Speakers distortion

[Krunok]

Yeh, but the better solution would be to lower the level at 120Hz exactly as much as the room "amplifies" it (i.e. classic room mode correction), meanwhile using an inverted phase 120Hz tone to reduce the distortion component of the 60Hz tone exactly as much as the transducer produces it.

Exactly. I just didn't mention the "classic room correction" step.

I was also going to ask, how would you tackle IM distortion?

Huh, no idea so far..

 

[Krunok]

Yeah that was my thought too. I'd like to know how effectively distortion is being measured before moving on to something else.

That was the initial idea. Thoughts about possibility to correct some of the distortion components came under the assumption distortion can be measured in a sufficiently precise manner. RTA MMM is giving quite consistent results for me, have you tried it?

 

[Cosmik]

Cone breakup and nonlinearity can be characterized for various signals and amplitudes. What I am not at all sure about is if you can really, practically, apply an electrical correction to solve a mechanical problem like that, and what it would do to the sound we hear. Given a cone entering modal operation, if you apply compensating signal to counter the mode, can you do that and still create the proper output? Not clear to me either way, but again not something I have thought about.

There may be distortion mechanisms in a speaker that can be compensated for through the use of lookup tables e.g. where motion feedback can correct certain types of distortion, then what it does could be learned offline and stored as a multidimensional lookup table (I think a neural network would be a possible way of doing this, where the signal's history is an input and, of course, as with an FIR filter, the signal's future can also be an input if latency is permissible). If the speaker's characteristics change through ageing or coil temperature or ambient temperature, the accuracy will suffer, of course.

But if the mechanism is cone break-up, I am surprised that anyone would imagine that it is correctable after they have thought about it! (myself included). Two ways of looking at it:

1. The fact you are getting distortion products from cone break-up means that the speaker is not doing what it is supposed to be doing in the first place. You are no longer in direct uni-dimensional control of what it does, so why should there be any reason to believe that it will obey your instructions if you feed it compensating signals? It might be possible to find a compensating signal that works, but you will not be able to derive it from a single-pass measurement: it would need you to iteratively train the system to do it through trial and error- and this is only systematically possible if the function is differentiable. Is the function differentiable?
2. In economics, 'pushing on a piece of string' is a phrase often used to illustrate what happens when you reach the end of your ability to control the economy, and everyone intuitively knows what that means. If your speaker cone refuses to move quickly enough to remain pistonic and begins to flex instead, that seems like a system that is beginning to resemble the behaviour of that metaphorical piece of string. Intuitively, there is no reason to believe there is a signal you can feed into one end of it that will produce the correct behaviour. Even if you manage to find one that produces an aggregate correction at one position in space, it won't be correct (possibly worse) at another position because the flexing produces peculiar dispersion & lobing effects.

 

[Krunok]

But if the mechanism is cone break-up, I am surprised that anyone would imagine that it is correctable after they have thought about it! (myself included). Two ways of looking at it:

1. The fact you are getting distortion products from cone break-up means that the speaker is not doing what it is supposed to be doing in the first place. You are no longer in direct uni-dimensional control of what it does, so why should there be any reason to believe that it will obey your instructions if you feed it compensating signals? It might be possible to find a compensating signal that works, but you will not be able to derive it from a single-pass measurement: it would need you to iteratively train the system to do it through trial and error- and this is only systematically possible if the function is differentiable. Is the function differentiable?
2. In economics, 'pushing on a piece of string' is a phrase often used to illustrate what happens when you reach the end of your ability to control the economy, and everyone intuitively knows what that means. If your speaker cone refuses to move quickly enough to remain pistonic and begins to flex instead, that seems like a system that is beginning to resemble the behaviour of that metaphorical piece of string. Intuitively, there is no reason to believe there is a signal you can feed into one end of it that will produce the correct behaviour. Even if you manage to find one that produces an aggregate correction at one position in space, it won't be correct (possibly worse) at another position because the flexing produces peculiar dispersion & lobing effects.

Sure, distortion coming from cone break-up cannot be compensated this way. But, with the drivers made of modern materials (carbon, kevlar, etc. ) I would imagine that cone break-up happens only on very loud levels. Or..?

 

[Krunok]

There is also no doubt that it would be much better if "distortion response" compensation is done based on measurements in anechoic envrionment. The same goes for frequency response EQ.

