Variphase Alternator

[TomG]

FinnSawyer (GeoM) proposed an alternative alternator design:

http://www.fieldlines.com/story/2006/4/21/16237/9933

On balance, sadly, I don't think it's got much of a chance of competing against the standard axial-flux machines which are the current state-of-the-art around here. BUT, it was extremely clever and inventive. It took me a long time to figure out how it worked, and doing so gave me an idea.

Why do all the phases in an alt have to be wound the same? Obviously the size of the hole in the middle is limited to the size of the magnets we're using, and we can't make them too fat, or they won't fit. Adding more turns requires us to use smaller gauge wire. But why do they all have to have the same balance of turns vs. wire size?

I'm pretty sure that all the coils in a given phase need to be the same, otherwise one coil will cut in before the others, and then the resistance of the unproductive coils will just be dead weight.

But different phases could have different turns vs. wire size, so long as we rectify them individually, and attach them to the battery bank in parallel.

This is the curve from an alt with 4 phases. Every coil in it is the same size and has the same cross-sectional area.

Mmmm, that's some good curve matchin'!

One problem I can see is that the phases with early cutin are going to cook at high RPMs, because the current through them will keep increasing, and they're made of lots of turns of very thin wire. So something will be needed to limit the current as the RPM increases. Maybe an inductor would work? Nice and easy, and would limit the heat being dumped into the stator to a sensible value. And at high RPMs we don't need to care about dropping the efficiency of the early cutin phases, because we've got the rest of the phases to compensate.

Also, at low RPMs, when we've only got one phase working, we might expect the alt to run less smoothly than a 3-phase alt. It should smooth out at higher RPMs, as more phases reach cutin.

[Flux]

This idea has been around for years.

"I'm pretty sure that all the coils in a given phase need to be the same, otherwise one coil will cut in before the others, and then the resistance of the unproductive coils will just be dead weight."

That is not true, but there would be little reason for using different numbers of turns in the coils of a phase.

You have spotted the basic limitation of this approach, the high voltage, high resistance coils run inefficiently in high winds and either fry or rob power from the lower voltage coils. You could devise methods of switching out the high resistance windings in high wind. The switching is messy but possible. SWWP had some success with this on the Air 303 without switching but you would need to switch on bigger machines.

This falls into a similar category to having two alternators, one suited to low wind and one to high wind. The scheme is perfectly practical but you need more material overall for the same output.

In low winds you only have a fraction of the winding active but at low current this is not a big issue, unless you use many steps you need a certain measure of inefficiency to match the very flat prop characteristic.

Similarly in high wind you only have part of the winding carrying the full load and you really could use all the material you have at this point to get a good efficiency when the prop curve is steep and you can match it reasonably.

Keep on with the idea, you can certainly do far better than the usual simple compromise. If you can simplify the switching to something acceptable you are on to a winner.

Perhaps those totally afraid of an electronic solution would accept electronics in switching circuits when they will not accept it in a converter. My attempts to solve this one required more complexity in the switching than the associated converter to do it another way so I went the converter route.

If you are prepared to accept something a bit better than the conventional but with little of the true benefit of a proper cube law match then the switching may be quite modest, but even star/delta seems to be considered rocket science by most, although it gives a fair improvement and star Jerry would be simpler and even better.

Keep us posted on your progress.

Flux

[Flux]

I forgot to mention that the inductance trick does limit the current in the high voltage coils, it involves more cost, more copper loss and more weight but it is not impossible. It does little to improve the efficiency in higher winds, but it is better than nothing. I didn't pursue it but it may be a good starting point.

Flux

[wooferhound]

You could always run the output of each individual phase through it's own bridge rectifier. But in the end I believe you would be looking at some burnt up stators.

http://www.fieldlines.com/story/2008/5/28/225946/561

[TomG]

As I said: "...so long as we rectify them individually, and attach them to the battery bank in parallel."

Keeping them in parallel means the phases with early cutin never have to carry the current from the later phases. We just need to stop them frying from their self-generated current.

[TomG]

Actually, it does seem to do quite a good job.

Here are the results from a 3-phase alt whose 3 phases cut in at roughly 68, 100 and 200 RPM.

Without inductors it gets up to about 200RPM before the efficiency drops to 50%.

The power curve without inductors:

And the same alt, but with inductors added to the first 2 phases. Now it can get to about 450RPM and harvest twice the useful power before the efficiency drops to 50%:

I don't know about the practical cost and size of inductors. I know there's an equivalent circuit called a gyrator which uses a capacitor and an op-amp instead, but I don't know about its suitability for this application.

