• Author / Uploaded
• Eric Moskus

Citation preview

The Black Box Eric Moskus & Nathan Sibon, Zach Bruin, Josh Roberts, Chris Nottoli, Melissa King November 26th, 2013 Acoustical Testing I: Dr. Dominique Chéenne Dr. Lauren Ronsse

1

Abstract: The objective of The Black Box was to accurately describe the components contained inside of a locked case, using any available equipment. The case itself contained two inputs (labeled “C” and “D”) and two outputs (labeled “A” and “B”). It was found that between input “C” and output “B” was a graphic equalizer, a downward expander with a 4:1 ratio, and a compressor/limiter with an ∞:1 ratio and a threshold at -20 dBu. Output “A” was configured to act as an antenna, introducing radio frequency (RF) interference and noise into that output. Input “D” was entirely disconnected from the signal chain within the case.

Introduction: The Black Box was the last of three projects to be completed for the Acoustical Testing I course at Columbia College Chicago. The aim of the project was to accurately describe the components contained inside of a locked case (hereafter referred to as, “the black box”).

Initial Listening Tests: The black box (pictured in Fig. 1) has two inputs (“C” and “D”) and two outputs (“A” and “B”). Initial listening tests were conducted to determine the fundamental connections between each input and output. Various signals (sine tones, white noise, etc.) were generated from a TerraSonde Audio Toolbox (pictured in Fig. A-1 in Appendix A) and routed into the black box input(s) via an XLR cable. The signal was then routed from the black box output(s) into a QSC K8 loudspeaker, so that the output(s) of the black box could be actively monitored. This process was repeated for every combination of inputs and outputs (e.g. input “C” to output “B”, input “D” and “C” to output “B”, etc.) to determine which pathways allowed signal to pass into the loudspeaker.

2

Fig. 1: Image of the black box with inputs "C" and "D" and outputs "A" and "B". The power supply is located on the left.

These listening tests revealed that the only working connection was between input “C” and output “B”. A subsequent listening test using music as a source also revealed what sounded like a noise gate between input “C” and output “B”. Only the transients and loudest sections of the music were output from the loudspeaker. Regardless of input, the output of “A” was only noise, and the original signal being sent in from the loudspeaker was lost. It was noticed, however, that a sine tone played through input “C” could faintly be heard over the noisy output of “A”. This only occurred using high output levels of the Audio Toolbox. This was believed at the time to be an effect of crosstalk between the channels. When the only input used was “D” no signal could be heard out of “B”, and “A” only output its characteristic noise. From the listening tests it was believed input “D” was not connected to the signal chain within the black box.

3

Output “A”: Initial listening tests demonstrated that the output of “A” was noise. To better understand the quality of that noise, a series of nine different frequency response graphs were constructed from input “C” and “D” to output “A”, using the TEF20 unit and Sound Lab software. Each frequency response test was performed at a slightly higher output level than the previous, so that the variation of frequency response over amplitude could be observed. Fig. 2 shows the overlay of these nine separate frequency response graphs.

Fig. 2: Pictured is an overlay of nine separate frequency response graphs over varying amplitude, from inputs "A" and "D" to output "A". Created using the TEF20 unit and Sound Lab software. The output knob on the TEF20 unit was used to increase the amplitude for each of the nine tests. Each frequency response was then overlayed to create this graph. The imprecise method of turning the knob means that the difference in amplitude for each test is different. Amplitude levels on the y-axis of graph were not calibrated and can only be used in relative terms. Overall, the graph shows little variation in frequency response with changing amplitude. It also shows several distinct peaks in frequency response (60 Hz, 180 Hz, etc.). Parameters used during these tests can be seen in Table B-1 in Appendix B.

4

The small variation in frequency response for varying amplitudes that is seen in Fig. 2 indicates that output “A” is largely independent of input level. It was also believed that the frequency response peaks (beginning with a peak around 60 Hz) may have been a result of ground hum or RF interference from components in the surrounding room. The crosstalk-effect heard in the listening test also indicated interference entering output “A”. Subsequently, a small radio transmitter (pictured in Fig. A-4 in Appendix A) was brought in to see if the noise observed on output “A” could be increased or altered when RF was transmitted into the black box. The signal generator (pictured in Fig. A-3 in Appendix A) was connected directly to the XLR input of the radio transmitter, so that sine tones could be transmitted from the radio transmitter antenna into output “A” of the black box. Output “A” was monitored in real-time through SpectraPLUS. The signal flow diagram of this setup is shown in Fig. 3.

Fig. 3: Signal chain used during the radio transmitter tests. It represents two physically isolated signal chains. However, the antenna of the radio transmitter was placed almost directly against the black box.

