The birth of a 15 inch woofer

Hello Ultima Thule

Thanks, I am looking forward to seeing that as well.

The drivers are China-made subwoofers for car stereo. I do not know much about them. The frame however, is made by Hengtai. They make huge amounts of frames, you see them all around the world. Lots of JBL, AE and several pro audio brands use them.
 
Hi there again, and thanks for your reply! :)

Yes your right, on a general note I have seen quite a few speaker basket/frame designs which resembles B&C, 18Sound etc while browsing through Aliexpress, Alibaba etc., don't know if they are hengtai or knock-off of knock-off as is so prevalent over there. lol
However, the basket you have chosen looks good to me too, me eyes are also on how open they are both above and below the spider plus a few other things as well.

One idea I had is, if a certain basket design isn't exactly what one wants, such as, often the air flow below spider with a rather poor design and constricted, would be to add spacers for each bolt attaching basket to the magnet drive on the order of perhaps couple of centimetres or so which separates the magnet and basket from each other and also eases up the airflow below the spider, this also allows for more flexibility to play with longer excursion.

A typical type of frame that looks "good" for the most part but seem to have no air vents directly beneath the spider is like the one in the attached picture.
Giving it a second thought I am wondering if they are intentionally forcing the air to vent only through the centre of the pole piece, maybe for PA it's more important to force as much air to pass through the VC for improved power handling while for HiFi stressed use it causes too much compression??

Regards Michael

ps. I noticed some of the pictures were linked, try instead to attach the pictures directly if possible, it's nowadays recommended here on DiyA in order to avoid Link Rot, where pictures disappears because the web site/URL link have for some reason ceased to exist.
 

Attachments

  • Finlemho-Professional-Audio-Speaker-Woofer-12-Inch-3-Inch-Voice-Coil-2PCS-For-Line-Array-DJ.jpg
    Finlemho-Professional-Audio-Speaker-Woofer-12-Inch-3-Inch-Voice-Coil-2PCS-For-Line-Array-DJ.jpg
    87.8 KB · Views: 624
Yes your right, on a general note I have seen quite a few speaker basket/frame designs which resembles B&C, 18Sound etc while browsing through Aliexpress, Alibaba etc., don't know if they are hengtai or knock-off of knock-off as is so prevalent over there. lol
However, the basket you have chosen looks good to me too, me eyes are also on how open they are both above and below the spider plus a few other things as well.

It is not unlikely that most of the pro manufacturers buy their frames from Hengtai as that factory is so huge. The fact that many manufacturers seem to use identical looking frames makes this even more likely. If one of those driver manufacturers owned the design, they would likely be involved in tons of legal cases.


One idea I had is, if a certain basket design isn't exactly what one wants, such as, often the air flow below spider with a rather poor design and constricted, would be to add spacers for each bolt attaching basket to the magnet drive on the order of perhaps couple of centimetres or so which separates the magnet and basket from each other and also eases up the airflow below the spider, this also allows for more flexibility to play with longer excursion.

Like this?

attachment.php


It is not for creating space or more vents, but to make a spacer between the 127mm mounting hole diameter and the motor OD of 104mm, and also to make support for the pole piece together with space for the voice coil.

(I uploaded the image now, I did not see the "upload-function in the quick reply box, but thanks for the tip.)


A typical type of frame that looks "good" for the most part but seem to have no air vents directly beneath the spider is like the one in the attached picture.
Giving it a second thought I am wondering if they are intentionally forcing the air to vent only through the centre of the pole piece, maybe for PA it's more important to force as much air to pass through the VC for improved power handling while for HiFi stressed use it causes too much compression??

Again, you are spot on. Many frames for pro audio also have vents with cooling fins, so when the motor is bolted to the frame, it acts like a forced cooling for the frame as well.

We do not need any of this in high end audio. The effect of this could on one hand be air noise, and on the other hand, it typically generates a mechanical resistance. Both of which are unwanted in a super transparent driver.
 

