The Speaker Study. By: Jay Bliefnick. Acoustical Testing 1. Attn: Dr. Dominique Chéenne, Dr. Lauren Ronsse. Group Members:

Transcription

1 The Speaker Study By: Jay Bliefnick Acoustical Testing 1 Attn: Dr. Dominique Chéenne, Dr. Lauren Ronsse Group Members: Hannah Knorr, Michael Hanson, Matt Johnson, Miles Possing, & Ming Yu 11/27/13

2 Table of Contents Abstract... 3 Introduction. 3 Anechoic Testing Setup....4 Non-Anechoic Testing Setup... 9 Frequency Response Analysis Crossover Frequency Analysis..12 Time Response Analysis. 14 Polar Response Analysis. 15 Non-Anechoic Response Comparison...16 Earthworks Microphone Comparison.. 19 Conclusion

3 Abstract In the speaker study, students were required to thoroughly analyze a speaker in an anechoic environment. An Event 8XL reference monitor was selected as the test speaker, and all tests were conducted using the Goldline TEF-20 system and an Electrovoice RE-55 microphone. The frequency response of the speaker was measured in a variety of orientations and frequency resolutions to determine an accurate reading across the entire spectrum. The crossover point between the two components was identified, and the time response was tested at the highest time resolution TEF allowed. The full-frequency polar response was also found using an automated turntable. The speaker s frequency and time responses were also taken outside of an anechoic environment to compare the two sound fields. Finally, the anechoic frequency and time response tests were completed using a second microphone: an Earthworks M-50 condenser. Through these tests, the performance of this speaker was verified across a wide variety of conditions. Introduction The goal of the speaker study was to determine as much about a selected test speaker as possible, whether this be parameters or possible test conditions. The first round of data was collected in the anechoic chamber (Figure 1) at Columbia College, which provided a reflectionfree environment in which to test. This allowed for direct analysis of the test speaker with as little room interaction as possible. Figure 1: Anechoic Chamber at Columbia College. The speaker was placed in the approximate center of the chamber with the microphone placed on-axis, 1 m away. This was the first speaker-microphone orientation utilized. 3

4 The anechoic chamber was actually recently remodeled to repair flood damage caused several years ago. Over the summer, the temporary ceiling absorption was removed, and in its place three layers of Roxul sound-rated insulation were installed. Two layers of RHT-80 (the densest panels) and one layer of RHT-60 were fastened to the ceiling, with all side areas being filled as well. The final layer of RHT-40 (least dense) was suspended ~12 lower on a wire support system. The finished ceiling not only looked better, but also provided a much more even frequency response in the room, without the extreme notch at 1 khz seen before the new construction (caused by the previous wedge-style absorption panels). The one issue that still remained with the chamber was the relatively small dimensions of x 10 2 x 8 1 (Volume = 1137 ft 3 ), which created some modal influence. With the anechoic chamber providing an accurate method of isolating the speaker s properties, it was important to research the comparative responses in a non-anechoic environment. The open area in the LL01 classroom provided the necessary conditions for these tests. The frequency response was taken and compared to show how the influence of the room (reverberation) affected the results. The time response was also tested, utilizing multiple reflection panels. These created peaks which could be measured and quantified into distances, then compared with the anechoic results. Finally, the anechoic tests were run once again using a second microphone to compare the two and confirm the response of the speaker. Anechoic Testing Setup The first round of testing was conducted in the aforementioned anechoic chamber. The Event 8XL studio reference monitor was selected for study due to its consistent response, which could be expected to produce accurate results. The Electrovoice RE-55 was selected for use as the primary microphone. This dynamic, omni-directional microphone allowed for precise measurements to be taken without the need for phantom power. This combination of speaker 4

5 and microphone would be used for all tests, except for two additional tests taken at the end of the study with an Earthworks M-50 condenser microphone. In the anechoic chamber, four main features of the speaker were investigated: the frequency response, the crossover frequency, the time response, and the polar directivity pattern. These were completed under a variety of setup orientations and testing parameters to thoroughly analyze all potential issues. These tests were completed using the Goldline TEF-20 analysis system installed on workstation computer 2, adjacent the anechoic chamber. This was accomplished by using the multiple patch bays connecting the facility (Figure 2). Computer Running Station 2 Central Anechoic Event 8XL TEF 20 Software Patch Bay Patch Bay Snake EV RE-55 Figure 2: Signal chain used to connect the speaker and microphone in the anechoic chamber to the computer using the patching system. The process was simplified by using the same number for each step: input 1 for all microphone connections and output 3 for speaker connections. The first speaker attribute tested in the anechoic chamber was the frequency response. This was completed using three setup positions, two independent methods, and a variety of parameter settings. The microphone-speaker relationship remained constant for the tests at 1 m, on-axis with the speaker s center. This orientation was chosen due to it being the industry-wide standard. For the initial placement, the speaker was set in the center of the chamber facing the microphone parallel to the sidewalls. The second setup rotated the system clockwise approximately 30 degrees, whereas the final position placed the system about 30 degrees counterclockwise from the original arrangement (Figure 3). These tests provided a look into the influence the room had on the measurements and how much modal interaction was present. In the initial placement, the frequency response was gathered two separate ways: as a single test across the entire spectrum, and by breaking the range into multiple tests, compiling the graph piecemeal. To begin, a single sweep was run from 20 Hz 20 khz using a sweep time 5