Surely there are other simple current-limiting circuits?

[finnsawyer]

"On balance, sadly, I don't think it's got much of a chance of competing against the standard axial-flux machines which are the current state-of-the-art around here. BUT, it was extremely clever and inventive. It took me a long time to figure out how it worked, and doing so gave me an idea."

Did you expect me to let this go unchallenged?  I think you may be guilty of fuzzy thinking here, namely assuming that one could build my 12 magnet 18 coil design on the same rotor as the 12 magnet 9 coil three phase.  Can't be done using the same magnets and same size coils.  In the case of the three phase with let's say round magnets of 1 inch diameter and coils 2 inches in diameter a rotor diameter of six inches should work fine.  Obviously you can't fit 18 of those coils and 12 magnets on that size rotor.  With my design the magnets are spaced every 30 degrees as before and the coils are spaced every 20 degrees.  The coils and magnets are brought in toward the center until a magnet is just poised to start passing over the edge of a coil.  At that point the coil center and magnet center are 10 degrees apart.  The distance between centers is then 1.5 inches, the sum of the half radius of the magnet, the half radius of the coil, and the width of the copper windings.  So we can write Sin(5) = .75/Radius.  Radius = 8.6 inches.  This is nearly 3 times the radius of the three phase.  The upshot is that each voltage pulse occurs in only 10 degrees of rotation rather than 30 degrees for the three phase, and as you have said yourself in the past, the speed of travel of a magnet past a coil does matter.  It takes 30 degrees of rotation in the case of the three phase for a magnet to pass from being centered over a coil to no part of the magnet being over any part of the coil, but only 10 degrees for that action to take place with my design.  With the three phase a coil interacts with two magnets at once so it gets a double hit (twice the voltage), but there are only 6 coils active at any given time, whereas with my design there are 12 coils getting a triple hit (three times the voltage due to the triple speed).  Let me see:

    3 coils per phase times 2V for the double hit times 1.73 for the phase interaction gives 10.38V for the three phase, and 12 times 3 times V gives 36V for my design.  If you don't get it now you never will.

There are some who would take issue with the requirement that the inner coil diameter be no less than the magnet diameter.  Personally, I'm neutral on the issue.  I'd also like to point out that with my design there is no particular requirement for the maximum coil diameter as as long as the 10 degree separation is maintained (for the example given).

[TomG]

Did you expect me to let this go unchallenged?

I really didn't But I'm not trolling; it's my genuine opinion at this point. If I've misunderstood it, your further clarification will be greatly appreciated.

The alt you specify would look like this then?

Fig.1

(With a 0.2" error due to poor draftsmanship)

And the rectifier arrangement looks like this?

Fig.2

In the standard alt, 3 coils per phase, in series, gives R_phase = 3*R_coil

and star wiring gives R_line = 6*R_coil

In your alt, 6 coils per phase, in series, gives R_phase = 6*R_coil

and series wiring gives R_line = 18*R_coil

The standard alt you described would have a power curve like this:

Fig.3

Your alt (with the same size coils and magnets, much wider rotor, etc) would have a power curve like this:

Fig.4

And yes, that does include the effects of your multiple cutins, or "breakpoints" as you called them. You can just about see them on the left hand side of the curve if we zoom in:

Fig.5

The first cutin occurs at 40RPM, and second at 60RPM, and the third at 120RPM.

It looks like the benefits aren't enough to overcome the cost of the increased resistance due to having more coils, and putting them in series...

OK, where did I go wrong?

[finnsawyer]

I wonder why you didn't just continue this stuff in the original thread.

O.K., you need to increase the diameter so no magnets and coils overlap except the ones that are centered, but it's a good representation.

The rectification is all wrong.  The coils are all wound the same way.  Simply connect them in series and connect that to one full wave bridge rectifier.  Well, I shouldn't say all wrong.  I think I know where you are coming from.  It's that split version I mentioned, but I think you can do that with two diodes just like a center tapped transformer output.  But I don't want to go there now.  Please stick with the simple case.

