Testing the pro’s audio
Costa Mesa, CA—You may have not heard of QSC Audio, but you’ve heard its sound. Amplifiers, digital audio processors, and loudspeakers from the company provide sound at private parties, movie theaters, conventions, concerts, and sporting events. QSC products bring the sound to venues such as the Sydney Opera House in Australia and the football stadium at the University of Alabama.
At QSC’s headquarters in Costa Mesa, engineers develop and evaluate the company’s products. They also provide test-engineering support for the company’s production line, where amplifiers and digital audio processors undergo many tests that you might expect to see only during development. The loudspeakers and the transducers inside those products also go through a series of tests in the company’s sound labs.
Test engineer Shaw Somarel supports production test of amplifiers and digital audio processors. He develops and maintains automated test stations that perform more than 100 measurements on each amplifier or digital audio processor. The test stations, like the rest of the line, must support many of the company’s electronic products.
The production line is a high-mix, medium-volume, build-to-order line. An amplifier, for example, starts as a bare PCB (printed-circuit board). A bar code identifies the board for parts placement, test, final assembly, and packing. Eric Andersen, QSC’s VP of manufacturing, explained that the line supports 84 different amplifier models without needing a setup for each board. All materials for all models are available at all times at the point of use. The networked assembly system downloads placement programs from a database as needed based on the bar codes.
When an amplifier board enters the production line, solder paste is silkscreened onto the PCB’s pads. A pick-and-place machine then assembles the surface-mount components on the board at about 10 parts/s. Next, through-hole machines add axial and radial leaded parts. Assemblers then manually add large capacitors and inductors, which are often needed in high-power amplifiers (QSC’s most powerful amplifiers produce 4000 W/channel on each of two channels).
Once a board has its surface-mount components, it goes through a reflow oven. “We design our boards so that they can all use a single reflow heat profile,” said Andersen. A wave-solder machine solders the through-hole parts to the board. During wave solder, fixtures hold the components in place to minimize flexing and lead stress. Following wave solder, assemblers add heat sinks to the power transistors. Assemblers then install a fully populated board into a chassis, which is then ready for test.
The first test an amplifier undergoes is a thermal test, which verifies that the amplifier’s thermal-shutdown circuits work. A fixture shorts the amplifier’s outputs while a container around the chassis restricts airflow. The amplifier heats until it reaches a set temperature point that activates the thermal-shutdown circuit, which mutes the amplifier’s outputs. As it cools below the set temperature point, the amplifier will recover and resume operating.
After the thermal test, the amplifier is ready for final test. There, an automated test station downloads the amplifier’s test procedure from a server on the company’s network. Somarel wrote the test station’s control software in Visual Basic .NET. He stores the test parameters in Excel spreadsheets.
An amplifier test station consists of a PC, a PLC (programmable logic controller), a Variac, a Fluke DMM (digital multimeter), a Fluke oscilloscope, an Audio Precision audio analyzer, and switch modules. The test station also contains power resistors capable of handling an amplifier’s power output. Figure 1 shows the power resistors mounted in the test rack. The PLC controls contactors for configuration of the load resistors and AC mains power. It also controls relays that perform audio signal routing to the amplifier under test. The load resistors connect to the amplifier’s output through a custom speaker-cable assembly.
Dissipating power as high as 4000 W requires several power resistors wired in parallel. Figure 2 shows how the power resistors form 8-Ω, 4-Ω, and 2-Ω loads used for testing both linear and switching (class D) amplifiers.
Each final test consists of more than 100 measurements such as frequency response, rated output power, THD (total harmonic distortion), SNR (signal-to-noise ratio), CMRR (common-mode rejection ratio), output signal phase, and output short-circuit current. The audio analyzer makes all the output signal measurements. It uses a proprietary communications bus that talks to a PCI card in the computer. The DMM (connected through GPIB) measures the AC mains voltage, the current supplying the amplifier, and the amplifier’s output current under load. During each test, the PC, which controls the Variac through an RS-232 connection, regulates the Variac’s output to the rated line voltage of the amplifier under test.
The test station has a 100-MHz analog oscilloscope that’s not under computer control. Technicians use it to adjust the bias spikes at an amplifier’s frequency crossover points. “We can probably use a digital oscilloscope,” said Somarel, “but analog oscilloscopes typically give a more detailed trace display.”
Somarel is developing a new amplifier test station for amplifiers that won’t need an oscilloscope. He plans to replace the current Audio Precision System One audio analyzer with an APx585 model that has an oscilloscope integrated into its application software. The new analyzer has eight channels as opposed to two channels, which will reduce test time because the analyzer will be able to test all of an amplifier’s channels at once (QSC Audio’s amplifiers have as many as eight channels). Somarel expects at least a 50% reduction in test time with the new test station.
Somarel has already integrated the eight-channel audio analyzer into test stations for QSC’s digital-audio products. One of the digital products, called Q-Sys, can control an entire audio system over Ethernet. Each Q-Sys controller can route digitized audio in a 512x512 matrix among amplifiers and audio sources such as microphones.