When the signal generator was set at maximum output, the sine tones being transmitted via the radio transmitter could be seen being imposed on the output of “A”. It was shown that RF signals could be broadcast and demodulated on output “A”, proving that it was acting as an antenna, and was constantly picking up RF interference on its channel. Fig. 4 shows the transmitted signal being imposed on the noise of output “A”.

5

Fig. 4: This graph shows a spectral analysis of output “A” while sine tones were being transmitted from the nearby radio transmitter. The axes are not shown on the graph. Frequency is represented in the x-axis, time in the y-axis, and amplitude is represented by color. The green line to the right of the arrow represents the output of the signal generator being transmitted via RF into the output of “A”. The output frequency of the signal generator was manually oscillated during this test, hence the curvature of the line.

Input “D”: Initial listening tests using the QSC K8 indicated the possibility that input “D” was not connected to the signal chain within the box. To verify this, signal was output of the Audio Toolbox into input “D” and out of output “B”. Output “B” was then connected to a millivoltmeter (pictured in A-1 in Appendix A). The millivoltmeter registered zero voltage coming from the black box's output, even despite increased input signal level. Thus, it was determined that input “D” was completely disconnected from the signal chain within the black box.

6

From Input “C” to Output “B”: Frequency Response: Using a TEF20 unit (see Fig. A-1 in Appendix A) and the accompanying SoundLab software, a series of thirteen different frequency response graphs were constructed from input “C” to output “D”. After each frequency response graph was generated the output level on the TEF20 unit was slightly increased in the same manner as that used to construct Fig. 2. The overlay of all these frequency response graphs can be seen in Fig. 3.

Fig. 3: Pictured is an overlay of thirteen frequency response graphs over varying amplitude from input "C" to output "B". Created using the TEF20 unit and Sound Lab software. The output knob on the TEF20 unit was used to incrementally increase the amplitude for each test. Each frequency response test was then overlayed into one graph. The imprecise method of turning the knob means that the difference in amplitude for each test is different. Amplitude levels on the y-axis of graph were not calibrated and can only be used in relative terms.

7

Fig. 3 shows all the characteristic frequency response peaks and dips of an equalizer. It was also determined that each peak and dip corresponded closely with ⅓-octave band center frequencies (e.g. 100 Hz, 125 Hz, 160 Hz). Graphic equalizers contain center frequencies corresponding to ⅓-octave band center frequencies. Thus, it was determined that there was a graphic equalizer between input “C” and output “B”, and that its shape matches the shape of the equalization seen in Fig. 3.

Transfer Functions: Fig. 3 also shows a flattening of frequency response at higher output levels, which is characteristic of a compressor/limiter. In order to attain a more precise view of what was happening to the signal with increased amplitude, transfer function graphs were constructed. The signal generator was used as the sound source and the signal through the black box, from input “C” to output “B”, was monitored on the millivoltmeter. The reference signal directly from the signal generator was also monitored on the millivoltmeter. The full signal flow for the test is shown in Fig. 4, and an image of the equipment used during these tests can be seen in Fig. A-2 in Appendix A.

Fig. 4: This is the signal flow used during construction of the transfer function graph seen in Fig. 5. The level directly from the signal generator was monitored on the millivoltmeter. The signal through input "C" and output"A" was also monitored on the millivoltmeter.

8

To construct the transfer function graphs, the level of the direct, reference signal from the signal generator was incrementally increased by .5 dBV, and the corresponding voltage of the signal through the black box was recorded at each increment. These points were then plotted and the process was repeated for a total of eight different frequencies. They are all overlayed and graphed in Fig. 5.

Fig. 5: The graph above is a comparison of input level (in dBV) to output level (in dBV) for eight different frequencies being run through input "C" and output "B". The signal generator was used as the sound source, and the millivoltmeter was used to record levels in dBV. To construct the graph, the level of the direct, reference signal from the signal generator was incrementally increased by .5 dBV, and the corresponding voltage of the signal through the black box was recorded at each increment. These points were then plotted, and the process was repeated for several different frequencies. The straight red line on the graph represents a linear, 1:1 response, which was simply used a reference. ∆x between different frequencies was determined to be an effect of the graphic equalizer.

