See this to build intuition.
Unfortunately no polar data there visible 😀
Crazy stuff, not sure if there is any need to get further into how it actually happens as we can evaluate "sound" of various drivers from the frequency responses, just like in the video.
The key idea is visible there, standard excitation in electrical domain leads to very different acoustic response depending on the moving parts physical properties. Logically what ever the excitation signal is the driver adds the frequency response to it. Precondition the signal one way or another before driver, notch peaks down, to get flat frequency response in acoustic domain. Likewise, if there is 10db breakup peak,you need to notch it 10db electrically, not 3 or 6 or 12, but 10db 🙂 Its the same up until physical properties like stiffness of surround changes with too much excursion.
Crazy stuff, not sure if there is any need to get further into how it actually happens as we can evaluate "sound" of various drivers from the frequency responses, just like in the video.
The key idea is visible there, standard excitation in electrical domain leads to very different acoustic response depending on the moving parts physical properties. Logically what ever the excitation signal is the driver adds the frequency response to it. Precondition the signal one way or another before driver, notch peaks down, to get flat frequency response in acoustic domain. Likewise, if there is 10db breakup peak,you need to notch it 10db electrically, not 3 or 6 or 12, but 10db 🙂 Its the same up until physical properties like stiffness of surround changes with too much excursion.
Last edited:
I've posted this before, but for the sake of interest: a before & after HD2 + HD3 comparison of a Seas aluminium/magnesium alloy dome, with the ~26KHz mode notched by a high-impedance parallel LC inserted in series with the driver. Not mine I hasten to add -this is from a project by a chap in Germany, done quite a number of years ago now (10+ IIRC). You can get similar results with other tweeters of this general type, the detail differences depending on the quality of the motor design & baseline HD levels.
Attachments
^ more precisely the amplification by bellmode stays the same, say 6db, but the notch filter increases impedance in the circuit reducing distortion current, so net result is less distortion measured in acoustic domain.
For example, play 1kHz tone on a woofer, non-linearity in the woofer motor, say Le(x), makes third order harmonic component appear as Voltage, this would be 3kHz distortion tone. This voltage source is in series with the circuit just like your power amplifier outputing the original 1kHz tone voltage.
Lets think the 3kHz distortion tone voltage is -40db under the 1kHz tone in electric domain but we have breakup peak at 3kHz and you'd measure distortion with mic -34db because the mode boosts it 6db acoustically. Now your distortion graph would show elevated 3rd order distortion for 1kHz. Btw. if you measured off-axis it could be -46db or something, and would show reduced distortion 😉
Well, if impedance in the circuit is high at 3kHz, the distortion that appeared as voltage makes less of a current into the circuit, which makes less force in the motor, which is less acoustic sound, at 3kHz. If circuit impedance was doubled at 3kHz, the current would drop in half and compensate the breakup amplification resulting -40db distortion also in acoustic domain. Breakup "amplification" is the same as before because physical structure stayed intact, but as circuit impedance was increased, current reduced, which makes less force in voice coil, and the net result is less distortion measured in acoustic domain than before.
---
Sorry about spamming, its just fun for me to think about the stuff and writing helps thinking so you all are going to get collateral damage 😀 please do comment if it feels wrong or something, or if there is anything to comment about. I'm not professional on this so don't take it gospel, I'm just connecting pieces of information together to something that seems logical and output it here without second thought, edit it afterwards if there is time left, or keep on posting as long as there is something to post on the subject 😀 have fun
For example, play 1kHz tone on a woofer, non-linearity in the woofer motor, say Le(x), makes third order harmonic component appear as Voltage, this would be 3kHz distortion tone. This voltage source is in series with the circuit just like your power amplifier outputing the original 1kHz tone voltage.
Lets think the 3kHz distortion tone voltage is -40db under the 1kHz tone in electric domain but we have breakup peak at 3kHz and you'd measure distortion with mic -34db because the mode boosts it 6db acoustically. Now your distortion graph would show elevated 3rd order distortion for 1kHz. Btw. if you measured off-axis it could be -46db or something, and would show reduced distortion 😉
Well, if impedance in the circuit is high at 3kHz, the distortion that appeared as voltage makes less of a current into the circuit, which makes less force in the motor, which is less acoustic sound, at 3kHz. If circuit impedance was doubled at 3kHz, the current would drop in half and compensate the breakup amplification resulting -40db distortion also in acoustic domain. Breakup "amplification" is the same as before because physical structure stayed intact, but as circuit impedance was increased, current reduced, which makes less force in voice coil, and the net result is less distortion measured in acoustic domain than before.
