Controling Cone Breakup In An Active System

[Deward Hastings]

Re: Controling Cone Breakup In An Active System

If the cone breakups are suppressed 30 dB below the tweeter SPL, they have no noticeable effect upon the summed response. They are masked by the tweeter response.

Therein are two of your flawed assumptions. You assume that the effect of cone breakup is confined to or resides primarily in the breakup region itself. And you assume that the distorted signal represented by the peak will be effectively masked by the undistorted signal from the tweeter, rather than appearing as distortion of that signal.

There is a world of evidence that simple harmonic distortion at levels up to 1 percent or more is either not perceived, or where perceived is regarded favorably (SET, for eample). The same is not true for intermodulation or spectral contamination at that level, which is both perceived and regarded unfavorably.

[daryl]

Re: Controling Cone Breakup In An Active System

Thanks Deward Hastings for agreeing with my post.

I am now going to repay your kind gesture of agreement by dissagreeing with what you have to say (you should have waited to see if I agree with yours before you agreed with mine, now your gonna have to backpedal).

Simple frequency response curves make it appear that breakup is indeed “out of the passband” . . . as if the peaks (and dips) on the graph are its only effect.

A simple frequency/phase transfer function curve is THE MOST IMPORTANT and REVEALING test in audio.

Frequency/phase curves and impulse/step curves are both a systems linear transfer function and are used to diagnose LINEAR issues only.

The transfer function does indeed show if resonances are out of the passband and as far as LINEAR issues are concerned the peaks and dips on the graph ARE the ONLY effects of cone resonance.

And simple harmonic distortion measurements, particularly the way they are commonly made with a swept filter, miss completely the IM and “spectral contamination” distortions which are, I believe, the primary “signature” of cone breakup. Even a couple percent of IM or SC distortion is not visible in frequency response or harmonic sweeps (although the non-linearity revealed by harmonic distortion is a reasonably good proxy for IM in some kinds of electronic circuits).

Simple harmonic distortion measurements show the non-linearity of a system.

The simplicity of tests is designed to make imperfections look WORSE not better.

Even in controlled listening tests test-tones reveal system imperfections far more clearly than actual music by magnitudes.

There are only three main categories of reproduction error: linear, non-linear and noise.

All imperfections MUST fit into one of these bins.

Harmonic distortion, intermodulation distortion and spectral contamination are all in the non-linear family.

They are not separate or individual.

One in most cases is a reasonable proxy for another and which you use is dependant upon which applies best to the effect you wish to see.

There are special cases however where HD will not properly show the IM or SC issues.

For instance in the speacial case of inductance modulation due to cone excursion you need to use IM or SC methods to view it because a single-tone HD won't show you the effect because the inductance isn't high enough to cause modulation at frequencies low enough to cause suitable cone excusion and cone excursion isn't high enough at frequencies high enough to be effected by inductance modulation.

Thus for inductance modulation measurement you need a low tone to cause excursion and a high tone to be modulated (an IM or SC measurement).

This is not the case when looking for the non-linearity associated with cone resonance since any tone which drives the cone into non-linearity will be distorted by that non-liniarity and clearly show the signifigance of the problem.

Also the simple HD plot makes the problem look WORSE than it actually is because of it's extreme spectral density which concentrates all of the energy into the problem frequency where with actual music the energy would be dispersed throughout the spectrum.

Shure IM and SC occur but they are indicated perfectly by the HD test and that is why it is used.

[Jeff B.]

Re: Controling Cone Breakup In An Active System

55 Watt Bi-Amplified Amplifier Module


$85 each. $70 each for four or more. Well isn't that cool.

So who's gonna get in there and figure out how to change the crossover frequency? :D

FYI - MCM's Sizzling Summer Catalog has these amps at $69.61 ea. Just refer to -
Catalog: SC13
Source code: 8SC208
Item Number: 50-6277
to get them at this price.

Jeff B.

[Deward Hastings]

Re: Controling Cone Breakup In An Active System

Simple harmonic distortion measurements show the non-linearity of a system.

