Transitioning Audio Tests from R&D to the Production Line

Steve Temme discusses the transition from R&D to production testing in this March 2023 issue of AudioXpress. Production line audio testing poses many challenges such as noisy environments, harsh operating conditions, high throughput, relative limits, and more. In this article, Steve Temme shares his observations and outlines the main considerations to ensure a successful operation.

Full Article

 

Introduction to Production Line Audio Testing Article

Your product design is finalized, your prototypes are close to perfect, and your audio product is ready to move to production. But how do you ensure that the product hitting the retail shelves sounds exactly as you intended?
Your production line is probably thousands of miles away, with time zone and language complications. Pandemic-related restrictions make it hard to visit to configure, supervise, and train factory personnel. Furthermore, production is often outsourced to a contract manufacturer, adding another layer of complexity to both culture and information transfer.
Every production facility has one priority: to build products that meet the specifications with the greatest efficiency and highest yield possible for optimum profitability. Think about what that means for end-of-line tests. Their goal is to have as many units pass as possible. Tests must be fast to avoid slowing down the production line, and the test environment is generally noisy, often with variable temperature and humidity. The quest for high yields and manufacturing efficiency, along with challenging operating conditions and sometimes poor training, can result in unreliable measurements. When this happens, the only loser is you and your brand’s reputation.
I’ll never forget being told about a production line operator who increased pass-rate by opening the test box during measurements so that the factory background noise would mask the high-order harmonics to fool the perceptual Rub & Buzz algorithm. This might improve his yield, but it’s not so good when these products ship to customers! While this is one of the more outrageous violations of end-of-line test protocol, I have also seen many other questionable measurement practices over the years. In this article, I’ll try to de-mystify production tests and look at some simple measures that you can take to ensure a smooth transition.

Full Article

Automotive Max SPL Measurements

Measuring Automotive Max SPL ArticleIn this short article, Steve Temme discusses measurement of automotive Max SPL, and introduces the efforts of the Audio Engineering Society (AES) technical committee working on automotive audio to standardize the way essential attributes of complex automotive audio systems are measured across the industry. He explains why Max SPL measurements are important, defines this measurement, and describes the standardized measurement procedure suggested by the committee. Test configuration and physical setup is discussed, and example results presented.

Full Article

 

 

 

Full article text:

Measuring Automotive Max SPL
By Steve Temme Listen, Inc.
I am currently a participant in an Audio Engineering Society (AES) technical committee working group on automotive audio. This diverse group of about a dozen worldwide experts has focused on trying to standardize the way essential attributes of complex automotive audio systems are measured across the industry. Three specific measurements have been our initial focus: Frequency Response, Max SPL, and Impulsive Distortion. The committee’s proposals for measurements were presented for feedback at the AES Fall Online 2021 conference in a session titled “In-Car Acoustic Measurements.”

I presented our work on Max SPL Measurements, Hans Lahti (Harman) presented Frequency Response, and Stefan Irrgan (Klippel) presented Impulsive Distortion; the session was chaired by Jayant Datta. Here, I will describe our proposed method for Max SPL measurements.

Let’s start with why this is important. People need to be able to compare how loud an infotainment system can play in a car— manufacturers like to quote this in specifications, and consumers enjoy bragging rights about the sound level of their car stereo. Max SPL is defined as the maximum sound pressure level (SPL) that a car’s infotainment system can reproduce inside the cabin with the windows, sunroof, and convertible top closed. There are many ways this can be measured, but to keep it simple, two different measurements are recommended—overall Max SPL and Max SPL Spectrum regardless of distortion level. The reason we don’t take into account distortion when we measure the Max SPL is because it is difficult to characterize distortion in a modern-day infotainment system—these devices frequently contain much signal processing, and this makes them unsuitable for playing back the sine wave stimuli that are typically used for harmonic distortion measurements.

First, let’s examine the physical test setup. Our proposed test configuration replicates the position of an average person’s head in the driver’s seat using a precisely and specifically positioned six- microphone array in the driver’s seat. The height and the angle of the seat, the positioning of the microphones with respect to the seat, and the height and the angle of the microphones are clearly defined to ensure standardized measurements across all vehicles.

The sound system settings on the head unit—the tone control and fader—are set to the factory default setting; in most cases this is neutral or flat with no equalization. The head unit’s volume control is set to its maximum level using the volume control knob or digital user interface equivalent (e.g., volume level slider). Overall Max SPL can be measured using a microphone array with the six microphone signals power averaged by analog or digital means and connected to either a conventional or software-based sound level meter that can measure true RMS and be C-weighted, as described in the IEC-61672 standard. However, if a software-based system is used for measuring the Max SPL Spectrum, it is simpler to also measure the overall Max SPL through the software. Figure 1 shows a test configuration that makes both measurements simultaneously using SoundCheck software, and an AmpConnect 621 audio interface.

