Report KD 708 about a 22-pole PM-generator available

[Adriaan Kragten]

Report KD 708 can be copied for free from my website: www.kdwindturbines.nl at the menu KD-reports. The title of this report is: "Ideas about a 22-pole permanent magnet generator for the VIRYA-2.7 windmill using a 4-pole, 3-phase, 1.5 kW asynchronous motor frame size 90 and 27 1/2 neodymium magnets size 40 * 10 * 4 mm". To give an impression of this generator, figure 1 out of KD 708 is added as an attachment.

[SparWeb]

Thank you for the report Adriaan.

I think I'd like to study those slots for magnets in detail.  With a gap on the sides of each magnet, less flux is diverted in a short-circuit into the rotor.  Can you confirm for me that each magnet is oriented with its "N" pole facing outward, forcing the iron between magnets to have a "S" face?  That seems to be a very efficient use of magnet material.

[Adriaan Kragten]

Yes, each magnet is mounted with the north pole to the outside. The small grooves beside the magnets have two functions. The first function is to make that a south pole has the same width (10 mm) as a north pole. The second function is to prevent internal magnetic short-circuit through the iron of the armature at the sides of the magnets. But even with these grooves, some magnetic flux will be lost. However, in chapter 4 of KD 708, I have calculated the flux density in the stator spokes and found that it is 1,79 T, so higher than saturation. So even with rather thin 4 mm magnets, a very strong magnetic flux is realized. The magnet costs of this generator are only about € 28 which is very low for a PM-generator with this armature volume. As the core heads are very small, the amount of copper which is needed for the winding is also smaller than for a normal 4-pole generator. The small coil heads also results is less copper losses so in a higher efficiency. Mounting of the coils is also very easy as the generator has a 1-layer winding with no crossing coil heads. So I think that this generator can be rather cheap if it would be mass produced.

[SparWeb]

Indeed, the cost advantage become obvious pretty quickly.  I see it as a path for motor conversion, for my own purposes.  Your additional comments about the economy of the windings is also well taken.  The intent of such a generator's output is likely to only be for rectifying to DC, correct?

[Adriaan Kragten]

In chapter 5 of KD 708, it is described how the optimum winding can be found such that it can be used for 24 V battery charging if the winding is rectified in star and for 12 V battery charging if the winding is rectified in delta. I have used the VIRYA-2.7 rotor as an example to describe the optimum matching. One can also chose voltages of 48 V and 24 V and this results in double the number of turns per coil. So I only describe the use of this 22-pole generator for low voltage battery charging.

However, the generator has a rather high frequency and it might be possible to drive a small 3-phase asynchronous motor if the generator has the correct high voltage winding. The frequency is 50 Hz for a rotational speed of only 272.7 rpm. But I think that even the smallest pump will have a too large motor for this 22-pole generator. So if one wants to drive a pump, one can better use the big 34-pole generator as described in report KD 614 for the VIRYA-5 combined to a 1.1 kW pump motor or the very big 46-pole generator as described in report KD 624 for the VIRYA-6.5 combined to a 2.2 kW pump motor.

[SparWeb]

Again, your reports have proven to be inspiring to me.

I have built some models in FEMM to examine the effectiveness of this magnet arrangement, and on motor-conversions I find that the number of magnets actually needed can be cut in half.  Some other features of this method must be kept in mind, such as the increased sensitivity to the gap, but I've been able to find some very cost-effective combinations.  It works just as well on 4- and 6- pole generators as it does on your 11-pole example.

Thank you!

[bigrockcandymountain]

What you just modelled there looks very much like my motor conversion. I have magnets for both n and s poles though.  So a question.  How thin can you go with  neo magnets and still saturate the stator? Of course it depends on stator thickness but let's say your 3 hp stator or something like that.

I did 1/2" wide magnets 1/4" thick, so the air gap could be really small.  The corners only stick up .005 " higher than the middle if the mags. 

[MattM]

What happens when you bridge magnetic fields using synchronous reluctance?

