Thanks Tinsel, I need to stress though for those considering building that the principal of the motor is the slight difference in phase of the currents as well as their variations, and the position the charging coil is placed is very dependent on the - pulse width - coil inductances - charging capacitance - rotational speed Vs magnet spacing, and if we want high rotational speed or low rpm torque.

The design considerations are

1. How much power we want the motor to use. 
2. How fast we want it to spin.
3. How much torque it has and where.

For more Torque at lower rpm we can design the motor to do that by using more inductance more capacitance and closer to 50% duty which will limit the speed of the motor unless it is computer or manually controlled to maximize all possible parameters for different situations.

For more speed we need higher coil switching frequencies so it's narrower pulse widths - less inductance - less capacitance - higher voltages and timing tweeks.

Certain things are limiting by nature.

Anyway I am close to getting two 1 Ohm resistors on load at 1800 rpm, I'll post a quick clip of how I do tonight after a rearrangement and tune up. Twice I re installed the magnets the wrong way around for the sensor strip placements. arrrggg. Then wondered why it wouldn't work. 

Too small of a pulse width compared to the inductance starves the motor of it's ability to reuse the inductive energy released as well as keep taking from the supply while under load, I think. The pulse width needs to be not so long as to waste energy but not too short for the inductance to prevent proper current flow. That's when tuning on the Dyno gets to be fun  .

Cheers

Thank you for pointing out that diode SeaMonkey, it is kind of important, with the changes I made I got a bit confused I'll confess.
Anyway I think I only need one where (D10) is now in the revised drawing, for some reason I can't modify the other post to replace the drawing.

Also I think the change I made in preparation for the second motor coil tricked me, I wound a new coil from four wires of 0.5 mm twisted wire but because I put it on the same former it slipped my mind, so it has more or less inductance which is affecting things, I'll use two strands for each charging circuit.

Now after looking at the currents and switching between 220 and 440 uF I can see almost exactly what to do. If I want to try for maximum power I need to design around a certain duty cycle % of "on" time closer to 50% at the target rpm range and with a bigger capacitor, if I want to use more power.

The coils being close seems not to be a problem but I'll do a few specific tests and measure the inductance of the new charging coil then re-position the charging coil again this time with the pulse width set to give me the correct duty at the right rpm band. It's a lot better but still not right. This time I'll position it back on the other side again between MC1-2 and point "A", so that the timing of the maximum of the current in the charging coils is neutral with respect to the magnet pass with the correct pulse width at the right rpm to give 45% duty in the target rpm range with no boost.

I think I've found the optimum pulse width for the motor coil so the rest I figure around that. It's a little bit of trial and error but it's difficult to know if an idea will work some times without trying it, I need to remember that the motor coils are meant to do the work and the charging coil is just a bonus so it's core tip should be a bit further away from the rotor magnets I think.

New drawing below, I just put in a 6A01 diode which is a 6 amp 100 volt part.

I keep getting new idea's which can side track me at times.

No probs Conrad, Your questions are very valid.

1) Yes you could use a higher voltage supply or a variable DC power supply, that's just what I like to use because I have solar charged batteries and I want to be able to apply more or less power without needing to change the pulse width.

2) Yes that's right, the way things look is that the prolonged currents in the coils attracts the next opposite magnet and with the charging coil slightly out of phase there is a slight neutralizing of the cogging drag because of that.

So I think the real torque happens when events go like this--  the motor coils repulse a north magnet in the "on" time then the prolonged currents through the motor coils during the "off" time attracts the next opposite polarity magnet, the same happens with the charging coil just a tad later and when the motor coils push the rotor so fast that the charging coil phase is too late to push a south it must then attract the next north if anything. The extra attraction of the motor coils to the next magnet by the prolonged currents would neutralize the drag and the drag of the south magnet leaving the motor coils can only force more current through the charging coil. I must end the theory of operation with a big I think because I can only say what I think. The currents also become sinusoidal looking at some points I'll include a shot here shortly if you check back.

I try to look at it like pipes of fluid and the coils are elastic bags, the capacitors are reservoirs the diodes are check valves the voltage is pressure (potential is level in reservoir) and the switches are sluice gates, they need to chop the water off.