For both the overall Max SPL and Max SPL Spectrum measurements, a broadband (20Hz to 20kHz) monophonic pink noise stimulus is used. It has a crest factor of 15dB and is played for 30 seconds to make sure the system can sustain that level continuously. This is played at maximum volume to ensure the system is tested at the loudest signal the car will play. The sound source may come from any source—a memory stick, a CD, or Bluetooth from a smartphone or auxiliary line in. The average SPL in dB(C) is measured for 30 seconds. This is called a Leq measurement, and it takes the spatial average of the six-microphone array, power averaged, to get the overall Max SPL level (Figure 2).

The Max SPL Spectrum is measured using a real-time analyzer set to 1/12 octave resolution, 30 second linear averaging time and no waiting. This enables us to measure the level versus frequency irrespective of the human ear’s perception. The Max SPL is recorded at each microphone simultaneously from 20Hz to 20kHz and the power average calculated (Figure 2).

Listen offers a pre-written SoundCheck test sequence that measures both the Max SPL Spectrum and a single, power averaged value for Max SPL in line with the working group’s proposed guidelines. This enables consumers and manufacturers to measure the maximum overall SPL and maximum SPL versus frequency that a car’s infotainment system can reproduce inside its cabin. The sequence uses the method and test configuration with a six-microphone array in either the driver or passenger seats. It takes advantage of Listen’s 6-in, 2-out AmpConnect 621 audio interface, which seamlessly integrates with the software-based multichannel analyzer to measure, display, and average the results from the six microphones in real time, and power average them to calculate Max SPL. This sequence may be downloaded free of charge from Listen’s website. More details about these measurements, and the other measurement proposals developed by the technical committee, will be presented at the 2022 AES International Conference on Automotive Audio, June 8-10, in Dearborn, MI.

 

Further information on the AES Technical Committee on Automotive Audio, including a link to the working group’s draft white paper on can be found here: https://www.aes.org/technical/aa/

More about measuring automotive Max SPL.

Q&A with Steve Temme – Loudspeaker Industry Sourcebook 2022

Want to get to know Listen Founder and President, Steve Temme, a little better? Each year, Loudspeaker Industry Sourcebook features interviews with industry leaders in which they ask their opinions on current events, issues and trends in the audio industry. Read what Steve Temme had to say.
Full Article

End of Line Distortion Measurements

Steve Temme discusses the importance of detecting manufacturing-induced defects such as Rub & Buzz and Loose Particles during end-of-line testing, and explains the various algorithms that are used. He compares conventional and perceptual metrics for the measurement of Rub & Buzz, including Listen’s new enhanced Perceptual Rub & Buzz algorithm, and discusses why it can be beneficial to use both conventional and perceptual measurements in tandem.

Full Article

The Evolution of Production Line Rub & Buzz Measurements

Steve Temme discusses the evolution of production line Rub & Buzz measurements in this April 2022 issue of AudioXpress. Starting with  simultaneous analysis of higher order harmonics, he explains the progression of improvements including the greater accuracy offered by normalized distortion measurements, and progressing to the introduction of perceptual metrics. He covers the introduction of the first perceptual distortion algorithm introduced in 2011, and the newest enhancements to this which offer the repeatability necessary for successful end-of-line perceptual distortion measurement, where the reduction in false rejects and resulting higher yields add significant value to speaker and headphone manufacturers.

Full Article

How to Measure Free Field Speaker Response without an Anechoic Chamber

Did you knowSimulated Free field measurement you can make free-field measurements without an anechoic chamber? In the March 2021 issue of VoiceCoil, Steve Temme explains his unique method for achieving this. The article explains how the ‘splice’ method results in a full range frequency response from a combination of near-field and windowed far-field measurement, and compares the results with anechoic chamber measurements and the manufacturer’s published response curve.

Full Article

Listen Celebrates 25 Years

2020 is Listen’s 25th anniversary! When Listen sold its first product in 1995, it was the first soundcard-based audio measurement system, and its competition was often human listeners. Now, with an installed base of well over 10,000 systems and the industry-wide adoption of both soundcard-based architecture and highly automated sequence-based measurement, Listen is still leading the way with new algorithms and product development to meet the changing needs of the marketplace. In this article, reprinted from Loudspeaker Industry Sourcebook, Listen founder and president, Steve Temme, shares his story.