So, I don't know what to make of your curves at the moment.  You are right about the resistance:  6R at 10.38V versus 18R at 36V.  If you were to match the loads.  That is 6R or 18R, you would find you get a total power from the three phase of 8.98/R and 36/R from the single phase (V^2/R).  My design would produce 4 times the power, as well as 4 times the power into the load.  O.K. that's an ideal case, and somewhat artificial.  When charging a battery things would be different.  The alternator has to overcome the battery voltage, but my design would reach that voltage (cut-in) at a lower rpm, and start charging earlier.  With the higher resistance the current would be lower.  What to make of that.  Well, the value of resistance is not sacred.  We could wind the coils of my design (same amount of copper) two in hand instead of with a single strand.  That would reduce the resistance to 4.5R and the voltage to 18V.  Note that the alternator would still produce the same power into a matched load.  Neat, hey?  When charging a battery it would still reach cut-in at a lower rpm, but now with the lower resistance more current will flow, so my design can provide more power with a lower cut-in rpm.  I would have thought that with the lower cut-in rpm the VAWT people would be interested, but such does not appear to be the case.    

 

[finnsawyer]

Redo the second figure so that all the coils are in a line "vertically".  Bring the "center" tap out from the middle of the blue stack.  That becomes the circuit ground and goes to the negative terminal of the battery (usually).  Connect the two ends of the stacked coils each to the anode of a rectifier diode.  Connect the cathode end of each diode to the positive terminal of the battery.  That is, the arrow of each diode should point toward the battery (the direction of positive current flow).  That is the circuit you should analyze for the break point.  Of course, you need to get the phasing of the coils correct in each colored stack and between the colors.  Simply put, the phasing should be same for the overall stack as for the simpler case of just connecting the coils together sequentially, so that as a simple alternator without the center tap it would have the same behavior.

Note that with the center tap the resistance between lead and ground has been cut in half.  But we also have less voltage, less than half, in fact, because the contribution of the blue stack has been cut in half.  But that's what causes the break point, since the "cut-in" rpm for the reduced blue stack must be higher than for the other two groups of coils.  

[TomG]

I wonder why you didn't just continue this stuff in the original thread.

Yeah, I've kinda hijacked my own diary, haven't I? Do you want me to repost this stuff onto the end of your diary entry?

The rectification is all wrong.  The coils are all wound the same way.  Simply connect them in series and connect that to one full wave bridge rectifier.  Well, I shouldn't say all wrong.  I think I know where you are coming from.  It's that split version I mentioned, but I think you can do that with two diodes just like a center tapped transformer output.  But I don't want to go there now.  Please stick with the simple case.

Gladly! I haven't bothered redrawing that - hopefully even I can't mess up "all in series, one rectifier".

So, I don't know what to make of your curves at the moment.

Ignore them - they're completely wrong, because the rectifier was all wrong. I've redone it, and the power curve now looks like this:

Fig.6

which is a much lower cut-in. Let's look at the more sensible, low-end range...

Fig.7

That's a mighty low cut-in! If we wind the standard alt with more turns of finer wire, to get the same low cut-in as this, the power curve looks like this:

Fig.8

Wow! Now that they both cut-in at about 28RPM, yours is far better! It produces over twice the power, and doesn't drop to 50% efficiency until 55RPM, compared to about 48RPM for the standard design.

I would have thought that with the lower cut-in rpm the VAWT people would be interested, but such does not appear to be the case.

From this analysis, they should be. For the price of the extra copper, your design seems to handle the low RPMs much better than the standard design.

So, I hereby retract my earlier statement: for the low RPM range, your design outperforms the standard axial flux alt.

(Don't you just love science? If this was any other subject, we could argue forever and settle nothing.)

[ghurd]

Not that I completely follow all of Finn's alt design,

but it is the 1st time I mostly(?) understood what Finn GeoM was talking about with it.

G-

[Stonebrain]

'Don't you just love science?'

Yes I do.

In science you can't just produce nice curves without explication.

I cannot guess the way you made these curves.That's why these curves don't mean a thing to me,and so,for me they don't prove anything about geoms alternator.

Allready,Finnsawyers design is causing a headage to me,but these mysterious curves make it worse.

cheers,

stonebrain

[TomG]

Like this?

Fig.9

'Cos I can't see how that would work. The waveforms of the two sides look like this:

Fig.10

Both go negative at some point - how do they charge the battery on that half of the cycle? I can get it to work with 6 diodes, but not with 2.

[TomG]

Would you like a copy of my spreadsheets and simulations? I'm not planning to copyright them or anything  You'll need OpenOffice, a recent JavaSDK, and JFreeChart to get them working. You'll also have to give me a while to annotate them, or it'll be too much hassle for you to figure them out.

[vawtman]

TomG, You seem to have gathered enough info to build a possible working device.Finn has also but until you guys build a functioning device you will never know if it works.Nobody else will build it for you i wouldn't think.

 Just theories floating in space.