---
Sorry about spamming, its just fun for me to think about the stuff and writing helps thinking so you all are going to get collateral damage 😀 please do comment if it feels wrong or something, or if there is anything to comment about. I'm not professional on this so don't take it gospel, I'm just connecting pieces of information together to something that seems logical and output it here without second thought, edit it afterwards if there is time left, or keep on posting as long as there is something to post on the subject 😀 have fun
Last edited:
I am very confused why you say this?
It can be very clearly seen in distortion graphs?
As we are all getting older, I personally very much doubt that must of us are capable of hearing any real issues above roughly 8-10kHz.
Relative to the rest of the signals, not just on its own and separately.
That being said, with the majority of aluminum domes, this is well under control.
Distortion throws harmonics. The part of the driver spectrum they land on may be flat, or if affected by resonance may change the level of the harmonic which alters the distortion measurement (even though the fundamental may have more energy than the measurement suggests, which is also a linear error).
Breakup is itself a linear resonance with acoustic effects. If a harmonic falls on a typical breakup peak, distortion will measure lesser or greater depending on what axis you measure.
Breakup is itself a linear resonance with acoustic effects. If a harmonic falls on a typical breakup peak, distortion will measure lesser or greater depending on what axis you measure.
AFAICT, the Q of a break-up mode will be at least partly tied together with the 'softness' of the load impedance as seen from the point of view of the speaker terminals. At a glance, the entire metal dome is surrounded by an electrical shock absorber, so if there's a problem with resonance at break-up, the braking load slowing down the edge of the dome is either way too high (zero ohms or negative output impedance) causing reflection instead of absorption, or not enough (kilo-ohms).
Most of the time I would guess it's too close to zero, because most amplifiers are designed to be fast low distortion types, and DSP or active filters are often used with the idea that distortion will be lower than with passive components. (I did that and only later discovered that passive components could provide additional improvements.)
Most of the time I would guess it's too close to zero, because most amplifiers are designed to be fast low distortion types, and DSP or active filters are often used with the idea that distortion will be lower than with passive components. (I did that and only later discovered that passive components could provide additional improvements.)
Any electrical braking would be due to back-EMF voltage making current which opposes the movement, but if you check out math and explanation here https://www.edn.com/loudspeaker-operation-the-superiority-of-current-drive-over-voltage-drive/ under title "The assumed control of cone motion" the electrical braking, or damping, would happen only around drivers main resonance, where back-EMF induced current makes force that opposes the movement. Breakup happens much above this, and the back-EMF current has phase shift to it and no braking happens, at worst it might be more than 90deg and back-EMF would actually boost some.
Any braking current would need as low circuit impedance as possible to be most effective. On cone/dome breakup its not useful current but harmful, so we should increase circuit impedance to reduce the current around breakup for better sound.
You can analyze circuit impedance from drivers perspective, like in Purifi paper, and figure out how it plays out. You'd want high impedance for back-EMF above main resonance, and low impedance on the resonance to avoid effects in frequency response. Or any combination of, that best suits your application and what you have in mind. With active system and DSP one can tailor frequency response independently of circuit impedance, and tailor circuit impedance with passive components to reduce distortion, to what ever within reason. With passive speaker its double duty with passive parts and some kind of compromise on both.
Well, how big of a deal all this is? perhaps not that big in grand scheme of things but it is something to optimize if there is possibility, if target is top performance 🙂
Any braking current would need as low circuit impedance as possible to be most effective. On cone/dome breakup its not useful current but harmful, so we should increase circuit impedance to reduce the current around breakup for better sound.
Yeah this is opposite, high output impedance amplifier would make high impedance circuit and prevent back-EMF distortion currents flowing, also would prevent the electronic damping at drivers resonance, which would result as peaking frequency response. Unless, DSP was used to reduce the peak. Also enclosure can be designed to reduce peak, also passive components can do it. This is not very marketable, as mixing and matching speakers and amps would not work very well, introduce big changes in frequency responses, or be too expensive to eliminate it, so its not done much by commercial business. But we as diyers can tailor and do what ever we wish, utilize electrical damping, utilize distortion reduction as per other constrains on the project.
