100 Things #91: Measurement of Intermodulation Distortion

Intermodulation Distortion measurements are a great alternative to harmonic distortion for measuring narrowband devices such as hearing aids and communication devices. In such devices, harmonic distortion measurements tend to underestimate the distortion as the higher-order harmonics fall outside the pass band of the device. In this short video, Steve Temme demonstrates and explains the two IM distortion measurement options in SoundCheck – intermodulation distortion and frequency distortion and discusses how they can be used for low frequency speaker measurements, narrowband devices and microphones.

Measurement of Intermodulation Distortion

Learn more

Read on about more analysis capabilities in SoundCheck.

Video Script:

Although harmonic distortion is perhaps the most commonly measured distortion metric, it’s often not ideal for measuring narrowband devices such as hearing aids and communication devices. These products often have a high frequency cut-off around 3-5 KHz, so the higher-order harmonics fall outside the pass band of the device, so harmonic distortion measurements often underestimate the distortion.

A useful alternative we offer in SoundCheck is intermodulation distortion. Intermodulation distortion relies on the interactions between two simultaneous pure tones to produce measurable intermodulation products. These measurements actually present a more realistic representation of real-world signals such as speech and music that are rich with intermodulation products than the single tone used in harmonic distortion

SoundCheck offers two intermodulation distortion measurement options – Intermodulation Distortion and Difference Frequency Distortion. For Intermodulation Distortion, we superimpose a sweeping frequency tone against a fixed frequency tone. For Difference Frequency measurements, we use a stimulus consisting of two sweeping tones separated by a specified frequency interval, which can be a fixed difference or a fixed ratio. These are fully customizable.

In both cases, the two signals interact to produce intermodulation products. With Intermodulation Distortion, these are equal to the sum and difference of the upper frequency and integer multiples of the lower frequency. Difference Frequency distortion, only considers the components that are the difference and multiples of the difference, between the excitation frequencies.

Each type has its own specific applications. For example, Intermodulation distortion is mostly used for loudspeaker measurements, particularly at low frequencies, and Difference Frequency distortion is ideal for testing narrowband devices as the frequencies can be chosen so that the intermodulation products mostly fall within the pass band. This is easy to do in SoundCheck – simply configure your two test stimuli, and select your analysis – either Intermodulation Distortion, or Difference Frequency Distortion – in the analysis editor.

Intermodulation distortion is also a valuable technique for measuring microphones. Usually, the harmonic distortion from the source speaker playing the test tone is greater than the harmonic distortion that you are trying to measure from the microphone. However, if separate test tones are fed individually to two separate loudspeakers, the loudspeaker’s harmonic distortion has no influence on the measured intermodulation frequency components, enabling accurate measurement of the microphone’s intermodulation distortion.

To learn more about intermodulation and other types of distortion, check out our website, and stay tuned for a new in-depth seminar on distortion.

100 Things #87: Make Non-Coherent Distortion Measurements

Did you know that you’ve been able to make distortion measurements in SoundCheck with real-world signals such as speech and music since 2006? This is a valuable technique for testing modern devices with on-board DSP that filters out signals such as sine waves and noise. Non-coherent distortion measurements offer excellent correlation with perception and are easily implemented in SoundCheck. Steve Temme explains this technique in this short video.

Make Non-Coherent Distortion Measurements

Read more about making non-coherent distortion measurements

The 2006 AES paper on non-coherent distortion measurements is available to read from our technical papers library. This paper details all of the important considerations for making these measurements, including using a multitone versus music for a stimulus signal, understanding distortion measurement results, and more.

Video Script:

We talk a lot about harmonic distortion and transient distortion, but did you know SoundCheck also offers non-coherent distortion measurements? In fact, I believe we were the first audio measurement company to include this option.

Non-coherent distortion is a broadband distortion metric that includes harmonic and intermodulation distortion as well as noise. It offers better correlation to perception than harmonic or intermodulation distortion alone, and it can be used with real-world test signals such as speech and music as long as there is enough energy in the frequency range of interest. Otherwise, you might just be measuring background noise. I usually make these measurements in the nearfield to reduce background noise by placing the microphone close to the loudspeaker. This is particularly useful for the many modern devices that feature DSP that treats pure tones as noise and tries to filter them out.

Non-Coherent Distortion is a normalized cross-correlation measurement that determines the degree to which the system output is linearly related to the system input.

