I linked this Lars Risbo blog post already. It clearly says that a serial notch filter is best for resonance taming, it will reduce distortion much better than parallel filter, because it affects motor's hysteresis https://purifi-audio.com/blog/app-notes-2/low-distortion-filter-for-ptt6-5x04-naa-11
andrxylo, a single coil in series with woofer will only compensate baffle step or work as lowpass (1st order electric). You will need both and preferably a serial notch (LCR parallel in series with woofer) for resonances. Sorry I am not competent to suggest more details.
andrxylo, a single coil in series with woofer will only compensate baffle step or work as lowpass (1st order electric). You will need both and preferably a serial notch (LCR parallel in series with woofer) for resonances. Sorry I am not competent to suggest more details.
Last edited:
Just to make things more confusing, a filter for rigid cone resonance taming is usually called a parallel notch, because its LCR components are paralleled, even though the filter itself is in series with the driver. The one that doesn't help with that is usually called a series notch, because its components are in series with each other, but the filter is in parallel with the driver.
Member
Joined 2003
The point is to ensure that the driver "sees" a high impedance at the frequencies of resonance, which provides electro-mechanical damping. The effect is not subtle, I've found about 10dB reduction in H3 as a result of this type of filter implementation. Other "pro tip" for passive crossovers that I can share, is that odd order filters perform and sound better than even order filters, for the same reason that the notch provides.
With modern software its pretty easy to compare the effects of the passive filter components, to a digital filter with the same transfer function. Here's one example of a 4" midrange driver with rigid cone, breakup peaking around 8-9kHz. Raw distortion measurement shows the effect on 3rd Harmonic.
With a Digital Filter low pass implementation, only the fundamental is reduced, distortion profile is unchanged.
The same filter was implemented with a parallel notch (or series if you ask Lars..). You can see the reduction in 3rd harmonic from 2-3kHz is about 10dB, not insignificant.
The magic notch filter uses only low cost components of small values, this improvement can be implemented on most drivers for only a few $. If you wonder at all - Yes, it is audible!
Member
Joined 2003
Yes, it matters, and yes - parallel filter as I show above is the way to provide high impedance at the driver voice coil to provide the damping and reduce distortion. Resistor value should be fairly large 30 ohm +, and it needs to be the last set of components in the filter before the driver. It is all well detailed in the article from Purifi. Series filter does very little in this regard.
I agree with this. When audiophiles listen to their system and it doesn't sound "right" they think "distortion". Well it certainly is that, it's a logical conclusion. But many audiophiles then think of distortion as measured by THD or IMD as being indicative of what they hear, but that is not a logical conclusion. The vast majority of what we hear as "distortion" is linear distortion which THD is not even attempting to measure - let alone it being meaningless even for nonlinear distortion. In loudspeakers nonlinear distortion is a secondary effect and can be made to be insignificant.
Last edited:
Because time and time again experts have found that subjective listening tests, if not highly controlled, lead to highly erroneous conclusions. I have personally witnessed this to such an extent that I don't trust anyone's subject claims about sound quality unless they were done with laboratory precision. Hence, the need for "research"!!
Last edited:
Cone breakup is a serious issue. The first thing I looked at when selecting a woofer was how well it handled cone breakup - I would never use any driver into this region. I pretty much ignored everything else because this factor alone was the Achilles heal of driver/system design. And even when I used the best drivers I could find taming the cone breakup was the single most complicated issue with the crossover design. The breakup always seemed to find a way to break through and affect the system response. It's a real PITA.
Member
Joined 2003
I don't trust the subjective opinions of others either, and in a many cases I question the objective data as well. At least with my own measurements in my own home, I know what I did entirely and what the conditions of my measurements were, and am in full control of my own subjective evaluation. It is "self-research" as there are so many conflicting sources of information, some of which is wildly incorrect if you try and research anything via online sources such as this forum. There can easily be the same trust issues with research, conclusions made before any experiments started, and carefully crafted experiments to specifically support the conclusions already made. But that's going a bit OT.
The greatest benefit I've found with FSAF, is the ability to separate the original audio from the residual, so you don't have to second guess any small nuances in <0.1% of the audio listened. Those small signals become very apparent and more easily identified when you can hear them standing alone. Of course, the cause of the error in audio requires more detailed evaluation, as there can be many sources from driver motor nonlinearity, suspension nonlinearity, cone breakup, cabinet resonances, and then down to room interaction / reverb and other unwanted background noise. FSAF total distortion chart only shows spectrum content for the entire audio measurement which could be 5-15 seconds, a lot happens in that time frame so more detailed analysis of the residual audio through listening and also spectrogram can be very useful. Using FSAF as a tool for evaluation of cone breakup control is only one useful application of many.