6 of 180 seconds. This provided a frequency resolution of 18.6, which was good enough for an initial set of data, but was not accurate enough in the low end. To generate a frequency response with an adequate resolution for all ranges, the test was completed once more, only this time in octave intervals. By using this method, a high resolution could be maintained while limiting the amount of time required to run the tests. Overall, seven tests with approximately 2-minute sweep times were conducted in octave intervals (with the low frequency test accounting for several octaves). This method produced similar data to that gathered using the single sweep, but with a higher degree of accuracy. Finally, a long 5000-second single-sweep was attempted, which pushed the TEF system to its maximum frequency resolution of 2.0 Hz. Figure 3: Three speaker orientations used in the anechoic chamber. The multiple positions allowed the investigation of modal properties within the space. Using the second and third speaker-microphone positions, the long and short singlesweep tests were completed. Given the uniformity between the piecemeal response and the long sweep test conducted at position one, it was decided that the single-sweep tests would suffice. The responses were largely the same, especially above 500 Hz, but as expected, the chamber s modal properties and absorption created variances in the mid and low frequencies. The second property of the speaker investigated was the crossover frequency between the woofer and the tweeter. As the Event 8XL was a bi-amped powered speaker with a mono input, each component had its own frequency response. These would blend into a unified field after a 6

7 sufficient distance, but at close ranges, the two pieces could be isolated. The crossover was tested in two ways, generating somewhat contradictory results. The first method for testing the crossover frequency involved blocking the sound generated from one component and then testing the generating source. Each test was conducted 1 from each source, on-axis. To test the woofer, the tweeter was blocked with a layer of Auralex Acoustics SonoFiber absorber panels (Figure 4). To test the tweeter, the same process was repeated for covering the woofer using the same absorbers (Figure 5). The first test using this method demonstrated the under whelming performance of the absorbers under this application, so an additional layer was added underneath to improve the results. Figure 4: To test the woofer, the tweeter was blocked using an Auralex Acoustics Sonofiber absorber panel. Figure 5: The same process was used to test the tweeter. Additional layers of absorption were added beneath the visible Auralex panel. The techniques of testing the two components worked as intended, and accurate fullfrequency responses were taken of both. It was found that the crossover range was 1 khz 5 khz (generous approximation), so those frequencies were retested at a higher resolution (5 Hz) for the woofer and tweeter. Their graphs were overlaid to see where the inflection point was. The one issue with this process was the fact that the two components had different sensitivities, creating level variances at the 1 testing distance. Due to these factors, the crossover frequency could only be narrowed to between 2 khz and 3 khz. 7

8 The second method of testing the crossover frequency did not involve any form of absorption or any isolating materials. Instead, the microphone was place 1 from the face of the speaker, on-axis, looking directly between the woofer and the tweeter. The same, limited frequency range test was conducted and demonstrated interesting results. The crossover frequency appeared to be lower, at approximately 1.8 khz utilizing this process. With the frequency response and crossover frequency tests complete, the next parameter to study was the time response of the speaker. Speaker orientation one was utilized for time response tests. The most important factor in obtaining accurate time response data was to maximize the time resolution within the TEF system. To achieve this level of precision in the time domain, resolution in the frequency domain needed to be sacrificed. Therefore, the time response tests were conducted encompassing the full 20 Hz 20 khz range. The sweep time was also limited to.77 seconds to improve results. The tests were run with two sweep time/sample amount parameter settings. The first test had a sweep time of 10.9 s using 512 samples, and the second had a sweep time of 175 ms with 8192 samples. These produced identical results, which was expected considering: Time Resolution = Sweep Time Total Samples For both examples, the time resolution of the system was 21.3 μs, the maximum allowable resolution in the TEF system. The effect of the speaker and microphone stand were also investigated utilizing absorption coverings. They were seen to have negligible influence. The final speaker parameter tested within the anechoic chamber was the polar directivity response. This was accomplished using the automated turntable (seen clearly in Figure 4), which was operated through TEF. The test consisted of multiple frequency response tests (as performed before), conducted at a selected rotation degree step, to obtain a full 360-degree speaker response. The time length and sample size of each sweep could be chosen, which would 8

