Combine RF and Audio Cell-Phone Tests

The audio circuitry in a digital mobile phone contains analog and digital components ( Fig. 1). You’ll find it beneficial to test these components at the board level, especially if your manufacturing process is not mature, so you can find process-induced faults before sealing the board inside a handset. But because a phone can pass board test yet not work in the real world, you’ll need a final functional test to verify that a consumer will hear quality sound.

Figure 1. Cell-phone functional blocks include data converters, voice-processing circuitry, and RF transmitter and receiver stages.

Analog phones offered few audio-test options—they required simultaneous call-processing and audio tests at the final, functional-test stage to verify the performance and tuning of the modulator. For example, US standards require that a Supervisory Audio Tone (SAT) transmitted from the base station be echoed back from the analog mobile unit. This tone is a constant 5970, 6000, or 6030 Hz and is crucial for signaling between the base station and mobile unit. Your analog phone tests had to verify that your mobile phone could echo or regenerate this tone, modulate it together with voice signals, and transmit the combined signal with minimal phase or frequency errors. Your tester also had to verify the presence of the Signaling Tone, a short 10-kHz tone, in the mobile unit’s RF transmission.

You have more leeway in the test of a digital phone’s audio circuitry. Although customers expect better sound quality from digital phones, audio specs for digital phones are generally more lenient than those for analog ones. Specifications for CDMA handsets1 are silent on the topic of audio quality. GSM standards2 do contain requirements for frequency response, loudness, echo return loss, and distortion, but these requirements are not tied to RF performance. In fact, most audio measurements in the GSM specification ignore the RF path in the mobile phone completely, and simply require looping the signal between the speaker and microphone, through the ADC and DAC.

As an example, GSM 11.10 (Section II.11.1.1) requires manufacturers to source eight different tones into the phone’s microphone, then read the tones as digital data coming from the ADC. This approach checks for basic functionality of the microphone, amplifier, and ADC, but not for properties such as frequency response and sensitivity. One infers from the GSM spec that audio quality is primarily an issue of the phone’s design—if you choose adequate audio components you’ll automatically get adequate audio quality—and not of the manufacturing or tuning process.

Yet audio quality is far too important to ignore completely in manufacturing final test. Acoustic problems can arise as a result of errors during the surface mount or assembly process. You can choose from two strategies for testing the audio components in a digital phone during final test: You can test a small loop within the audio subsystem, or you can verify the entire transmit and receive path. Because so few test standards govern digital tests, you have a great deal of flexibility in setting up your tests; the equipment, process strategy, and even the test specs are changeable.

Final-Test Choices
The first option—the audio loopback test—checks whether the speaker, microphone, and initial parts of the audio path are operating. The test setup requires only an audio generator/analyzer, a battery, and an artificial head (Fig. 2).

Figure 2. A cell phone’s audio circuits can be tested in loopback mode using only an artificial head and audio-frequency instruments; a separate test can check RF components.

The test loops a signal back between the microphone and speaker. The loopback may occur inside the voice codec or close to it, testing the audio path up to the DAC and ADC. GSM standards, for example, specify a loop from the microphone, through the ADC and DAC, and back to the speaker. Typically, you would transmit a command to put the phone in test mode, and then place the phone on an artificial head. (GSM standards specify where you must place the artificial ear and mouth relative to the phone’s earpiece and mouthpiece, but again, CDMA standards are silent on the topic.) The artificial head will then source sound into the mobile unit’s microphone at a certain sound pressure, and it will measure the echo from the speaker. Additionally, data may be sent and received from electrical connections to the phone. You never need to place a call to perform this type of test, so no RF equipment is necessary.

The loopback approach works well if you perform audio testing at the same station as human interface (keypad/display) testing. Because this station does not include an RF test set, looping back the audio signal is the only feasible way to test the phone’s acoustics. The loopback approach, in which you separate RF tests from other tests, may help you avoid bottlenecks; if the two sets of tests require different test times, you can assign a different number of test stations for each, making more efficient use of resources (Fig 3).

Figure 3. (a) You can combine audio and RF tests, but (b) you might find you’ll require less equipment by separating them.