When I was doing room EQ for my speakers I made significant corrections only in the 20-300Hz region. I barely touched 300-600Hz region (1-2dB) and after 600Hz I didn't correct at all.

 

[DonH56]

I suspect actual cone breakup cannot be compensated and tend to agree with @Cosmik. Other distortions (nonlinearities) before that point, I would think so, and have as evidence my original DIY sub from ca. 1980 and current Rythmik subs. All use servo control (feedback) and measurably reduce distortion.

 

[Floyd Toole]

I suspect actual cone breakup cannot be compensated and tend to agree with @Cosmik. Other distortions (nonlinearities) before that point, I would think so, and have as evidence my original DIY sub from ca. 1980 and current Rythmik subs. All use servo control (feedback) and measurably reduce distortion.

It is unfortunate that "cone breakup" is used to describe all non-pistonic motion of cones and domes. The low-order flexural modes respond to minimum-phase equalization, just as low-frequency room resonances do. Standard design principles aim to push the first flexural mode above the crossover frequency, which these days is very possible. When true chaotic "breakup" occurs one hopes that it is above the operating frequency range of the transducers - that is good design as I understand it from my engineering colleagues.

 

[DonH56]

It is unfortunate that "cone breakup" is used to describe all non-pistonic motion of cones and domes. The low-order flexural modes respond to minimum-phase equalization, just as low-frequency room resonances do. Standard design principles aim to push the first flexural mode above the crossover frequency, which these days is very possible. When true chaotic "breakup" occurs one hopes that it is above the operating frequency range of the transducers - that is good design as I understand it from my engineering colleagues.

Very good point; I was mixing together the flex modes and true "breakup". I dug up my old acoustics book and it describes the different modes (great reference, but these days my eyes glaze over with any text that begins with wave equations then delves into multi-dimensional Fourier analysis before getting to the pretty pictures of cone modes). I had forgotten some of them, or at least not thought about them in ~forever. And everything I have studied (not all that extensive) agrees with you and your engineers (natch) -- good designs push all modes above the crossover, and breakup (true quasi-random breakup) above audibility (or at least way above crossover so effectively the same thing). Much easier with modern cone materials and coatings; I remember hauling in the driver for my sub (from an Infinity IRS -- heavy poly cone) to compare with an ancient paper-coned driver my professor had so we could view modes and breakup in the lab. Fun stuff!

 

[Cosmik]

Linkwitz tests cones for breakup distortion:
http://www.linkwitzlab.com/mid_dist.htm

And an interesting thread on the consequences of cone breakup with loads of interesting info:
https://www.diyaudio.com/forums/multi-way/327183-cone-break-quantify-measurements.html

Basically, you don't want it. Going three-way (or more) and using DSP will help you a lot compared to struggling with two-way passive.

 

[DonH56]

Didn't Siegfried pass away late last year? Not sure, but seems like I read about it, what a loss... Hope they keep the website going!

 

[Floyd Toole]

Yes, Siegfried died of cancer, at home. He will be missed - one of the rational minds in audio. I discuss some of his ideas in my book.

 

[DonH56]

Yes, Siegfried died of cancer, at home. He will be missed - one of the rational minds in audio. I discuss some of his ideas in my book.

Ah, thanks Floyd, I was afraid of that... We've lost a lot of good folk personally and professionally the past couple of years, within our family and among the music and audio (and engineering in general) industries. Seems like a lot of the people I looked up to over the years have passed on as my wife and I are hitting "that age". You take care of yourself, sir!

 

[Deleted member 2944]

Didn't Siegfried pass away late last year? Not sure, but seems like I read about it, what a loss... Hope they keep the website going!

https://www.audiosciencereview.com/...peaker-designer-passes-away.4526/#post-101716

Dave.

 

[DonH56]

Thanks Dave. You'd think I'd remember better since I commented in that thread; senility (or just too %^!$ many work hours!)

 

[Blumlein 88]

I had never paid attention to the REW RTA which allows you to add distortion and control the phase of that distortion.
So in terms of the idea of cancelling out distortion, I used a UMIK 1 at 12 inches from an LSR 305. And added distortion while varying phase.