Full Article

Practical Testing of Voice Controlled Devices

Author: Steve Temme.  Reprinted from the Jan 2020 issue of AudioXpress.

This article discusses tools and techniques that are available to accurately measure the audio performance of voice-controlled and connected devices under the many various real-world conditions they may be used. It covers basic acoustic measurements such as frequency and distortion response, which have always been carried out on conventional wired systems, and the more complex real-world tests that apply specifically to voice-activated devices, along withthe techniques and standards that may be used.
Full Article

 

 

 

 

 

An Interview with Steve Temme

Author: Shannon Becker.  Reprinted from the March 2018 issue of Audio Xpress.

Shannon Becker interviews Steve Temme about how he founded Listen, Inc. and grew it from a start-up into an audio measurement leader.
Full Article

Practical Measurement of Bluetooth and Lightning Headphones

Picture of Bluetooth Headphone Testing Article reprint

Bluetooth Headphone Testing Article

Author: Daniel Knighten.  Reprinted from the July 2017 issue of Voice Coil.

In this article, Dan Knighten discusses Bluetooth headphone testing and Lightning headphone testing, specifically how to overcome the challenges of measuring headphones with wireless and digital interfaces such as Bluetooth, Lightning and USB-C to make the same measurements as on conventional wired headphones.

Full Article

 

 

 

Article Text

Practical Measurement of Bluetooth and Lightning Headphones

For decades, headphones have been passive devices with a direct, analog interface. Today, we are seeing a proliferation of headphones with wireless Bluetooth interfaces and various kinds of new and often proprietary digital interfaces. These new interfaces include Apple’s Lightning port and USB-C. In all cases these headphones present unique challenges to measurement because they cannot be directly connected to traditional test and measurement systems. In this article, we will explore how to overcome these interface challenges in order to make standard measurements on devices with nonstandard audio interfaces.

Closed Loop and Open Loop Testing
To begin, let’s define what we mean by open and closed loop testing. The test configuration for conventional headphone measurements, as seen in Figure 1, is what we term a “closed loop measurement.” This traditional type of audio measurement has been done for years with
all types of transducers (e.g., loudspeakers, headphones, microphones, etc.), and audio measurement systems can make these measurements without problems. The test signal passes from the audio interface through the speaker/headphone where it is converted to sound
pressure. Then, it goes through the microphone where it is converted back to voltage for analysis. The entire path from input to output is on the same interface, usually in the same domain (analog), and most critically, the analysis system’s input and output sample rate are
perfectly synchronous. The entire measurement from signal generation to capture of the device response simultaneously occurs with just a small amount of input to output delay added by the speed of sound.

In headphones with Bluetooth, Lightning, or other unconventional digital audio interfaces, this loop is broken. The input and output are on two different physical devices, which do not share a sample clock, and the signal goes through one or more analog to digital conversion stages. The delay from input to output of the device is likely quite long, compared to classic analog headphones. In fact, the delay might effectively be infinite. In the case of Lightning-connected headphones, there is currently no third-party solution available for injecting a test stimulus into a Lightning port. In Bluetooth systems, the connection is intrinsically non-synchronous. Bluetooth does not provide a synchronous sample clock across the wireless connection and instead relies on asynchronous sample rate conversion and various other techniques to maintain a glitch-free audio stream. This is what we mean by “open loop.” A closed loop system has a closed loop, synchronous signal chain.

An open loop system does not have a continuous or synchronous signal chain. However, SoundCheck makes it possible to measure all conventional parameters of a device, even when those devices are open loop devices, with a variety of tools including:
• Triggered acquisition—support for capturing measurements on playback devices
• File analysis—the ability to analyze signals captured by recording devices
• Resampling—conventional asynchronous sample rate conversion
• Frequency shift—the capability to align signals between non-synchronous systems

Let’s explore how these tools are applied in some typical test scenarios.

Bluetooth Testing
Figure 2 shows a typical Bluetooth setup. Since we are testing the Bluetooth headset using a Bluetooth interface, it is nominally a closed loop scenario. The audio signal comes out of the analog interface and is transmitted via the Bluetooth interface to the headset. It is then played
by the headset and picked up by the ear couplers where it is returned to the analog interface and computer for analysis. However, what makes this an open loop test is that Bluetooth does not transmit a sample clock and, therefore, the receiver and transmitter are asynchronous.

In this case, we will use frequency shift to align or synchronize the stimulus and response waveforms. Frequency shift uses a stationary reference tone to precisely find the difference in sample rate between two waveforms. Once the exact difference in sample rate is found, one waveform is then resampled with reference to the other. Frequency shift enables precise, conventional measurements to be made on Bluetooth devices despite their asynchronous nature.

