decogging motor conversion tutorial

[dinges]

After a discussion on Otherpower IRC I decided to write a small tutorial on the decogging of motorconversions.

It's NOT a manual on how to do motorconversions; Zubbly's manuals and stories deal with that. In this document I tried to delve a little deeper into the actual decogging of the generator; the several methods and how to determine the correct amount of skew or offset.

It's not intended as a cookbook but the goal was to give as much practical information, illustrated by as many drawings as possible.

The images in the file have been made available in my IRC gallery, in case anyone should want to use them as illustration on this board.

It is presented 'as is'.

The PDF file (1 MB) can be found here:

http://www.otherpower.com/images/scimages/3538/decogging_tutorial_V1.pdf

The individual images can be found here:

http://www.anotherpower.com/gallery/decogging

I hope it will be useful to those contemplating motorconversions but shying away because of the perceived difficulty of decogging.

Regards,

Peter.

[Warrior]

Peter,

Excellent Writeup!!!

Thanks,

Warrior

[Slingshot]

Thanks,

That's a useful tutorial.

[Ungrounded Lightning Rod]

Let me add a variation to the "offset method":

 - All the N poles are evenly spaced.

 - All the S poles are ALSO evenly spaced.

 - But the S poles are offset from the "all poles are evenly spaced" position by half a stator tooth, all in the same direction (rotationally speaking).

The group of N pole magnets and the group of S pole magnets are each balanced (if there's more than one of each).  So the whole assembly is also balanced.

[dinges]

You mean like this, ULR?

Hadn't thought of that one yet. Yes, the rotor would be balanced. That would be a big plus. Only worry I have (may be not correct though) is that the stator will neither see a N and S pole at the same time, but either an N or a S. Wouldn't this reduce the fieldstrength of the magnetic field that the stator sees?

Anyway, will add your suggestion to the pdf-file. Thanks.

[SparWeb]

Thank you, Peter, for the write-up.

As usual, when it's written clearly, the theory seems plainly obvious.

[Ungrounded Lightning Rod]

You mean like this, ULR?

Yep.

Yes, the rotor would be balanced. That would be a big plus. Only worry I have (may be not correct though) is that the stator will neither see a N and S pole at the same time, but either an N or a S. Wouldn't this reduce the fieldstrength of the magnetic field that the stator sees?

Think of it as seeing more gap.  Sometimes the extra gap is under Ns, sometimes under Ss, sometimes split between them in various ratios.

Anyway, will add your suggestion to the pdf-file.

It will look a lot better if the stator is more reasonable (especially for a four-pole):

 - Narrower slots, wider poles.

 - Magnets wider than the slots.

 - Slots wide at the bottom and narrow near the rotor/stator gap.  (T-shaped stator poles.)

 - Many more slots.  (At least 12 - probably some multiple of that - for a four-pole three-phase.)

[dinges]

It wasn't till after posting my previous reply that it (seemingly larger airgap) was mostly an artifact of a very simplified drawing. As you mentioned, with more and narrower stator slots it's hardly an issue.

The method you suggested has been added on page 19 & 20.

Do you know of any conversions that have been made using that method? If so, I'd like to know about them as it seems a method with a lot of advantages; ease of machining the magnet pockets (as opposed to CNC milling a helix...) and the inherent balance of the rotor.

[Ungrounded Lightning Rod]

Do you know of any conversions that have been made using that method?

Nope.

But I'm thinking of trying it.  It should be easy to do with a dremmel and a simple wooden jig (or even a hacksaw ala garbogens) rather than requiring a milling machine or lathe.

Main problem would be to avoid exposure to the aluminum dust from the rotor.

[Ungrounded Lightning Rod]

One downside:  The waveform would be slightly asymmetric, with one peak narrower and higher, the other broader and lower.

[Ungrounded Lightning Rod]

I take that back:  The peaks would be equal, slightly asymmetric, and mirror images of each other.

[Ungrounded Lightning Rod]

No that's wrong, too.

The windings are symmetrical so the two poles each produce the same waveform - the normal waveform for this pole geometry at half the voltage.

With one pole offset A degrees the output waveform will become the sum of one half-waveform with the other shifted by (A*(poles/2)) degrees.

This is still symmetrical, but it's hard to generalize about it in the time domain.  But it IS easy to genaralize about in the frequency domain, treating the waveform as a sum of a fundamental-frequency sine wave plus some harmonics.  (This is especially useful for motors because the output waveform is a strong fundamental plus a rapidly decaying series of odd harmonics.  The even harmonics canceled out due to the mirror symmetry of the waveform.)

The sum of two equal sine waves of voltage V offset by phase angle A is 2*v*cos((A*(poles/2))/2).  So a small offset will slightly reduce the voltage of the fundamental.  (For instance, a 5 degree offset in a four-pole motor will attenuate the fundamental by cos(5 degrees) = .9962, about half a percent.

The harmonics, however, get reduced by more, because the phase angle is larger for their higher frequency.  It is multiplied by their harmonic number:  The third harmonic voltage is 2*v*cos(3*A*(poles/2)/2), the fifth by 2*v*sin(5*A*(poles/2)/2), etc.  (For four poles and a 5 degree offset the third harmonic is attenuated by about 3.5%, the fifth by 9.5%, the seventh by 18%, etc.)  Eventually the sin function comes around back to 1 - and the associated harmonic may actually be attenuated less than the fundamental.  But by that point you're FAR out into the harmonic tail.  (For a 5 degree offset that's about the 26th harmonic, - which doesn't exists - but a slight angle error might give the 25th or 27th a half-percent boost over the fundamental.)

Net result is that the waveform will be symmetric, attenuated very slightly, and somewhat closer to a pure sine wave than the un-decogged rotor would have produced.

[jimovonz]

How effective would this be at de-cogging? I can see that it would reduce cogging but would it not just produce two lesser 'bumps' per pole rather than one big 'bump'? Obviously with more poles, different offsets could be used to divide it further but compared to skewing where there is a continuous gradient as it were, isn't this method still 'lumpy'?