what do people think about multi-tone testing in relation to the suitable frequency range for horn+compression driver combinations: https://josephcrowe.com/blogs/news/distortion-and-sound-quality-what-is-reasonable
Tones modulate in the presence of a non-linearity, so in a sense IMD is just another way of representing HD.
I've been making two-two and multi-tone measurements lately, and am getting convinced they are more important to SQ than THD.
I question if they are related to harmonic distortion.
IMD appears to show up quickly at lower SPL levels, and hasn't increased as fast as THD, with increasing drive levels.
Dunno yet, need for the wind to die down to test at some SPL outside, but IMD testing whatever the form (2-3-multitone) , so far seems to correlate with my subjective take on how low (not) to use a CD (on a given horn).
The relationship between modulation and non-linearity is fundamental in electronics and related disciplines.
Agreed, but when it comes down to how drivers react to complex signals, drivers with mechanical vibrations superimposed on the same diaphragm.... well, i think something is different than simpler non-linear electronic analysis.
For drivers, i think THD is different than IMD, and that assuming low THD equals low IMD is a mistake.
I may be full of bull, with more testing to do still, but so far THD doesn't seem to matter much...... whereas IMD makes an audible difference..
Perhaps you could start with why you feel that the electronics case is simpler, I don't see it that way.
THD measurements consider only the harmonically related frequency components, while anything else is considered rubbish / noise. Thus, the origin of the term THD+N, that is often used by amplifier manufacturers. With IMD, as the resulting components (sum and difference) need not be harmonically related to any of the tones used in the test, possibly making things less musical / more annoying.
Though it is generally thought that a system with low THD across the entire band of interest would have low IMD as well, it may not be as simple as this. For example, if a system is driven into (significant) non-linearity by a loud tone, it should follow (fairly easily) that any other soft tones being played back at the same time would also distort, even if their low amplitudes are individually incapable of producing audible distortion when played back on their own.
So, it is therefore not surprising that the two-tones used by IMD (which result in a composite signal) mimic music signals more closely than the single tones used for THD measurements. And, mechanical properties of the diaphragm, excursion movement and other thermodynamic phenomena inside the horn (and room) only complicate things further.
Though it is generally thought that a system with low THD across the entire band of interest would have low IMD as well, it may not be as simple as this. For example, if a system is driven into (significant) non-linearity by a loud tone, it should follow (fairly easily) that any other soft tones being played back at the same time would also distort, even if their low amplitudes are individually incapable of producing audible distortion when played back on their own.
So, it is therefore not surprising that the two-tones used by IMD (which result in a composite signal) mimic music signals more closely than the single tones used for THD measurements. And, mechanical properties of the diaphragm, excursion movement and other thermodynamic phenomena inside the horn (and room) only complicate things further.
Another distortion mechanism that won't necessarily be shown by THD measurement: https://purifi-audio.com/2020/04/28/dist/
You don't consider hysteresis a form of non-linearity?
Newvirus, while the effects of the two might be descibed as being different, that's not to say their source is. I think this is straying from the original point.
Newvirus, while the effects of the two might be descibed as being different, that's not to say their source is. I think this is straying from the original point.
It can cause non-linearity but its has dependancy on the past state of the signal so dosesn't fit into time invarient system theory.
I think you misunderstand why THD+N is specified. In most electronic devices (amplifiers, data converters, etc.) there is a noise floor. Some output that occurs even with no input (or input grounded).
For the sake of an example, let's say a device has a noise floor of -100db below full scale output. If you input a 1kHz tone at sufficient amplitude that the output is at full scale and this results in a harmonic output of -50db that's pretty easy to see. You could easily specify the THD as -50db and the THD+N would be about the same. Now lower the input signal such that the output signal is -60db below full scale. We would expect the harmonic generated to be 50db lower than the fundamental signal, so that would be -110db. However this is below the noise floor. We can't measure it. This is where THD+N specifications are useful. In cases where the harmonic distortion is so low that it is frequently below the noise floor (data converters are a good example) we generally prefer to specify THD+N.
Electronics seem rather one dimensionally simple, compared to the complexity off adding in multiple mechanical components...diaphragm, surround spider, coil, etc .
Sure, we can say all distortion is ultimately non-linearities....and that those non-linearities produce both THD and IMD...but of what value is that statement?
I'm interested first in distortion audibility, and then, if audible, which non-linearity is to blame for the audibility.
Frankly, I'm inclined to agree with E. Geddes' take on THD's lack of audibility. But I can't say the same for his take on IMD.
IMD, with multiple tone sum and difference interactions,...isn't just a single sine wave modulating itself......and so far for me, appears to be bearing audible fruit as i try to reduce it.
Here's the first and last sentences from one of the Purifi articles on distortion. I'm getting more in line with last bold text part. (emphasis mine)
THD, IMD
When we talk about distortion, there’s often a distinction made between Harmonic Distortion (HD) and Intermodulation Distortion (IMD). These aren’t two different types of distortion per se, but different ways in which the same distortion mechanism can manifest itself more or less saliently.