Attachments

  • 15 inch woofer rear.jpg
    15 inch woofer rear.jpg
    341.2 KB · Views: 1,473
With regards to the spacer thing I had in my mind, it's not really an adapter, just a way to separate the frame and the magnet a bit improving air movement beneath the spider, see the attached picture, the upper is the original I used from 18Sound and below is the modified version, and the mounting bolts keeping frame and magnet together goes through those individual spacers, and of course the VC has to be elongated to accommodate for the change in distance.
Hope this made it clearer what I tried to convey.
 

Attachments

  • 18TLW3000 drawing (with spacers).png
    18TLW3000 drawing (with spacers).png
    13.8 KB · Views: 567
I totally understood that.

The reason why mine is a bit more complicated is because the part on the back of that 18-sound driver, which is made by Hengthai BTW, has an inner diameter for the coil space of around 90-95mm, and they do not make a 3 inch version, so I kind of had to make my own.
 

ICG

Disabled Account
Joined 2007
- X-max (real x-max) in the 10-12mm range.
- 15 inch paper cone and paper dustcap with F0-1000Hz operating range.

The two things are mutually exclusive. If you want it to play that far up, you don't want it to make any more than ~3mm maximum! actual excursion because the huge excursion make it impossible to reproduce any clean midrange. Huge distortion and very rough, doppler effect ridden midrange without resolution and stage reproduction are the consequence.

Your voice coil and motor isn't possible to build reasonably either. You will realize a double voice coil will make the Mms increase lot, expect it to be at least 40-50, but rather more likely 70g increase. To put in '0' for inductance may be very convenient for 'designing' the driver but in series the inductance will not allow 1000Hz with a reasonable spl, it will drop much sooner. The double VC also lowers the magnetic flux a lot because of the 2nd voice coil gap, you will need a much stronger magnet to compensate for it.

Aside from that, in parallel the parameters will be a lot different.
 
Last edited:
The two things are mutually exclusive. If you want it to play that far up, you don't want it to make any more than ~3mm maximum! actual excursion because the huge excursion make it impossible to reproduce any clean midrange. Huge distortion and very rough, doppler effect ridden midrange without resolution and stage reproduction are the consequence.

This has been talked about a lot by the anti long excursion club.

Around 1,1m/s, the pitch will change by around 1Hz at 300Hz, and this is pretty much the treshold where people start to not be able to hear a pitch difference in any case, even with a single sine tone and some buttons to shift back and fourth between the tones. This equals around 5mm of excursion at 40Hz in our case. This again equals 112,6dB in the pass band for one driver, or 118,6 for two drivers (2 pi/1m, which in room is often not that far from the truth.).

In my book, making a driver is about selecting an ideal working range, and adding a kind of "safety valve". I mean, when you listen to music, you would not be interested in running the system on the limit of breakdown, and that is what x-max defines. Around 5mm we have less than 2% deviation in the force factor. That means we should have extremely low distortion from the motor combined with inaudible doppler effect etc. This should be the absolute sweetspot of the driver.

So how do you design a driver that has this level of tolerances in the motor at 5mm, and still manage to make it completely "break down" above 5mm? And why would you do that and throw away the rather useful 6-9dB of useful headroom?

Remember also, that for a typical 12 inch midbass driver at 200Hz, this equals 1mm of excursion and just 113dB for two drivers for the same amount of doppler. Most people would think the doppler effect is no longer a problem with this solution, but if you use the 12 inch up to 2,5mm at 200Hz, you would get potentially more audible doppler there as well.

Your voice coil and motor isn't possible to build reasonably either. You will realize a double voice coil will make the Mms increase lot, expect it to be at least 40-50, but rather more likely 70g increase.

I believe this has been answered before, but I can take it again:
The total mass of the windings is 39,9 grams.
The total mass of the voice coil former is 6 grams.