9 determine the frequency resolution for each test, and thus the accuracy of the end result. The rotation step size could also be selected: from 22.5 degrees, down to a tiny 2.5 degrees. Both of these parameter selections could improve the resolution of the eventual polar response, but the sacrifice was the time required to complete the full test. As a compromise to achieve valid data with a relatively short test cycle, a sweep length of 300 seconds per test with a 15-degree step size was chosen. This resulted in a total test time of approximately 2.5 hours. The process went through without a hitch, but considering concurrent tests were being conducted in the adjacent Real World Transmission Loss Chamber, accuracy needed to be verified. An identical test was run after hours, while the building was empty, and showed nearly indistinguishable results. Once reliable data was gathered from the two polar tests, a third was conducted utilizing a much higher frequency resolution and smaller step size. The sweep time for each test was raised to 600 seconds, while the rotation step size was decreased to 10 degrees. These changes inflated the overall test time to more than 6.5 hours. The analysis was again run while no one was utilizing LL01, ensuring isolation. Eventually the process was complete and an even more accurate polar response was obtained. Non-Anechoic Testing Setup In addition to testing the speaker within the anechoic chamber, the response from a nonanechoic environment was needed to provide comparison. The open area in the LL01 classroom served as the testing venue for this analysis. This was made possible due to the incredibly high signal to noise ratio within TEF, allowing accurate measurements even in noisy environments. The speaker was set upon the same stand and the Electrovoice RE-55 was set up as before. First, the frequency response of the speaker was tested using the single-sweep method. Utilizing a sweep length of 300 seconds, a frequency resolution of 11 Hz was achieved: accurate 9

10 enough for this test. The response graph was similar in general shape, but with the addition of the room interaction, it was obvious the speaker was no longer isolated. The time response was the other parameter tested in the non-anechoic environment. The same parameters were used as before to ensure the highest time resolution possible. The first test was conducted in the exact setup as the frequency test. The time response resembled the anechoic results, save for a second peak seen after the initial decay. It was suspected that this peak was caused by a reflection off of the floor (the only close surface). To test this theory, absorption was applied to the floor between the speaker and microphone (Figure 7). Using the same parameters, the second peak disappeared, confirming the reflection from the floor. To push these results further, another reflecting surface was added into the equation, this time in the form of a moveable solid panel (Figure 6). This time three tests were completed: no absorption, absorption on the floor, and absorption on the reflector panel. With no absorption, two additional peaks were seen after the direct sound. When the absorption was applied to both the floor and the panel, their respective peaks on the time response graph disappeared. This confirmed the hypothesis that the time response peaks were due to close reflecting surfaces. The distances of these components were then measured and compared against the computed values. Figure 6: Non-anechoic speaker tests with reflector panel. Frequency response and initial time response conducted without the panel. Figure 7: To confirm the reflection source, absorption was added to the floor (and eventually the reflector panel), which eliminated the peaks. 10

11 Frequency Response Analysis The frequency response was computed in a variety of ways, and in three separate orientations. In general, all graphs were quite similar, especially above 500 Hz, but some variances could be seen in the lower registers. The first position was tested in two different manners: as a single sweep and piecemeal using multiple individual sweeps (Figures 8 & 9). Figure 8: Frequency response of the speaker in position 1 using the piecemeal method. The different colors represent the individual tests conducted in octave bands. The frequency resolution was much higher than had a single sweep been run, given the time spent on the tests. Figure 9: Frequency response of the speaker in position 1 using the single sweep method. The maximum TEF sweep time was used to acquire the highest frequency resolution. The two responses appear identical. 11

12 Once accurate data was confirmed for speaker position one, the setup was rotated and retested in positions two and three. By placing the system in different orientations within the anechoic chamber, the influence from the room itself could be ascertained using the same second single sweep method. Figures 10 and 11 demonstrate the variance between the initial positioning. Test setup two showed general similarities to the first graph, but the peaks/troughs from 300 Hz to 500 Hz were not congruent. Test setup three, however, displayed extreme differences between the placements within the chamber. In this position, the microphone was the closest to the sidewall, which could have contributed to the altered response. Figure 10: Anechoic frequency test position 2. Figure 11: Anechoic frequency test position 3. Crossover Frequency Analysis The crossover frequency of the speaker was determined between the tweeter and the woofer in two separate manners: by physically isolating each component and by testing directly between the two. These methods produced two different results for the possible crossover frequency, which needed explanation. By isolating the components of the speaker, the individual frequency response graphs from the tweeter and woofer could be found (Figures 12 & 13). These displayed a region of 1 khz 5 khz where the woofer began to decrease in power and the tweeter began to increase. 12