Before you laugh, the idea is to cut magnet count using opposing flat plates separated by an air gap.  They use notched solid core plates to bend flux - across multiple poles simultaneously - on some high efficiency motors, so why not on a PMG?  Maybe instead of 22 magnets you could cut it down to 1 big one or a small number of pairs with the magnetic flow front to back.  Some people find many small magnets cheap to obtain, so likewise you could easily use a high number of smaller magnets, too.

[SparWeb]

BRCM,
I did a range of sizes and types.
You can (hopefully) click to expand the size to fill your screen.
If not, you can open the original file in a new tab (right-click, "open image in new tab") to see the full size.

1/8" thick magnets, 5 per pole.


1/4" thick magnets, 5 per pole.


3/8" thick magnets, 5 per pole.


3/8" thick magnets, 6 per pole.


1/2" thick magnets, 6 per pole.

[SparWeb]

Matt,
I don't think I follow your idea.

If it helps, I went to KJ magnetics website and priced the magnets for each of the 5 combinations I show above. 

L       W       Thk          KJ #      Qty      $       Total
0.50   0.50   0.125      B882    150   1.22   183.00
0.50   0.50   0.250      B884    150   2.09   313.50
0.75   0.50   0.250      BC84    105   2.99   313.95 (seems to be the most cost effective for the flux attained)
0.50   0.50   0.375      B886    150   2.97   445.50
0.38   0.38   0.375      B666    252   1.82   458.64
0.38   0.38   0.500      B668    252   2.34   589.68

Note that the last 2 combinations would cog terribly with the magnets so well aligned with the stator teeth.  I believe cogging would be low in the 5-magnet combinations.  I include the others for the sake of comparison and the work was easy to add, once I got started.

[Adriaan Kragten]

Again, your reports have proven to be inspiring to me.

I have built some models in FEMM to examine the effectiveness of this magnet arrangement, and on motor-conversions I find that the number of magnets actually needed can be cut in half.  Some other features of this method must be kept in mind, such as the increased sensitivity to the gap, but I've been able to find some very cost-effective combinations.  It works just as well on 4- and 6- pole generators as it does on your 11-pole example.

Thank you!

(Attachment Link)

Nice picture. In this picture you have five magnets for north poles opposed to six stator poles. So the fluctuation of the sticking torque will be flattened strongly for the north poles. However, you have only one south pole opposed to six stator poles and this south pole will have a position for which the magnetic flux flows easiest. So the south pole will cause a certain peak on the sticking torque. This problem can be solved if you make grooves in the remaining material such that you also get five south poles with the same width as the north poles. The final result is that you have totally 20 armature poles for a stator with 24 stator poles. The number of preference positions will be less than for 22 armature poles and 24 stator poles but the fluctuation of the sticking torque may still be low enough.

[bigrockcandymountain]

Wow that is a nice comparison.  What colour on those pictures represents a saturated stator?  Im surprised how even the flux is, even in the non magnet poles.    Looks like someone needs a milling machine.

[Adriaan Kragten]

Again, your reports have proven to be inspiring to me.

I have built some models in FEMM to examine the effectiveness of this magnet arrangement, and on motor-conversions I find that the number of magnets actually needed can be cut in half.  Some other features of this method must be kept in mind, such as the increased sensitivity to the gap, but I've been able to find some very cost-effective combinations.  It works just as well on 4- and 6- pole generators as it does on your 11-pole example.

Thank you!

(Attachment Link)

To my opinion there is a mistake in the geometry of the stator stamping, at least if I compare the given geometry with the geometry of the stator stampings of Kienle & Spiess which I used for my generator. For these stampings, the stator spokes heave a constant width and the stator grooves are 15° tapered to get the maximum cross sectional area for the copper wires. The flux density in spokes with a constant width will be almost constant.

[bigrockcandymountain]

Cms magnetics still has the 1 1/2" x 1/2" x 1/4" n42 for $1.49 each.
That makes them 1/4 the price you quoted for the 3/4" long 1/4 thick.