As a second option, you can perform audio tests with a call in progress. To check the phone’s microphone and transmit path, an artificial mouth provides a signal into the phone, where it’s coded, processed, modulated, and sent out the antenna to an RF test set. The RF test set demodulates and decodes the signal, and an audio analyzer measures the output. Next, to test the phone’s speaker and receive path, the RF test set codes and modulates a signal and sends it over-the-air to the phone. The phone demodulates, decodes, and sends the signal out the speaker and into the artificial ear, and the audio analyzer measures it. This type of testing (Fig. 4) verifies the complete audio path and may be the only way to test a dual-mode digital/analog phone.

Figure 4. You can test the complete audio path using this sort of setup. Red arrows indicate the transmit path; green arrows indicate the receive path. This test requires an RF test set connected to the UUT’s antenna port.

In this scenario, you will most likely perform RF tests and audio tests on the same test station. Most manufacturers that use this audio-test strategy count on having very low measurement times—acoustic measurements that take only a few seconds to complete. If these manufacturers also can keep display and keypad test times low (by using machine vision and robotics), all three types of tests may be combined into one station, reducing handling time between test stations, and also reducing the amount of effort needed to test phones. In addition, by using a single test station, you won’t need to duplicate equipment such as computers, fixtures, power supplies, and multimeters.

A disadvantage of testing audio components with a call in progress is that audio test cannot be run independently—RF instrumentation must be present. Even when faced with bottlenecks, you might not be able to add extra test stations because of the high cost of the RF instruments (Fig. 5).

Figure 5. (a) Separate test stages are appropriate for long audio tests and manual lines. (b) If audio tests are short, you can save equipment and perform a more thorough test by combining RF and audio tests, leaving the keypad and display for a separate station. (c) Vendors with fully automated lines may combine all three tests in one station.

Figure 6. (a) Fixtures for RF test only can consist of a simple shielded box with connectors and a nest for the phone. (b) A fixture for both audio and RF test must contain complex electronic and mechanical components. You must pay careful attention to maintain RF integrity as well as audio isolation.

In addition, fixturing is more complicated for a combined RF/audio test stage. The fixture must be isolated for both RF and acoustic noise, and the artificial head must be placed inside this fixture and connected to instruments with shielded cables and connectors. Audio connectors and PCBs that are not filtered properly may cause RF leakage, so you need to ensure the RF integrity of the fixture. Extra mechanical parts must be included to support the artificial head and additional electronics, and these must be designed to prevent echoes and leakage, for both the RF and the audio signals (Fig. 6).

If you’re testing GSM handsets instead of CDMA versions, you’ll face an additional challenge testing the audio circuitry with a call in progress. According to the GSM 11.10 specification (Section II.11.1.7.1), “The normal operation of the speech codec results in modulation of the amplitude of pure tone signals at a frequency that depends on the frequency of the tone.” So, if GSM audio circuitry is tested with a single tone, the measured signal will be distorted, and it will be difficult to determine the quality of the audio components—you’ll have no good way of knowing whether the distortion is natural or a result of poor-quality components.

A multitone or simulated-speech measurement is more effective for GSM, but the test equipment is more costly and the tests take longer. The audio subsystem embedded in an RF test set generally is not capable of complex audio measurements, so you will need additional stand-alone audio analyzers and arbitrary sources. You also will need to invest time to develop software to generate and measure the proper multitone or simulated-speech signals.

Audio Measurement Comparisons
For audio tests on any type of cell phone, you’ll need to decide what type of audio signal to use and how to measure it. You can choose from three methods (Fig. 7): the single-tone method, the multitone method, and the multitone-with-DSP method.

Figure 7. You can choose from three audio-test methods: (a) a single tone, measured by an audio analyzer in the RF test set, (b) multiple tones, applied and measured one by one with an audio analyzer, or (c) multiple tones, measured all at once using an audio digital signal processor.

Generally, if a digital phone’s design has been verified, the acoustic subsystem in the phone can be tested with a single-tone go/no-go test. This test simply determines whether the components have been assembled correctly.