There’s a lot of complex math behind this – if you want to know more about that you can read our 2006 AES paper. Here, I’m just going to show you a quick demonstration.

Configuring non-coherent distortion in SoundCheck is a simple checkbox in the transfer function analysis editor.

I have a good speaker, and a speaker that exhibits some fairly significant distortion. Let’s look at the good speaker first. I’m going to play a short excerpt of Bird On A Wire at 90dB SPL by Jennifer Warnes – this song is widely used as a test track as it has good dynamic range.

And if you look at the results, you can see non-coherent distortion in percent per square root Hertz (spectral density) versus frequency. Since non-coherent distortion uses a broadband test signal for measurement, there is no direct correlation to harmonic or intermodulation distortion in percent. Typically the distortion level appears much lower than harmonic or intermodulation distortion because the test signal energy is spread out over the entire frequency range and not a single frequency for measuring harmonic distortion.

Now I’m going to play the same song on a speaker that I know shows some fairly heavy distortion

Now, looking at these results, you can see the non-coherent distortion is considerably higher than the good unit, especially at low frequencies.

So that’s it. Non-coherent distortion offers a way of measuring transducers with real-world test signals that correlates well to listener perception. To learn more, check out our AES papers on the subject, or download our free test sequence for non-coherent distortion measurement.

100 Things #72: Transient Distortion Analysis With enhanced Loose Particles

Listen’s enhanced Loose Particles algorithm measures transient distortion independently of harmonic distortion, with correlation to human hearing and offering high accuracy, easy perception, and fast analysis. eLP is ideal for identifying transient distortion in any environment, like identifying rattling buttons in audio devices, and automotive Buzz, Squeak and Rattle (BSR) measurements. SoundCheck sequences can quantify measurements into an eLP event count, making limit setting reliable in production environments.

Measure transient distortion independently from Rub & Buzz distortion

Try it Yourself

You can try this algorithm for yourself! Visit our enhanced Loose Particles page for all things eLP including methodology, algorithm research, and more

Learn more about enhanced Loose Particles

Listen founder and president Steve Temme, and support manager Steve Tatarunis, co-wrote a technical article on practical impedance measurement. The paper dives into both single channel and dual channel measurement methods, considerations, techniques, measuring thiele-small parameters, and more.

Video Script:

Did you know that SoundCheck’s new enhanced Loose Particle algorithm measures transient distortion independently of harmonic or Rub & Buzz distortion? This has many applications ranging from production line driver testing to detecting rattling buttons, wires, fasteners and other loose parts being vibrated by the loudspeaker driver, for example, in a car door or smart speaker.

This is really useful to help you find the root cause of many production line problems. It’s also really easy to correlate with what you hear. Let’s take a look.

This is a very simple set-up I have here, pretty much what you might use on a production line – a measurement microphone, SoundCheck, and an AmpConnect 621 audio interface with a built-in amplifier. I have 2 loudspeakers, a good one, and one that exhibits significant transient distortion.

Here’s the good one…

And here’s the bad one…

Let’s look at how our new enhanced loose particle algorithm analyzes them. 

This one’s the good speaker. You can see a nice smooth response waveform.

And here’s the speaker with transient distortion – this actually has a loose solder bead trapped inside the voice coil gap. You can see these little glitches on the waveform, that’s the rattling solder bead.

In the first step of our algorithm, we remove the fundamental, leaving just the distortion artifacts. And we can play this through SoundCheck to listen to the recorded artifacts and correlate what you hear to the measured results. Here’s the good speaker… and here’s the bad one…. This really helps us understand how the measurement relates to what we hear, and helps us set limits. 

Next we apply an envelope analysis to get a time domain measurement… you can see the energy in the recorded waveform, plotted against time. You can see the transient defects as random bursts of energy. And you can also see that we have a lot more defects in the bad-sounding speaker than the good one.

Now we calculate the Prominence of each of these energy bursts. In this context, prominence describes the magnitude of the peak relative to the adjacent minimums. It’s actually an established mathematical concept used in other branches of science, but we’re the first to apply this concept to audio measurements. This was actually the result of extensive research at Listen, in which we determined that the prominence of a peak reflects the impulsiveness of the distortion artifacts more accurately than the absolute magnitude of the peaks. 