They have 2 countersunk holes for mounting and can be had N or S countersunk.  That makes the magnets less than $100 for this machine. 

[SparWeb]

Hello Adriaan,
You are correct, the slots probably are tapered as you describe them.  I attempted to measure the stator with wiring still inside.  I didn't want to disturb or damage the wiring so I have been reluctant to push too forcefully inside.  That's limited my ability to measure the stator dimensions, but your reasoning is sound and probably straight teeth is more normal.

One thing to consider about making computer models:

These models I'm making are "comparisons".  They are not expected to provide accurate absolute predictions.  What I'm doing will help see the value of differences and changes, choose one option vs another. 
Picking a specific point and reading a specific field strength value from the colour is probably not worth very much.

[SparWeb]

BigRock,
Great deals on magnets!

Like I said above, the colours are sort-of useful, but not accurate.  Just guesses. 
In pure iron Saturation is about 2 Tesla IIRC, but I think the practical steels are about 1.5T or 15,000 Gauss may be a good reference point. 
It stands to reason that materials used in motors are chosen for higher saturation point, so they may be closer to 2T. 

I can find info about these materials from Carpenter.  I don't know if their materials are the exact same ones used in motors, however.
https://www.carpentertechnology.com/en/product-solutions/?catId=225

Adriaan has been in contact with a manufacturer of stator laminate stampings.  He may be able to share some of that information (depending on what they permit to be shared).

[MattM]

Matt,
I don't think I follow your idea.

Low reluctance motors using stainless steel forms to curve magnetic fields into better paths seems like it the current rage with electric cars.  They can replace the copper windings inside an armature with solid cores that have carefully created channels to actually improve efficiency by virtually eliminating eddy currents.  It has me thinking the copper windings on an electric motor are literally the opposite of the radial PMGs, therefore you could craft the stator to do something similar in the former - only on the outside - to drop reluctance in the radial PMG.

The simulations you made seem to suggest there is a net 6-pole arrangement in the magnetic field.  There are ways to probably get there in non-traditional ways.  One example is simply six magnets around the circumference each arranged in alternating directions.  The second is two rows of quadrupoles laying in opposing directions, taking advantage of their net effect in 3-dimensions.  (My illustration is out of sync with the magnetic lines by 90º, but it takes too long to make a new one.)  The last is a single large magnet laying front to back, with three prong rotors connected to a steel outer ring and each rotor arm staggered 60º off the opposing rotor.  You can do literally any number of poles doing this as long as the front and back plates are synched appropriately.  Sorry I wandered off topic, but I came the other night after watching several YouTube videos and reading different theories on reluctance.

[SparWeb]

Hi Matt,
I see more, dimly...

The motor conversion that I'm considering is starting with a stock 6-pole motor, which is the reason you see 6 poles in the pattern.  But as Adriaan started off this thread with 22 poles, it looks like the idea that I'm chasing really isn't tied to any particular geometry. Also, the scope of a new motor design is much larger than the simple conversion that I plan to do, putting many constraints on what I can do, while Adriaan is free to optimize.  Especially if I choose not to re-wind the motor, as I usually do.  On that point, Adriaan has also proposed cost-conscious methods to make a complete motor project very practical.


I tried looking up the type of alternator that I think you're referring to...  what I found seems very peculiar and different from what I'm trying to do.

[SparWeb]

Here's another run with a more correct stator geometry.  These are 1/2" x 3/8" magnets again, like the 3rd example above.  This time I deliberately traced a path crossing the 6 teeth in one pole to represent the total flux passing through one coil group of one pole.  I think I could try to estimate the output of the generator from this, but I've tried before and wasn't convinced of the sanity of the results.





