|Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting|
The popularity of portable (and personal) music devices, such as phones and iPads have hugely increased interest (and market demand!)in headphones. As more and more loudspeaker engineers find themselves employed in the fast-growing headphone market, either through company diversification or changing jobs, it is important that the unique challenges of testing headphone packages are fully understood. Many of the characteristics that make for a good in-room listening experience with a loudspeaker--good frequency response, low distortion, no rub & buzz or loose particles, and so on--also apply to headphones, and many of the principles of loudspeaker testing apply. However, there are some major differences and additional issues that need to be considered. These include couplers and associated correction curves, acoustic seal, fixturing, and additional tests such as L/R tracking.
In this article I discuss the issues that are common to testing all types of headphones.
SIMILARITIES AND DIFFERENCES
First, I will describe the similarities in testing loudspeakers and headphones.
The setup essentially consists of an electroacoustic measurement system, some kind of ear simulator containing a reference microphone, and the de vice under test. A test signal is sent to the transducer (headphone), which in turn is measured by a reference micro phone in a coupler.
The basic measurements made on headphones are very similar to those made on loudspeakers. These include frequency response, phase (polarity), distortion (THD and rub & buzz), and impedance. In both cases, the test signal is usually a swept sine wave, and the level can vary. Some set the drive level to achieve a certain sound pres sure level at a given frequency; others choose the level that equates to 1mW of power. Certain products may necessitate testing the frequency response at one level and performing a second, higher level test for distortion.
The primary difference in the test setup between a loudspeaker and a headphone is the method in which the transducer interacts with the microphone. Whereas loudspeakers are tested in open air, a headphone or earphone must be presented with an acoustic load that simulates the human ear. It is common to compare the left and right channel frequency response.
Large differences at certain frequencies can be very audible in a stereo device, even though the individual responses may be within specification. Some times, electrical characteristics such as crosstalk may also be measured.
Before beginning to test head phones, you need to consider two major factors: correction curves and the acoustic seal. These both have an effect on the frequency response, and the latter also affects the repeatability of measurements.
Coupler Correction Curves
Loudspeaker engineers are familiar with the ideal frequency response for a loudspeaker measured in the free field being a flat line (Fig. 1A). For head phones, however, this is not the case.
Headphone measurements are taken at what is known as the Drum Reference Point (DRP)-a point representing the human eardrum (Fig. 2 shows where this is on a Head & Torso Simulator (HATS)). If you were to measure the same loudspeaker that produced the flat free-field response curve in Fig. 1A at the DRP, the frequency response would look like Fig. 1B. In other words, for a headphone to sound like a loud speaker with a flat frequency response, it must produce a frequency response curve as shown in Fig. 1B.
This frequency response curve is a correction curve, or transfer function that represents the effects of the head, torso, pinna, ear canal, and ear simulator. To further complicate matters, different correction curves are applied according to whether your measurements are made in the free field (an echoic room) or diffuse field (reverberation room) (Fig. 3). For the most part, as with loudspeaker measurements, the free field is used. Typically when making measurements, the sub traction of the correction curve from the actual measurement can be carried out in your test software, so that your output frequency response is shown as the familiar straight line.
Another issue that needs to be ad dressed when testing headphones is the acoustic seal, or leakage. Realistic headphone measurements (using a HATS or other ear simulator) have a certain degree of leakage because the headphone does not fit tightly to the pinna. This has an effect on the frequency response, with a demonstrable loss at low frequencies (Fig. 4). Although realistic, it affects the repeat ability of measurement. In the R&D lab, this is compensated by repeating the measurement multiple times, removing and repositioning the headphone between each measurement and aver aging; on the production line different couplers and fixtures are used to offer a more repeatable seal-I explain these in more detail below.
DIFFERENT TYPES OF HEADPHONES
Before looking at test configurations for headphone testing, I will examine exactly what I am measuring. Head phone is a broad term that covers several different designs of product, each with its unique testing challenges.
Headphones fall broadly into four categories (Fig. 5): circum-aural (a large cup that completely surrounds the ear and pinna), supra-aural (an earpad that sits on the pinna), earbuds (also known as supra-concha), where the transducer rests at the entrance to the ear canal, and in-ear, where the sound port sits inside the ear canal. Although the measurements that need to be made are the same for each of these, the fixturing-a critical part of the test setup-is different for each type.
Figure 6 shows a headphone test set up (suitable for all types of headphones) using a PC- and soundcard-based measurement system. Although you may, of course, also use hardware-based systems, a soundcard offers more than sufficient accuracy for testing head phones, and such a system is usually less expensive. As you can see, the only part of this setup that will be unfamiliar to those accustomed to loudspeaker measurements is the "black box" that represents a device that simulates the ear (and replaces the microphone in a loudspeaker setup).
WHAT'S IN THE BLACK BOX?
The equipment in the black box can be a multitude of things, depending on whether you are testing in the R&D lab or on the production line, your bud get, and the type of headphone you are testing. Essentially it will be the microphone-containing device you have selected which simulates the human ear with some degree of accuracy, and appropriate fixturing to ensure repeat able results with the particular head phones you are testing. This is perhaps the most complex part of this setup, and the one that poses the most challenges to engineers. In order to select the most appropriate ear simulation device, the first question is whether the test should be a realistic simulation of actual use (commonly performed in development) or a highly repeatable test capable of differentiating defective units from good ones (production).
