Stored energy and coloration of soft cones

[Ludvig]

People are often talking about that paper and poly cones suffer from stored energy and coloration due to their soft cones.

Stored energy:
There are however paper/poly drivers that have a very flat response. If we should regard a driver to have minimum phase behaviour in its passband, how can they have stored energy problems when the response is flat?

Coloration:
Is the coloration intruduced by some kind of "cone distortion" independent of motor linearity or what does this coloration come from?

Could these two behaviours be amplitude dependent so they are quiet and clean at lower SPL:s and starts to store energy and distort the sound at higher SPL because the cones can not handle the high accelerations?

On the contrary, the whole Pro market uses paper cones almost without exceptions for their high SPL applications, which points in a direction that paper cones do not suffer at high SPL:s.

Just a couple of thougts that maybe someone have something to say about.

[ThomasW]

There are 2 schools of thought regarding cone break up modes (something all drivers have).

First is the "we don't care" school. They ignore the modes and pretend they don't impact the performance.

The other school chooses to use drivers where their operation is most pistonic (linear). This means avoiding cone modes in the passband.

If I get a chance later today I'll post some graphics showing cone modes in drivers where the FR plots don't show them

[dlr]

I'm curious where you come by this

There are 2 schools of thought regarding cone break up modes (something all drivers have).

First is the "we don't care" school. They ignore the modes and pretend they don't impact the performance.

The other school chooses to use drivers where their operation is most pistonic (linear). This means avoiding cone modes in the passband.

If I get a chance later today I'll post some graphics showing cone modes in drivers where the FR plots don't show them

With regard to linear distortion, everything is in the frequency response. The CSD, step, impulse, all can be calculated from the FR. If there's a resonance, it will be there. If it's "hidden", the measurement is in error somewhere.

Any specific mode of a resonance won't be evident, but the fact that it exists has to show in the FR, CSD, step or impulse.

This was an interesting and detailed set of threads on this over at the Mad board some few months ago.

[atalio]

Stored energy:
There are however paper/poly drivers that have a very flat response. If we should regard a driver to have minimum phase behaviour in its passband, how can they have stored energy problems when the response is flat?

Coloration:
Is the coloration intruduced by some kind of "cone distortion" independent of motor linearity or what does this coloration come from?

Could these two behaviours be amplitude dependent so they are quiet and clean at lower SPL:s and starts to store energy and distort the sound at higher SPL because the cones can not handle the high accelerations?

Excellent observations. I spent some time on the Mad board about this issue. IMO, the sonic problems you allude to in lossy cones is real, is caused by the cone and not the motor, and is not easily discernable in a frequency response plot.

I do not believe that the steady-state frequency response or even a conventionally-plotted CSD can reliably convey all the information necessary to assess lossy cones. I also believe that drivers are essentially minimum phase, but small dynamic non-linearities can cause audible problems that are not easily visually detected in the measurements due to their low levels. Just as it would be quite tough to determine harmonic distortion from viewing a time or frequency domain plot of a driver signal, it may also be tough to determine the sonic "floppiness" of a driver using these visual methods.

What we need is a specialized measurement technique that can better isolate the lossy components even when they are quite small, relatively speaking.

[Dennis H]

While frequency irregularities may not show up in the far field, they are obvious in the near field. One of Jon's favorite methods is to hold the mic almost touching the cone and move it across the cone. The frequency where that starts changing the response is where the cone is starting to be non-pistonic. Now some will ask what is the difference as long as it's flat in the far field? The difference will usually show up in a non-linear distortion measurement of a pistonic vs a non-pistonic cone. About CSD's, they aren't particularly useful as they don't give any information that isn't in the frequency response -- just another way of looking at the same data.

[atalio]

Near field versus far field. That's great and it seems intuitively spot-on. Going further, it would be nice to devise a specialized distortion measurement - a graph or frequency dependent parameter - that could more succinctly isolate the floppiness from the other typical distortion mechanisms.

"The difference will usually show up in a non-linear distortion measurement of a pistonic vs a non-pistonic cone."

That's my feeling, too. I argued this point on the Mad board some time ago, but it was not met with a lot of enthusiasm. If you're changing the time envelope of the signal in a way that does not precisely follow a linear filter envelope, than non-linear distortion will occur. Some types of time alterations can be linear, e.g. a linear filter's effect on the onset of a step response. Others may not be, e.g. slew rate limiting, certain types of settling behavior, or strange acoustic phenomena. The difference between linear and non-linear is often not easily detected by viewing the time waveform. It may require sensitive distortion measurements, as you've stated. Conventional, well-behaved cone resonance would be considered minimum phase and linear.