QSC’s digital-audio products send digitized 24-bit audio using CobraNet, an audio protocol that runs over 100-Mbps Ethernet. QSC’s engineers modified one of the company’s current digital products that converts the CobraNet audio into an AES (Audio Engineering Society) digital stream so that the audio analyzer can process the digital audio.
“The CobraNet audio stream is demuxed and converted into an I²S digital stream for digital signal processing,” said Somarel. “The stream is then routed to a digital transceiver, which converts I²S to an AES audio stream. Only two channels of the CobraNet audio stream are processed at a time, because the Audio Precision APx585 can only read two channels of AES audio signal.”
Speakers make the sound
No sound system is complete without speakers, and QSC Audio manufactures a wide range of them in Costa Mesa. Most of the speaker testing takes place in two labs: the 4-pi lab and the 2-pi lab. In these labs, engineers evaluate complete speakers and transducers.
In the 4-pi lab, loudspeaker engineer John Brodie tests complete speakers and their cabinets’ internal waveguides. The room is large enough for Brodie to measure a speaker’s first response—the sound that reaches your ear directly from the speaker, before reflections. In the 4-pi lab, first-response audio reaches two microphones about 15 ms before reflected sound.
Each of the two microphones hangs from the ceiling at a 90° angle to each other. The speaker under test sits on a turntable that sits on a commercial lift. The lift raises the speaker to the level of the microphones. Lasers on the wall let Brodie align the speaker under test with the microphones.
To test a speaker, Brodie excites it with an amplified log-swept-sine or multitone signal. The microphones connect to a test system. “We use a system called SoundCheck by Listen Inc. SoundCheck uses a 24-bit PC sound card as a digitizer, digitizing sound from the microphones through preamps.”
Brodie noted that the test system requires some setup time. It has a library of steps that define the sound card, the test site calibration data, and the speaker impedance. The library lets the test system perform math such as FFTs (fast Fourier transforms) on captured audio and generate the stimulus signals.
Brodie demonstrated the system with a QSC model HPR 122 amplified speaker. He used a log-sine stimulus tone. (Listen to an audio file of the tone.) The SoundCheck system then performed an FFT on the first-response audio, analyzing it for frequency response, power, and distortion.
As part of a speaker evaluation, Brodie needs to measure a speaker’s maximum power and frequency. He tests a speaker with up to 100 Vrms across its terminals, then steps down to –12 dB from that point. He then calculates relative change in level to find the amount of compression. “If a speaker port tube is too small,” he said, “you get wind turbulence and have to redesign it.”
Brodie also uses the response from the two microphones to characterize the speaker’s projection profile. “A speaker’s waveguide design may be for 90° dispersion pattern. If it comes out 80° or 100°, we will redesign the waveguide.” To get the dispersion pattern, Brodie uses the turntable to rotate the speaker in 5° increments.
The audio analyzer software needs site calibration data so it can mathematically eliminate reflections and other effects from measured sound. To get calibration data, Brodie took a speaker outside and played it loud enough to get measurements without reflections (hopefully without bothering the neighbors). He then used that baseline measurement and subtracted reflections as part of the calibration.
If you look up at one of the walls in the 4-pi lab, you’ll see a set of wood panels where the center panel holds a transducer. Behind the panels, Luis Esparza evaluates transducers in the 2-pi lab. The 2-pi lab is located high above the floor of the 4-pi lab to minimize reflections.
Figure 3 illustrates the setup Esparza uses to measure the dispersion pattern of a transducer. A microphone, supported by ½-in. copper pipe assembly, can pivot 90° around the transducer at a 1-m distance. From inside the 2-pi lab, Esparza moves the pipe, changing the angle of the microphone relative to the transducer under test. He moves the microphone in 5° increments and records a transducer’s off-axis response. Esparza uses a PC-based audio analyzer called Clio from Audiomatica. It consists of a signal-conditioning box that connects to a 24-bit PCI sound card. He uses the card to produce stepped-sine sweeps and MLS (medium-length sequence) measurements that measure a transducer’s transfer function.
Figure 4 shows a plot that Esparza produced on a transducer’s transfer function. The graph has several plots, each of a different color where each color represents frequency response with the microphone at a different angle relative to the transducer. In this test, Esparza made measurements from 0° to 45° from center in 15° increments.
Esparza uses such plots to find a transducer’s directivity response, from which he determines where a speaker’s crossover network should cross from its low-frequency transducer (woofer) to its high-frequency transducer (tweeter). In the example illustrated in Figure 4, Esparza would set the crossover frequency at about 2 kHz.
Esparza also makes plane-wave measurements where he measures the response of compression drivers. (Compression drivers are efficient transducers that produce high-frequency sound waves.) Normally, a compression driver is designed to emit sound as a plane wave. To make these measurements, Esparza attaches a compression driver to a waveguide tube that channels all of the sound to a microphone (Figure 5). He can measure the plane wave directly from the transducer without dispersion affecting the measurement using a swept-sine or MLS signal. “We get phase and magnitude measurements at the same time, and we can measure distortion and power compression using the same test fixture,” he said.