Last edited:
I agree that cone breakup is a huge issue. The first thing I do with any speaker is address the breakup somehow, and going back to the uncontrolled breakup always hurts my ears. Breakup makes harsh sounds far too harsh, and causes wolf notes even if the breakup is after the crossover. If your crossover is so steep or low that breakup is inconsequential, then you are paying a premium for extra drivers in order to use them at a fraction of their usable BW, or extra crossover parts, or extra fancy drivers. However a simple notch filter is a huge upgrade for a lot of existing speakers.
It is amusing for me to see a lot of current speakers, notably Polk which have the filleted surround crevice and other features presumably to control cone breakup but which still fail to do that. There are so many drivers on Parts Express and others which tout "eliminated breakup" but nothing about the breakup is eliminated as can be seen clearly on the charts, despite showing goop applied on the surround. It's actually infuriating the amount of gaslighting that goes on surrounding cone breakup and resonances (and every other driver parameter apparently). How far over a Q of 1 must the corner frequency be in order to qualify as breakup?
There is cone breakup as in bell modes, but there is also what happens when the cone geometry buckles under excessive vibration and loses it's geometric rigidity. Imagine a curved piece of paper buckling and then snapping back into place. The moment you overcome the rigidity provided by the curve in the paper, it becomes a very weak and nonlinear surface. This usually only occurs at the extremes of loudness but since the cone vibration is magnified by bell modes a poorly damped cone can reach this point sooner.
It is amusing for me to see a lot of current speakers, notably Polk which have the filleted surround crevice and other features presumably to control cone breakup but which still fail to do that. There are so many drivers on Parts Express and others which tout "eliminated breakup" but nothing about the breakup is eliminated as can be seen clearly on the charts, despite showing goop applied on the surround. It's actually infuriating the amount of gaslighting that goes on surrounding cone breakup and resonances (and every other driver parameter apparently). How far over a Q of 1 must the corner frequency be in order to qualify as breakup?
There is cone breakup as in bell modes, but there is also what happens when the cone geometry buckles under excessive vibration and loses it's geometric rigidity. Imagine a curved piece of paper buckling and then snapping back into place. The moment you overcome the rigidity provided by the curve in the paper, it becomes a very weak and nonlinear surface. This usually only occurs at the extremes of loudness but since the cone vibration is magnified by bell modes a poorly damped cone can reach this point sooner.
Right 👍 -that's why I've started writing the rather clumsy 'parallel notch in series with the driver' & similar so it's at least clear what the configuration is, since without there's usually confusion about whether series or parallel refers to the configuration of the components or their placement WRT the drive unit.
This is a HUGE topic and there is some truth and loadsa BS in this thread.
ALL cones breakup. Good sounding ones breakup in a controlled manner so they don't show resonances, either in the frequency response, THD or the Waterfall.
In most units, you can see the first breakup as a little blip in the impedance curve. At that frequency, the surround tries to misbehave. Getting it to behave is an important part of drive unit design.
Below that frequency, the cone is 'pistonic'. Above that, the 'piston' part of the cone starts to shrink. Jordan had the right idea but his maths was wonky. The 'rest of the cone' is 'breaking up". Good sounding plastic cones 'break up' differently from good sounding paper cones but the important point is the shrinking piston part and the well behaved 'breakup' for the 'rest of the cone'.
Hard cones don't behave like this. They go wild and have a big yucky peak(s) that shows in the frequency response, THD and Waterfall.
IM not so HO, metal cone speakers are not good speakers.
Domes for treble units can be hard cos the resonances can be made so high they don't matter.
BTW, rjb,
It was Peter Fryer of Wharfedale who first used Laser Interferometry to look at cone/dome behaviour. Celestion and then Wharfedale did the first SCanned Laser Plots using a Doppler instrument. JBL came later. Peter went to B&W but I'm not sure he carried out much work there himself. KEF first used this juju in their development of the units for the KHT 2005 Eggs
The same people were involved in all the UK efforts. It's a small industry. Dunno about JBL
I saw a commercial Doppler scanning instrument circa 2008 but it was crude compared to our efforts.
ALL cones breakup. Good sounding ones breakup in a controlled manner so they don't show resonances, either in the frequency response, THD or the Waterfall.