The simplicity of tests is designed to make imperfections look WORSE not better.

This argument goes back at least 40 years that I know of, to when "HiFi" manufacturers generally decided to report harmonic distortion but not IM distortion because the IM figures made their equipment look worse.

One in most cases is a reasonable proxy for another and which you use is dependant upon which applies best to the effect you wish to see.

There are special cases however where HD will not properly show the IM or SC issues.

I submit that cone breakup IS one of those "special cases" . . .

This is not the case when looking for the non-linearity associated with cone resonance since any tone which drives the cone into non-linearity will be distorted by that non-liniarity and clearly show the signifigance of the problem.

Well . . . we do see harmonic distortion peaking at the subharmonic frequencies of the breakup. More important, though, we do not measure harmonic distortion in the breakup region, because our equipment typically does not reach high enough to identify it, and we, of course, cannot hear those harmonics. We rarely see two-tone or multi-tone tests taken in the breakup region which do produce measurable distortion components at frequencies which we can measure and do hear.

[daryl]

Re: Controling Cone Breakup In An Active System

Hi Deward,

I was just about to add an addendum to my previous post because I have thought of a possible mechanism for distortion requiring IM measurement.

Well . . . we do see harmonic distortion peaking at the subharmonic frequencies of the breakup. More important, though, we do not measure harmonic distortion in the breakup region, because our equipment typically does not reach high enough to identify it, and we, of course, cannot hear those harmonics. We rarely see two-tone or multi-tone tests taken in the breakup region which do produce measurable distortion components at frequencies which we can measure and do hear.

I believe harmonic distortion peaking at factors of cone resonance frequencies would be due to motor non-linearity generating harmonic energy at cone resonance frequencies.

As far as whether -30db is enough attenuation it would depend upon how much drive is necessary to cause the cone to distort at it's resonance frequencies and how much energy within a program is available to excite it after you consider the highest likely spectral density a program might have in the cone resonance region.

If you suppress a cones resonance freqeuncy 30db and if (big IF) the level of energy recieved at it's resonance frequency is still sufficient to raise it's HD to say 10% (-20db) the system overall will only show 0.3% (-50db) HD due to the 30db attenuation of the signal to the woofer at it's resonance frequency and the resulting 0.3% HD might be attributed to the tweeter.

Now if and that's another a big IF which is dependant upon the first big if, the cone being pushed into distortion also causes frequencies within it's passband to be distorted by the same 10% or something similar, then you could certainly say that 30db is not enough attenuation.

I am skeptical though because of the likely spectral density a program might have in a cones resonance region coupled with the already existing 30db attenuation and then how much of the distortion at the resonance frequency is likely to translate to distortion within the cones passband.

But I do see a possibility and think it would be warranted to conduct two-tone linearity tests with the tones being cone resonance frequency('s) and one tenth the cone resonance frequency('s).

But it is also important to gauge the likely spectral density at the cone resonance frequencies so you don't come to the wrong conclusions.

[neildavis]

Re: Controling Cone Breakup In An Active System

FYI - MCM's Sizzling Summer Catalog has these amps at $69.61 ea. Just refer to -
Catalog: SC13
Source code: 8SC208
Item Number: 50-6277
to get them at this price.

Jeff B.

Very good price, considering all that you get.

I've got a question about the lowpass crossover section. Is the schematic wrong, or is the design wrong, or am I reading something incorrectly? See the link for details:

http://techtalk.parts-express.com/sh...er#post1507920

[Deward Hastings]

Re: Controling Cone Breakup In An Active System

I believe harmonic distortion peaking at factors of cone resonance frequencies would be due to motor non-linearity generating harmonic energy at cone resonance frequencies.