The flux in the N and S poles is roughly equal (about 1.5 milliWebers).  The normal field strength is the same in both, but that figure is very sensitive to how I draw the lines, while the total flux is not.
If you look carefully at the shading of the teeth in each, you can see that the N pole has more flux in the middle teeth and less in the outer two.  The S pole has much more evenly-distributed flux in all 6 teeth.
It all averages out in the total flux through each pole (also indicating a closed circuit where everything that goes through one point comes back to it, without leakage).  So at this level there seems to be some sanity.

[Adriaan Kragten]

Here's another run with a more correct stator geometry.  These are 1/2" x 3/8" magnets again, like the 3rd example above.  This time I deliberately traced a path crossing the 6 teeth in one pole to represent the total flux passing through one coil group of one pole.  I think I could try to estimate the output of the generator from this, but I've tried before and wasn't convinced of the sanity of the results.





The flux in the N and S poles is roughly equal (about 1.5 milliWebers).  The normal field strength is the same in both, but that figure is very sensitive to how I draw the lines, while the total flux is not.
If you look carefully at the shading of the teeth in each, you can see that the N pole has more flux in the middle teeth and less in the outer two.  The S pole has much more evenly-distributed flux in all 6 teeth.
It all averages out in the total flux through each pole (also indicating a closed circuit where everything that goes through one point comes back to it, without leakage).  So at this level there seems to be some sanity.

Now you show a picture of a stator with 36 grooves so with 36 poles. For the north poles you have five magnets opposed to six stator pole but for the south poles you have three big poles. These big south poles will cause a strong preference position every 10°. So you should make grooves in the south poles such that each south pole also has five small bars with the same width as the magnet width of the north poles. In this case you have totally 30 armature poles for a stator with 36 stator poles. The armature pole pitch is then 12° and the stator pole pitch is 10°. So the difference is 2°. This means that you will get 360 / 2 = 180 preference positions per revolution. This is equal to 30 * 36 / 6. I am not sure if this is enough to flatten the peak on the clogging torque. For the original generator with 22 armature poles and 24 stator poles, there are 264 preference positions per revolution and this is certainly enough.

In your pictures it can be seen that there is a strong magnetic flux in all stator spokes but also at six positions at the bridge in between the bottom of the stator grooves and the outside of the stator. This is because the magnetic flux coming out of three spokes has to pass through this bridge. For the original 22-pole generator there is only saturation in the spokes because only half of the magnetic flux flowing through a spoke is flowing through the bridge. Therefore it would have been better to use a stamping with a larger inside diameter and smaller bridges like used for a 6-pole motor. However, this wasn't possible because a stamping with 24 stator poles of frame size 80 is only supplied for a 4-pole motor.

[SparWeb]

I am not sure if this is enough to flatten the peak on the clogging torque.

Neither am I.  My guess is a series of four weak preference positions 2 degrees apart followed by one strong preference position.  This series of 4-weak-1-strong would repeat every 10 degrees.  Cutting 4 slots in the S faces might reduce the intensity of the strong position, but may increase the preference of the other 4.

[MattM]

Your cogging comes from eddy currents.  Eliminate them to remove reluctance.

[Adriaan Kragten]

I am not sure if this is enough to flatten the peak on the clogging torque.

Neither am I.  My guess is a series of four weak preference positions 2 degrees apart followed by one strong preference position.  This series of 4-weak-1-strong would repeat every 10 degrees.  Cutting 4 slots in the S faces might reduce the intensity of the strong position, but may increase the preference of the other 4.

I think that you are right. There is a difference in between the original 22-pole generator and this 6-pole generator with divided poles. For the original 22-pole generator, the flow of the magnetic flux in between a north and a south pole is the same for all poles. If there is a preference position when an armature pole is just opposite a stator pole, this is the case for only two of the 22 poles and it happens every 1.3636°. So you will get 360 / 1.3636 = 264 identical preference positions per revolution. But for the 6-pole generator with poles divided in five bars there are six points where the magnetic through the iron of the armature is maximal. This must also cause some non uniformity of the magnetic flux in the stator.

[SparWeb]

This must also cause some non uniformity of the magnetic flux in the stator.