"About CSD's, they aren't particularly useful as they don't give any information that isn't in the frequency response -- just another way of looking at the same data."

I agree, assuming minimum phase behavior (which is generally true for drivers, but not loudspeakers). Strictly speaking, I think even drivers depart from minimum phase, to some very small degree, and that shows up as a form of distortion. Some would argue about driver mimimum phase in more black-and-white terms, but I see it as minimally off-white and very dark gray when nonlinearities enter the picture.

[cotdt]

How about...

Input:
1. Nearfield response measured from center of woofer
2. Nearfield response measured from edge of woofer

Output:
Output would be the difference between the two. This would indicate how floppy it is. One specific test is the Peerless Exclusive versus the Seas Excel Magnesium, since they both perform great when it comes to nonlinear distortion, yet they sound completely different. Obviously, the Peerless Exclusive has a very soft cone by the finger test, but it is just as detailed-sounding as the Seas Excel. I think this might be the test that shows why they sound completely different.

[Jonasz]

cotdt: I've never heard the Exclusive series but I listen to the mag-Excels daily. If the Peerless is as detailed as the mags but sounds completely different, how would you describe the differences? Do the Peerless have a "warmer" character or is it less transparent?

[dlr]

I have to disagree with some assumptions being made

The distortion of the cone (floppy or otherwise) will show up as linear distortion, not non-linear distortion, since that distortion is fully signal-dependent, not due to the displacement of the coil in the field, the primary source of non-linear distortion. The fact that the level of linear distortion changes with the applied signal strength doesn't make it non-linear distortion.

The breakup region is also purely minimum-phase. I have used CALSOD to model breakup regions of 10" drivers with massively large breakup anomolies. CALSOD uses MPEs (minimum phase elements) as the building block for all non-linearities (deviation from flat response) to create a model that matches the FR of a driver. In every case, the summed magnitude and phase can be made to exactly match the measured response, to beyond 20KHz. Sometimes it's taken 30-40 of these MPEs highly optimized in narrow frequency bands to accurate represent FR. When that is complete, the phase is calculated and will match precisely as well, without failure.

Others at the Mad board (primarily Feyz supported by Paul V) showed the mathematical derivation as to why this is so, but I am not one capable of reproducing it.

The issue of near-field and far-field is also not fully supportable. I'm not disagreeing to disagree, but objectively I reach a different conclusion. Consider that all drivers of whatever kind radiate energy from all points essentially at different time delays other than when isolating a concentric ring on a dynamic driver when measuring on-axis only. Even when called pistonic, the speed of sound in the driver is finite, thus there is a delay. So the output in the far-field is the integration of these points. In fact, the delay in air can and will mean that the signal emitted from points farther from the former, but nearer the measurement point, will arrive sooner than the signal emitted at the former. The geometry will have a large influence on this, regardless of the material. But the material will alter it as well, of course, since the speed of sound in materials will differ.

At low frequencies this is not of much consequence, but becomes increasingly so as the frequency increases.

But this is still not the whole story. A driver has to create a an acousitc wave. I am not versed in wave mechanics, but I have discussed this with John K who is, quite some time ago. He has pointed out that there is a point in close proximity to the driver where the wave has not fully established. It is in essence a transition region, so a measurement there is not fully indicative of what the final wave motion will be. This to me makes extreme near-field measurement suspect as to ultimate impact.

The difficulty is still in defining what would constitute a complete set of measurments. Making large numbers of distortion tests with various signals and amplitudes are time consuming and never fully complete, but in the end if for a specific set of tests, distortion is what it is, no matter the material or geometry.

I suspect that much of the differences in perception has as much to do with off-axis response, as the cone materials tend to dictate geometry and we all know just how much of a role geometry plays. There's also the issue of the stop-band that IMO is too often ignored. We hear much more into the low level than many may think, at least that's my feeling. I'm amazed at subtleties that we can discern.

WRT to CSDs, they are good for determing where and to what length a resonance ridge occurs. At times a single ridge may show that it is actually composed of two or more ridges that build to create an apparent single ridge. It is also useful in differentiating resonance from other aspects. My recent tests of the RS52 shows to me at least that the 2K peak in the FR response is not a true resonance at that frequency as had been originally discussed here and elsewhere. In this case it can be shown that it is due to the vector summing of the integrated output from the dome due to time delays, even though the driver is well within it's "pistonic" region. I was able to essentially increase the midband sensitivity by 2db due to the phase interference reduction, no different than one sees in a hard-domed tweeter.