The test involves using a simple audio generator/analyzer to generate a single 1-kHz tone into the phone under test’s mouthpiece. The analyzer then measures the tone at the RF test set’s demodulated output or at the phone’s earpiece. Some RF testers include a simple audio subsystem that can source a single tone and measure its amplitude, frequency, and distortion. This single-tone test is well suited for a combined RF and audio test stage, because the same piece of equipment can be used for all measurements. And, the measurement time is very short—on the order of a few seconds.

But several problems can slip by a simple single-tone go/no-go audio test. For example, a microphone could become damaged during the assembly of a phone’s plastic case. Or, a case could be incorrectly aligned, muffling the sound or causing buzzing noises. And, even if you use high-quality components, some batches will undoubtedly fail to meet specifications. To uncover these sorts of problems, you will need more comprehensive audio tests.

A second, more complex method involves a multitone measurement, sourcing several tones and measuring each in turn. This kind of test is recommended in the GSM 11.10 specification. Section II.11.1.7.1 specifies that “Measurements are made at one-twelfth-octave intervals . . . for frequencies from 100 Hz to 4000 Hz inclusive.”

Measuring a large set of audible frequencies will give you a good picture of the frequency response of the audio components. Test tones whose frequencies lie at the high and low extremes of the range should be barely audible; tones between 2 kHz and 3 kHz should exhibit peak amplitudes (Fig. 8). Here again, you can use an audio analyzer to perform the tests. An audio analyzer can measure harmonic distortion to check whether your phones resonate at certain frequencies. You also can use the analyzer to check signal-to-noise ratio for any static or background noise that affects the audio quality.

Figure 8. A typical cell-phone speaker’s audio response peaks between 2 and 3 kHz.

With the third method, you use an audio digitizer with DSP capability. You program the audio source to simultaneously produce all the tones needed in the audio test. During test, the burst of tones takes less time to produce than does a sweep of individual frequencies. The digitizer acquires the audio signal from the phone under test and processes it using an FFT algorithm. The results show amplitude for each of the test frequencies. Users also can use the acquired information to compute SINAD, harmonic distortion, and signal-to-noise ratio for the telephone.

When performing this test, choose the frequencies carefully. This is important when performing tests while a call is in progress. For example, since the GSM vocoder will modulate the signal’s amplitude at a frequency dependent on the tone’s frequency, this effect will be increased if you use frequencies which are multiples of each other. The modulation of a 1-kHz signal would be added to the modulation of a 2-kHz signal, making your results even less reliable. If you wish to measure the amplitudes of all frequencies mentioned in the GSM specification, for example, you may break the frequencies up into several groups of multitone signals that will not resonate.

Digitizing a multitone signal shortens test time considerably. An eight-tone measurement may take 8 s or more with the swept method, but only 1 s with the FFT method. A tester can generate enough tones to measure a noise- or speech-like signal in the same amount of time that another instrument could measure a single-tone signal.

The measurement is fast enough that a combined RF/audio test takes about the same amount of time as the RF-only tests represented in the left-hand portion of Figure 5a. Or, because the audio DSP is independent from any RF equipment, you can perform the audio tests separately (or with the keypad/display tests). Besides saving time and increasing flexibility, the DSP-based test offers better accuracy than single-tone or swept-tone tests. With a DSP, you can measure frequency response in a noisy room, and the DSP will measure the strength of the audio signal only at the frequencies of interest. Finally, sourcing multiple tones simultaneously avoids the distortion that a GSM codec imparts on a single-tone signal (depending on the frequencies chosen), allowing the measurements to reflect the true performance of the audio path. T&MW

FOOTNOTES
1. ANSI/TIA/EIA-98-B, Recommended Minimum Performance Standards for Dual-Mode Wideband Spread-Spectrum Cellular Mobile Stations, July 1997. (7/10/98 version available from Global Engineering Documents, Englewood, CO, 800-854-7179, global. ihs.com.)
2. European Telecommunications Standards Institute, ETSI Technical Specification GSM 11.10 Mobile Station Conformity Specifications, April 1993. ETSI, Sophia Antipolis, Cedex, France, +33-4-92-94-42-22, www.etsi.org.

Dawn DuPriest joined Hewlett-Packard as an application engineer in 1995. For the last two years, she has developed software and customer applications for mobile phone functional test. She graduated with a B.S. in Computer Engineering from Case Western Reserve University.