The transfer function graph in Fig. 5 shows the amplitude-flattening effect of the compressor/limiter on signals at higher input levels. It was determined that the compressor was acting as a limiter with an ∞:1 ratio, because regardless of output level the signal would not rise above the compressor/limiter threshold. The threshold of the limiter was determined to be around -29 dBV (or -26.8 dBu), as that was considered the average input level at which the compressor began to work 9

on the different frequencies. It was also noticed that low frequencies appeared to be subject to a higher limiter threshold than that for higher frequencies. For instance, Fig. 5 shows that the limiter flattens high frequencies to around -30 dBV output, while a low frequency, such as 63 Hz, flattens closer to -20 dBV output. This phenomenon can also be seen in the graph shown in Fig. 3, where the threshold of the limiter can be seen to rise at lower frequencies. It was believed that the only explanation for this phenomenon was a shelving equalizer, coming after both the limiter and the graphic equalizer in the signal chain. This was believed to be the only way in which low frequencies could be boosted above the limiter threshold. Unexpectedly, Fig. 5 also helped explain why it sounded as if there was a noise gate between input “C” and output “B” during the listening tests. A consistent slope of four was calculated for the different frequencies, as they rise from the noise floor to the point where they hit the limiter threshold (as seen in Fig. 5). A downward expander, working at a 1:4 ratio, explains the consistent slope of the lines in Fig. 5, as well as the sound of the noise gate during listening tests. It was also determined that the downward expander threshold must be at or above the compressor threshold, because the downward expander works on the signal at every point between the noise floor and limiter threshold.

Signal Flow: The signal flow of components in the black box was confined to being between input “C” and output “B”. This is because output “A” was determined to be acting as an antenna, and input “D” was determined to be entirely disconnected from the signal flow. The graphic equalizer was determined to be the first component between input “C” and output “B”. If the graphic equalizer came after the downward expander then the lines of slope four seen in Fig. 5 would not have been consistent. If the graphic equalizer came after the limiter then the equalization should retain its shape after the limiter

10

threshold is reached, though Fig. 3 shows that this is not the case. It was also determined that a shelving EQ was the last component in the signal chain because it was believed to have raised low frequency content over the limiter threshold. The order between the downward expander and limiter could not be determined. The actual signal found inside the black box is shown in Fig. 6.

Fig. 6: This signal flow chart shows the actual component flow inside of the black box. This was determined exactly only after opening the black box.

Fig. 6 shows that there was no shelving equalization after the limiter between input “C” and output “B”. The higher limiter threshold at low frequencies (as observed in Fig. 3 and Fig. 5) was actually due to an impedance mismatch between systems, and could have been remedied with the use of a resistor in the signal path. As predicted, the shape of the graphic equalizer resembled the shape of the equalization seen in Fig. 3. The ratio's for both the downward expander and the limiter were accurately found. However, the actual threshold of the limiter (-20 dBu) was slightly off the estimated (-26.8 dBu) limiter threshold. All other components were accurately found.

11

Conclusion: Through a series of various tests, the components inside of the black box, as well as the basic settings for those components, were found. It was determined that between input “C” and output “B” was a graphic equalizer, a downward expander with a 4:1 ratio, and a compressor/limiter with an ∞:1 ratio and a threshold at -20 dBu. Output “A” was configured to act as an antenna, introducing radio frequency (RF) interference and noise into that output. Input “D” was entirely disconnected from the signal chain within the case. Several extra tests could have been performed to attain greater accuracy in results. If the electrical impedance of the devices had been considered then the mistaken claim of a shelving equalizer could have been avoided, and the low frequency swell in response at high amplitudes could have been negated. In turn, this would have allowed the threshold of the compressor to be calculated more accurately. The attack time of the limiter could also be calculated by sending impulses of various lengths into the black box. If an impulse bypasses the limiter then its duration is shorter than the attack time of the compressor. The length of the impulse could then be increased until it is caught by the limiter, and this length would be the attack time of the limiter.

12

Appendices Appendix A – Pictures of Setup and Equipment

Fig. A-1: Pictured is the TerraSonde Audio Toolbox, which was used to generate different types of signals (sine wave, white noise, etc.) for various tests. The TEF20 unit is also pictured. It was used to generate the frequency response graphs seen in Fig. 2 and Fig. 3.

Fig. A-2: Pictured is the signal generator and millivoltmeter used primarily in constructing the transfer function graphs seen in Fig. 5. The red line on the millivoltmeter (on the left) represents the level of the the 1kHz tone through the box while the black line on the millivoltmeter (on the right) represents the dry, reference signal directly from the signal generator.

13

Fig. A-3: This is a picture of the radio transmitter used to transmit sine tones into output "A". The transmitter unit utilizes an XLR input, and broadcasts on F.M. 87.3.

Appendix B – Frequency Response Parameters

Frequency Response Start Frequency: Stop Frequency: Sweep Time: Sweep Rate: Frequency Resolution: Time Resolution: Distance Resolution: Bandwidth: Number of Samples:

Parameters 20 Hz 20,000 Hz 59.733 s 334.5 Hz/s 18.3 Hz 54.68 ms 61.79 feet 18.3 Hz 8192

Table B-1: Frequency response parameters used to construct the frequency response graphs seen in Fig. 2 and Fig. 3. All parameters were kept the same for each test, and all tests performed with the TEF20 unit in the Sound Lab software.

14