I took apart a Peerless 850128 (CSC217R, moving parts identical to HDS205 except for 4 vs. 2 layer VC) and a Vifa TC14 proto with fiberglass membrane. The trick is to soak them in hot soapy water for a couple of hours. The glue for the surround/basket and spider/basket joints turns whitisch and you can carefully peel off the parts (it's not PVA because it does not really turn soft, maybe some latex dispersion). With the CSC, you'd better take off the lacquered paper dust cap first, which is done by short soaking and use of a scalpel blade.
Anyway, in the Vifa, the bottom end of both windings where ending right on top of each other. On the top end, the 2nd winding ended 1.2 mm before the 1st, which is a lot considering there is only +/-3 mm overhang.
The CSC has a nominal overhang of +/-5.5 mm. The 1st and 2nd windings end flush on both sides. The 3rd and 4th are flush with them on the bottom end but on the top side, the 3rd ends about 2 mm early, the 4th 3.5 mm.
This will effectively make for an assymmetrical Bxl-curve. With the pole piece not being undercut and the pole plate being extended by no more than 0.5 - 1 mm, the stray field will be more extended on the bottom side, hence the assymetries of B-field and winding (l) distribution would appear to add.
I have never seens this strange winding technique in any driver with a vented spider, i.e. one where one could look at the top end of the VC without taking the thing apart.
Anyway, in the Vifa, the bottom end of both windings where ending right on top of each other. On the top end, the 2nd winding ended 1.2 mm before the 1st, which is a lot considering there is only +/-3 mm overhang.
The CSC has a nominal overhang of +/-5.5 mm. The 1st and 2nd windings end flush on both sides. The 3rd and 4th are flush with them on the bottom end but on the top side, the 3rd ends about 2 mm early, the 4th 3.5 mm.
This will effectively make for an assymmetrical Bxl-curve. With the pole piece not being undercut and the pole plate being extended by no more than 0.5 - 1 mm, the stray field will be more extended on the bottom side, hence the assymetries of B-field and winding (l) distribution would appear to add.
I have never seens this strange winding technique in any driver with a vented spider, i.e. one where one could look at the top end of the VC without taking the thing apart.
simulation of B-field
I took the driver apart, measured the geometry of the parts and calculated the field with Maxwell 9 SV. I assumed 1008 steel and a Ceramic 5 magnet (Ceramic 8 does not make a huge difference).
Here's a 2D-rotational symmetric plot of B. The rotational axis is the non-existent pole piece vent.
I took the driver apart, measured the geometry of the parts and calculated the field with Maxwell 9 SV. I assumed 1008 steel and a Ceramic 5 magnet (Ceramic 8 does not make a huge difference).
Here's a 2D-rotational symmetric plot of B. The rotational axis is the non-existent pole piece vent.
Attachments
field along VC +/- possible excursion
The green graph is the B-field along the center of the gap (i.e. r=16.875 mm) and from -14 mm to + 14 mm above the center of the top plate.
Hence, 14 mm = vertical center of gap and center of voice coil,
5.5 and 22.5 mm = first and last winding of VC when in rest position,
0 and 28 mm = first and last winding of VC when displaced by +/- x_lin
The green graph is the B-field along the center of the gap (i.e. r=16.875 mm) and from -14 mm to + 14 mm above the center of the top plate.
Hence, 14 mm = vertical center of gap and center of voice coil,
5.5 and 22.5 mm = first and last winding of VC when in rest position,
0 and 28 mm = first and last winding of VC when displaced by +/- x_lin
Attachments
The purple graph is the B-field along the same line, but with top and bottom reversed. This is essentially a symmetry check. As you can see, the symmetry is not too good. The stray field below the gap is stronger than the stray field above the gap.
It is interesting to see that you can get a flat B curve only along the inner 3.9 mm of the 6 mm gap. At the top and bottom end of the gap, the field is already down to 0.8 T = - 15% of the plateau.
It is interesting to see that you can get a flat B curve only along the inner 3.9 mm of the 6 mm gap. At the top and bottom end of the gap, the field is already down to 0.8 T = - 15% of the plateau.
how the asymmetrical VC works as a band aid to fix the asymmetric slopes
Even without having done the integrations of the force along various positions of the VC (if anyone's mastered the Maxwell calculator menu, drop me a line), I am beginning to understand how it works:
I was intially confused because I was looking at absolute B and wondered why the top end of the VC, which has a lower density of turns than the center and bottom end, would be placed in the area of the weakest field.
But the secret is in looking at delta force integrated over the full length of the VC. Consider the VC to be in the rest position, i.e. extending from 5.5 to 22.5 mm, initially. Now, let it move inwards (downwards) so that it extends from 0 to 17 mm. As it travels these 5.5 mm downwards, the lower end sees a slowly decreasing field, but the upper end sees a strongly increasing field. So it is the asymmetry in the rate of change of the stray field that is causing potential problems. Now if the upper end of the VC has a lower winding density, this offsets the strong increase of B.