I think that's largely true . . . but there are other mechanisms at work as well the effects of which we may not be successfully measuring. I speculate, for example, that the notch at 4kHz is the first breakup mode (and perhaps the only one that can be treated for practical purpose as a simple resonance). I speculate that is where the center of the cone moves one direction and the rim the other . . . the first "concentric mode" breakup . . . and the resulting cancellation of radiation produces the null. If one were "correcting" that in linear space one would use a peaking filter, not a notch.

I just have a nagging suspicion that when it is doing that it is not properly reproducing music at 500-1000 Hz. (or anywhere else for that matter). The loading on the voice coil (and the surround) from the flapping cone cannot be without deleterious influence . . . and I don't see how a single tone harmonic distortion test would measure that.

[Jeff B.]

Re: Controling Cone Breakup In An Active System

Very good price, considering all that you get.

I've got a question about the lowpass crossover section. Is the schematic wrong, or is the design wrong, or am I reading something incorrectly? See the link for details:

http://techtalk.parts-express.com/sh...er#post1507920

Well, I saw that post, but I can't really say what is correct for certain. I can say, however, that I modeled things as a 4th order LR at 3khz and went from there, and my final measured response matched what I was expecting and did not have a peak at 3kHz in the summed response. Other than that, I don't really know the facts.

[jkim]

Re: Controling Cone Breakup In An Active System

Therein are two of your flawed assumptions. You assume that the effect of cone breakup is confined to or resides primarily in the breakup region itself. And you assume that the distorted signal represented by the peak will be effectively masked by the undistorted signal from the tweeter, rather than appearing as distortion of that signal.

No, I didn't make any flawed assumptions. Of course, there are assumptions in there, but they are not flawed but instead 'assumptions yet to be tested.' What I'm saying is, input signals at breakup frequencies are masked by the tweeter response when they are attenuated enough (e.g., 30 dB down). But nonlinear distortions caused by the cone resonance, being excited by lower frequency signals, are not always masked this way (e.g., masked by harmonics in music), no matter how much the input signals at breakups are suppressed.

That's my point. This is a generally accepted concept and the basis of my take on this issue. What you're proposing needs to be backed up by some tests. For example, an assumpstion underlying the above concept is, the level of breakup-induced nonlinear distortions are independent of the level of input signals fed at breakup frequencies. If they are related, that is, indirect excitation of cone resonance by the driver's nonlinear distortion is significantly augmented by direct excitation by (even really low-level) fundamental input tones at cone resonance frequencies, it'll be worth suppressing breakup nodes as much as possible. I'd not contend that this is not a plausible scenario. But I haven't seen a test result that supports this hypothesis.

There is a world of evidence that simple harmonic distortion at levels up to 1 percent or more is either not perceived, or where perceived is regarded favorably (SET, for eample). The same is not true for intermodulation or spectral contamination at that level, which is both perceived and regarded unfavorably.

Higher odd order harmonics are generally known to be obtrusive even if they are at very low levels. My statement holds regardless of a type of nonlinear distortions. That is, nonlinear distorions due to the driver structure, whether they are simple harmonics or IMD, are constant terms added to the output whereas the level of harmonics contained in music (or more generally 'overtones' which aren't necessarily simple harmonics) keeps changing during music passages. This is why distortions due to the driver structure are not always masked by natural overtones in music.

[Deward Hastings]

Re: Controling Cone Breakup In An Active System

For example, an assumpstion underlying the above concept is, the level of breakup-induced nonlinear distortions are independent of the level of input signals fed at breakup frequencies.

There is no such assumption. IM and SC distortion varies with the levels of the interacting signals. A wildly flapping cone causes more distortion of other signal on the cone than a mildly flapping cone.

That is, nonlinear distorions due to the driver structure, whether they are simple harmonics or IMD, are constant terms added to the output whereas the level of harmonics contained in music (or more generally 'overtones' which aren't necessarily simple harmonics) keeps changing during music passages. This is why distortions due to the driver structure are not always masked by natural overtones in music.

You misunderstand both distortion and "overtones". Distortion components are not "constant", they vary with the driving signal. And "overtones" are

[jkim]

Re: Controling Cone Breakup In An Active System

You misunderstand both distortion and "overtones". Distortion components are not "constant", they vary with the driving signal. And "overtones" are harmonics.