This includes both of the coils, and the entire extended voice coil former. That leaves us with 64,1g for the cone, dust cap, suspension, lead out wires and glue to hit a 110g target.


To put in '0' for inductance may be very convenient for 'designing' the driver but in series the inductance will not allow 1000Hz with a reasonable spl, it will drop much sooner. The double VC also lowers the magnetic flux a lot because of the 2nd voice coil gap, you will need a much stronger magnet to compensate for it.

It does not make any difference if the coils are in series or if they are in parallell. The inductance is always a relative to the coil resistance.

Seas has some subwoofers called "Extreme". They are in a 2 layer and a 4 layer version. Their inductance is respectively 0,73 and 2,33mH. This results in a difference in the relative output at 1kHz of 3dB.

My driver on the other hand, has a far less dense winding, and it does not "see" much steel. In addition to that, it has some really heavy shorting rings. This results in an increase in power of around 13% at 1kHz compared to 200Hz. That equals a rise of 0,5dB at 1kHz. I am considering shaving off a bit on those rings in order to reduce the capacitive coupling to the coil itself, just to reduce loss.

We will get some inductance, you always do, but for practical purposes, the inductance of a driver is a more complex thing than what you get from an inductor. The capacitive part is also a source of a small loss. The small signal sensitivity at 1kHz will therefore probably not increase even with the shorting rings kept as they are. We are actually likely to measure close to 0mH at 1kHz.


Aside from that, in parallel the parameters will be a lot different.

Yes, on paper some of them will be different, but it makes no difference as the changes are only following the 1/2 BL for 1/4 Re you would expect from a driver with 1/4 the impedance and otherwise identical performance. The efficiency will be the same, and the response in a given box will be the same. The only difference will be the voltage sensitivity.
 

ICG

Disabled Account
Joined 2007
This has been talked about a lot by the anti long excursion club.

Around 1,1m/s, the pitch will change by around 1Hz at 300Hz, and this is pretty much the treshold where people start to not be able to hear a pitch difference in any case, even with a single sine tone and some buttons to shift back and fourth between the tones. This equals around 5mm of excursion at 40Hz in our case. This again equals 112,6dB in the pass band for one driver, or 118,6 for two drivers (2 pi/1m, which in room is often not that far from the truth.).

Well, these numbers are quite deceiving because you assume a 0° listening angle and only the direct sound and, on top of that, that only one or two signal(s) is(are) reproduced at the same time. That is quite rarely the case (well, actually practically never). The main problem is not the pitch, that's only the icing on the cake. The problem is the one signal will be modulated with the other, altering the impulse response because the source of the sound moves with the excursion. With more complex signals (music), it gets worse the response is very important for the location and not only the 0° radiated sound is affected, the other angles and therefore reflections in the room too, so it does not matter if you listen to the speakers just at 0° to you. The modulation is perceived and the human ear is actually pretty sensitive to that. The cone breakup and partial oscillation of the membrane (and asymmetrical movement inwards/outwards of the partial movement because different push/pull flex of the cone) can't be prevented either, it is a non-linear distortion. Such light cones combined with that high excursions are very prone to it. Look at some Klippel measurement videos. The surround and the cone can be coated and/or stiffened modify or dampen it but both affects the fine dynamic and resolution aswell. The combination of the several effects acting on it are adding up and the loss in resolution and stage location is audible at lower excursions already. You can try that with fullrange or bass-mid drivers and tweeter, it definitely happens at lower excursions than your proposed 5mm. Especally if you take into account that the driver is a BR driver and the excursion reduction because of the port is already in place there, the excursion will be higher at 60-90Hz.

In my book, making a driver is about selecting an ideal working range, and adding a kind of "safety valve". I mean, when you listen to music, you would not be interested in running the system on the limit of breakdown, and that is what x-max defines. Around 5mm we have less than 2% deviation in the force factor. That means we should have extremely low distortion from the motor combined with inaudible doppler effect etc. This should be the absolute sweetspot of the driver.