You can analyze circuit impedance from drivers perspective, like in Purifi paper, and figure out how it plays out. You'd want high impedance for back-EMF above main resonance, and low impedance on the resonance to avoid effects in frequency response. Or any combination of, that best suits your application and what you have in mind. With active system and DSP one can tailor frequency response independently of circuit impedance, and tailor circuit impedance with passive components to reduce distortion, to what ever within reason. With passive speaker its double duty with passive parts and some kind of compromise on both.
Well, how big of a deal all this is? perhaps not that big in grand scheme of things but it is something to optimize if there is possibility, if target is top performance 🙂
Last edited:
We are starting to talk around each other, or at least there is some miscommunication, because I can't follow anymore.
A picture tells more than a lot of words, so here some random driver with a break-up resonance and its distortion.
You can clearly see that the peak is reflected in the distortion graph at freq of peak divided by the harmonic number.
Yeah, allenb wrote about that H3 of 6.5kHz happens to land at the 20kHz where there is dome resonance, and gets boosted by it, hence peak for H3 at 6.5kHz.
6.5kHz third harmonic happens to land on the resonance, but for example 6kHz H3 does not, hence its lower in the graph. Both probably have same distortion mehcanism happening, and same distortion amount in electrical domain, same distortion current due to motor or what ever, but the resonance boosts the other in acoustic domain and it shows a peak in acoustic distortion measurement.
If you EQ that peak with DSP, the distortion graph would not change. EQ the peak with increasing impedance, and the distortion graphs straighten out.
So, its not breakup distortion, but distortion from lower frequencies get amplified by the breakup.
If you measured the distortion off axis, you'd see distortion drop on the 6.5kHz, because the breakup boost of the dome is likely only on-axis, might be a dip off-axis, reducing also the distortion component of 6.5kHz.
6.5kHz third harmonic happens to land on the resonance, but for example 6kHz H3 does not, hence its lower in the graph. Both probably have same distortion mehcanism happening, and same distortion amount in electrical domain, same distortion current due to motor or what ever, but the resonance boosts the other in acoustic domain and it shows a peak in acoustic distortion measurement.
If you EQ that peak with DSP, the distortion graph would not change. EQ the peak with increasing impedance, and the distortion graphs straighten out.
So, its not breakup distortion, but distortion from lower frequencies get amplified by the breakup.
If you measured the distortion off axis, you'd see distortion drop on the 6.5kHz, because the breakup boost of the dome is likely only on-axis, might be a dip off-axis, reducing also the distortion component of 6.5kHz.
just gojng out of my living room listenning session : Apalachian Speing - Copland conductor- Fast mvt. I am listening a little more spl level with classic than rock. Here with my spl meter , the lowest was 65 dB and the louder 85 dB. Damit the dynamic level seems compressed at seeuing the number, not so at listening the music. Is a 16/44 NOS DAC the culpritt of dynamic width between the lowest and the biggest spl notes, or that is simply te reccording ? I assume it is not the same at chhosing a tweeter ? (Seems to me that spl levels are not so high according the real life in venues... conductors can feel 130 dB !)
Am I agging ? Thought about that levels and tweeters ?
Am I agging ? Thought about that levels and tweeters ?
This is also my experience. Also with just a regular passive filter btw I think?
(I just haven't made any passive filters in a loooooooong time)
In fact, this whole discussion feels a bit like opening old dusty boxes again.
Because on this forum people had this exactly same discussion many times before.
Especially when John Krutke, Linkwitz and others were active. (have a bit of a brain fart atm when it comes to names)
That being said, I never really had thoughts about the "why".
Mostly because I have never found very satisfying explanations on real scientific literature (AES, IEEE, HAL etc etc etc)
Do you also have measurements of all things and suggestions?
Because it makes me extremely curious!
Anyway, this doesn't seem to be much of a deal with many tweeters. (I guess it's much easier to control the break up)
Woofers on the other side, is a whole different story.
Sometimes huge spike around 800-3000Hz (and often around 2kHz or so).
