Single rotor, axial flux test results

[Beaufort]

Two different coil shapes were tested on this small alternator to investigate why they did not seem to match expected power output and efficiency.  The rotor and magnet spacing was kept the same.  The stator held the coil faces approx. 0.080" from the magnets.  

Rotor Plate:  ¼" thick, mild steel ring (9.75" OD, 5.19" ID)

Magnets:  Total of 24, 1" dia x ½ thick, N42, arranged in 12 poles in a figure-8  as shown in the pictures

68 turn coil:  14 AWG wound 0.5"-0.55" thick, leg width 0.8".  Note the overlap on both sides of the round magnet when passing over the center.

http://www.fieldlines.com/images/scimages/11755/68t.jpg ">

67 turn coil:  16 AWG wound 0.4" thick, leg width 0.62".  The inside wedge shape was opened up to avoid overlap.

http://www.fieldlines.com/images/scimages/11755/67t.jpg ">

The stator had an arm attached that pressed down on a digital scale, 9" from the center of rotation for the alternator.  The single rotor rides on a spindle that passes through a bearing at each end of the test setup.  The scale is zeroed out before each run to eliminate the off-center mass of the lever arm.  The rotor is driven with 2 V-belts by a 36V electric golf cart motor, with several selectable speeds.  The 3-phase alternator output is rectified and wired to a deep cycle battery.  Alternator speed is checked with a photo tachometer and a reflective target on the alternator-side drive pulley.

Power is calculated according to the formula given at the following link:

http://www.epi-eng.com/piston_engine_technology/power_and_torque.htm

Where P (HP) = (Torque (lb-ft) * RPM)/5252

68T coil (Star configured):  

  21.7 VDC at 380 RPM which gives 0.0571 V/RPM, or 12V cut-in at 210 RPM.

  No load test:

  69 grams at 380 RPM (6W) and 220 grams at 890 RPM (46W)

68T coil (Delta configured):  

  7.61 VDC at 240 RPM which gives 0.0317 V/RPM

67T coil (Star configured):

    13.48 VDC at 219 RPM which gives 0.0616 V/RPM, or 12 V cut-in at 195 RPM.

   No load test:

  80 grams at 219 RPM (4W) and 145 grams at 650 RPM (22W)

67T coil (Delta configured):

   14.7 VDC at 412 RPM which gives 0.0357 V/RPM

   No load test:

   160 grams at 412 RPM (15W)

Here are the data tables used to determine the relative efficiency of each configuration:

And finally a chart of efficiency versus RPM for both sets.  

Conclusions:  These sets could use more data points, but it appears that the average of the efficiencies is a bit higher for the 67T coil versus the 68T coil (compare 68T Star to 67T Star, and 68T Delta to 67T Delta).  The reasons are open to debate, but I'm guessing that the slightly thinner coil helps given the single rotor design.

[prasadbodas2000]

That is a very good experimental way to compare performance on various coil configurations.

I am not sure about the precision of the stator torque measurements esp. when the load is in just a few grams and where the stator must have been floating but supported sideways somehow.

Even then the interesting finding is that when both the legs of the coil do not overlap magnet at same time, it seems to have give 8 (star) to 12 (delta) % more volts per revolution.

[willib]

Hi Beaufort

Your coils seem to be evenly spaced, but your magnets are not?

It also looks like your magnets are way too far apart?

I have a suggestion if you are willing to try it out.

On a 9 inch outside diameter circle , place 16 poles on the rotor.evenly spaced

It works out to 22.5 degrees .

http://www.fieldlines.com/story/2010/1/1/10337/89726

Then try a coil in there.

You can adjust it in and out to see how the output changes because of the relationship between the coil and the magnets.  

[Flux]

Now you have sorted out your input power formula the results seem much as I would expect.

On very low power efficiency will be low and most of the load will be friction. Your results may indicate that your thicker wire is close to the maximum size before eddy loss starts to become excessive.

You reach a maximum when you are clear of the friction losses and then at high power it falls again due to copper loss from resistance. Star will give lower efficiency in this comparison as it is running at low speed. If you did the star tests at 20v it would compare directly and would probably be the highest but even then it is not a direct comparison as the rectifier drop is not the same in each case.

Jerry will come out better than delta so things are fairly normal.

At 12v with 1.5v rectifier drop you will be hard pushed to get efficiencies over 70% and in fact these alternators running into a rectifier load never achieve the very high efficiencies possible into a balanced 3 phase heater load.