So, calculating the prominence results in a numerical magnitude for each event in the time envelope, and we then set a threshold –  that’s the level above which the event will be counted as a transient artifact.  I’m using 10dB – this is generally a good starting point, although it varies depending on the acceptable distortion for the product. And remember, to help with setting limits you can play back the filtered artifact waveform to correlate the prominence level with audibility.

Finally, we count the number of events that exceed the threshold over the measurement duration – you can see that the Loose Particle count for the bad speaker is 110, compared to 0 for the good speaker.

We use the loose particle count to set a pass/fail limit. This gives us reliable results under a variety of conditions because, in a typical fast production measurement, background noise events typically only occur once or twice compared to many loose particle transients. So as long as you set the limit according to your environment, you can always get reliable results.

So there we have it! This new transient distortion measurement, accurately and reliably detects transient distortion, analyzing it separately from harmonic and rub & buzz distortion. It’s accurate even with background noise, and easy to correlate to audibility. Limit-setting is easier than with other methods, and it also works well in applications beyond driver testing such as Buzz Squeak and Rattle measurements in cars, and rattling components in audio devices.

Introducing Enhanced Loose Particles (eLP)

Watch our launch seminar for Listen’s latest algorithm, enhanced Loose Particles (eLP). This new algorithm significantly improves production line transient distortion measurement. It offers high accuracy, easy correlation to human perception, and identifies transient distortion artifacts separately from harmonic distortion for easy troubleshooting. It’s also ideal for identifying rattling buttons in audio devices, and automotive Buzz, Squeak and Rattle (BSR) measurements. This new algorithm is based on original research, and uses techniques not previously used for distortion measurement. Check out this demonstration and detailed explanation of how this algorithm works by Listen president Steve Temme.

Introducing Enhanced Loose Particles

Read on to learn more

For all things Enhanced Loose Particles including listening examples, published articles, and details of the eLP methodology, check out our Enhanced Loose Particles page.

Listen’s new enhanced Loose Particle algorithm offers accurate transient distortion measurements, even in the presence of background noise. This algorithm is highly accurate, as well as easy to configure and set limits. Results are easily correlated to audibility by listening to the loose particles in the recorded waveform.

This algorithm uses the same time envelope analysis as our pioneering (2004) Loose Particles algorithm, but rather than filtering and counting pulses, additional analysis is applied to measure the prominence, or impulsiveness, of the detected artifacts. The algorithm calculates the prominence of each loose particle event in the waveform by comparing the magnitude of a peak relative to the surrounding minimums, which reflects the impulsiveness of the artifact more accurately than the absolute magnitude of the peaks.  A user-defined prominence threshold determines the level at which the peak will be counted, and a user-defined count of events over the time window determines the pass/fail threshold.

Objective results obtained by this method are easily correlated to subjective analysis as you can listen to the recorded waveform with the fundamental removed, hearing only the loose particles. This enables the prominence threshold to easily be set based on defect audibility.

This analysis method is inherently reliable in a factory environment as external background noise events typically only occur once or twice during a measurement, whereas many loose particle transients will occur during the same timeframe. The event count is user-determined, and is set according to the background noise in the measurement environment. Prominence threshold and loose particle count are the only parameters that the user needs to define, so limit setting is simple, well correlated to audibility, and can be configured to give reliable results even when background noise is present.

100 Things #24: ePRB (Perceptual Rub & Buzz Measurements)

Not all distortion is audible. Perceptual Rub & Buzz measurements identify only those devices with audible Rub & Buzz defects, rather than all Rub & Buzz defects. Basing end-of-line pass/fail decisions on this metric increases yield while maintaining customer satisfaction.

ePRB (Perceptual Rub & Buzz Measurements)

Learn more about enhanced Perceptual Rub & Buzz

Our full ePRB launch seminar goes in-depth with algorithm explanation, testing demonstrations, use cases, and more.

For all things Enhanced Perceptual Rub & Buzz including listening examples, published articles, and a sequence to try ePRB for yourself, check out our Enhanced Perceptual Rub & Buzz page.

Video Script:

While conventional Rub & Buzz measurements identify faulty products and provide valuable information about the health of your production line, they can sometimes result in devices with inaudible distortion artifacts being rejected. To maximize yields, perceptual distortion metrics enable the rejection of only devices that actually sound bad.

SoundCheck is the only audio test system with a  perceptual Rub & Buzz metric that offers excellent correlation with subjective perception and sufficient noise immunity for production line use. 