In most units, you can see the first breakup as a little blip in the impedance curve. At that frequency, the surround tries to misbehave. Getting it to behave is an important part of drive unit design.
Below that frequency, the cone is 'pistonic'. Above that, the 'piston' part of the cone starts to shrink. Jordan had the right idea but his maths was wonky. The 'rest of the cone' is 'breaking up". Good sounding plastic cones 'break up' differently from good sounding paper cones but the important point is the shrinking piston part and the well behaved 'breakup' for the 'rest of the cone'.
Hard cones don't behave like this. They go wild and have a big yucky peak(s) that shows in the frequency response, THD and Waterfall.
IM not so HO, metal cone speakers are not good speakers.
Domes for treble units can be hard cos the resonances can be made so high they don't matter.
BTW, rjb,
It was Peter Fryer of Wharfedale who first used Laser Interferometry to look at cone/dome behaviour. Celestion and then Wharfedale did the first SCanned Laser Plots using a Doppler instrument. JBL came later. Peter went to B&W but I'm not sure he carried out much work there himself. KEF first used this juju in their development of the units for the KHT 2005 Eggs
The same people were involved in all the UK efforts. It's a small industry. Dunno about JBL
I saw a commercial Doppler scanning instrument circa 2008 but it was crude compared to our efforts.
Member
Joined 2003
Allow me to add a couple more "experiences" to back up the "theory" and re-post a listing of notch and notch-like filters I had made.
A few years ago I found online the series-notch-filter-in-parallel "bypass" but it had little effect on Lowther PM6A's ear-bleeding peak (whether caused by cone or whizzer) despite trying many different R values. Later I found the parallel-notch-filter-in-series-with-driver and got it in one try -- basically a 1st-order XO between two virtual PM6A in parallel. The notch frequency (simplified) formula works out to 5khz/sqrt(mH*uF) -- independent of impedance assuming a narrow span -- which is the same as a 2nd-order pass filter.
Thinking this way, filter as crossover a driver with itself, opened up many possibilities. One of my favorite speakers is a vintage mono-pair, a set of two drivers with visually the same build and the same sound except one went a bit higher/brighter. After a tone-sweep the brighter one could be "toned-down" with a notch or shelf filter, then both drivers EQed together to make a fine stereo pair.
Fullrange drivers tend to have a broad HF peak or plateau -- possibly by design to satisfy so-called "average home listening conditions" without speaker toe-in. I don't know if this is considered "cone breakup", but the fix is the same, just a very wide parallel-LCR notch doubling as BSC and preserving (relatively boosting) very high frequency above the notch as if it were a super-tweeter in parallel.
My only "flat" fullrange driver is Ted Jordan's JX92S, designed after he had figured out how to use cone breakup to advantage. Despite its appearance I would not consider it an aluminum cone, but exceedingly thin foil over a paper-like non-stiff substrate, that absolutely ought not be touched. Whereas, Mark Audio relatively stiff metal cones including MAOP could be massaged back to shape (and by design or otherwise, have monster HF peak and plateau).
A while ago before I took up speaker-diy during Covid, I had come across your...
A few years ago I found online the series-notch-filter-in-parallel "bypass" but it had little effect on Lowther PM6A's ear-bleeding peak (whether caused by cone or whizzer) despite trying many different R values. Later I found the parallel-notch-filter-in-series-with-driver and got it in one try -- basically a 1st-order XO between two virtual PM6A in parallel. The notch frequency (simplified) formula works out to 5khz/sqrt(mH*uF) -- independent of impedance assuming a narrow span -- which is the same as a 2nd-order pass filter.
Thinking this way, filter as crossover a driver with itself, opened up many possibilities. One of my favorite speakers is a vintage mono-pair, a set of two drivers with visually the same build and the same sound except one went a bit higher/brighter. After a tone-sweep the brighter one could be "toned-down" with a notch or shelf filter, then both drivers EQed together to make a fine stereo pair.
Fullrange drivers tend to have a broad HF peak or plateau -- possibly by design to satisfy so-called "average home listening conditions" without speaker toe-in. I don't know if this is considered "cone breakup", but the fix is the same, just a very wide parallel-LCR notch doubling as BSC and preserving (relatively boosting) very high frequency above the notch as if it were a super-tweeter in parallel.
My only "flat" fullrange driver is Ted Jordan's JX92S, designed after he had figured out how to use cone breakup to advantage. Despite its appearance I would not consider it an aluminum cone, but exceedingly thin foil over a paper-like non-stiff substrate, that absolutely ought not be touched. Whereas, Mark Audio relatively stiff metal cones including MAOP could be massaged back to shape (and by design or otherwise, have monster HF peak and plateau).