When compared to a conventional test, only two changes need to be made. First, a short, stationary tone is pre-pended to the stimulus signal (see the Sidebar article). Typically 1 kHz for as little as 250 ms, this signal provides the frequency reference that the frequency shift step needs to align the stimulus and response signals in an asynchronous test scenario. Second, a
post-processing, frequency shift step is inserted into the test sequence between the acquisition and analysis step.

The rest of the sequence is identical to a conventional headphone test sequence and all normal parameters including frequency response, THD, polarity, rub and buzz, and so forth can be measured.

Lightning Headphone Testing
Any device that does not provide an analog or digital input and output is intrinsically an “open loop” device from a test perspective. Headphones that use the Apple Lightning port for connection are considered open loop because Apple does not provide Lightning audio
output adapters. The only device that can currently play audio into a Lightning headset is an iPhone. Measuring Lightning headphones requires an iPhone or similar Apple device to be used to store and play back the test signal (see Figure 3). This creates several open loop testing
challenges. To test a Lightning connected headphone, we will use three specific tools: triggered acquisition, resampling, and frequency shift.

Again, our test sequence will use a short 1 kHz tone, pre-pended to the normal test stimulus but this time it serves two purposes. First, it triggers a record-only acquisition, so that the test is automatically triggered when playback of the test signal begins. It is also used as the reference tone for frequency shift. Also, if our playback device, the iPhone, is using a different sample rate to the audio interface, we may need to use a resampling step. Finally, frequency shift will again be used to synchronize the stimulus and response waveforms. After the response waveform is captured via a triggered acquisition step and has been resampled and frequency corrected, calculation of the desired measurement parameters can proceed as with any conventional headphone.

Pre-written test sequences for both Bluetooth and Lightning headphones are available at no charge from Listen’s website, www.listeninc.com.

Lightning Headphone Microphone Measurements
Since most headphones now also include a microphone, it is worth mentioning how the microphone on a Lightning- connected headset is tested. The test sequence and method
for this is a little more complex, although ultimately it is really just the converse of testing the earphones. Figure 4 shows a typical test configuration. The preparation of the test signal and the use of resampling and frequency shift steps are identical to testing the earphones of a Lightning connected headset.

The difference is that instead of playing back the stimulus through the earphones and using a triggered, record-only acquisition, the stimulus is instead generated using a calibrated speaker or mouth simulator and recorded on an iPhone. The recorded signal is then transferred back
to the computer hosting SoundCheck and analyzed using a recall step to import the waveform into memory from storage on the iPhone.

Conclusions
Bluetooth and Lightning interfaces add an additional level of complexity to testing that is not there with their analog counterparts. However, because the SoundCheck test system is completely agnostic about where the stimulus and the response waveform are generated and
captured, these tests can be carried out with relatively simple modifications to existing test sequences. In fact, it pretty much comes down to a simple modification to the stimulus signal and some additional post-processing steps prior to analysis—all of which are easily automated.
This enables easy characterization and measurement of Bluetooth, Lightning, USB-C, and future devices with advanced digital audio interfaces.

 

Preparation of the Stimulus Signal (sidebar)
In SoundCheck’s frequency-shift algorithm, a Fast Fourier Transform (FFT) is used to extremely accurately calculate the centroid of a stationary tone. The result of this calculation can then be used to align or synchronize two signals even if they are sampled at different rates. A short, stationary signal is necessary for the frequency shift algorithm to lock on to. This is easily achieved by pre-pending a 1 kHz, 250 ms sine wave to the stimulus signal. Since SoundCheck enables the creation of compound stimuli, this short, single-tone burst can be followed with absolutely any test signal (e.g., a Farina log sweep, noise, speech, or other non-sinusoidal
waveforms).
The short sine wave also serves as the trigger tone for triggered record, as is necessary for testing Lightning headphones. The trigger tone clearly identifies the start of the signal. Care must be taken to set an appropriate trigger level. If it is too low, ambient noise can cause false
triggering; too high and it will never trigger. The trigger level should be set so that it is above the ambient noise and below DUT output level. The Multimeter virtual instrument is an ideal tool for finding the optimal trigger threshold.

A typical stimulus signal for open loop headphone test is shown in Figure 1. This compound stimulus works for both Bluetooth headphones where the sine wave is used for frequency alignment and for Lightning headphones where it additionally serves as a trigger and reference for frequency shift.

 

Additional Headphone Test Resources

More about Bluetooth headphone testing

Headphone Testing main page