We led the way with research in this area here at Listen. In 2011, we launched the first Rub & Buzz algorithm built on a model of human hearing . This original model was well received due to its excellent correlation with subjective listening tests. However, like the human ear, performance was less repeatable in the presence of background noise. Many hours of original research over the past few years have resulted in our new Enhanced Perceptual Rub & Buzz algorithm. This  combines our original methods with proprietary noise suppression and a refined perceptual model to exceed the performance of the human ear. In other words, it replicates the hearing characteristics of a human ear in a quiet environment, even in the presence of background noise.

This means that it offers the excellent noise immunity needed for a wide range of manufacturing environments.

Let me show you what I mean.

These graphs show 3 speakers, a good, bad and borderline loudspeaker, each measured 10 times using 3 different perceptual Rub & Buzz algorithms; our original 2011 version, our new ePRB algorithm, and a perceptual algorithm from another audio measurement company. You can see that while all three indicate the relative magnitude of the distortion, in both our original algorithm and the ‘other’ algorithm, the variation in the repeated measurements for each speaker is inconsistent enough that it would be problematic on the production line. Our new algorithm, the middle graph, clearly shows much greater repeatability between measurements. This offers a high level of confidence in the results and makes it easier to set limits.

Correlation to subjective tests is also improved through the use of more comprehensive masking curves that include additional factors to more  accurately replicate the human ear’s behavior and reveal increased detail in the ear’s highly sensitive 500Hz – 2kHz range. 

Perceptual distortion metrics are a valuable end-of-line test addition as they  increase yield by passing products with inaudible distortion. That said, in most cases, it is desirable to also measure normalized Rub & Buzz and Loose particles, as these are convenient ways of monitoring your production line for early warning of any problems that could eventually lead to a returned product from a customer.

All three distortion measurements, along with a whole host of other end-of-line parameters can be made in SoundCheck simultaneously, using the same stepped sine sweep stimulus signal. In other words, there is no increase in test time when you add perceptual metrics to your end-of-line test. Check out our website for more detailed information, demo sequences to try this out, and more.

100 Things #21: Normalized Rub & Buzz Measurements

In this short video, Steve Temme explains how SoundCheck’s unique normalized Rub & Buzz measurement method offers greater accuracy, high speed, and excellent noise immunity for production line testing. Additionally, analysis of each harmonic separately yields valuable information to help identify the exact production line fault.

Normalized Rub & Buzz Measurements

Learn more about normalized rub & buzz measurements

This technical note on Harmonic Distortion Measurement, authored by Listen president Steve Temme, explains the relationships between sampling rate, stimulus frequency and measured harmonics for both normalized and conventional harmonic distortion measurements.

Video Script:

Higher order harmonic distortion is an important production line measurement for identifying Rub & Buzz defects introduced in the manufacturing process. SoundCheck uses unique algorithms for measuring Rub & Buzz distortion and  ‘Normalized Rub & Buzz.’  Let’s look at how these differ from conventional methods.

There are 2 commonly used methods for measuring Rub & Buzz. The ‘tracking high pass filter’ method uses a high pass filter to remove the lower harmonics and sums the remainder. This combines all the harmonics plus transient distortion caused by loose particles into a single metric. While this gives an overall indication of distortion level, it reveals nothing about the cause of the distortion. It is also susceptible to background noise, which can result in false rejects on the production line.

A more accurate, but slower method, sequentially measures individual harmonics using a tracking filter that moves from one harmonic to  the next. While this provides accurate and detailed information on each harmonic, it is not fast enough to use on automated production lines.

SoundCheck’s unique Rub & Buzz distortion methods, which date back to the mid 1990s,  offer more accurate measurements, high speed and excellent noise immunity for production line testing. Furthermore, they analyze each harmonic separately which provides a wealth of additional information to help identify the exact production line fault

Let’s take a look at the unique features that make up the SoundCheck algorithms. 

Firstly, our proprietary Harmonictrak algorithm applies advanced filtering techniques to exclude the noise between the harmonics. All the harmonics are measured discretely and simultaneously. This is extremely fast compared to other methods, and also highly immune to background noise. Furthermore, because it measures each harmonic separately, it provides valuable information about the precise reason that the device failed the test. For example, high levels of harmonic distortion in the 10-15th order range usually indicate a rubbing voice coil, whereas distortion at the 50th harmonic and above is frequently caused by vibrating voice coil lead wires hitting the cone. 

