For the PDF version of this feature, click here.

Mobile phone production testing has been increasing in complexity with the continued addition of enhanced user features that require production verification. Together, with the push to reduce test time and cost, these trends are multiplying the challenges for today's test engineers. Consequently, designers are coming to rely more on the design qualification process to simplify testing.

However, most handset manufacturers still operate under 100% testing requirement. Additionally, no design qualification process, regardless of how rigorous it may be, can guarantee 100% defect-free components — particularly if those components are electromechanical transducers, which can become defective during the assembly or handling process. Thus, the need for audio quality measurements remains a fundamental part of the ever-growing test suites for mobile phone manufacturers.

Implementing a through-the-air audio quality test system generally involves some form of system level characterization, which can also be useful for any application that requires audio characterization, such as entertainment systems and MP3 players.

There are two basic categories of audio quality tests:

• Audio only test: These tests can be done at preassembly, at the circuit board level or after final assembly. This configuration usually sends the stimulus signal from the audio analyzer through audio components on the device under test (DUT) and back to the audio analyzer for measurement (e.g., an audio loopback mode test). The test measures audio quality parameters such as distortion and frequency response.

• Combined audio/RF test: This test is usually conducted after final assembly and involves verifying audio stimulus signal over the complete RF transmit-receive path.

It requires a communications analyzer with audio analyzer capabilities.

Combined audio/RF tests usually provide an accurate picture of the overall DUT, verifying correct modulation/demodulation, as well as the integrity of the audio signal. However, they don't necessarily reveal the condition of the electromechanical transducers, i.e., the speaker and the microphone. In fact, it's possible for a transducer with mechanical defects to produce valid results in combined audio/RF tests.

Traditional audio tests are still essential to establish the condition of the transducers.

Through-the-air audio test measurement parameters, such as total harmonic distortion (THD) or total harmonic distortion plus noise (THD+N), can reveal subtle transducer defects that may be hard to detect with other methods. They also provide the manufacturer with an objective measure of audio quality for total quality control purposes.

A good test fixture is essential to attaining repeatable, meaningful results. The most important function of the fixture is shielding ambient acoustical noise, which means the fixture should be enclosed in a material that acts as a noise shield. To reduce noise as much as possible, the enclosure should be made with material that would produce maximum transmission loss. Transmission loss is governed by mass law, approximated here as^[3]:

TL = 20 log (ms × ƒ) - 48

where TL = random coincidence transmission loss (dB),

ms = mass per unit area (kg/m²),

ƒ = frequency of the sound wave (Hz).

A test fixture enclosure can be made from a variety of materials. In general, the denser the material, the greater the transmission loss. Here, either steel or aluminum would be the practical choice.

Mass law is affected by the coincidence effect at higher frequencies and by the resonance effect at lower frequencies^[3]. Both effects need to be considered when designing the fixture.

Coincidence effect: High-frequency waves cause ripples or “bending” waves that travel-longitudinally along the wall of the fixture enclosure. The frequencies of these ripples differ from those of incident waves, except at a certain frequency called the coincidence or critical frequency (fc). At this critical frequency, the sound energy is transferred efficiently through the walls of the fixture enclosure and the transmission loss described by mass law no longer holds. The value of fc is described by^[3]:

ƒc = A/t

where A = constant of material (Hz-mm) (see Table 1 for values),

t = material thickness in (mm).

The goal is to have an fc that's beyond any frequency of interest in the test. Selecting a thin wall with a high A value material raises the critical frequency, as shown in Table 2. For the frequency range of 200 Hz to 4 kHz, aluminum or steel walls that are 3 mm thick or less would suffice.

Resonance effect: Assuming the fixture is rectangular, consider the two facing wallsof the fixture enclosure. Imagine that a sound wave originates from one wall as if there is a speaker in it. This sound wave is reflected back by the facing wall^[1]. If the distance between the two facing walls is half the wavelength, then the reflected wave will be in phase with the incident wave, resulting in a classic resonant or standing wave.

For each pair of facing walls inside the fixture, the resonant frequency ƒ is

calculated as^[1]:

ƒ = (c/(2l)) n

where c = speed of sound (~344m/s), l = distance between two facing walls (m) and n = 1, 2, 3, etc. (order of harmonic).

For a rectangular fixture, the resonance frequency is estimated as^[4]:

where n_x is the harmonic resonance between the x-wall and its facing wall, i.e., n_x is 1 for the first harmonic, and l_x, l_y, and l_z are the length, width and height of the fixture in meters.