The asymmetrical winding technique helps to symmetrize the inductance curve, too, even if this is probably a second order effect. As the coils moves down, there is more steel enclosed in the VC, acting as a core. With the lower winding density, the increase in inductance with inwards travel is not as marked as it would be with uniform winding density.
Even without having done the integrations of the force along various positions of the VC (if anyone's mastered the Maxwell calculator menu, drop me a line), I am beginning to understand how it works:
I was intially confused because I was looking at absolute B and wondered why the top end of the VC, which has a lower density of turns than the center and bottom end, would be placed in the area of the weakest field.
But the secret is in looking at delta force integrated over the full length of the VC. Consider the VC to be in the rest position, i.e. extending from 5.5 to 22.5 mm, initially. Now, let it move inwards (downwards) so that it extends from 0 to 17 mm. As it travels these 5.5 mm downwards, the lower end sees a slowly decreasing field, but the upper end sees a strongly increasing field. So it is the asymmetry in the rate of change of the stray field that is causing potential problems. Now if the upper end of the VC has a lower winding density, this offsets the strong increase of B.
The asymmetrical winding technique helps to symmetrize the inductance curve, too, even if this is probably a second order effect. As the coils moves down, there is more steel enclosed in the VC, acting as a core. With the lower winding density, the increase in inductance with inwards travel is not as marked as it would be with uniform winding density.
possible mods
First of all, I shaved 0.5 mm off the radius of the pole piece to make room for a copper cylinder. The caused the plateau to drop from 0.93 T to 0.81 T. At the same time, the flat region became wider, 5 mm compared to 3.9 mm before. Essentially, I took the top off the B curve, the sides didn't get much better.
Next I simulated what would happen if I reduced the magnet height from 20 to 15 mm. I happen to have a stash of 5 mm dia, 10 mm long Neo 42 magnets. On the quick, I took Neo35 in the simulation (the difference is like 10-15%) and 5 mm 1010 steel washer used as a spacer.
First thing to notice is that the field in the gap dropped from 0.8 to 0.4 T, even with a solid ring of 5 mm of Nd! The length of the Nd magnet does not contribute too much, but the area does. If I were to buy new Nd magnets, I'd probably go for 5 mm length and greater diameter and use two 5 mm steel spacers.
Anyway, the field now looks much more symmetrical!
First of all, I shaved 0.5 mm off the radius of the pole piece to make room for a copper cylinder. The caused the plateau to drop from 0.93 T to 0.81 T. At the same time, the flat region became wider, 5 mm compared to 3.9 mm before. Essentially, I took the top off the B curve, the sides didn't get much better.
Next I simulated what would happen if I reduced the magnet height from 20 to 15 mm. I happen to have a stash of 5 mm dia, 10 mm long Neo 42 magnets. On the quick, I took Neo35 in the simulation (the difference is like 10-15%) and 5 mm 1010 steel washer used as a spacer.
First thing to notice is that the field in the gap dropped from 0.8 to 0.4 T, even with a solid ring of 5 mm of Nd! The length of the Nd magnet does not contribute too much, but the area does. If I were to buy new Nd magnets, I'd probably go for 5 mm length and greater diameter and use two 5 mm steel spacers.
Anyway, the field now looks much more symmetrical!
Attachments
curve symmetry
As you can see, the top 50% of the B curve look a lot better now. The outer stray field can probably be symmetrized by carving out the pole piece, similar to the construction of the Usher 9950 tweeter (cross-section at Northcreek) or the XLS subwoofers (there is also a cross sectional view some place on the d-s-t site).
Getting the field back up to slightly below 1 T will take an awful number of Nd magnets. I should consider reusing the original 20 mm magnets and simply glue a 5 mm steel plate onto the top of the pole piece to obtain the extented pole piece.
Does anybody know how much magnetization a ceramic magnet looses when the return circuit is opened and closed again (assuming it was orginally charged with the circuit in place)?
As you can see, the top 50% of the B curve look a lot better now. The outer stray field can probably be symmetrized by carving out the pole piece, similar to the construction of the Usher 9950 tweeter (cross-section at Northcreek) or the XLS subwoofers (there is also a cross sectional view some place on the d-s-t site).
Getting the field back up to slightly below 1 T will take an awful number of Nd magnets. I should consider reusing the original 20 mm magnets and simply glue a 5 mm steel plate onto the top of the pole piece to obtain the extented pole piece.
Does anybody know how much magnetization a ceramic magnet looses when the return circuit is opened and closed again (assuming it was orginally charged with the circuit in place)?
Attachments
Excellent info! And Maxwell V9 is nice, isn't it? The 3D with transient is really cool, but $$$!
