By, Bob Green, Senior Market Development Manager, Glenn Fasnacht, Engineering Project Manager, Alain Lainé, Project Engineering Manager, Keithley Instruments, Inc.

Increasingly in the wireless industry, manufacturers are monitoring the performance of the individual components that comprise finished products. The potential losses caused by a defective component escalate exponentially for each stage of manufacture where the component goes undetected. Conversely, while manufacturers strive to detect failures as early in a build cycle as possible, they also try to use the minimum testing necessary in order to maximize profitability.

In the case of wireless phones, product quality and user satisfaction depend heavily on the speaker (earpiece) and microphone. These electromechanical transducers interface the human operator directly, and can be affected by a variety of failure modes that aren't typical of purely electronic components.

Test Criteria
The most basic operating parameters tested with speakers and microphones include:
Frequency Response. The range of frequencies over which a transducer can produce or respond to sound waves is usually specified with some figure of "flatness" (for example, 100 Hz to 10 kHz ۭ bB). Voice frequencies important in telecom applications extend from approximately 300 Hz to 3 kHz.
Distortion. Any non-linearity in the transfer function of a circuit or electromechanical device (with the possible exception of amplitude) causes the output waveform to differ from the input waveform, a process known as "distortion." Examples include total harmonic distortion (THD), intermodulation distortion (IMD), phase distortion, crossover distortion, and clipping. Of these, THD is probably the most often measured in telecom audio circuits.
Peak Spectrum Values. Different failure modes of circuitry or individual components may alter the frequency content of an input signal. These changes can be discerned by comparing plots of the input stimulus vs. the output of a DUT to the same plot for a known good unit.
Noise. Unwanted content can be added to a signal of interest through the coupling of stray electromagnetic fields in the environment, mechanical defects in electromechanical transducers, or as a consequence of circuit gain and component characteristics.
Ringer tone. While not critical to audio quality, manufacturers nonetheless can specify the ringer tone's volume, frequency, and other characteristics.

The Nature of Audio Testing
Generally, high-end audio components and systems are tested over the full range of human hearing (or beyond) in order to detect distortion or other anomalies that amount to mere fractions of a percent. The process is often slow, and performed in a laboratory environment. In contrast, production testing processes for wireless phones, pagers, and related telecom devices must be fast, cost-effective, and suitable for use by technicians. The range of frequencies may be limited to voice frequencies, and audio may be restricted to a single-channel. Thus, wireless audio testing can be accomplished with simpler, less costly equipment. Some test capabilities of particular value in audio production testing include:
•Audio waveform generation from 20 Hz to 20 kHz, including swept sourcing
•Audio shaping filters, including programmable high- and low-pass types
•Identification and measurement of harmonic levels
•Measurement of THD, THD+Noise, and SINAD
•Wide band or narrow band noise measurement
•Ability to interface required measurement transducers
•A standard computer interface for uploading programs and data
•Small footprint and low profile for space-efficient rack mounting

Figure 1. Model 2015-P Audio DMM and Model KPCI-3108 Data Acquisition Card are examples of instrumentation approaches for production audio testing.

Recently, Keithley Instruments introduced the third in a series of audio digital multimeters specifically designed for audio test (Figure 1). A standard DMM measures AC and DC volts, amps, ohms, temperature, frequency, and related parameters. In contrast, the audio DMM concept includes state-of-the-art DSP technology, math functions, distortion measurements, swept audio frequency output, and other functions unique to audio waveform acquisition and analysis. The FFT method is used to measure distortion and other frequency-related parameters, which is more direct and much faster than analog measurement methods.

A second solution for testing wireless products is structured around use of a personal computer and a medium- to high-end data acquisition plug-in card (also shown in Figure 1). An ideal card for this application would provide high speed, low noise, multi-function measurement and control, a wide range of gain amplifier settings, and at least 16-bit A/D conversion. Like the audio DMM, such a card should be able to generate complex, high quality waveforms in the audio spectrum. Digital I/O and triggering would also be desirable as a means of synchronizing complex tests with other test equipment. A PC-based system based on a data acquisition card obviously does not offer the "pre-packaged" functionality of the audio DMM. However, the PC-based system can offer great testing versatility through the use of appropriate software and brute processing power, and even satisfy some applications more easily than with a DMM-based instrument.