That does not work that way. The force factor isn't the only problem, the mechanic of the whole oscillating system is, the membrane does not do the same as the coil! And if you want to use 5mm excursion, why then the '1000W' and dual coil? The long VC former will start tumbling movement with the soft suspension, even small asymmectrical moves and low angle errors lead to a surprisingly wide movement of the end of the coil, it will likely result in the need of a much wider air gap or a 2nd spider at the back, like most dual coil drivers have. Besides that, if you use only the middle of the coils (-> low excursion), the overlap of the front coil will burn up sooner or later because it can't disperse the heat to the pole plate. A possible solution would be a good heat conductive VC former, like aluminium but that would introduce much higher losses because of the eddy current it brings with it.

So how do you design a driver that has this level of tolerances in the motor at 5mm, and still manage to make it completely "break down" above 5mm? And why would you do that and throw away the rather useful 6-9dB of useful headroom?

You don't have that 'useful headroom' because of the various effects mentioned. You cannot use the power capabilities for the low frequencies, you cannot use the excursion if you want to maintain the hifi-esque sound quality. And you want to know how to do it? That's easy, you don't. Take a smaller driver (one or more 6-12" mid or bass-mid) and put a bass or subwoofer driver under it. Or use ie. 2x12". Almost all problems solved at once and is with complete dsp, extra amp channel and effort on the enclosure still a ton cheaper and not as flawed with the numerous compromises than your driver.

Did you calculate how much you'll have to spend on prototype drivers? Or how much the resulting driver would cost if it would go into production?

Why do you think there is none of such drivers on the market? All of the technologies are available for 20 years or even longer. Do you think noone but you would have thought about such a driver? The practical use of it is quite limited, you can get the same result by much

Remember also, that for a typical 12 inch midbass driver at 200Hz, this equals 1mm of excursion and just 113dB for two drivers for the same amount of doppler. Most people would think the doppler effect is no longer a problem with this solution, but if you use the 12 inch up to 2,5mm at 200Hz, you would get potentially more audible doppler there as well.

Uhm, that's not correct. You've taken the excursion of one 12" driver at 113dB but for two it's at least 6dB more before it comes into effect (+3dB for the double cone area and +3dB for double the power and then some for the larger membrane surface (860cm² on your driver vs. roughly 1050cm² of 2x12", which is ~25% more!)).

I believe this has been answered before, but I can take it again:
The total mass of the windings is 39,9 grams.
The total mass of the voice coil former is 6 grams.

This includes both of the coils, and the entire extended voice coil former. That leaves us with 64,1g for the cone, dust cap, suspension, lead out wires and glue to hit a 110g target.

You forgot the surround and the air load and the very likely need of a 2nd spider. And your paper membrane is extremely likely not stiff enough for the excursion, it will bend, fold or tear. I pretty much doubt such a membrane will endure it without having carbon fibre (or similar) reinforced paper pulp or some kind of mechanical imprinted structure/shapes like ridges, rings or similar.

It does not make any difference if the coils are in series or if they are in parallell. The inductance is always a relative to the coil resistance.

No, that's not correct, that only applies if your amp is a current source. Almost all amps nowadays are voltage sources though! In series the inductance adds and increases the impedance much earlier, it works as a low pass filter, it increases the impedance and lowers the power, effectively dropping the response on the upper end.

Seas has some subwoofers called "Extreme". They are in a 2 layer and a 4 layer version. Their inductance is respectively 0,73 and 2,33mH. This results in a difference in the relative output at 1kHz of 3dB.

Yes, because they have

a. a very good inductance control in place over more than the whole winding length with extremely low impedance rise in the single coil version and
b. the 500mH cannot be correct, that's (if!) the value of a single coil. If you compare the Seas single and dual coil version, you'll realize that the Le of the dual VC is more than 3x of the single coil version (0,73mH vs 2,33mH). The same happens if you serial both coils, the inductance will rise a lot more because they share the same core (coupled magnet field). If the 0,5mH are correct (which I doubt but let's assume it is), expect the two coils in series to rise the inductance around the same for your driver.

