Everything breaks up at some point, even if the response is completely smooth everywhere.
That can only happen once breakup starts. However, the surround may have different behaviour than the cone (it doesn't always track in-phase and may have its own resonant behavior).
Well, the HD5 rise at ~1KHz is a direct consequence of the 5KHz bell mode, so if you stamp on that hard enough, i.e. if you cross sufficiently low, and sufficiently steeply to shunt it to a very low level, then the distortion amplification it causes lower down the passband should disappear in proportion. Based on Zaph's measures, the W18EX has a very similar distortion profile. So I'd expect 1.4KHz, LR6 or LR8 should do OK. Lower would be better if the tweeter can handle it, I completely agree, especially since as noted the W22 is starting to run into trouble at about 2KHz.
Last edited:
Sure. 'Break up' is basically a result of either radial (bell) or concentric modes, or a combination of the two. All cones have it, conventional metal woofers, midbass / midrange units are typically designed to create high Q modes at higher frequencies, so the cone is purely pistonic lower down.
Soft cones usually start to have low Q resonances at a lower level; being lower Q / better damped the amplitude usually isn't as great. Some feel they loose resolution as a result; YMMV on that one. The individual HD levels can be a good guide in combination with the FR etc., and even CSD has some uses in identifying modes not always as obvious on the FR graph (the data is actually the same, just presented in a different visual format).
Hard paper cones, kevlar, carbon etc. typically fall somewhere in between, in terms of general characteristics. You can modify these to a greater or lesser extent via varying geometry, thickness, coatings, mixes etc. so there's never a 'one size fits all' scenario -these are just very rough / broad trends, not gospel truth.
Yes, but you have to cross lower than that. I'm not sure if we're talking about different things, but if you cross at 1.4 kHz you WILL get the peaky distortion output at ~4-5 kHz. The thing is, the distortion ramp-up that starts around 600 Hz is basically the inherent ("natural"?) distortion of the motor that sounds louder because the resulting frequency of that distortion ends up at a point where the driver is extremely sensitive. You CANNOT get rid of that distortion ramp-up, or the distortion peak itself except by blocking signal to the driver where the HDx is higher. You can see similar peaks of HD2 and HD3 as their fractions of the 5 kHz breakup (HD2 peaks around 2ish kHz, HD3 peaks at 5/3=1600 Hz, HD4 peaks at 5/4=~1200Hz)
EDIT: Another way to put it - let's say you had a 96 dB/octave LP filter, essentially "brickwall" at 1 kHz. You still play 900 Hz at full magnitude with not much more above that, in terms of direct signal, but guess what, you will still be getting a substantial amount of 900 * 5 = 4.8 kHz output from the HD5 distortion there. If you play a pure sine wave at 1 kHz with this driver, you will see on an analyzer the added H5 at not less than 0.1% (assuming signal above 2.83 V).
Last edited:
What driver is more than 40 dB down at the cone resonance? I have not seen many (any?) driver that has substantially lower response at the cone resonance compared to the baseline sensitivity in typical passband.
Maybe some niche driver with a very heavy and well-damped cone?
But to answer your question, you would only see that if you measured frequency response using a broad-spectrum signal, maybe pink noise or even white noise. Is this something that is readily done? I'm familiar only with a sine sweep (i.e. single tone) to get response.
Last edited:
All diaphragms have a first mode resonance (which we call breakup). Only in tweeters can the first mode resonance be pushed above audible, and only then with very stiff light membranes for a diaphragm.
Once we are above the first mode resonance, then other resonant modes will appear. Metal, ceramic, and other very stiff cones usually have a very pronounced breakup, and the upper level resonances are usually less pronounced than that first mode resonance. With these kind of drivers, it is best to operate the driver below this first mode breakup frequency.
Plant fibers, wool, silk, flax, papyrus, wood pulp, and other fibers are very useful cone materials because a good driver designer can control the first mode resonance and make it well damped. Polymer cones also. The first breakup mode might happen at a pretty low frequency, but the driver is designed to be operated both below and above the cone resonance. A well designed paper cone can have a smooth frequency response even though the cone is not moving as a piston... there is a lot of flexing, bending, and twisting of the cone due to resonance.
To me the most amazing example of this is a really good silk dome tweeter. The dome operates in a state of breakup over its entire range, but when we make a CSD waterfall plot of the driver, it looks very clean... How? because the diaphragm is very well damped over all of its resonant frequencies.
Sometimes fibers are pressed randomly into a pulp or felt to make the cone (what we generally call a "paper" cone), and other times they are woven together into an organized patern. A woven diaphragm will have a different damping characteristic than a random felt. Damping is caused by internal friction in the material, and when fibers slide past each other, there is more opportunity for friction.
Coatings can also increase damping, and the surround can be a significant contributor to damping. You may notice that some drivers have a significant overlap where the surround meets the cone. Most of the time a big overlap is create more damping at the edge of the cone.
What about those old fashioned elliptical drivers?
A bell rings with 4 nodes and 4 anti nodes, stick an odd number of weights ( three blobs of pva ) radially and it'll struggle to ring. I used to experiment with spirals of pva, can't say I could tell any difference. Don't headphone drivers tame break up with radial/spiral ridges. I'm not sure if the ceramic drivers in mini X speakers break up in the audio range. I'll use them for tweeters someday.
A bell rings with 4 nodes and 4 anti nodes, stick an odd number of weights ( three blobs of pva ) radially and it'll struggle to ring. I used to experiment with spirals of pva, can't say I could tell any difference. Don't headphone drivers tame break up with radial/spiral ridges. I'm not sure if the ceramic drivers in mini X speakers break up in the audio range. I'll use them for tweeters someday.