When choosing the fixture's design, it's important to keep in mind that the resonance-frequency depends on the compound ratio of fixture length to height to width. Unfortunately, there's no one standard ratio, although a perfect cube should be avoided because one frequency can resonate between any facing walls. One of the commonly used ratios (R. Walker, BBC, 1996) is the following^[4]:

Another major challenge in test fixture design is to minimize cross-coupling, which is a result of the stimulus signal at the test speaker directly coupling into the test microphone instead of going through the DUT. The best remedy for this is to use an acoustic coupler to direct the sound from the test speaker to the DUT microphone. If the signal path through the acoustic coupler is sealed properly, it can significantly minimize the stimulus signal leakage that causes cross-coupling.

It's also important to minimize the ambient signal levels inside the fixture by adding material that will absorb or dampen sound. A variety of acoustic damping materials is commercially available with varying levels of sound absorption or reduction, depending on material type and thickness.

Transducer selection is the next important aspect of fixture setup. Frequency response is one of the most important specifications for the test speaker and the test microphone. Frequency response should be flat (within ±3 dB) for the frequencies of interest, usually 200 Hz to 4 kHz. The test microphone will generally require a pre-amplifier, which should also have a flat frequency response for the frequencies of interest and typically have a gain of around 20 dB. Other specifications, such as power level, noise performance and sound pressure level should be considered based on the specifications of the DUT.

Because mobile phone transducers are by necessity physically small, their sound production quality is somewhat limited. Therefore, specialized expensive transducers usually aren't necessary. A variety of adequate, low- cost transducers is readily available.

The size of the test microphone and the test speaker should generally be small to allow for convenient mounting inside the test fixture. The sound pressure level decreases 6 dB for every doubling of the distance, so it's critical to pay attention to the distance (d) between the test speaker and the DUT microphone, and the test microphone and the DUT speaker when mounting these items inside the fixture. These distances should be equal and, in most cases, d should be from 2 mm to 15 mm.

Although each component in the system has a set of manufacturer's specifications, the overall system should be treated as a black box and characterized as a system. Characterizing a selected group of parts (e.g., speaker and microphone of the test fixture, without the DUT) can also reveal useful insights into the system's overall behavior by verifying the performance of the subsystem.

Three basic procedures are generally applicable for any type of characterization:

1. Determine system noise.

2. Determine a test signal level for a chosen mid-band frequency.

3. Use this signal level as a reference level to quantify frequency response.

As an example, consider part of the system, the speaker and the microphone (with pre-amplifier), as shown in Figure 2. The following test equipment will be needed for characterization:

• A DMM/audio analyzer.

• Sound pressure level (SPL) meter (preferably with 30 dB to 120 dB range).

A simple way to start is to tape the speaker on a wall and align the microphone directly in front of it. The microphone can be mounted on a supporting stand, but avoid using a microphone support that could deflect sound, e.g., a facing wall. To ensure the same sound pressure levels, the SPL meter's microphone should also be placed at distance d (in the same physical location as the microphone occupied) when measuring SPL.

System noise: With source of the audio analyzer off, measure VRMS using the DMM/audio analyzer and the sound pressure using the SPL meter. This provides a measure of the noise level in the system and the environment. In a reasonably quiet room, the sound pressure should be around 40 dB, and VRMS can be a few millivolts as the microphone picks up and amplifies ambient noise.

Test signal levels: Generally, it's safer to start with low signal levels to avoid damaging the device^[2]. Also, note that the higher the test signal, the more potential it has to create cross-coupling effect and produce resonance conditions inside the test fixture.

Set the audio analyzer output to 1 kHz, 50 mV, and measure VRMS and SPL. Then, increase the stimulus by 6 dB (doubled to 100 mV). Notice that the VRMS measured should be doubled and the SPL measured should be increased by 6 dB — indicating the linear operating region of this subsystem. Repeat this process at least three or four times to verify the linear operating region.

If we arbitrarily assume the DMM/audio analyzer output has no limitation and keep repeating this process, we will reach a point where a 6 dB increase in the stimulus is no longer met by a 6 dB increase in measured signal, indicating a non-linear region. In practice, we may not reach this region because the instrument's output has finite levels. However, this is generally not a concern as we are more interested in lower signal levels.

After verifying the linear region or at least part of it, choose a stimulus signal that will produce measured signals at least 20 dB above noise; i.e., if the noise SPL is 40 dB, the stimulus signal level chosen should produce at least 60 dB of SPL. In most manufacturing test applications, the test signal ranges from tens of millivolts to a few hundred millivolts, producing SPLs from 65 dB to 85 dB.