Before going deeper into the construction of this driver, you should verify how much your parameters will change with single/parallel/serial voice coil wiring in place. I'm afraid you're missing few details of the whole picture in your driver development.

My driver on the other hand, has a far less dense winding, and it does not "see" much steel. In addition to that, it has some really heavy shorting rings. This results in an increase in power of around 13% at 1kHz compared to 200Hz. That equals a rise of 0,5dB at 1kHz. I am considering shaving off a bit on those rings in order to reduce the capacitive coupling to the coil itself, just to reduce loss.

We will get some inductance, you always do, but for practical purposes, the inductance of a driver is a more complex thing than what you get from an inductor. The capacitive part is also a source of a small loss. The small signal sensitivity at 1kHz will therefore probably not increase even with the shorting rings kept as they are. We are actually likely to measure close to 0mH at 1kHz.
[...]
Yes, on paper some of them will be different, but it makes no difference as the changes are only following the 1/2 BL for 1/4 Re you would expect from a driver with 1/4 the impedance and otherwise identical performance. The efficiency will be the same, and the response in a given box will be the same. The only difference will be the voltage sensitivity.

Well, theory and examples of existing drivers rise serious doubts regarding that.
 
The upper cutoff frequency caused by Lvc of the two identical coils does of course stay the same whether they are in series or in parallel. Period !

Fc of the Re-Lvc combination is given by

fc=Re/(2*Pi*Lvc)

Both values vary fourfold by changing from parallel to series connection and therefore their ration stays the same. Q.E.D.


Regards

Charles
 
First, as phase_accurate says, it doesn`t mather if the coils are in parallell or in series.

Second, the doppler effect. If this is an big issue, how about 6,5" and 8" thats more normally used in hifi speakers? Their freq range often about 30-2khz?
Here the doppler effect will be an bigger issue than the 15" at the same SPL.
So I think we can forget about this issue also.

I don`t say it is an easy job for an 15" to play up to 1khz.
But you have to look at the real issues. Cone break up. Intermodulation in motor. Suspension. How much mms do it need, preferably light for the upper range.
On the other hand there has been build 15" for long time that do this job quite good.
TAD 2402 monitor. JBL Everest series. JBL K2. JBL M2 and so on.
In the K2 the 15" is crossed at 900hz. I don`t see much complaints about these kind of speakers all having 15" crossed in the 7-900hz region.


The biggest trouble is actually that when you first listen to big High Efficiency speakers done right, you never go back! :)

Her is som documents on speaker efficiency:
SPEAKER EFFICIENCY
 
Things might appear to be quiet these days, but I can assure you they are not.

During the past few days I have made some important reviews of the suspension. The 204mm spider with a 3 inch coil gives a pretty hefty 56mm width of the suspension area. This is all good for the ability to move, but I am a bit worried that it might resonate, especially since it will be a very soft suspension for its size. Resonances originate from the mass and compliance relationship in the spider, so an idea that might work is to reduce the diameter a bit, and then make an adapter ring for the spider landing with an oval inner opening. That way, the width of the spider will be progressive, leading to an effective resonance distribution over a wide range of frequencies, where each resonance has extremely low Q. This approach should remove resonances and at the same time keep the Cms linear over a wide excursion range.

As I wrote before, the spider will handle the significant part of the Cms, while the surround will be far softer. There are several reasons for this.

First off all, the spider is easier to control than the surround, especially if we need some added coating on parts of the surround that has to be done manually. That means the variation between samples will be small.