<article body> https://purifi-audio.com/2019/05/02/distortion-the-sound-that-dare-not-speak-its-name/
Even though there are no standardised methods of presenting loudspeaker IMD measurements on paper, it is quite crucial to learn to think about loudspeaker nonlinearities in terms of intermodulation rather than harmonic distortion.
Hweb, I do not misunderstand why THD+N specifications are given, but just wanted to say that while making THD measurements, anything that is harmonically unrelated to the fundamental is considered noise, and therefore taking the ratio between the useful (fundamental) and useless things (distortion and noise) gives you an idea of how fidelitous your system is. Also, what you said is more or less the same thing.
THD+N from Tektronix
Yes, hysteresis definitely causes the system to not belong to the linear or time-invariant category, but independence from past states is termed as memorylessness.
When a pure tone distorts, the power in the harmonics (even in the worst-case) is going to be lesser than that in the fundamental, and is therefore likely to remain masked (musically) until significant power is applied. However, during inter-modulation, different tones are multiplied to generate beat frequencies with amplitudes proportional to the product of the amplitudes of the original tones, with no masking guarantee. Besides, there are also going to be higher order inter-modulation products due to the beat frequencies.
So, yes, it is indeed true that the source of these distortions is the same, but with a little thinking it should be possible to see why IMD could rise (and possibly become audible) earlier in power.
Consider a simple simulation where a simple tone of 1kHz and a dual tone of 1kHz and 10kHz (of same power) are passed through a non-linear system with a sinusoidal transfer function (y=sin x). The input (top) is in the time domain with the output (bottom) in the frequency domain. Now, AllenB, which one do you think would be audible earlier in power, considering the higher order products of IMD (green trace) ?
THD+N from Tektronix
Yes, hysteresis definitely causes the system to not belong to the linear or time-invariant category, but independence from past states is termed as memorylessness.
When a pure tone distorts, the power in the harmonics (even in the worst-case) is going to be lesser than that in the fundamental, and is therefore likely to remain masked (musically) until significant power is applied. However, during inter-modulation, different tones are multiplied to generate beat frequencies with amplitudes proportional to the product of the amplitudes of the original tones, with no masking guarantee. Besides, there are also going to be higher order inter-modulation products due to the beat frequencies.
So, yes, it is indeed true that the source of these distortions is the same, but with a little thinking it should be possible to see why IMD could rise (and possibly become audible) earlier in power.
Consider a simple simulation where a simple tone of 1kHz and a dual tone of 1kHz and 10kHz (of same power) are passed through a non-linear system with a sinusoidal transfer function (y=sin x). The input (top) is in the time domain with the output (bottom) in the frequency domain. Now, AllenB, which one do you think would be audible earlier in power, considering the higher order products of IMD (green trace) ?
Attachments
Last edited:
We're comparing apples to oranges here.
Harmonic distortion, by definition, has a single-tone input.
IMD tests must be multi-tone.
Direct comparison between the two, therefore, isn't close to straightforward.
Purifi make the interesting point that some harmonic distortion mechanisms might result in cancellations in the harmonics (eg, if the suspension and motor produce equal amounts of even-order HD, but in opposite directions), where those cancellations might not occur with inter-modulation tests. ie, IMD is more revealing.
A question, then: can anyone point to some measurements which show high IMD and low HD, or vice-versa?
Chris
Harmonic distortion, by definition, has a single-tone input.
IMD tests must be multi-tone.
Direct comparison between the two, therefore, isn't close to straightforward.
Purifi make the interesting point that some harmonic distortion mechanisms might result in cancellations in the harmonics (eg, if the suspension and motor produce equal amounts of even-order HD, but in opposite directions), where those cancellations might not occur with inter-modulation tests. ie, IMD is more revealing.
A question, then: can anyone point to some measurements which show high IMD and low HD, or vice-versa?
Chris
The free text states "The construction is a high impact ABS will glass filling added for additional structural support and rigidity. The acoustic loading offers controlled response down to 1200Hz".
My guess is that the 400Hz is a simple typing / translation mistake - 400Hz would require a substantially bigger device 🙂
What is your definition of close listening? A large horn and 15" woofer might have difficulty at close range sounding right.
I think the horn (Dayton H812) that I'm using is as large as I can go: it gives decent LF loading at the expense of modertate HF beaming.
With anything much larger, I think the beaming would be too much
The final build will be listened to at about 2m.
I currently listen to the test rig at about 1m 🙂
The test rig has the 15s recessed about 15cm, relative to the mouth of the mid horn. I needed that adjustment to make the sound cohesive.
The mid/HF transition seems OK, despite the HF horn being ~10cm shorter than the mid. A quirk of my setup or hearing sensitivity or music choices or something.