Author: Brian Fallon. Reprinted from the April 2012 issue of Voice Coil.
In part 2 of this 2-part series, Brian Fallon discusses the challenges and practicalities of measuring headphones with built-in electronics such as digital and Bluetooth headphones. He covers connectivity, frequency limitations, test signals, measurement units and specialty measurements such as noise cancellation.
Note: This article was written a few years back and there have been many advancements in the SoundCheck software and other accessories for measuring advanced headphones.
Read the first few paragraphs of Headphone Testing Part 2 Article
“Headphone Testing (Part 1)” published in the December issue of Voice Coil, covered the basics of analog headphone testing: correction curves and fixturing, choosing hardware (ear simulators and couplers), the different requirements for testing in R&D vs. production, and the various essential measurements such as frequency response and distortion.
Analog headphones are relatively straightforward to test because there are only two electro-acoustic transducers to measure. Headphones with built-in electronics, such as digital headphones (including Bluetooth and USB), and noise-cancelling headphones are harder to test because the electronics and transducers need to be tested together as a complete system. In this article, test considerations for such headphone systems and the practicalities of testing them are discussed.
Testing Digital Headphones
Headphones with digital connectivity add complexity to audio testing. In addition to testing the acoustic transducers, the digital circuitry must also be considered. The fundamentals of analog headphone testing (the use of artificial ear simulators and couplers, the principals of repeatability vs. realism, and the tests that characterize the device) are the same, but the test signals must pass through the headphone electronics, which can greatly influence the test results. The principal distinction of digital headphones is that they contain a D/A converter and often DSP circuitry as well as a headphone amplifier. This means that unlike an analog headphone, where measurements are being conducted only on the electroacoustic elements, the measurements for digital headphones are being performed on the whole system which comprises everything from the digital signal to the acoustic output of the transducers. While it is certainly possible to isolate and test each of these components on its own, it is also very important to understand the intricacies of testing the complete device.
The initial challenge of testing digital headphones is managing connectivity. The test system must be able to communicate directly with the device. A software-based system is ideal because it can communicate directly through the computer’s USB interface to the USB headphone, which will appear in the operating system along with other audio devices. Test signals can be sent digitally and the acoustic signals can be analyzed synchronously. If a hardware-based system is used, an extra program is usually required to connect the test system to the device under test.
Bluetooth headphones, however, require an additional interface for the computer to connect to the device under test. This may be a hardware Bluetooth communication box or a simple Bluetooth dongle, either built into the computer or externally connected by USB (see Figure 1). Bluetooth interfaces cause transmission delays in the audio chain. The test system must be able to account for these delays in order to take meaningful measurements. Some test systems can use an autodelay algorithm that looks at the system’s impulse response to calculate the delay and remove it from the measurement, if necessary.
It is also important to be aware of frequency range limitations when designing tests for Bluetooth devices. Bluetooth devices typically operate at low-sampling rates of either 8 kHz (narrow band) or 16 kHz (wide band). These sampling rates limit the frequency range (because of the Nyquist frequency) to significantly narrower than analog headphones or even USB headphones. For example, a Bluetooth device with an 8-kHz sampling rate will only play audio up to slightly less than 4 kHz. Such limitations need to be considered when designing the test specifications. It can also be interesting to test the Bluetooth device beyond its cut-off frequency to see how well its anti-aliasing filter suppresses out-of-band signals.
Bluetooth presents a further challenge in that sine waves are not always transmitted accurately. When this occurs, alternative stimulus signals must be considered. Broadband noise is one possibility, but because of noise suppression circuits in some devices, this may not be a practical solution. A multitone signal, where several frequencies are played simultaneously, is another option. This produces a very fast frequency response measurement and is immune to the sudden dropouts that can occur in Bluetooth transmission. The downside of this test signal is that traditional harmonic distortion cannot be measured. Yet another possibility is the use of speech or music signals. These real-world audio signals transmit very well over Bluetooth, but their downsides are that they typically require long-term averaging and cannot be used for harmonic distortion measurements. If distortion measurement is required, non-coherent distortion may be measured using any test signal. This technique compares the input and output power spectrums to measure the non-coherent power and calculate the distortion plus noise (see Figure 2).
Traditional analog headphones are tested with a stimulus level that is rated in terms of voltage or power. The sensitivity is also specified in these units such as dBSPL / mW. When testing headphones with USB or Bluetooth connectivity the stimulus is simply a digital signal whose level can be expressed in terms of digital full scale. The sensitivity is, therefore, expressed in dBSPL/FS. Manufacturers sometimes choose to relate these signal levels back to voltage, which can be done if the characteristics of the D/A circuit are known. In such cases, the gain of the built-in headphone amplifier chip must also be accounted for. Another method used for relating these digital units back to the analog domain is through the use of a codec (see Figure 3). A-law and M-law are two codecs widely-used in Bluetooth applications that can translate the digital units into “virtual volts.” These codecs are most commonly used in telecommunications, especially for headsets.
Check out our main Headphone Testing page for up-to-date information on testing advanced headphones.