anti-nodes of response when fully coherent columns are utilized. As mentioned previously, these fully coherent columns operate similarly to the vertical line array systems utilized in legacy DFAN excitation methods, where columns of vertically coherent concert mid-high and subwoofer enclosures are used to produce an acoustic field. These traditional methods, in addition to using columns which share a drive signal, employ additional summation when separate low- frequency speakers, or subwoofers, are used. Though the high frequencies are as incoherent as possible, the low frequencies are almost completely coherent in typical configurations.

Figure 17 – Acceleration Response Average, Solar Arrays

In Figure 17, the response on the solar arrays, is 3.3228g RMS higher when coherent columns of acoustic devices are used. It was observed that measured responses would correlate to the impact-modal driven predictions when similarly diffuse excitation was achieved, and columns of coherent devices produce results which cannot be as accurately correlated to these predictions.

An example of this is the 148 Hz mode, which is shifted slightly from the previously recorded mode in the survey at 146 Hz. In both 32 and 16 drive configurations, this mode is accurately excited. In the column-coherent configuration, where the entire column of acoustic devices acts as a unified compression wave, nodal and anti-nodal ‘hotspots’ in the field seem to cause excitation variations at very specific points in the field.

Figure 18 – 148 Hz mode Excited by Column Coherent Drive (Left) vs. Incoherent Drive (Right)

In Figure 18, the column coherent method of excitation (left) leaves the mode relatively unexcited in the solar arrays, but well excited in the structure, while the incoherent method seems to equally excite both. The mode prediction in the modal survey was important in this comparison, as the mode exists and can be seen in the data presented in Figure 14- Solar Array vs. Acoustic Excitation Modes. It is possible that the planar nature of the stack-coherent operation creates an anti-node in the areas where the solar array models occupy, but more research is needed, as there were no microphones directly occupying this area.

In lieu of additional simulation or follow-up research to confirm this theory, the data was reprocessed to look at coherence both horizontally around the circle and vertically as well. In Figure 19, the leftmost plot shows the data from eight control microphones located horizontally around the incoherent (32 drive) configuration, processing one control microphone for each of the eight columns. This is typical of measurements from legacy DFAN tests which aim to prove that coherent concert speakers are sufficiently incoherent from stack-to-stack using drive mixing matrices to achieve full level. The plot on the right includes all (32) microphones, four per each of the eight columns. This plot shows a naturally higher coherence between vertically spaced pairs, but is still well within expectation of incoherent drives with the bulk of the energy less than 50% coherent.

Figure 19 – Coherence Lateral Controls vs All Controls, 32 Drive

Figure 20 plots the same control microphones as the previous figure, but for the 8-drive, column coherent configuration. The difference between the horizontally configured control microphones (left) and the microphones that cover all vertical and horizontal locations is stark, especially in the lower frequencies where larger structures are especially susceptible to standing waves.

Figure 20 – Coherence Lateral controls vs All Controls, 8 Drive

Conclusions

Though further research is warranted, this study seems to validate the authors previously published claims that maximal incoherence creates the most predictable and diffuse field possible. (Schick, et al., 2021) Comparisons between the (32) octave band control microphone locations and the (12) legacy placed microphones (for MIMO narrowband control) saw a tighter grouping in the latter in all configurations and tests where the incoherence was highest. This contrasts with narrowband control methods that require phase alignment between control locations, which can create certain field conditions at those points, but are unable to confirm such conditions elsewhere in the field and do not typically offer the benefit of monitoring microphones.

The variation in excitation where column coherent drive is used is a particular concern of the authors and seems to highlight the pitfalls of this type of control method. This type of configuration can be expected to produce results which vary significantly across the field in ways that cannot be accounted for and should be avoided whenever possible. Since it is impossible to populate the space occupied by the test article with microphones to verify uniformity of field, the method that provides the most confidence without additional measurement should be used.

In conclusion, it is suggested that field uniformity and diffusivity are key goals for a successful test because they allow the end-user to predict the structural response more accurately for any sized test article. Since the acoustic field is not mechanically coupled to this mechanically analyzed test, continuing real-world testing and qualification of commercial payloads for launch have shown that, for practical purposes, it is more important for the field to be uniform and predictable than to attempt to simulate a payload fairing environment or arbitrarily limit the field to conditions to emulate the performance characteristics of legacy reverberant chambers. Furthermore, since many launch vehicles are often enveloped into acoustic validation tests, modeling fairings becomes a complex and futile exercise, which can deviate from the original goal of an acoustic test – to provide a uniform, predictable, and repeatable acoustic field.

Biographies

Though further research is warranted, this study seems to validate the authors previously published claims that maximal incoherence creates the most predictable and diffuse field possible. (Schick, et al., 2021) Comparisons between the (32) octave band control microphone locations and the (12) legacy placed microphones (for MIMO narrowband control) saw a tighter grouping in the latter in all configurations and tests where the incoherence was highest. This contrasts with narrowband control methods that require phase alignment between control locations, which can create certain field conditions at those points, but are unable to confirm such conditions elsewhere in the field and do not typically offer the benefit of monitoring microphones.

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.

Thomas Hoffmann is global 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 completed 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.

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 and aerospace applications. Ed was educated originally at Charterhouse in Surrey, and then at St John’s College, Cambridge, where he graduated with honors.

Dale Schick is an Application Manager for m+p International Inc. He has experience in environmental testing and research as applied to Aerospace and over 14 years of practical experience in the industry.

Bibliography

American National Standard, 2004. ANSI S1.11 Specification for Octave-band and Fractional- Octave-Band Analog and Digital Filters. Melville: Acoustical Society of America.

Friedlander, E., Schick, D. & Hoffman, T., 2022. Qualification of DFAN Field & Acoustic Control Methods. Annapolis, MD, AIAA, pp. 5-12.

Friedlander, E., Schick, D., Hoffman, T. & Shorter, P., 2023. Validation of Novel Direct Field Acoustic Test Workflow by Simulation and Measurement. Toulouse, France, ECCSMET, pp. 3-4.

Kolaini, A. R., Doty, B. & Chang, Z., 2012. Impact of Acoustic Standing Waves on Structural Responses: Reverberant Acoustic, Pasadena: Jet Propulsion Laboratory, California Institute of Technology.

Larkin, P., 2014. Developments in Direct Field Acoustic Testing. Sound & Vibration, November, pp. 7-8.

Roe, R. R. J., 2016. NASA-HDBK-7010. s.l.:NASA.

Schick, D., Friedlander, E., Hoffman, T. & Kinsella, E., 2021. Verifying a Novel Approach to Acoustic Testing: Acoustic Devices & Control Systems. Virtual, s.n.

Underwood, M. A. P., 2021. Comparisons of the Structural Response of a Test Article Excited by

DFAT(TM) Diffuse and Non-Diffuse Acoustic Fields. Orlando, s.n.

 

Read Full Article

Explore More

╱╲