13 This region was further investigated by testing this specific frequency range at a higher resolution, and then plotted on the same graph. This resulted in a potential crossover range of 2 khz 3 khz. The exact frequency could not be narrowed further due to the relative levels of the two components: it was impossible to plot the two responses equally (Figure 14). Woofer Response Tweeter Response Figure 12: Isolated woofer response. Figure 13: Isolated tweeter response. Figure 14: Combined tweeter Lower graph was with and woofer response. additional layers of absorption. The second method for testing the crossover frequency placed the microphone directly between the tweeter and the woofer, 1 from the speaker plane. By selecting this condition, the relative levels of the components could be seen simultaneously (Figure 15). They showed an obvious dip in the frequency response of ~20 db at 1.8 khz. This was well below the values the previous tests generated. The phase relationship was also plotted over the range, and it displayed a shift at the exact same frequency. It would seem that this solved the crossover frequency question. However, by conducting research online, the manual for the Event 8XL was discovered, revealing the published crossover value of 2.6 khz. This value confirms the first crossover tests, but did not answer the issues raised by the second test. One possible explanation could be that 1.8 khz was where the two components cancelled each other out due to phase 13

14 interference. On either side of this frequency, either the tweeter or woofer would dominate, limiting the phase problems, but at that frequency, a depression was formed. Considering the fact that the tweeter had a higher sensitivity, by this reasoning it would make sense that the crossover frequency was raised above 1.8 khz. Frequency Response Phase Response Figure 15: Crossover frequency test with microphone placed 1 from speaker plane between the tweeter and the woofer. The white line was the frequency response; the red line was the phase response. The graph shows the extreme dip at 1.8 khz as well as the phase shift at the same frequency. Time Response Analysis The time response of the speaker was tested in the anechoic chamber to determine the maximum time resolution within the TEF-20 system, and to provide an adequate baseline to compare the non-anechoic tests against. A time resolution of 21.3 μs was achieved by using a full-frequency range, a sweep time of 175 ms and 8192 total samples (Figure 16). More tests were conducted, but the time response graphs were identical. The tests showed an even decay pattern, after the reception of the direct sound at 3.1 ms. This equated to 1.06 meters, based on a 343 m/s speed of sound. Beyond the initial decay, the time response remained consistently low, as expected. This demonstrated that there were no sources of reflections within the anechoic chamber that influenced time response results. 14

15 Figure 16: Anechoic chamber time response. The smooth decay curve was due to the reflection-free environment. Polar Response Analysis The polar response of the speaker was collected over the course of three separate tests. The first two were conducted using identical testing parameters: 300 s sweep times with a rotation step size of 15 degrees. For the third test, the resolution was increased for both the sweeps and the steps: 600 s sweep times and 10-degree step sizes were used. Given the higher accuracy, the third test was primarily used to evaluate the polar response (Figure 17). Below 250 Hz, the speaker was very omni-directional: showing less than a 10 db difference between 0 and 180 degree orientations. At 56 Hz, the level of the speaker response began diminishing, as the frequency graph predicted. Between 500 Hz and 1 khz, the rearward projection remained strong, but lobes began to form, indicating discrepancies in the response. Above 2 khz, the speaker became much more directional, as the difference between 0 and 90 degrees was greater than 10 db. The rear response was also substantially decreased with even more pattern deviations. As the frequency increased, the directivity of the speaker became even more pronounced, with the response at 16 khz being 20 db greater on-axis than off. This narrowing of the polar response was attributed to the tweeter becoming the dominant source component, as they are generally more directional than woofers. 15

16 Figure 17: Polar response of Event 8XL in an anechoic environment. Polar test 3 was plotted (highest resolutions) in octave bands across the entire spectrum. Non-Anechoic Response Comparison With the speaker thoroughly studied under anechoic conditions, the next step was to test the frequency and time responses under free-field conditions. Because the crossover frequency was determined using close microphone techniques, a re-test was unnecessary, and the polar response was wired to only be completed in the anechoic chamber. The frequency response was the first test to be run, setup with no modifications to the test environment. There was no absorption placed near the test, and the additional reflector panel was set to the side. Using a sweep time of 300 seconds, a frequency resolution of 11 Hz was achieved. The graph displayed the same basic response pattern as before, but with significantly more deviations (Figure 18). 16