 Mark

[TomG]

Info, yes. And given the resources, I'd have built a prototype by now - I want to know what works! But given what I currently have available, I build simulations and test out theories. A lot (but not all, I realise) of my questions can be addressed without picking up a multimeter.

I'm not arguing for the primacy of theory over practice here! Just making do with what I've got at the moment.

[vawtman]

I'm lucky to have great resources.A new freind from this sight Don from Ideal Fabrications and my uncle Ace.Who litterally shipped his shop up here to (Wi)from(CA)San Jose area.Ace, machined parts for planes back in the day.

 Between them they can make steel dance.

 I'm havin fun and should have updates shortly.

 If you need help let me know.My email is posted

 Mark

[TomG]

I burn with jealousy. :-/ Just had a look through your diaries - it certainly looks like you're having fun! Some nice stuff there...

In inner-city UK, envying you the wide-open spaces...

[finnsawyer]

Good work!  You decided to rewind the three phase, whereas I have preferred in my discussions to rewind my design by going to the two in hand of the same size wire.  I suppose one could continue that to a shorter length of wire to give my design the same cut-in as the three phase.  You might try that approach as well.  Or take the simple case of three in hand.  That would lower the voltage to 12V and give only a slightly lower cut-in than the three phase, and 2R for the resistance.  Another thing.  I have claimed that my design should provide four times the power into its matched load than the three phase into its.  If you accept that statement that would be a good check on your rewind as the power provided by an alternator into its matched load does not change, if the total amount of copper remains the same as the copper wire size and length (and resistance) are changed.  If you find that is not the case, then we have some work to do.  I think the VAWT people should accept your offer of your simulation (spreadsheet).  Actually, anybody that has built a PM axial alternator should take your spreadsheet and see how its predictions square with their results.  Are our Hosts listening?

To Ghurd and those with the headaches:  I'm sorry if my explanations have been inadequate, but Faraday's law is not intuitive, and there are fairly complex interactions involved.  Have a nice headache free day!

[finnsawyer]

When a lead voltage goes below battery voltage plus .7 volts, the diode in that lead ceases to conduct current.  Simply replace it by an open circuit for that part of the cycle.  There will be periods of time where no current flows into the battery.  That, unfortunately, is a characteristic of single phase rectification.  That's why I advised you to study single phase rectification.

This is probably not the most efficient use of the magnets unless the break point is what you really want.  But I'm not sure how useful that would be in practice, since it seems to happen late (at too high an rpm).  But it sure is interesting.

Love that graph!

[TomG]

Having looked into it briefly, it seems like it should be fairly easy to make suitable inductors. An old transformer core for the iron, and a very reasonable amount of copper (probably less than 1 coil's worth, depending on the resistance you can put up with), should get you into the milliHenrys range. That's enough to substantially choke the current when the frequency starts to rise.

I really wish I could wind some test coils for this (inductor idea). Trying to do the maths is horrible. Has anyone got either the necessary bits, or a decent power-electrics simulator that they could use to make some graphs of this? How much can we limit a coil, with how little copper?

[Stonebrain]

Well,I think I don't have the patience or time to study your spread sheets and simulations.

But the least I would exoect is information about configuration the curve is referring too.

"Ignore them - they're completely wrong, because the rectifier was all wrong. I've redone it, and the power curve now looks like this:"

Ok,I do my best.You mean you're referring to the 'right'(as opposed to 'wrong') configuration.I've been thinking about this,and finally I guess you mean you're talking about a 3p alternator with the 3 phases in serie.

If this indeed gives much better results I must give my congratulations.After star and delta,you find out the 3phase in series wiring,and this would be -much- better performant at low rpm.Great news!

I'd like to hear what flux thinks about this,though.

Of course I may have understood things wrong.And thjat's what I mean.If you give a curve you must be clear what it's about.

cheers,

stonebrain

[TomG]

OK, I haven't got a full reply to this yet, but I thought I'd post the current version of my standard-alt spreadsheet, to see what improvements people suggest:

http://www.otherpower.com/images/scimages/4272/PublicAlt.xls

I've only really used it for 3-phase at the moment, so changing it from the default 9/12 settings might result in odd behaviour. No guarantees!

One thing I know I'd like to add is a "switch" for Star/Delta/Jerry wiring, but I haven't done it yet.

[TomG]

Fig.6 was the 3-phase, series-wired, 18 coil, 12 magnet/rotor, alt, of Finn's design, with (IIRC) 100 turns/coil of 3mm^2 CSA (about AWG#13, or 3-in-hand AWG#17), with magnets 5cm x 2cm x 1cm, with airgap flux of 0.6T, coil temperature of 30°C, and reactive impedance factor of 1.3.