SoundCheck also offers an enhanced version of this called Normalized Rub & Buzz. In this version, we compare the harmonic levels to the fundamental level at their measured frequency before their ratio is plotted, rather than the fundamental level at the excitation frequency. This is more accurate, because it removes the effect of the non-flat frequency response from the distortion. This makes it easier to see the peaks in the distortion response independent of the peaks and dips in the fundamental response, and easier to set limits.

Both these measurement options can use a stepped sine wave test stimulus, so they can be measured simultaneously with other end-of-line parameters such as frequency response, THD and even perceptual Rub & Buzz. While the main objective of Rub & Buzz detection is to identify faulty-sounding speakers, the individual harmonic analysis offered by SoundCheck’s algorithms makes them also valuable for continually monitoring production line performance for any drift in characteristics that could lead to product failure with customers.

Although these methods are now 25 years old, they still out-perform all other methods for measuring higher order harmonic distortion on the production line. More details on these unique algorithms can be found in the published papers section of Listen’s website.

100 Things #15: Loose Particle Detection for Identifying Transducer Manufacturing Defects

Loose Particle Detection is a technique for analyzing random transient distortion in the time domain. It is an important measurement for end of line transducer QC as it identifies manufacturing defects caused by foreign particles such as glue or magnet fragments trapped in the gap behind the diaphragm or dust cap. This algorithm This differs from Rub & Buzz measurements which analyze the higher order harmonics in the frequency domain to detect periodic distortion.

Note: This video was made in 2022 and highlights Listen’s original Loose Particle Algorithm. This was updated in the 2023 release of SoundCheck (Version 21) to the new Enhanced Loose Particle Algorithm. While the principal of time-domain analysis remains, the new algorithm utilizes a new metric, Prominence, to offer results that better correlate to audibility. Check out the information on the new algorithm listed in the links below.

Loose Particle Detection for Transient Distortion Analysis

Learn more about SoundCheck’s Loose Particles (transient distortion) analysis

Watch the new 100 Things video introducing the updated 2023 version of this algorithm.

Watch the video introduction to Enhanced Loose Particles – Listen’s newest Loose Particle Detection Algorithm.

Read about the Enhanced Loose Particles algorithm.

Read the AES paper published about this new algorithm for transient distortion detection.

 

Video Script: Loose Particle Detection for Identifying Transducer Manufacturing Defects

SoundCheck’s unique Loose Particle Detection Algorithm analyzes random transient distortion in the time domain. This identifies manufacturing defects caused by foreign particles such as glue or magnet fragments trapped in the gap behind the diaphragm or dust cap –  otherwise known as ‘crap in the gap’. Although this algorithm has been in SoundCheck since 2005, I’m always surprised how few people are aware of it, since it’s so valuable for improving manufacturing yield.

Let’s first look at the types of distortion we see on the production line. Conventional Rub & Buzz measurements analyze periodic distortion by looking at higher order harmonics in the frequency domain. These usually indicate a rubbing voicecoil or similar. However, with loose particles, analysis in the frequency domain does not yield useful information.

Loose particle analysis examines transient distortion in the time domain. This identifies random clicking, popping or rattling noises made by particles trapped behind the diaphragm or dust cap.  We can visualize this by looking at both the time and frequency analysis of the signal together. Here, in the frequency domain, you can see the periodic distortion, in other words, Rub & Buzz. And here, in the time domain, you can see the random transients, or loose particles.

Let’s take a closer look at the algorithm.

Loose particle detection uses a swept sine stimulus and time-envelope analysis of the waveform response to capture the transients. Detection thresholds are set for the magnitude, duration, and number of transients to quantify their level and number over time. Limits and filters are applied to separate device transients from background noise. Fine tuning with these tools customizes the algorithm for your product and manufacturing environment.

Measuring the different types of distortion enables rapid troubleshooting and correction of manufacturing problems, and SoundCheck is the only production line measurement system that can accurately distinguish between periodic and random transients.

Loose particle detection is also useful for identifying loose or rattling components such buttons, keyboards, wires and fasteners. In the digital domain, it can catch Bluetooth dropouts, clipping and other digital signal processing artifacts. This functionality is available in all editions of SoundCheck – check it out!