I agree wholeheartedly. My first really successful "EQ of a full-range" was to tame the 5khz peak of Hifi-bird/Isred/Michael's Audio "drum paper" 5.5-inch -- I had meant to listen to a track or two of Bach Goldberg (Celine Frisch harpsichord) but stood listening for 70 minutes to all 30 variations, every note/hand-movement/musical-gesture.My experience is that only a minority of FR drivers cannot be improved by some sort of EQ on the top, but I think it's a relatively small price to pay.
A while ago before I took up speaker-diy during Covid, I had come across your...
Last edited:
Don’t ask me how come, but I still have a few very old Audax and Focal units with magnesium powdered domes (magnesium white reflects light better than black 😉 ) stemming from KEF labs. We are talking late eighties… so I think there were more people experimenting with laser interferometry these days.
Those guys are discussing the cone breakup distortion from about minute 25.
And the main takeway for those designing the speakers is that you cannot measure it, you cannot put a number on it, and you cannot compare 2 different speakers which one will have worse audible cone breakup by measuring them. You can measure that cone breakup exists and where, but you cannot put a value on it.
I think if you design a speaker for yourself or you plan to sell 100s of them this is not a problem.
The desktop speakers I am currently designing for myself based on Tang Band W6 driver I've started in October and I'm still far from done. I already have over 100 hours of dedicated listening tests and maybe 300+ hours of casual listening on those unfinished speakers, so I know very well where the problems are, what I want to fix, and what to listen for to compare if the problem is gone. It's not like it's hard labor trying to make money, the unfinished speakers already sound miles ahead of anything I had before and anything I heard at any friends house, I think I am shooting into $5k sound quality.
I think if you design a speaker for yourself or you plan to sell 100s of them this is not a problem.
The desktop speakers I am currently designing for myself based on Tang Band W6 driver I've started in October and I'm still far from done. I already have over 100 hours of dedicated listening tests and maybe 300+ hours of casual listening on those unfinished speakers, so I know very well where the problems are, what I want to fix, and what to listen for to compare if the problem is gone. It's not like it's hard labor trying to make money, the unfinished speakers already sound miles ahead of anything I had before and anything I heard at any friends house, I think I am shooting into $5k sound quality.
Surround, or the inability of the surround to control a [relatively] soft cone's edge resonance? I usually find it's the cone that's the culprit rather than the surround -it's never an exact comparison, but if you took, for example, a Seas ER15 (reed paper cone) & compare it to its (now sadly defunct) L15 aluminium cone sister model, the former shows a small but obvious fluctuation in the impedance curve around the anticipated 1KHz, whereas the latter's curve remains constant, suggesting it's the cone edge rather than the suspension itself that's the underlying cause, and the poor old surround is simply unable to damp it properly.
Badly designed ones aren't, in the same way badly designed drivers of any kind aren't. It's not difficult to use a quality metal cone unit though -given the option, I tend to prefer them when used properly (used in the piston band, stopband resonances controlled where needed, driver not used where distortion amplification is an issue etc.). To each their own of course.
I don't have a spare hour & twenty to sit through the video, but I've cast an ear over the 25mins - 35mins section & it a/ seems as solid as you'd expect given the source, and b/ I think you may be drawing some mistaken conclusions from it, because they don't appear to have said any such thing -a good job too, since as-written, it's nonsense (no offense).
Er -yes you can. Quite comprehensively, even with inexpensive gear.
Ditto. That's actually one of the easier things to measure & observe. We don't have frequency response, impedance response, harmonic distortion & IM distortion plots etc. just for jollies you know. 😉
Isn't that a bit inconsistant with the previous sentence claiming you can't measure it? Yes, you can measure the frequency at which major resonances ('breakup') occur, you can measure their amplitude (since it's part of their frequency response!), you can measure the individual harmonic and IM distortion products, and their levels relative to the fundamental either in dB or as a percentage. So it's actually quite easy to put a 'value' on it. You may get variations in opinion about what extent is audible and where, and how this can be most effectively dealt with. But it's certainly not difficult to identify it and its position, effects & quantities of those.
Last edited:
@Scottmoose as I am a total noob, here is a question, not a hidden statement - suppose you get 2 completely different speakers with different drivers and they happen to measure the cone breakup in the same area at the same value as you described above, may it happen that one of those speakers will actually sound much worse?