You may possibly have shown that the thicker wire is close to the limit but in general these results are within experimental error. I wouldn't set too much value on the changes of coil size and shape.

With a single rotor, flux goes just about everywhere other than where where it should so small changes in coil geometry are meaningless. Even with a dual rotor there is no clear cut case of best coil shape as long as you avoid the ridiculous.

At least it is good to see scientific tests that include input power, load tests with no measurement of input power are almost meaningless.

Flux

[Beaufort]

Great summary...I didn't want to put anything definitive out there to mislead anyone and there are plenty with more experience.  

So is a dual rotor alternator more efficient than a single rotor for a given power output?  (assume that the magnets and copper are adjusted to suit the lower/scattered flux density).  I haven't done any dual axials this way, but that might be next.

The theoretical models I have for these configuration show efficiencies over 70% for RPM under 300, and that model accounts for resistance losses and diode drop.  What I usually do is to backsolve the flux density from the initial no-load reading and it seems like the model matches well for Star power output, but the efficiencies are all higher than what I tested.  So I still don't feel like I can model total performance without adding a fudge factor for something else...or maybe my model is wrong.  

Yes, testing without measuring input is critical if one wants to put a set of wings on these alternators.  That's what shocked me with the 68T coils; the output matched my model but the input was way too high.  I'm sure a set of blades on it would have stalled and the overall result would have been poor.  It seems to make more sense to take the mechanical input power for an alternator (after testing it and getting the formula) and to match this to blade mechanical power using a known Cp/TSR plot.  I've read here that most people rely too much on electrical output and miss the mechanical matching.

[Beaufort]

The stator had a rigid spindle attached to it that passed through a set of pillow blocks on the ends.  It was supported sideways quite well in order to pull the belt drive tight, but it easily rotated to allow the torque measurement.  I put some foam at the contact point between the arm and the scale to absorb vibration (my pulley was off a bit), so there was a gauge error of about +/-20 g as it ran and jumped around.

[Beaufort]

Good catch.  The CAD model has the mangets off slightly at the top because I was rotating them around to see how they sweep over the coils.  I didn't manage to catch it before doing the screen shots...thanks for pointing that out.  They should be on standard 30 degree radials for the 12-pole design.

That's a good thread you posted on round magnets.  The spacing on this design was driven by stuffing the wedge coils in there on the 68T, and the magnets seemed like they were far enough apart.  I had a previous round 16 pole dual axial design where I put the magnets too close together and it didn't perform well.  I've read in other places that the magnet spacing should be greater than or equal to the air gap (on a dual axial).  As Flux said below, single rotor flux is a different bird and that's a big reason why I wanted to understand it better.  Single rotor has some nice benefits in terms of mechanical simplicity, drawbacks in other areas.

[Flux]

With these alternators the efficiency mainly depends on the ratio of internal resistance to load resistance. For a given amount of magnet the single rotor will link the turns much less effectively especially if the stator is reasonably thick so you can expect a dual rotor to be more efficient for a given amount of magnet.

There are factors when rectified loads are used that I can't account for with simple modelling I find that in real life the effective resistance comes about to be about 1.3 times the measured resistance. I have also found for some reason using low volt drop diodes doesn't reduce losses as much as it ought. Higher voltage versions such as 24 or 48v seem to come out more efficient even if you add more diodes in series to get a sensible comparison. Whatever you do the efficiency with a rectifier never reaches that of an alternator loaded with resistors. In that case you can get the very high theoretical values when the load resistance is low compared with the internal resistance.

I have also found a factor where the type of bearing seems to affect efficiency in the lightly loaded region above cut in whereas it doesn't seem to show so much in the friction loss below cut in. This could be significant in the size of machine you are testing and I have a suspicion that it mainly shows on this type of dynamometer loading, I can't say that I have noticed this bearing effect when wind driven but measurements under those conditions are less likely to be accurate.

I have to confess that I have struggled to get efficiency much over 75% with rectifier loads when the theory seems to predict something approaching 90% and that is with dual rotors.

Flux

[willib]

Beaufort,

No , magnet spacing has nothing to do with the air gap.

Magnet spacing has to do with opposite poles being over the legs of a single coil at the same time. But the poles should also alternate which coil is to be energized depending on which phase is being energized.

In a single phase machine this happens x number of times during a single rotation of a machine . X being the number of coils .

In a three phase machine opposite poles will encounter the two legs of the same coil/phase one third of the time during a rotation .

You have the cad program ,draw a 9 inch circle put 16 1 inch mags around the perimeter at 22.5 degrees apart , then draw 12 circular coils , 2 inches in dia. and rotate the poles around and you will see just how a three phase machine should work.