It is also slightly easier to control the X-position of the coil if the spider is the dominant compliance. This might not always be true, but the selected assembly method means the coil will be fixed in the gap, and the spider will be mounted in the frame with only the coil in position. The spider will be glued to the voice coil former as it is in its resting position. Then the cone will be added in a similar way, but as the cone has its own mass, it will be a source for some error. We therefore need to support the cone mechanically during assembly to make sure we do not assemble the suspension under load. A 0,5mm error in the cone assembly will therefore translate to less than 0,15mm for the entire assembly.

As the spider is far wider than the surround (around 40mm vs around 20mm), it will keep its compliance over a broader excursion range. Since the spider is the dominant compliance, and also is the part that offers the longest excursion, we will get a more linear total suspension. This also reduces the drivers tendency to activate jumping resonances (as described in an earlier post).

The mass balance of the driver has been described before. However, the center of gravity has not been talked too much about. The center of gravity on the voice coil itself will be towards the back of the voice coil former, but not by much. It represents around 45g behind the spider. The rest of the mass will be around 60g as the spider and glue probably represents around 5g. The mass of the cone will increase towards the front. The length of the assembly behind the spider is around 110mm, while the length of the assembly in front of the spider is around 88mm. This means the balance of the mass in front of, and behind the spider, is almost perfectly in balance, leaving just a few grams resting on the surround. Having a more rugged spider makes sense as it will carry most of the drivers mass.

The oval spider ring will be mounted such that the driver is intended to have the oval standing vertically when the driver is mounted. The spiders horizontal width will be around 163mm, while its vertical width will be around 175mm. This gives a spider that is less prone to sag over time.
 
I mave made some renderings to better show what is going on here.

Her is the oval fixing ring for the spider exposed:

attachment.php



Here you can see the exposed motor parts:

attachment.php



And a side view of the entire thing:

attachment.php
 

Attachments

  • Spider assembly.png
    Spider assembly.png
    283.7 KB · Views: 1,385
  • Exposed motor.png
    Exposed motor.png
    134.2 KB · Views: 1,326
  • Woofer side view.png
    Woofer side view.png
    116.8 KB · Views: 888
Maybe I'm not understanding the spider ring right, but wouldn't it create odd stretching and potentially mechanical noise?

Yes, it does create some odd stretching, but the spider on this driver, who will normally work below 10mm of excursion, is actually wider than the one on the 18N862, which is intended for 19mm excursion. So the odd stretching will be extremely limited.

Also the material of a spider is made to be distorted without emitting any sound on its own.

A very small percentage of the total excursion goes to the outer roll of the spider. On the 0 and 180 degree points, there are 5 rolls, so 100% of the motion will be distributed around 20% on each roll. On the 90 and 270 degree points, there are only 4 rolls. So around 25% of the motion will be distributed on the 4 rolls. But as the motion in the 4.th roll is limited by the fixing ring at 0 and 180, the 5.th roll will hardly move at all. This means the suspension movement will be relatively even around the spider, and the odd stretching will be very limited.

The spider will, off course, be fully glued to the surface of the spider ring in order to make sure it does not flap around making any kind of noise.
 
I believe I did not explain the resonance control ring on the spider properly.

The distance from the voice coil to the fixed outer edge of the spider, the spider compliance, its mass, and its inner loss, are the parameters that decide the resonant frequencies of the spider. By changing one of them (except the loss) you will change the resonance frequency of the spider. A normal round spider has the same resonance freqyency at all angles.

By adding this oval ring, and gluing the spider completely to it, the spider will have lower mass, shorter distance and lower compliance at some angles than others, all resulting in higher resonance frequencies.

Having several different resonance frequencies in one connected system, not being in phase, makes the system cancel out all resonances, leaving only tiny parts of energy for each resonance. The loss factor is more or less unchanged. So it can effectively remove the tiny remains of resonances left from the reactive parts of the spider.

This is especially important for this spider since it is the dominant part of the suspension, it is a bass + midrange driver, and it has very long excursion suspension, low moving mass, all leading to a large, light weight and highly compliant spider that would normally be prone to resonate.