I see no reason for any of this to be different in a cone.

I know this was long, but it's not a simple topic.

dlr

[TacoD]

Wow thank you for this lecture Dave!

[atalio]

An example: I don't believe slew-induced distortion will show any marked anomalies in the phase response, by visual inspection. Similarly, two frequency plots that match "precisely" could belie radically different distortion profiles.

[Dennis H]

The distortion of the cone (floppy or otherwise) will show up as linear distortion, not non-linear distortion, since that distortion is fully signal-dependent, not due to the displacement of the coil in the field, the primary source of non-linear distortion. The fact that the level of linear distortion changes with the applied signal strength doesn't make it non-linear distortion.

That is incorrect.

Non-linear distortion is defined as the production of any frequency not in the original signal. Flexing of the cone produces frequencies not in the signal. As an extreme example the ringing of a Seas metal at several kHz can be stimulated by a signal at 1/3 that frequency. This is a cone phenomenon, not a motor phenomenon.

The same thing happens with soft cones but they show a broad region of breakup so there is a broad region of non-linear distortion. It may be stronger or weaker than the metal cone's breakup depending on the exact cones used but it will be there and it will be broad. Distortion curves of the vintage Seas drivers show this clearly. For a while they built the same unit with both metal and paper cones -- same motor, same basket.

Near field testing is not good for plotting the frequency response for the reasons you and John K noted. Nobody uses it that way. However, it's very good for finding breakups where part of the cone may be flexing forward and part may be flexing back. Yes, the fundamental will add minimum phase in the far field but the flexing implies that extra frequencies are being created and that is non-linear distortion.

W21EX Paper: http://www.seas.no/excel_line/excel/E004.PDF
W21EX Mag: http://www.seas.no/excel_line/excel/E013.PDF

[dlr]

Qualified agreement

An example: I don't believe slew-induced distortion will show any marked anomalies in the phase response, by visual inspection. Similarly, two frequency plots that match "precisely" could belie radically different distortion profiles.

First, I don't know precisely what you man by slew-induced distortion, please clarify.

Yes, if two precisely (to use your phrase) matched drivers have the same FR curve, they have can have quite different distortion profiles, but the qualifier is that they first have to match well into the stop-band on both sides to be considered even close to being matched (practically never the case), the distortion would have to be measured at various power levels (likely to show differences) and the odd-order non-linear distortion would likely be different even if the motors were the same (as is the case from many manufacturers).

Of course, the off-axis measurements would have to match precisely as well to quality as being equal. Using two different materials pretty much quarantees that this situatation cannot be achieved in the real world.

Since non-linear motor distortion from two different drivers that use exactly the same motor can exist and the linear distortion will be different, how can one say the problem is simply the cone material? It's not a black/white issue.

Further, it may well be that the overall distortion levels are similar, but the distribution of the distortion will likley be different between the two, yielding a case where they sound "different", but one not necessarily "better" than the other except for personal perception and preferences.

Some folks say that all hard-domed tweeters are too harsh, though they are certainly pistonic to a far higher point than softer domes.

Some folks have said that hard coned woofers, midwoofers or midranges have too much of a "metal" sound, but nowadays the Seas line is concentrated in drivers that have absolutely horrendous breakup. They also now have extremely low distortion, ...if the designer handles the crossover well... and uses the driver appropriately.

Should we write off the Seas Excel line because it has horrible linear distortion results at the extreme prior to crossover design?

Some folks prefer OW1 tweeters to some other drivers, even though the OW1 may measureably be worse in its distortion characteristics. So in that case, lacking full and complete distortion measurements, one might conlcude that the OW1 is the superior driver and make inferences as to details of its characteristics, when it is in fact the odd-order non-linear distortion that is perceived as "better". This has certainly been discussed here and at the Mad board specifically on the OW1, as some beieve that odd-order non-linear distortion is said to give a false perception of "air and detail".

Some folks prefer a smoother, less "etched" or "clinical" sound that may be provided by a softer diaphragm. Who's to say which is better?

So should we write off the OW1 because its measurements are inferior to some other tweeters?

It's all about tradeoffs. Categorically dismissing any particular diaphragm material is more in the realm of personal preferences to a point.

dlr

[dlr]

Yes, I lumped one into the wrong category.