So take a look at he screen shot.
THD of .046% with the highest being the 3rd harmonic at .027%. Look at the level and phase under the distortion control. Oddly you need +106 degrees at -51 db for 2nd harmonic and -60 degrees 3rd harmonic at -48 db for lowest values on the 2nd and 3rd harmonics. I'll need to try this in different positions. It at least shows in principal it is possible to reduce distortion by adding into the original signal.

Without adding distortion the THD was .41%, 2nd harmonic .25% and 3rd harmonic .31%. So that reduces 2nd harmonic to -72 db vs -52 db.

 

[pkane]

I had never paid attention to the REW RTA which allows you to add distortion and control the phase of that distortion.
So in terms of the idea of cancelling out distortion, I used a UMIK 1 at 12 inches from an LSR 305. And added distortion while varying phase.

So take a look at he screen shot.
THD of .046% with the highest being the 3rd harmonic at .027%. Look at the level and phase under the distortion control. Oddly you need +106 degrees at -51 db for 2nd harmonic and -60 degrees 3rd harmonic at -48 db for lowest values on the 2nd and 3rd harmonics. I'll need to try this in different positions. It at least shows in principal it is possible to reduce distortion by adding into the original signal.

Without adding distortion the THD was .41%, 2nd harmonic .25% and 3rd harmonic .31%. So that reduces 2nd harmonic to -72 db vs -52 db.
View attachment 20106

Never tried to adjust the phase in harmonic generator tool. I assume it works with single tones only by adding the appropriate harmonics based on the main frequency. If so, this harmonic cancellation wouldn't work with anything more complex, like music, that contains more than a single frequency.

As I understand it, harmonic distortion is a symptom rather than the cause. It's the result of a non-linear transfer function being applied to audio signal. Cancelling individual harmonics may help make things look prettier in a test, but I assume the real cure is to measure and to apply the inverse of the nonlinear transform. I see there was a similar discussion here previously:

https://www.audiosciencereview.com/forum/index.php?threads/distortion-cancellation-experiment.1843/

ESS DAC chips have some sort of configurable non-linear compensation built-in. I'd be curious to find out how that works, and if there's a way to do something similar to correct for the 'whole system' nonlinear transform, including the speakers.

 

[DonH56]

I had never paid attention to the REW RTA which allows you to add distortion and control the phase of that distortion.
So in terms of the idea of cancelling out distortion, I used a UMIK 1 at 12 inches from an LSR 305. And added distortion while varying phase.

So take a look at he screen shot.
THD of .046% with the highest being the 3rd harmonic at .027%. Look at the level and phase under the distortion control. Oddly you need +106 degrees at -51 db for 2nd harmonic and -60 degrees 3rd harmonic at -48 db for lowest values on the 2nd and 3rd harmonics. I'll need to try this in different positions. It at least shows in principal it is possible to reduce distortion by adding into the original signal.

Without adding distortion the THD was .41%, 2nd harmonic .25% and 3rd harmonic .31%. So that reduces 2nd harmonic to -72 db vs -52 db.
View attachment 20106

Nice work. One catch in applying distortion reduction in real time is that you must deal with the phase over frequency and multiple signals. That's what the phase-plane compensation method does (for ADCs; I also applied it to DACs (new at the time), and think Dan got a paper out of that work -- not me, I was at one of those "don't let the secrets out" companies though may have been a co-author, too long ago) but making it work for the general case is challenging. The big advantage of the technique is that, once you have the table generated, it is relatively signal-agnostic. The disadvantage is the table gets big and processing it on the fly tough, at least for GS/s converters, but it'd be interesting to play around with the concept at audio rates.

 

[Blumlein 88]

Another trivial post. I managed to get 2nd THD down to .0066% (-82 db) and 3rd to .0081%. In fact it was lower than this. It was low enough ambient noise would effect it some. I saw it drift as low as .0046% (-86.7 db) a few times, but couldn't manage to get a screenshot of it.

 

[Erik]

JBL vs ML at 82dB at the listening position.

View attachment 19841

Here are LSR305 THD measurements from thxstandard.com (the site is closed now)

 

[AnalogSteph]

Here are LSR305 THD measurements from thxstandard.com (the site is closed now)

View attachment 20215

Interesting. Looks like there's some cancellation of 3rd going on in the bass there... probably excursion-related distortion is mixing with L(I) distortion. The LF driver would appear to have a frequency response peak at around 3-3.5 kHz. Not entirely sure what's going on up top.