You misunderstood what I meant by "constant" terms. Very hard to communicate if you tend to misunderstand this much. I did not mean that distortion components are at constant levels whatever the signals are. They vary, of course, depending on the input signals. What I meant is, given a fixed set of fundamental tone inputs, a profile of nonlinear distortions due to the driver structure is constant terms added to the output, whereas a profile of natural overtones contained in music keep changing even if two sets of fundamental tones in complex music passages are nearly identical.

This is a reason why a driver with particular motor structure and cone material can consistently exhibit its unique tonality across different types of music, compared to another driver with different distortion profiles, even if their frequency responses are matched as closely as possible. This phenomenon has been reported by a few people, including Jeff and John Krutke, who performed an experiment using two closely matched speaker systems built with different drivers.

And "overtone" is a more general term that includes "harmonic." Refer to the wikipedia definition. A musical instrument's timbre is not only determined by its unique profile of harmonics but by its profile of overtones, which also include non-harmonic elements.

[wingnut]

Re: Controling Cone Breakup In An Active System

There is no such assumption. IM and SC distortion varies with the levels of the interacting signals. A wildly flapping cone causes more distortion of other signal on the cone than a mildly flapping cone.



You misunderstand both distortion and "overtones". Distortion components are not "constant", they vary with the driving signal. And "overtones" are

Deward, my cones flap wildly all the time:D but I don't seem to notice all their FIM distortion. No disrespect, but I think you're picking nits (not to mention thinking too much) Obviously some distortion is much less offensive than other kinds, like the kind paper woofs and tube amps are famous for (sorry, forgot that all drivers should sound the same if a non-inferior xover is used).

[daryl]

Re: Controling Cone Breakup In An Active System

Deward you seem to be mixing up your non-linear errors with your linear errors.

Linear errors are those which alter the frequency/phase response or time domain characteristics of the signal.

Linear errors cannot cause energy to appear at frequencies where it did not exist before, only the relative magnitude and phase of already existing componets may be altered.

Non-linear errors on the other hand are those which add new frequency componets to a signal.

The new frequency componets however must have a mathematical relationship with the already existing frequency componets of the signal.

They must be multiples of existing frequency componets or sums/differences of sets of frequency componets or combinations of both.

If an error adds new frequency componets without a mathematical relationship to already existing ones then it falls into the final error category of the three main error categories noise.

Do you not find it even a little odd that when suspension or voice coil asymmetries produce distortion 60 dB below reference it’s called “non-linear” and regarded with great seriousness, while a cone flapping so wildly that it produces output 15 dB above reference (that clearly was not in the original) is regarded as “linear” and nothing to worry about? That’s a factor of over 10,000 difference in magnitude . . . and your concern is about the smaller, and you dismiss the larger, effect? Why not just equalize it to flat and listen to it? If it's "linear" that should sound fine . . .

If "suspension or voice coil asymmetries produce distortion 60 dB below reference" you have non-linear distortion since that is what results from asymmetry and you would detect this with a non-linearity measurement such as HD, IM or SC.

On the other hand if you have a resonance which produces a 15db peak as viewed on a frequency response chart this can only be described as a linear error since non-linearities cannot be shown on a frequency response chart.

Even though you know the cone will be extremely non-linear at it's resonant peaks the two are completely separate issues.

You view the resonance on a chart of linear transfer function (frequency/phase chart or step/impulse chart).

If the deflection of the cone material were completely linear the resonances and their peaks would still occur but no harmonic or intermodulation distortion would result.

Conversely if the cone material were way less stiff (but just as non-linear) and you measured only below the resonance frequency you would have the harmonic and intermodulation distortion without the presence of resonance.

The two are mutually exclusive and the fact that they occur together is due to the physics of the situation as described in a previous post in this thread.
http://techtalk.parts-express.com/sh...0&postcount=39

Yes you want to equalize out the peak in order set the linear transfer function to a suitable shape which will sum properly with the other driver(s) and you also want the problem areas attenuated sufficiently that the cone is not excited into distortion.