That is incorrect.

Non-linear distortion is defined as the production of any frequency not in the original signal.

You're right, I lumped it into the wrong category. Thanks for the correction.

Near field testing is not good for plotting the frequency response for the reasons you and John K noted. Nobody uses it that way. However, it's very good for finding breakups where part of the cone may be flexing forward and part may be flexing back. Yes, the fundamental will add minimum phase in the far field but the flexing implies that extra frequencies are being created and that is non-linear distortion.

Agreed that flexing is distortion. My point is not that it isn't distortion or is one form or another, rather that it takes distortion testing to determine the type and level. Ascribing bad behaviour of a cone to the point of being dismissive without knowledge of the actual distortion characteristics is not warranted, IMO. It still comes down to the application and needs of the designer.

Personally, I can't see how I would use the knowledge that there is some specific breakup found in this mannter if I had measured distortion results for that driver. I'm more interested in knowing the broad indications of the classic measurements as Mark is making them.

YMMV

[Piotr]

Hi folks, been lurking here for some time but now it´s time to participate. Since dlr and I have discussed this topic before on madisound I couldn´t resist to jump in.

Cone distortion is very real and is non-linear (and of course there is linear cone distortion as well). Since a soft cone is flexing more than a rigid cone there must be some losses and these losses can not be linear. The flexing of a ideal soft cone will have a behaviour similar to the suspension IOW a spring. So even if the response curve are perfectly flat there will be distortion products that would never be there should we make the cone infinately rigid.

What´s even more problematic is that soft cones generally reach a point where the distortion suddenly increases and degrades performance a lot. A swedish engineer claim to have been the first to describe and baptise this phenomena decades ago, "avalanche distortion".

The distortion that can be seen in example the excel drivers are of a slightly different character. I´m talking about the increase in HD that can be seen that is related to the major breakup point. In this case the distortion is not directly cone generated but motorgenerated and amplified by the peak at breakup.

Of course a more rigid cone as those used by Seas Excel and Accuton will have some minor flexing even well below breakup, and therefore suffer from similar distortion as a soft cone, only less so. The more rigid the cone the less distortion from it.

/Peter

[Dennis H]

Ascribing bad behaviour of a cone to the point of being dismissive without knowledge of the actual distortion characteristics is not warranted, IMO.

I don't recall dismissing any kind of driver. That's why we measure them.

Personally, I can't see how I would use the knowledge that there is some specific breakup found in this mannter if I had measured distortion results for that driver. I'm more interested in knowing the broad indications of the classic measurements as Mark is making them.

Mark's testing takes a lot of time. You can run a nearfield RTA while moving the mic across the cone and it only takes a minute. The frequency where the curves start to diverge is a likely upper frequency limit. Often you'll see an impedance bobble near the same frequency and that's further confirmation. Once you've picked the drivers you want to use and have some idea of the useful frequency limits, you can do Mark's kind of testing with a pretty good idea where the problem areas may be.

[dlr]

I think you had meant to post to me

Just pointing out that I think that you had meant to reply to my post.

[dlr]

I should be more explicit at times

I don't recall dismissing any kind of driver. That's why we measure them.

Yes, we're in agreement, I hadn't meant to infer it to you. It was meant for the general thread on "floppy" cones. It seems pretty dismissive to me.

Mark's testing takes a lot of time. You can run a nearfield RTA while moving the mic across the cone and it only takes a minute. The frequency where the curves start to diverge is a likely upper frequency limit. Often you'll see an impedance bobble near the same frequency and that's further confirmation. Once you've picked the drivers you want to use and have some idea of the useful frequency limits, you can do Mark's kind of testing with a pretty good idea where the problem areas may be.

But that's why I like manufacturers that supply distortion curves and am very interested in Mark's tests that usually apply to a line of drivers.

If it's significant enough to blip the impedance curve, there's almost always a corresponding FR anomolie. Most FR anomolies that are not broadband (such as those generated by resonances) can in almost all cases be ascribed to a resonance, so I don't find that aspect as useful, other than knowing that there's an anomolie there. Maybe it's a cone mode, maybe it's basket air volume related, that's harder to pinpoint. Those tend to me more in the middle of the passband anyway, so you can't do much with them, of course.

So for my preference, I study the FR response, then hope to find a true set of distortion tests since that can't be determined by studying the FR. Narrow FR anomolies that aren't reflected in the impedance curve are usually indicative of some kind of cone mode or suround termination mismatch. Steady-state FR anomolies make themselves apparent enough for my use outside of true distortion measurements.