Frequency response: Once the stimulus or test signal level is chosen, we can characterize the frequency response of the system. Performing a frequency sweep with the audio analyzer over the chosen frequency range will produce a picture of how the system behaves at those frequencies.

In some cases, the system may exhibit frequency-dependent gain. In that case, it's important to verify that it isn't caused by saturation or resonance of any component. The sweep may be repeated at different signal levels. Generally, for the entire linear region, the gain should remain constant for a given frequency. Having this frequency response data is crucial to choosing the test frequency range and interpreting the test results.

Every test system is unique, so a “golden phone” is essential as a benchmark against which to compare subsequent test results. The golden phone is a DUT with known good performance and functionality, or which is otherwise determined to be a suitable reference DUT.

The overall test system can be characterized following the same general procedure described in the speaker and microphone example. During the system characterization procedure, the cross-coupling should be verified by performing a frequency sweep with the DUT power off or audio loopback mode off.

Once the characterization process is complete, design and test engineers can typically work together to define the specific test criteria. Having a complete set of characterization data with a golden phone will simplify this task significantly.

If a system's transfer function is perfectly linear, its response to a sinusoidal stimulus signal will be identical in shape with the stimulus; i.e., in the frequency domain, both the stimulus and the response signals will have the same frequency (f). However, if the system isn't perfectly linear, any non-linearity will show up as energy at harmonics of the fundamental stimulus frequency^[2]. Distortion measurements are one of the most widely used methods of measuring non-linearity, which, in the case of a system with electromechanical transducers, could mean a defective transducer.

THD+N is the most commonly used distortion parameter, because it measures the linearity of the DUT while taking into account the effects of both harmonic distortion and noise^[2]. A low THD+N value not only indicates that the harmonic distortion of the DUT is low, but that noise in the system is also low.

The frequency-dependent nature of the system and the DUT makes it advisable to use a frequency sweep to measure distortion over a selected frequency range. In the example described in this article, a 10-point frequency sweep took only 542 ms to complete, including the GPIB bus-transfer time. Figure 4a illustrates simultaneous THD+N and VRMS frequency response measurements.

Some basic steps need to be followed to create an audio test system capable of fast, multifrequency audio quality measurements using inexpensive components. Although conceptually simple, this type of traditional audio test can reveal subtle component defects that other test methods can overlook.

Although not addressed here, it's desirable to incorporate and design the same test fixture with additional RF design parameters to take RF measurements as well as audio. This will reduce the device handling time and can significantly reduce overall test time.

1. D. Halliday, R. Resnick, and J. Walker, Fundamentals of Physics, 4th ed., Hoboken, NJ; John Wiley & Sons, 1993.

2. B. Metzler, Audio Measurement Handbook, 1st ed., Beaverton, OR; Audio Precision Inc., 1993.

3. A.C.C. Warnock. (1985, July). Factors Affecting Sound Transmission Loss; Canadian Building Digest [Online]. Available: http://irc.nrc-cnrc.gc.ca/cbd/cbd239e.html.

4. S. Linkwitz. (2005, June). Room Acoustics; Linkwitz Lab, Corte Madera, CA [Online]. Available: http://www.linkwitzlab.com/rooms.htm.

Joey Tun is applications engineer at Keithley Instruments Inc., Cleveland, Ohio.

Keithley's models 2015-P and 2016-P audio analyzing multimeters and 2015 and 2016 THD multimeters combine audio band quality measurements and analysis capability with a full-function 6 ½ digit digital multimeter (DMM). In addition to harmonic distortion measurements and measurements of individual harmonics, test engineers can make a broad range of measurements including voltage, resistance, current and frequency. The units generate pure tone signals for distortion measurements, and the -P versions compute peaks in the spectrum of the measured signal.

The series 2015 and 2016 multimeters are designed for the demands of audio device test engineers in high-speed production test applications. They can perform a 30-point frequency response test and simultaneously measure THD, THD+noise, or SINAD at each point in 1.1 seconds, including the time to process a GPIB command and to transfer the data to a PC. In addition, the series 2015 and 2016 can measure narrowband noise with the use of internal, programmable digital filters. They also can determine characteristics of the signal spectrum such as harmonics and spectral peaks.

With a complete DMM in the instrument, the series 2015 and 2016 total harmonic distortion and audio analyzing DMMs can function also as the basic measurement instrument in any test system, eliminating the need for a separate DMM.