The 2-pi lab also has a laser vibrometer that Esparza uses to measure the physical vibration of a transducer or speaker cabinet. The laser scans across the surface of the speaker cone or cabinet wall. “We want to see how the cone moves,” he said. “We once had a frequency response issue at 3 kHz on a 6-in. cone. The outer edge of the cone moved more than the inner edge of the cone. Because of that measurement, we had the compression-driver manufacturer change its geometry, which corrected its frequency response.”
Esparza also uses the vibrometer when testing a cabinet. He provides measurement data for cabinet designers to improve cabinet construction and optimize the locations of braces.
The 2-pi lab also has a transducer distortion analyzer manufactured by Klippel (see photo of Luis Esparza on p. 25). The Klippel analyzer is a laser-based instrument that calculates a transducer’s nonlinear behavior by measuring the physical motion as a function of input current, which lets Esparza find how much of the input energy comes out of the transducer. The system measures nonlinearity of a transducer’s cone assembly, magnet, and structure.
After performing measurements in the 2-pi and 4-pi labs, QSC’s engineers can decide if a speaker is ready for reliability testing. That’s where Chris Davies, manager of the Loudspeaker Systems Group, takes over.
Engineers in Davies’ group test a speaker’s or transducer’s reliability by running it at its full rated power or higher in order to find out how much power it can handle. But running speakers at full power is rather loud and requires an insulated room to keep the sound from annoying QSC’s neighbors.
QSC’s reliability lab consists of two steel walk-in containers located behind the building. One holds a double-insulated chamber in which speakers and transducers under test operate at full power or higher for in excess of 100 hr. The second container holds the control equipment: CD players, a QSControl.net networked audio system with I/O and amplifiers, and an Agilent Technologies data-acquisition system. All equipment runs under software written in National Instruments LabView. Patch panels let Davies connect the amplifiers to the speaker under test in the adjacent container.
Davies uses the custom software to select an input source (a CD player or any of a variety of noise signals) and digitally route its output to the appropriate amplifiers. During a test, the data-acquisition system monitors voltage and current at the speaker under test and notifies the design team of any failures. The data-acquisition system measures current by measuring the voltage across a shunt resistor.
A test can run for as little as 2 hr or for more than 100 hr. After a 2-hr test, the system will give the speaker a rest and start again. Power levels start low and then ramp up. “We expect failures because we drive the speakers beyond the design limits, looking to see when they break,” said Davies.
QSC Audio’s engineers test audio products in the lab and in production to ensure that listeners get the most from musicians, DJs, movies, conventions, and concerts, and then musicians and DJs put the products to the real test on stage. At permanent installations such as theaters, convention centers, and stadiums, everyone can hear the results of those efforts.
QSC's "Passionate about sound" slogan applies throughout the company, but nowhere more than in company founder Pat Quilter's office. On the day of my visit, Quilter was designing a new amplifier, but his office is full of vintage audio equipment.
When you enter Quilter's office, you're in a museum of audio. Not only does he have some of his original guitar amplifiers, but he has equipment that's 100 years old. You can see the "Victrola" horn record player in the photo, which dates to about 1910. Quilter also has a record player that uses cylinders developed by Thomas Edison rather than flat disks (Ref. 1). "The problem with the cylinders," he noted" was that the master would wear out in 25 or so plays, so the artist had to constantly rerecord the music." The flat discs that Quilter played were molds from a first-generation master, which could reproduce many more copies. "They eventually figured out that they could put recordings on both sides of the disc," he said. "They also figured out how to shrink the copy just enough for it to drop out of the mold."
"Edison felt that the cylinder was technically superior to the flat disc," noted Quilter, "because it had constant-velocity grooves. But his marketing people were being bashed by the Victor company with its flat 78-rpm discs. Edison's people wanted him to develop a better flat disk, but instead, he developed another proprietary vertical-recording cylinder. A design flaw was that the molding imperfections were in the same plane as the music and thus anything on the master comes through loud and clear as where the 78s somewhat subdue the noise."
Quilter then demonstrated the recording, which produced a noticeably higher sound quality than the 78-rpm flat discs. He described it as the "Betamax" of early recordings as opposed to VHS, which became the standard for home video tape despite its lesser quality. All the record players operate without electricity. They're all mechanically powered.
In the 1920s, radio came along and people realized that "not all recorded music had to sound like a Victrola." Quilter explained that engineers at the Bell Telephone Lab developed a folded 6-ft.-long horn that fit inside a cabinet and it could reach down to frequencies of almost 100 Hz. It did a great job of amplifying the music from an electrically recorded flat disc. The sound quality was considerably better. Electric playback had become the norm by the late 1920s, he explained.
Quilter also has several vintage radios, two of which are just behind his right shoulder in the photo. His radio collection includes a 1937 Zenith tube radio that also still works. It has a knob that lets you change bands from AM to shortwave using a mechanical knob on the front panel. When you change bands, you change a set of dials around the frequency pointer that show band frequencies.
1. Cylinder Preservation and Restoration project, University of California, Santa Barbara. cylinders.library.ucsb.edu.