Typical Audio DMM Application — Speaker Testing
Loudspeakers can suffer from visible as well as concealed defects that occur during manufacturing or shipping. Gross problems, such as a torn cone or open voice coil, are easy to identify visually, aurally, or with instruments. However, more subtle defects, such as a warped cone, damaged magnet, or rubbing voice coil, may escape detection by conventional DMM-type measurements. Such problems may also go unnoticed in "listening" tests, especially if the test signal does not excite the device at a frequency that reveals the fault. Furthermore, the use of human operators to quantify audio quality is a slow, subjective, and inconsistent process, especially where defects result in low levels of distortion.

This test involves four speaker samples of the types used in portable telecom equipment. One of these speakers had a torn cone, while the frame of another had been bent so as to cause the voice coil to rub lightly against the speaker frame. Although a gross failure of a rubbing coil on the magnet structure can be fairly easy to detect, a minor misalignment may not be. In either case, a rubbing coil will require more electrical energy to move the cone because some of the energy is lost to friction of the rubbing coil.

Figure 2 shows a free air resonance frequency response plot of the four speakers as performed with the Audio DMM. Plot A represents the resonance curves of the good speakers. The resonant peak is high and narrow at 450 Hz. In Plot B, the speaker with a rubbing coil displays a resonant peak having a higher frequency (525 Hz) and lower amplitude. The speaker with a torn cone (Plot C) shows a resonant peak at 390 Hz.

Figure 2. Resonant Frequency Plots for Speakers

The same speakers subjected to a distortion response plot are shown in Figure 3. The speaker with rubbing coil speaker (A) has a much higher distortion level at low frequencies, while the torn cone (B) results in a spike at 4.4 kHz.

Figure 3. Distortion vs. Frequency Plots for Speakers

The graphs shown in Figures 2 and 3 were constructed with 50 data points, which required approximately two seconds to scan. It is interesting to note that an AC voltmeter would not have detected the torn cone, while a traditional THD meter would have had taken 10-20 seconds to perform the same test.

Sample DAQ Application — Final Testing
An example of a test that can be performed with the data acquisition card approach consists of measuring the overall performance of audio components in the complete product assembly, during final testing. In this test phase, not only are the microphone and speakers evaluated, but also the CODEC (Coder/Decoder). These tests are performed during normal transmit/receive operation of the phone and require the generation of multiple tones, and evaluation of audio signals. The test setup is relatively simple, as shown in Figure 4.

Figure 4. PC-based test setup, with cell phone, ear transducer, and microphone transducer.

An "artificial mouth" and an "artificial ear" (an external speaker and microphone) are placed in front of the phone's microphone and speaker. The mouth is connected to the D/A circuitry of the data acquisition card, which generates the required stimulus. The required steps to complete this test include initializing the board, computing the waveform to be generated, and then acquiring the signal. A power spectrum and total harmonic distortion can then be computed on this signal, and each harmonic compared in frequency and amplitude to its counterpart in the generated wave. The result of a test conducted with this type of setup is also shown in Figure 5. Dual tones (800 Hz and 1100 Hz) were fed to a good audio circuit and a faulty circuit. Note the higher incidence of sub-harmonics and noise in the lower plot.

Figure 5. Results of a dual-tone test showing increased harmoniccontent and noise for a faulty circuit.

Virtually any data acquisition package, such as TestPoint™, LabVIEW, LabWindows™/CVI, or HP VEE, or a language such as Visual Basic®, Visual C++, or Delphi, can be used to control the instruments. However, a mathematic library is necessary to compute the power spectrum. Some packages also contain this capability. Where necessary, libraries can be obtained from companies such as Quinn-Curtis, Inc. and others if it is not contained in the acquisition package.

Clearly, audio testing in production environments requires a feature set not easily realized with conventional audio test equipment. Speed, versatility, ease of use, and space efficiency are high on the list of priorities for such applications. This article has discussed two methods for such testing. The Audio DMM is a relatively new development that automates many standard tests, and uses sophisticated signal processing to yield signal amplitude, distortion, resonance, FFT, and related types of audio measurements. The PC-based data acquisition approach requires somewhat more in the way of software configuration, but offers the test engineer an extra measure of versatility to handle special needs.