For steel, it's probably better to estimate 1010 grade of 12L14; those seem to be most common with overseas production (I'm 99.9% sure the steel in those drivers comes from India/China). Also, probably want to look at Y30 grade magnet; Y35 or Ceramic 5 isn't used too much.
One thing to note with Neo - this is one case where height is important. I've found if the Neo is thin (say less than 6mm) a good amount of the flux simply shorts back around itself, through the air. Thicker magnets are needed here, as well as more area.
You can definitely see the impact of cutting back to add a copper shorting sleeve on the pole - a ~10% loss in flux. This is why Peerless and many others will simply place a ring down below the top plate, and leave the gap itself untreated.
Lastly, about the winding length differences. A difference of 1.2mm is pretty big, but not that uncommon; typically you'll see between 2 and 4 turns differences, unless you ride the voice coil winders really hard to actually wind them properly stacked (and truncated).
Dan Wiggins
Adire Audio
For steel, it's probably better to estimate 1010 grade of 12L14; those seem to be most common with overseas production (I'm 99.9% sure the steel in those drivers comes from India/China). Also, probably want to look at Y30 grade magnet; Y35 or Ceramic 5 isn't used too much.
One thing to note with Neo - this is one case where height is important. I've found if the Neo is thin (say less than 6mm) a good amount of the flux simply shorts back around itself, through the air. Thicker magnets are needed here, as well as more area.
You can definitely see the impact of cutting back to add a copper shorting sleeve on the pole - a ~10% loss in flux. This is why Peerless and many others will simply place a ring down below the top plate, and leave the gap itself untreated.
Lastly, about the winding length differences. A difference of 1.2mm is pretty big, but not that uncommon; typically you'll see between 2 and 4 turns differences, unless you ride the voice coil winders really hard to actually wind them properly stacked (and truncated).
Dan Wiggins
Adire Audio
Re: how the asymmetrical VC works as a band aid to fix the asymmetric slopes
This is the foundation for TC's LMT motor stuff - add extra windings/layers at the end of the voice coil, and a few extra in the middle as well, to kind of create a reverse XBL^2 motor; rather than two or more high flux regions, you have two or more high L regions.
Downside, of course, is incrreased Mms, Le, and loss of B because of the thicker gap required for the extra layers...
Dan Wiggins
Adire Audio
I missed this earlier... Yes, many people forget that BL is both B and L - flux and voice coil - and that it is the total integrated flux over the voice coil. Making the windings assymmetric can linearize or at leas symmetrize the BL curve.capslock said:
This is the foundation for TC's LMT motor stuff - add extra windings/layers at the end of the voice coil, and a few extra in the middle as well, to kind of create a reverse XBL^2 motor; rather than two or more high flux regions, you have two or more high L regions.
Downside, of course, is incrreased Mms, Le, and loss of B because of the thicker gap required for the extra layers...
Dan Wiggins
Adire Audio
DanWiggins said:
Thanks for the kind words, but I have merely begun to dabble with this stuff. I chose Maxwell because the geometry editor of FEMM drove me nuts. Now it's Maxwell's calculator that is driving me nuts, and the help texts are really cryptic.
I will have to find the data on Y30. It doesn't seem to fit neither into the American system (Ceramic 1, 3, 5, 8) nor the DIN system (e.g. HF28/16).
Greetings,
Eric
magnetic properties of hard ferrites
http://www.kaiven.com/property.asp
Dan, I am a little confused. Acording to this link, Y30 is actually the same thing as C5. Y35 is some strange stuff with slightly higher remancence, but lower coercivity. Clostest US equivalent is C11, which I have never heard of.
=> So I guess I am ok to continue my simulations of the original structures with C5.
But from a cost standpoint, I would be great if I could recycle the ferrite magnets instead of using Nd. Problem is, I don't want to go into building a charger.
Do you happen to know if there is any significant loss of magnetization if I take a motor that was charged after assembly, take out the back plate and reassembly after modification?
Greetings,
Eric
http://www.kaiven.com/property.asp
Dan, I am a little confused. Acording to this link, Y30 is actually the same thing as C5. Y35 is some strange stuff with slightly higher remancence, but lower coercivity. Clostest US equivalent is C11, which I have never heard of.
=> So I guess I am ok to continue my simulations of the original structures with C5.
But from a cost standpoint, I would be great if I could recycle the ferrite magnets instead of using Nd. Problem is, I don't want to go into building a charger.
Do you happen to know if there is any significant loss of magnetization if I take a motor that was charged after assembly, take out the back plate and reassembly after modification?