[TomG]

Yuck Done like that, you're losing half your power. Certainly not the most efficient use of magnets, and I don't think the breakpoint behaviour will make up for it. Series wiring with full rectification seems a better bet.

If we weren't worried about the breakpoints, wouldn't that free up the layout restraints a bit?

Re: Graph.  JFreeChart is made of win. So much easier than writing my own!

[finnsawyer]

Not necessarily half the power.  We have cut the effective resistance in half.  The fact that they make center tapped power transformers just to use this circuit tells me it's not that bad.  Why don't you run the numbers on it, and see what you get?  You basically have three states, top diode conducting, bottom diode conducting, and no diode conducting.  That might not lend itself to a spreadsheet, but you certainly can write a program to do it.  By the way, any single phase rectified power supply will have times of no conduction for the diodes.  That would be true at all rpms for my design and is also true for the three phase for some range of rpms after cut-in.  So, it's probably something worthwhile for you to investigate.

[finnsawyer]

I took a look at your last comment to my Diary, and thought I'd take little different tack here.

We've stated the three phase would put out 10.38V at a resistance of 6R, and my design would put out 36V at a resistance of 18R.  Let's put some real numbers to this:

     Three phase voltage = 20.76 volts, resistance = 1 ohm.

     Finnsawyer design/aka Mattson design voltage = 72 volts, resistance = 3 ohms.

We are putting power into a 12 volt battery with two .6 volt diode voltage drops in series.

     So, the current into the battery from the three phase is {20.76 - 13.2}/1, or 7.56 amps.  The power into the battery is then 90.72 watts.

     Similarly, we find the power into the battery with my design would be:

     12x(72 - 13.2)/3 = 235.2 watts.  Pretty consistent with your results.  Now let's assume we rewound the stator (two in hand winding) so that my design puts out 36 volts at .75 ohms resistance.  The power into the battery now becomes:

     12x(36-13.2)/0.75 = 364.8 watts. Hmm, let's now go to a three in hand winding.  The power into the battery becomes:

     12x(24-13.2)/0.33 = 392.7 watts.  What are we to make of these crazy results?  Well, our analysis is only approximate.  We should really make the voltages sine waves.  So, let's say the three phase voltage goes as 20.76xSINE(wt), while the voltage of my design goes as 72xSINE(3wt).  There is a time period when the voltage from my design is less than 13.2 volts.  Obviously at wt = 0 we get no voltage.  That corresponds to an angle of zero in the cycle of the waveform.  Conduction starts when

72xSINE(beta) = 13.2, from which we can find beta:

     beta = ARCSINE(13.2/72) = 10.56 degrees.  That is, the alternator conducts current from an angle of 10.56 degrees to an angle of 180 - 10.56 = 169.44 degrees during the first half cycle, and has a similar behavior for each half cycle thereafter.  For the other two cases we find beta is equal to 21.5 degrees and 33.4 degrees respectively.  So, in reality we should be determining the instantaneous current during the time the diodes conduct and multiplying that by the battery voltage to get the instantaneous power, and from that get the average power.

The three phase alternator is not totally immune from this effect, either.  Ideally each phase should conduct for one third of a cycle, 120 degrees.  In our example the three phase has a maximum or peak voltage of 20.76 volts at wt = 90 degrees.  So, that phase should provide voltage in excess of 13.2 volts from an angle of 30 degrees to 150 degrees.  It doesn't.  The phase voltage doesn't exceed 13.2 volts until wt (beta) is equal to 39.5 degrees.

The upshot is that all the power ratios should be suspect.  A more precise analysis is needed.  Nevertheless, I feel that my design should still perform better than the three phase having the same magnets and same size coils.    

     

[finnsawyer]

I want to stop something right here.  My design is NOT a three phase with the three phases wired in series.  Were you to do that with a balanced three phase output you would get zero voltage all the time.  Try it. Add the phase voltages up.  The design has the coils arranged in three groups that act in phase, but at any time one of the groups is loafing (not putting out any voltage).  Hope this helps.

[capthook]

TomG -

Thanks for posting the alt. spreadsheet!  I love plugging variables into an .xls to see what would happen.

Any updates/corrections/comments on the .xls?

Thanks again for creating the spreadsheet and posting it.....

It should be a must-download for everyone!

[capthook]

1 thing - there is no variable for magnet spacing?

What assumption did you make in this variable and where/how should one input a new value?

ie. .75" x .75" space 1 magnet apart or .75" spacing.

tx