[Deward Hastings]

Re: Controling Cone Breakup In An Active System

Deward, my cones flap wildly all the time:D

Inferior crossover, eh? :D

but I don't seem to notice all their FIM distortion.

That will save you a little money (good crossovers are not particularly expensive) and a lot of effort in speaker selection.

[Deward Hastings]

Re: Controling Cone Breakup In An Active System

Linear errors are those which alter the frequency/phase response or time domain characteristics of the signal.

Linear errors cannot cause energy to appear at frequencies where it did not exist before, only the relative magnitude and phase of already existing componets may be altered.

Non-linear errors on the other hand are those which add new frequency componets to a signal.

I do hear what you are saying, and recognize the distinction you are making, but many things are lost in this simplification. Cone breakup is not a simple frequency/phase response or time domain characteristic. If it were breakup could be "corrected" with a notch filter, and there would be no breakup associated sound in the passband, only at the breakup frequencies. I'm pretty sure you're not arguing that to be the case. The sound of cone breakup is not modeled by a peaking filter, even if the resulting FR curve appears identical to the FR curve in the breakup region, and there is no notch filter that can undo it. And there is another level of complication introduced when different parts of the cone are radiating quite differently.

Most of the "mechanical" characteristics of breakup are neither symmetrical nor linear. Even the relatively simple first concentric resonance of the cone is highly asymmetric, since the resonating cone is rigidly bounded at the center to the voice coil and loosely bounded on the outside by the surround. When the breakup expands to radial modes, and circulating modes at the perimeter of the cone, I think "linearity" is out the window . . . no filter is going to correct that, and more important no filter is going to correct what that does to other signal on the cone. Want "new frequencies"? Look at the IM and SC products, visible as sidebands throughout the spectrum.

Just as a curiosity question . . . when we do HD sweeps we use a moving filter that tracks the fundamental and makes sure we're only seeing the harmonic we're looking for. Has anyone done FR sweeps through the breakup region with a filter to demonstrate that everything coming off the cone is a nice pure sine wave at the driving frequency? You know, "linear"? I know that people have done multi-tone tests while sweeping through breakup . . . and that "new frequencies" *do* appear. That qualifies as "non-linear", right? It's relatively easy to produce lots of IM while maintaining relatively low HD . . . how does that fit in? What do we call it when speakers do it?

On the other hand if you have a resonance which produces a 15db peak as viewed on a frequency response chart this can only be described as a linear error since non-linearities cannot be shown on a frequency response chart.

Even though you know the cone will be extremely non-linear at it's resonant peaks the two are completely separate issues.

If the deflection of the cone material were completely linear the resonances and their peaks would still occur but no harmonic or intermodulation distortion would result.

If the deflection of the cone material were completely linear we probably wouldn't call it "breakup" . . . . We'd probably just call it a bell (actually there are drivers that do that, and we call it "ringing", not "breakup" (I'd even be comfortable calling what the RS180 does at 4kHz "ringing", coupled with radiation cancellation).

The two are mutually exclusive and the fact that they occur together is due to the physics of the situation as described in a previous post in this thread.

I find it hard to describe two phenomenon that have a common cause (cone breakup) as "mutually exclusive". I think that's letting the models dominate the reality . . .

Yes you want to equalize out the peak in order set the linear transfer function to a suitable shape which will sum properly with the other driver(s) and you also want the problem areas attenuated sufficiently that the cone is not excited into distortion.

OK. So we have that notch at 4kHz. which I speculate is the result of radiation cancellation from different parts of the cone moving in different directions. What do you do about it ? ? ?

Regardless any argument about terminology I'm going to stick with the conclusion that cone breakup has negative ramification not only at the breakup frequencies but throughout the spectrum, and that the best prescription is complete avoidance, accomplished by low and steep electrical filtering.