Greetings,
Eric
steel properties
http://www.precisionsteel.com/tech_data/chemical.asp?n_cat_id=1
So 1008 steel has < 0.1% carbon, 0.25-0.5% Mn, y 0.05% S,
1010 steel has 0.08 - 0.13% C
1012 steel has 0.1 - 0.15% C
The most common unalloyed steel in Europe is St37-2 (now called S235JR, material code 1.0037). According to www.metallograf.de, the typical composition is .08% C with 0.17% being the max they ever found. Mn is 0.35%, S 0.025%.
So for making small parts, St37 seems like a reasonable choice! Of course, I'd like to get my hands on a supply of Telar (Armco)....
I also found that most Chinese makers of cold-forged pole pieces, pole pieces and yokes advertise 1008 steel. So buying cheap car hifi or surplus speakers on ebay and salavaging the motor parts might also be an option.
http://www.precisionsteel.com/tech_data/chemical.asp?n_cat_id=1
So 1008 steel has < 0.1% carbon, 0.25-0.5% Mn, y 0.05% S,
1010 steel has 0.08 - 0.13% C
1012 steel has 0.1 - 0.15% C
The most common unalloyed steel in Europe is St37-2 (now called S235JR, material code 1.0037). According to www.metallograf.de, the typical composition is .08% C with 0.17% being the max they ever found. Mn is 0.35%, S 0.025%.
So for making small parts, St37 seems like a reasonable choice! Of course, I'd like to get my hands on a supply of Telar (Armco)....
I also found that most Chinese makers of cold-forged pole pieces, pole pieces and yokes advertise 1008 steel. So buying cheap car hifi or surplus speakers on ebay and salavaging the motor parts might also be an option.
pinkmouse said:
Al,
horseshoe magnets were made of AlNiCo which becomes demagnetized easily.
Somebody who has disassembled and reassembled a lot of drivers told me he'd never observed a noticable drop of Bxl or rise of Q_es, but we figured that it might well be due to the fact that the base of the pole piece is often in saturation, i.e. it will keep the flux in the gap constant even if there are small changes in the magnetization of the ferrite magnet. So it was not clear whether there actually is a loss of magnetization.
Personally, I don't mind a small change in Bxl, but it would worry me if the magnet was demagnetized to say 50%, because then it could potentially age.
optimized extended pole piece
This is the last iteration for now. I tried to hollow out my cylindrical pole piece, but it didn't make much difference. I kept getting a kink in my B curve about 3 mm above the top plate, so I started recessing the cylinder above the gap. In the end, I got something of a cross between an inverted T-pole and an extended cylindrical pole piece.
The symmetry is very nice now, but I don't like the curvature in the tails (the slopes should be as linear as possible). I suspect Maxwell is plotting the magnitude of B along the VC, but I am really only interested in the B-component perpendicular to the VC, so the drop off might be steeper.
Still, I suspect when I have to machine the outside surface of the pole piece, I might as well look into going for a full T design, i.e. undercutting the pole as well.
This is the last iteration for now. I tried to hollow out my cylindrical pole piece, but it didn't make much difference. I kept getting a kink in my B curve about 3 mm above the top plate, so I started recessing the cylinder above the gap. In the end, I got something of a cross between an inverted T-pole and an extended cylindrical pole piece.
The symmetry is very nice now, but I don't like the curvature in the tails (the slopes should be as linear as possible). I suspect Maxwell is plotting the magnitude of B along the VC, but I am really only interested in the B-component perpendicular to the VC, so the drop off might be steeper.
Still, I suspect when I have to machine the outside surface of the pole piece, I might as well look into going for a full T design, i.e. undercutting the pole as well.
Attachments
Interesting work Eric,
I've found your discussions of the disection and simulation of drivers very interesting. Nice work! I've disassembled an old AR 12" driver from the late 1970's since it needs a new spider and had some damage to the voice coil. I noticed that the motor is a very basic design and it could probably use some improvements. Here's the cone and voice coil:
http://www.classicspeakerpages.net/...id=5243&sub_topic_id=7029&mesg_id=&page=#7159
Here's Klipple data for a similar newer version made by Tonegen:
http://www.classicspeakerpages.net/...id=5243&sub_topic_id=7140&mesg_id=&page=#7350
To relate back to this thread, I'm thinking of experimenting with voice coil wind ideas in this thread.
Pete B.
I've found your discussions of the disection and simulation of drivers very interesting. Nice work! I've disassembled an old AR 12" driver from the late 1970's since it needs a new spider and had some damage to the voice coil. I noticed that the motor is a very basic design and it could probably use some improvements. Here's the cone and voice coil:
http://www.classicspeakerpages.net/...id=5243&sub_topic_id=7029&mesg_id=&page=#7159
Here's Klipple data for a similar newer version made by Tonegen:
http://www.classicspeakerpages.net/...id=5243&sub_topic_id=7140&mesg_id=&page=#7350
To relate back to this thread, I'm thinking of experimenting with voice coil wind ideas in this thread.
Pete B.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.