That's just how I handle it.

[dlr]

Very well put.

One point I would add is that the perfectly flat soft cone, though flexing, is also damping (ideally) in a manner to maintain flat response and/or better dispersion characteristics. Damping doesn't have to infer distortion. The "ideal" soft cone will damp such that there is no resonance established, yet the output is maintained linear to a higher frequency that would otherwise be non-linear due to phase interference.

All else being equal, a hard cone of the same geometry cannot be maintained linear to as high a frequency as can an optimally damped, flexing softer cone. This is more and more out of vogue due to improvments of hard(er) cones by other means and/or for other other desired results. An "ideal" soft cone would flex with optimal damping just as a closed-box with a Qt of 0.5 does. There can be advantage in the flex.

Soft dome tweeters flex, of course in a desireable way. The best ones flex with optimal damping that prevents resonances in the passband at the same time that they also allow the driver to reduce output nearer the dome tip at higher frequencies to prevent phase interference issues.

My point in all this is that blanket statements (not picking on anyone) about a driver based solely on the material in the diaphragm is a bit short sighted. Especially given that "paper" or "poly" drivers, though some might think them to be "soft", are usually not so anymore, given the use of doped paper (Scan-Speak)and drivers such as Dynaudios that are mineral filled polymers or other composite materials that can still exhibit exceptional qualities.

[dlr]

I meant the previous to be under Piotr's post

"Quick reply" goes to the top, I see.

My reply was still meant more for general consumption.

[Piotr]

dlr,

I agree with most of what you write.. I think.

I also feel that a flexing driver can have very fine performance when it comes to linear frequency response with minimal stored energy. They can of course also have very low distortion and can be used higher up in range due to better dispersion than the stiff one.

But still there will always be distortion products that a flexing cone sufferes from more than a rigid cone. The one I described earlier, the non-linear properties from the springlike action. This behaviour will casue both HD and IM. I have listened to system with very high performance using both kinds but I prefer rigid cones myself as I think that that extra bit of fine resolution can only be had with very stiff drivers.

/Peter

[dlr]

I'm still straddling the fence

dlr,

I agree with most of what you write.. I think.

I also feel that a flexing driver can have very fine performance when it comes to linear frequency response with minimal stored energy. They can of course also have very low distortion and can be used higher up in range due to better dispersion than the stiff one.

But still there will always be distortion products that a flexing cone sufferes from more than a rigid cone. The one I described earlier, the non-linear properties from the springlike action. This behaviour will casue both HD and IM. I have listened to system with very high performance using both kinds but I prefer rigid cones myself as I think that that extra bit of fine resolution can only be had with very stiff drivers.

/Peter

I've not been a fan of hard cones/domes, but having an Accuton C92-T6 has made me re-evaluate my position. I'm also going to be giving the domed RS52 mid a try. My preference has been for doped paper cone mids (13m/8640 and now the 12m/4631), but I do like some of what I hear in the Accuton. Maybe it comes down to size more than material in some aspects.

In the end, I still see much of it to be simply selecting the right driver for the intended application.You don't use 6-1/2" or 8" drivers for subwoofer duty and you recognize that an 8" 2-way is going to be limited in many ways. Likewise, the "cascade" breakdown (I like the term) is a case of choosing and using a driver optimally. Maybe it needs to be crossed a little higher to prevent excursions that trigger this. We make those choices all the time.

dlr

[dlr]

To be accurate, that's a C95-T6

I'd consider the C82-T8, but for some reason I don't trust the published SPL curve in the Madisound catalog. Different measurement system (?), no off-axis plots. Is it as flat as indicated?

[Piotr]

dlr,

Yes, I knew you meant C95 as I read some of your posts at madisound about your experience with it.

I also must say I do trust Accutons graphs and it probably is that flat. However the distortion in that little driver seem to be much higher than for example W15.

Have you tried the Excel drivers?

/Peter

[dlr]

Not the ones I'd like to try

dlr,

I also must say I do trust Accutons graphs and it probably is that flat. However the distortion in that little driver seem to be much higher than for example W15.

Have you tried the Excel drivers?

/Peter

I haven't seen distortion data for the larger Accutons. That I could use.

I've only tried a W17-E002 and didn't spend enough time trying to tame it. I think that the W18 series would be the ones to use as they seem to be improved over the W18 for a mid-woofer.