Yes, I'm seeing that, too.  Although it's just a simulation it is interesting to explore nonetheless.

I tried another sim, this time with little slots cut in the S pole faces.
Very little change in any of the global flux or the local field intensity at any point.

There is some value in this prediction because it offers a new choice to me during the motor-conversion process. 
If the first steps are to machine a deep channel in the N pole for the N magnets, then I can complete that phase of the rotor modification and re-assemble the machine to check.  If I find that the cogging is objectionable, especially that 4-step-1-step effect, then I can remove the rotor and cut the narrow slots on the S pole, which might improve the situation.

[SparWeb]

I believe some people can use FEMM to do time and motion simulations, but I haven't figured it out, myself.
There would be a lot more insight to gain if I could figure out a way to use these FEMM simulations to predict a torque on the rotor.

[bigrockcandymountain]

Sounds like a plan is coming together for a 6 pole. That is a good thing. 

I would do offset rings of magnets to prevent cogging, but that makes the machining very tricky. If you are building new rotor bushings, you could do 2 different ones i guess.  If you are using the original laminated rotor, you could probably do it with a vertical mill but it would be difficult. 

[SparWeb]

The plan that's developing will use the existing laminated rotor.  I do not believe that there will be a way to determine in advance where the rotor poles currently are within the induction bars cast into the rotor, and hopefully it does not matter.  The aluminum casting on the ends of the rotor is quite long, and it may offer a substantial amount of support during the machining process, so the rotor laminates are less likely to come apart.

I have tried machining down induction rotors before, with a wide range of results.  It worked on my Toshiba conversion, but on a rotor I worked on for a friend, the laminates came apart, and a fresh rotor was necessary.  On my Baldor conversion, I chose such large magnets that I had no doubt that the laminations would disintegrate, and just planned a new solid rotor from the start.

With this new strategy, much of the outer surface of the rotor can be kept. I have reason to expect it to remain intact as I machine pockets for the magnets.

This planned conversion deserves a new thread.  I'll start it once I'm committed to starting.  Here are some photos of the candidate motor:

[Adriaan Kragten]

I believe some people can use FEMM to do time and motion simulations, but I haven't figured it out, myself.
There would be a lot more insight to gain if I could figure out a way to use these FEMM simulations to predict a torque on the rotor.

I don't know how you can predict the torque from this FEMM simulations but there is a certain relation in between the strength of the magnetic flux in the air gap Bl and the maximum torque Q. For my normal 4-pole VIRYA generators with inclined magnet grooves, Bl is calculated by formula 5 of public report KD 341. The stator iron is about saturated if the calculated value of Bl is larger than 0.9 T, so in this case you get the maximum torque level for a certain armature geometry.

A certain value of Bl results in a certain tangential shearing stress tau (I can't write Greek letters on this forum). The tangential force F is the product of tau times the cylindrical area of the armature, so F = tau * pi * r * l in which r is the armature radius and l is the length of the stator stamping. The torque Q = F * r. So Q = tau * pi * r^2 * l. As tau is proportional to Bl, Q can be written as Q = C1 * Bl * pi * r^2 * l. Pi * r^2 * l is the armature volume lying within the stator stamping. So the maximum torque Q is proportional to the product of Bl times the armature volume.

The factor C1 can be determined if a certain generator is built and measured on a test rig. Extensive measurements for a 4-pole PM-generator frame size 90 are given in my public report KD 78. In figure 4 and 8 of this report you can see that the maximum torque is about 29 Nm if the generator winding is rectified in start. So it is almost independent of the chosen voltage even if the voltage is zero. So if you know the maximum torque level of this generator, you can calculate it for a bigger one if you compare the armature volumes and the Bl values. If both generators have the same Bl value, the maximum torque level of the bigger generator is only determined by the ratio in between the armature volumes. However, this method to predict the maximum torque level of a bigger (or smaller) generator can only be used if both generators have the same armature construction.