I've found that the holes in the coils work out nicely if they are equal to the distance between poles.

[Beaufort]

I understand where the magnets are supposed to line up over the legs of the coils, and they do line up in reality (the CAD model isn't correct as shown here).  However, there has been some discussion as to how far apart magnets can be:

http://www.thebackshed.com/windmill/FORUM1/forum_posts.asp?TID=1458&PN=1&TPN=2

I don't know if references to "the other site" are allowed here but this kind of knowledge needs to be shared.  The FEMM simulation there relates magnet spacing to air gap.  

[Beaufort]

Oh, and the latest Mr. Piggott book says that the most power can be gained from a given set of magnets by spacing them apart and using more copper; and less flux leakage between magnets.  That's what I was going for on this particular design.

[Beaufort]

OK, now we're getting somewhere.  I can pull my degrees out the trash now because I was convinced that there was little deviation from first principle formulas of driving voltage (and generated power).  I even went back and modeled the standard 4' dia, 8 magnet/6 coil machine in the books using a power curve you published a few years ago...and it didn't add up.  I'll have to go back and see the effect of bumping up the resistance by 1.3.

I've been following everything here for a number of years but this somehow hasn't come up in this way.  And one can be very mislead by just trusting those first principle calculations without any adjustment or testing (hence the repeated...follow the book designs or "else").  Well, I got into the weeds on the "else" part of it on this project...just trying to use up a bunch of 1" dia magnets.

[Flux]

I see no problem using up some 1" magnets.

Can I please comment on something earlier where you mention Hugh saying that by increasing the magnet spacing you can get more copper in and have less leakage flux and hence get more power out. This is very true and some of this applies to a single rotor but it was mainly aimed at dual rotor designs. The single rotor is magnetically very unsatisfactory and leakage flux is more or less inherent. You want flux to leave one magnet and pass directly through the stator to another magnet with as little as possible fringing between adjacent magnets on one disc. The very nature of a single rotor with no return path makes this an impossibility the path where you most need the flux to go has the longest length and invariably the lines take the shortest path.

The main thing with a single rotor is to keep the coils very close to the magnet face and to use thick magnets in relation to the coil thickness as long as you do this you can use any magnet but you must accept that the need to do this will always result in lots of leakage. Thick magnets close together result in leakage even on a dual rotor where the flux path is defined.

To be honest I wouldn't get bogged down too much about alternator efficiency, with a windmill the thing that mainly affects the output is the matching of the prop to the load, if you gain 10% electrical efficiency and in doing so you drag your prop off the top of its power curve you will loose out badly. Unless you use some load matching scheme you can't achieve a high electrical efficiency over any useful speed range. As long as you can match the prop then you can chase better electrical efficiency.

Flux

[martin1]

I have a question about 'increasing the magnet spacing so you can get more copper in'

Will you get more voltage if the width of the coil leg is greater than the width of the magnet or will the additional turns just add resistance?

[Flux]

All these machines work with distributed windings so the best way to think of things is in terms of flux linkage. Any turn that is bigger than the magnet will link all the flux at some point. Turns smaller than the magnet will not link the total flux but can still contribute some voltage. These turns are very short and have little resistance so you can make the hole smaller than the magnet and still gain but not to the point where you are reducing the loop area to silly levels. If you use rectangular magnets then making the holes trapezoidal with the large outer dimension the width of the magnet and the inner dimension smaller than the magnet seems to be a good compromise.

All turns bigger than the magnet contribute properly to the voltage but changing the magnet spacing does alter the voltage waveform and the peak to mean voltage ratio does change. Normally it has little effect.

Increasing the magnet spacing lets you have the same ( or similar) number of turns but lets you use thicker wire, this is where you gain from ( reduction in resistance). To benefit you must use the available space and wind your coils full until they touch.

Eventually if you went silly you would reach the point where the large increase in turn length would probably be worse than the gain in wire size as far as resistance is concerned and I have little doubt that there is an ideal spacing but certainly you can make modest increases in magnet spacing to advantage. With rectangular magnets, spacing of magnet width at the inner radius seems to be fine . You usually benefit by squeezing the coils at the centre by creeping under magnet size for the inner part of the hole. If you don't do this the coils tend to touch at the centre and leave wasted space round the outside.

This is all a compromise and as far as I know there has not been a detailed analysis done to find the exact optimum. Not that it really matters as we have to take available magnets which are probably not ideal shape anyway.

Flux