Exactly the frequency range you don't want any of these nasty issues.
edit: btw break-up spikes (or resonances rather) do show up in the impedance graph.
This can be seen by some full-range drivers sometimes, when this resonance has been left uncontrolled.
Last edited:
Btw guys did you see whatever the said stifness and lightness of metal dome, being Aluminum or Be, we always fal in the 0.3 to 0.4 Mms ?
As if the ligthness was not so an important factor but more an industrial process limitation ?
As if the ligthness was not so an important factor but more an industrial process limitation ?
You do realise that Mms is the moving mass of the entire driver, including coil & suspension, not just the diaphragm material? Unless everything is identical other than that dome material, Mms is not a useful means of comparing, as you don't actually know where the variations in mass are coming from. One of the advantages of materials like beryllium etc. is that they offer the potential of achieving things that can't necessarily be done with other materials, but this is a question of design priorities and details.
It isn't really a question of 'always' either. For instance, I'm eyeballing the Bliesma T34B-4 at the moment. Beryllium dome, Mms is rated at 0.26g, which is rather less than the 'always... 0.3 - 0.4 Mms' mentioned. And not only is it lighter, it has an Sd of 10.5cm^2, which is somewhat larger than most 1in dome tweeters, and it does it with a a narrow-suspension design (as you can see both physically & in the fact that HD2 doesn't rocket like a Saturn V < ~2KHz, which tends to happen with wide-surround designs), i.e. the dome itself genuinely is larger than most.
It isn't really a question of 'always' either. For instance, I'm eyeballing the Bliesma T34B-4 at the moment. Beryllium dome, Mms is rated at 0.26g, which is rather less than the 'always... 0.3 - 0.4 Mms' mentioned. And not only is it lighter, it has an Sd of 10.5cm^2, which is somewhat larger than most 1in dome tweeters, and it does it with a a narrow-suspension design (as you can see both physically & in the fact that HD2 doesn't rocket like a Saturn V < ~2KHz, which tends to happen with wide-surround designs), i.e. the dome itself genuinely is larger than most.
Last edited:
ah, indeed, if it includes the surrounds ! I knew it was just the cone/dome mass + its air load. Glad to know it involvres the weigth of the surrounds.
It is making a difference between foam, fabric, and polymers surrounds.
It is making a difference between foam, fabric, and polymers surrounds.
Recording. And medium used ( 16bit theorical max is 96db but... it'll ask for a reproducing system able to top those 96db over your noise floor ( acoustical background noise) which is around 45/55dbspl in typical 'quiet' domestic environnement... so we are tlaking about 140dbspl capability here... All ( almost) recordings are compressed one way or another.
The difference you noticed between rock and classical is typical dynamic range difference of both style ( 12/14db vs 20db).
Well, it depends how you look at this.
Because from a loudspeaker manufacturer point of view, I don't agree with this.
Especially these days (read; last 5-8 years), most of them all have active systems.
So there is no mixing and matching anymore, and with a few mouse clicks, you can change a DSP to whatever you want.
I think the answer to this is much, much more simple.
Most people are not aware of it, don't care about it or don't know anything about it.
The vast majority barely looks at any peaks, some that do don't go much further than saying "well it's outside the frequency band, so not important anymore".
At the same time when you do the measurements, you very clearly see the issues still back in the distortion graph.
People like Erin and others have VERY clearly shown how bad the design of some speakers are.
Although being in this field as a professional for so long, it didn't surprise me.
It only surprised my HOW bad it is sometimes. Way worse to many DIY projects I have seen in the last 20 years.
You have to see that as a compliment to the DIY community btw! 🙂
Last edited:
I have to dive back into the books again, John Eargle has a good chapter about this if I recall well.
But if I am not mistaken, the majority of the Mms for almost any driver (woofer and tweeter) is the voice coil + former.
It certainly is for (most) tweeters.
Going for different materials in tweeters only has an effect on the upper range, so the break-up/resonance frequency we are talking about.
Up till about 10-15kHz it's just full on piston mode.
Any abrupt irregularities in that range is just either (really) bad design or production/assembly problems.
In compression drivers they often stamp the dome to add stiffness as well.
I have never seen this in tweeters actually.
What sort of tweeters had you been embeeding into loudspeakers ? Could you be more explicit please ?