Agreed, I've found I can go to 3/4 of the breakup start freq with strong drivers, using steep linear phase xovers. Easily.
JBL knew how to do it (ridges). Of course, other manufacturers as well.
https://jblpro.com/en-US/site_elements/2206h-j-data-sheet look at that picture-perfect response.
I'm not certain if we're talking about different things or not. The spike in HD5 at 1KHz I was under the impression you were concerned about is only present because the unfiltered drive unit's 5KHz bell mode amplifies the motor distortion at that frequency (as in 1KHz centre). If you cross low and fast enough to attenuate the 5KHz bell mode sufficiently, the distortion amplification it causes lower down is significantly reduced, i.e. that 1KHz spike in HD5 goes away, as does the 1.25KHz spike in HD4, the 1.7KHz [nearly] spike in HD3 and so on. It's not necessary to cross below the peak in HD5, it's only necessary to shunt that bell mode sufficiently out of the way that the distortion amplification ceases to be significant, and HD in the passband reduces to the baseline level of the motor itself once mechanical resonances are removed from the picture. Notches can't do that: beating the damn thing into submission with a high order XO at a sufficiently low frequency will however.
Since the W18EX has almost exactly the same HD profile, but can be crossed successfully up to about 1.6KHz the same (purely as far as the HD performance goes) essentially applies to the W22EX, although as noted, I would prefer a somewhat lower filter frequency myself.
We are indeed talking about the same thing.
But I disagree with this. Unless there's something magical about "shunting that bell mode" away, the driver will always, always produce the high HD5 that is shown in the Hificompass measurement as long as it is playing in the frequency range where that HD5 peak exists (600 Hz to ~1500 Hz).
Do you disagree with my assertion that with a brickwall LP filter just above 1 kHz the driver output will still have a measurable 5 kHz component when playing a 1 kHz sine wave, to the tune of >0.1% if the 1 kHz signal is at 2.83 V?
Easiest part first: if the W22EX is brickwall filtered at 1KHz, then no: I couldn't give two hoots about its unfiltered distortion at 5KHz, which is not very high to start with.
For the rest, I will respectfully say that to the best of my knowledge and experience (along with a number of better qualified people) you're mistaken. The HD5 peak at 1KHz in the unfiltered measurements only exists because the bell mode resonance at 5KHz has not been addressed. By which I mean 'hammered to a sufficiently low level with high order filters'. If you do that, that 1KHz HD5 distortion spike that you see in the raw / unfiltered driver measurements goes away. Cause and effect. The thing is, raw HD measurements of the driver don't show the baseline motor distortion: they show the warts and all behaviour of the driver sans any filters. But that isn't what happens when filters are present and major resonant modes are effectively suppressed. Zaph explains this characteristic more pithily than I can in his Seas L18 writeup: Zaph|Audio -go about halfway down the page to the 'Distortion' subsection. He's speaking of different drivers, but the general principles at work are the same.
For the rest, I will respectfully say that to the best of my knowledge and experience (along with a number of better qualified people) you're mistaken. The HD5 peak at 1KHz in the unfiltered measurements only exists because the bell mode resonance at 5KHz has not been addressed. By which I mean 'hammered to a sufficiently low level with high order filters'. If you do that, that 1KHz HD5 distortion spike that you see in the raw / unfiltered driver measurements goes away. Cause and effect. The thing is, raw HD measurements of the driver don't show the baseline motor distortion: they show the warts and all behaviour of the driver sans any filters. But that isn't what happens when filters are present and major resonant modes are effectively suppressed. Zaph explains this characteristic more pithily than I can in his Seas L18 writeup: Zaph|Audio -go about halfway down the page to the 'Distortion' subsection. He's speaking of different drivers, but the general principles at work are the same.
Last edited:
Continue to agree.
Although i'd change " 'hammered to a sufficiently low level with high order filters' "
to: feathered to a sufficiently low level with high order filters, ala linear phase xovers
I don't understand. Zaph says that the L18 and W18 should be crossed no higher than 1.5 kHz to avoid the HD amplification, and he's speaking of only H3, let alone H4 or H5. This doesn't appear to prove your point to me (these drivers are smaller than the W22EX and have a higher breakup frequency as well).
Secondly, you do realize that the driver itself naturally generates HD5? Within the motor itself? And that that distortion product isn't affected whatsoever by any external filter components or DSP? How is a low pass outside the driver going to attenuate high frequencies being generated from the cone that aren't being fed from an outside signal to begin with?
To address the unfiltered distoriotn "which is not very high to begin with", why use the W22EX at all if -60 dB H5 is fine? You realize that is the level of performance you can get with a driver that is $32 (RS125), let alone $320?
I have a few ideas about cone breakup and tweeters.
8" cones tend to go to Hell around 3kHz. 5" cones around 7kHz. Do the Maths.
Robin Marshall was a talented musician, so had a good ear. Many fascinating ideas here:
Robin Marshall: A Modicum of Genius | Stereophile.com
I own a pair of his speakers:
I have moved on since his ideas, but his intuition was sound IMO.
8" cones tend to go to Hell around 3kHz. 5" cones around 7kHz. Do the Maths.
Robin Marshall was a talented musician, so had a good ear. Many fascinating ideas here:
Robin Marshall: A Modicum of Genius | Stereophile.com
I own a pair of his speakers:
I have moved on since his ideas, but his intuition was sound IMO.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.