17 This could be attributed to the influence of the room and the pattern of reflections that occurred under non-anechoic conditions. The reverberation characteristics of the space were mixed with the speaker s response resulting in a very disruptive graph, as opposed to the much smoother response gathered under more suitable test conditions (Figure 9). Figure 18: Frequency response taken in LL01 under non-anechoic conditions. The much more aggressive pattern could be attributed to the reverberation characteristics of the space. The time response under these test parameters was also gathered to determine if the source of these reflections. The same sweep time, time span, and sample size values were utilized as before, ensuring the highest time resolution possible. The first test was conducted as the frequency response was: with no absorption or reflector panel. The second test included absorption placed on the floor directly between the speaker and the microphone. Both graphs displayed the direct sound at 3.1 ms and initial decay, but only the test without the absorption displayed the second peak at 6.91 ms (Figure 19). By extrapolating these two numbers out, the distance for the direct sound was 1.06 m and the first reflection was 2.37 m. Using a tape measure, the distance from the front of the speaker to the front of the microphone was 1.01 m, and the path reflecting off of the floor was approximately 2.34 m. These values confirmed the source of the first reflection as the floor between the speaker and microphone. 17

18 No Absorption With Absorption Figure 19: Time response taken in LL01 under non-anechoic conditions. The blue line was with no absorption present, whereas the red line was with absorption between the speaker and microphone. To further validate the source of the reflections as seen on the time response graph, a reflector panel was set up adjacent to the speaker and microphone. It was placed parallel to the system, slightly more than a meter away from the direct sound path, which was expected to create a second peak slightly after the first reflection. Three tests were completed under these conditions: no absorption, absorption on the floor, and absorption on the panel (Figure 20). As before, with the absorption in place, the expected reflection peaks were eliminated. No Absorption Absorption on Floor Absorption on Panel Figure 20: Time response taken in LL01 under non-anechoic conditions. The black line was with no absorption present, the blue line was with absorption on the floor, and the orange line had absorption on the panel. 18

19 Earthworks Microphone Comparison To compare the response of more than just the speaker, a second microphone was employed to see if differences could be found. An Earthworks M-50 condenser microphone was chosen due to its exceptionally flat frequency response. Because it was a condenser, though, it needed phantom power to operate, so the Aphex preamplifier, located in the equipment rack, was hooked up. The one issue with the Earthworks was the small capsule size, because there was no proper way to properly calibrate the microphone. A standard 94 db at 1 khz calibrator was put to use, but since much of the sound was escaping due to the larger capsule slot, the microphone was not correctly set, resulting in lower than expected values. Regardless, the frequency and time responses were both taken for position one in the anechoic chamber. A 90 second sweep time was selected for the frequency response, resulting in a 37.5 Hz resolution (Figure 21). The graph showed similar features to the Electrovoice tests, but with an even smoother curve. The regions above 10 khz and below 70 Hz were especially different: the Earthworks displayed much stronger results. Figure 21: Frequency response of the speaker in position 1 in the anechoic chamber using the Earthworks M-50 microphone. This graphs displays many similarities, but is much smoother in the lows and highs. The time response was once again conducted with the highest time resolution allowed in TEF. This time, the response of the speaker was significantly quicker than ones previously taken 19

20 (Figure 22). It only took about 1.5 ms for the direct sound decay into the nominal level, as opposed to more than 2.5 ms for the Electrovoice microphone. This was likely due to the quicker transient response of the Earthworks, due the fact that it was a condenser. Figure 22: Time response of the speaker in position 1 in the anechoic chamber using the Earthworks M-50 microphone. The condenser microphone displayed a much quicker response than the dynamic. Conclusion After thoroughly testing the Event 8XL studio monitor, many parameters were ascertained: the frequency response, the crossover frequency, the time response, and the polar directivity pattern. All of these tests were conducted under anechoic conditions, but the frequency and time responses were also re-tested under non-anechoic conditions. The result was a full, accurate view of the performance of the speaker under a variety of testing situations. Although much was learned, further tests could reveal even more about the details of the speaker. The polar and crossover tests could be completed using the Earthworks microphone, which could also be used in the non-anechoic environment. For the crossover, the microphone could be placed once again between the tweeter and woofer, but tested at several distances to determine if that factored into where the frequency dip and phase switch occurred. The function of the port could also be investigated by testing the response of it at varying distances. 20