When developing a product, it is desirable to have a means of measuring under conditions that the end user will experience. A head and torso simulator provides this level of simulation, be cause it is equipped with artificial ears, which mimic the acoustic characteristics of the human ear, as well as artificial pinnae, which mimic the way a head phone would fit on a human. The fit of headphones and earphones can vary from person to person and even from one use to another, and can drastically change the user's listening experience.
Testing on a head and torso simulator can reveal this variability between fit and acoustic performance. This lack of repeatability is realistic and useful to understand how fit impacts the sound quality. Engineers will commonly take several measurements and average them to account for it.
Head and torso simulators are expensive (upwards of $20,000), so some R&D laboratories use less expensive (but with the expected performance trade-offs) alternatives such as cheek and ear simulators, couplers, simplified pinnae, and so on.
Head and torso simulators are not suitable for production line use for two reasons. As discussed, the fit may be slightly different each time, which makes it very difficult to achieve repeatable results on the production line, and also the cost is prohibitive.
This is overcome by designing a product with a known response using simulators (such as the Head & Torso) that accurately replicate a human, and then ensuring that all products coming off the production line match that ideal product. This means that different test methods can be used.
In production and QA it is important to use a fixture that is highly repeatable and can produce consistent results. An acoustic coupler, which is essentially a metal chamber that replicates the ear canal, is most commonly used. Couplers vary significantly in their complexity and cost. The most commonly used production coupler-a 2cc coupler, which is simply a 2 cubic centimeter cylindrical chamber-approximates the ear canal impedance.
More complex (and therefore more expensive) is an IEC 711 coupler, which has multiple internal chambers that more accurately replicate the acoustic impedance of a human ear. Although the characteristics of this coupler are close to the human ear in the lower frequency ranges, above 8kHz it is not well defined.
No manufacturer has produced a coupler that offers a truly accurate representation of human hearing above 8kHz because human ears vary more at high frequencies. For manufacturers who want to go one step further than a coupler, a pinna with simplified geometry is available (Fig. 7). Although not shaped like an ear, it is built to allow some degree of replication of the acoustic leakage that would occur with a real ear.
Even with the simpler geometry of a coupler, it is still difficult to achieve repeatable results, and most production line applications rely on custom fixturing to offer greater repeatability of mounting for a controlled seal. Usually a fixture is custom-built for a specific product to ensure repeatable attachment to the coupler. For circum-aural and supra-aural headphones, some sort of clamp is used to apply a consistent pressure, which aids repeatability.
Earbuds and in-ear earphones typically use a gasket of some sort to produce a seal, which produces more accurate low-frequency response and has the added benefit of reducing the influence of ambient noise on the test.
Now that you understand the test setup, I'll examine what you typically measure with a headphone. If you are buying a new test system for testing headphones, you should consider what and how you want to test before purchasing, because it has a bearing on the number of inputs and outputs you will need on your soundcard or test system. If you are testing a stereo pair simultaneously, you will need a two channel soundcard just for the acoustic tests. If you want to also simultaneously measure electrical characteristics such as impedance, you will need four channels. If you already have a two-channel system, you can measure impedance sequentially rather than simultaneously, although it will be a little slower.
A typical basic R&D test using a head and torso simulator for a stereo headphone pair would consist of playing a 1/12 octave stepped sine sweep from 20-20kHz and measuring the harmonic distortion and fundamental frequency response. Correction curves would compensate for the free field response from 0° incidence to the nose to the eardrum. You can then use post-processing to compare the left and right earphone responses and show the difference curve, both for magnitude and phase.
With a four-channel analyzer, you can expand this test with the addition of a couple of reference resistors to measure impedance of the headphones at the same time. Other options for more detailed testing would include measuring intermodulation and difference frequency distortion, or measuring maximum SPL using the distortion level to set the upper limit. Naturally, each company wants to test its products slightly differently, and most test and measurement systems offer the ability to customize tests.
A typical production line test would be similar, but you would need to apply a custom compensation curve, based on the coupler and fixturing.
On the production line, impedance, if measured, would typically be done simultaneously for speed reasons. Figure 8 shows an example test report for a R&D headphone test carried out in SoundCheck, a soundcard-based test system that is widely used both in the R&D lab and on the production line for measuring headphones.
IEC 60268-7: "Sound system equipment-Part 7: Headphones and ear phones" is probably the most comprehensive standard on headphone testing.
It discusses about how to classify ear phone types; e.g., supra-aural versus supra-concha describes various couplers and acoustic ear simulators, defines free-field versus diffuse field conditions, and explains how to test headphones.
For example, it specifies the standard test level to be 94dBSPL at 500Hz or 1mW at the earphones' rated impedance and defines how to measure the rated impedance. A copy of this standard costs approximately $200, and may be purchased from the IEC website.
This article has covered the basics of headphone testing, focusing on those issues that are applicable to all types of headphones.
These are each complex subjects requiring the use of different test signals and compensation for time delays, dropouts, and other characteristics uniquely inherent to these special types of headphone.
FIG. 9 A lab tech uses the setup described here with a head and torso simulator to carry out objective measurements of the audio quality of many different brands of headphones. The audio test forms part of a complete battery of tests which also examine noise cancellation and sound leakage. They also conduct additional tests that provide readers information on the comfort, usability, and design of headphones.
= = = =
Prev. | Next