The pressure from the momentum of the water when the gate slams shut forces the water to flow up through the diode and into a reservoir (flyback diode to return capacitor)
The pressure from the momentum of the water also fills the charge cap to a higher voltage. So when the motor coil is first switched, current flows from the charging cap before it flows from the supply. ( Current can't flow from the supply until the level in the charge cap tries to fall below the supply level).
The pressure from the return capacitor forces current through the charging coil and charging cap but not the motor coil.

Basically it looks like the magnets do interfere at times. But it should be all cancelling, the south pole leaving the motor coil would force current through the charging coil to push it away and that also adds to the charging cap voltage level to increase the bang of the next switch on. If I just cut the power to the coils from the supply it seems to want to keep running, it runs on for a while with no generator attached. (By run on I mean it takes a long time to stop, too long kind of thing than if it were all drag).

I only use two transistors because they are there to use for two coils. With separate coils I like to switch them separately, multiple strands on the one core I usually use just one transistor if it can take the current.

Cheers

P.S. Conrad you bring up a good point and I think I should draw radially segmented graph to show the on and off times for the transistor and current flow times for the two coils.

With the shots the yellow trace is the motor coil it's upside down and it is first, then the blue current trace is after it that's the charging coil current.
The two left shots show the difference between 220 uF and 440 uF for the charging cap, the bottom left is the abnormal one where the application of the extra 220 uF caused the current in the blue coil to be delayed longer and put the charging current way too far out of phase and made it look as if it comes first but it doesn't.

At certain times the current can not fall to zero in both coils it just goes up and down in value. The coils get a work out but they don't wear out. 



..

Yep it should work the same both ways and alternating pulses should be both more efficient and more powerful, whichever depending on the objective. Prolonging the currents after the switch off helps rather than hinders if we make our circuits to do it. For sure. We can do anything we want if it works my friend.  Even moreso if it's fun.

It's very similar to driving a Tesla coil, we use sharp edged pulses to keep a nice sine wave standing up and use the power of the sinusoidal currents driven by resonant rise. The impedance facing the recirculating currents is overcome by the way the magnetic field collapses and tries to short the coil by building potential.

In the video this time I show the charge capacitor on the yellow trace and the mosfet drain on the blue trace.

http://www.youtube.com/watch?v=vLiNk6yBWyY&feature=youtu.be

It's good to get confirmation of the efficiency you seen with one way pulses, but two way pulses can be just as efficient I think, the thing is there needs to be enough off time for the prolonged currents or wasting happens where more pulse width means less speed, prolonging the currents means the on time needs to be at least only 50% of the entire cycle so for one way DC pulses that is one pulse of 50% or less but for two way pulses that is two pulses of 25% or less. So the pulse width could get too short for the inductance maybe, something to consider.

At first I wanted to stop the currents quicker but then I figured why not let them do what they want and try to use it.

Cheers

@Conrad.....what is it with you and your P-channel mosfets? Do you just happen to have a box of them that you are trying to use up?
 

Flip that mosfet symbol over, put the Source to the negative rail and the Drain to the low side of the load, and use something like IRFP260 or IRFP460. N-channel mosfets are cheaper and perform better than the equivalent P-channel mosfets.

@TinselKoala: I hope I got it right with the N-Channel MOSFET. Note the difference in the connection of the Gate to the opto coupler and the reversed role of R1 and R2.

The intention is to bring the MOSFET (which drives the coil) in the on state whenever the LED in the opto coupler is shining.

I am a bit slow at electronics.

Greetings, Conrad

@Conrad: Now you can put your scope probe "ground" references at the negative rail where they belong, and use a second probe "tip" at the location on the bottom of the load where you now have the first probe's reference attached (mosfet Drain). The voltage difference between the probes will give the drop across the load and can be used to compute the current in the load.
Your gate voltage dropping network is workable I think, but the resistances are kind of low for my liking. You generally don't need such a low value to pull a mosfet gate down when it's supposed to turn off, so you might be wasting some power here. Also, it's sometimes nice to protect the gate with a pair of Zeners from gate to source, back-to-back, at the desired max gate voltage, like 12 or 15 volts. This will limit the max gate voltage and hopefully shunt overvoltage spikes away from the gate and sometimes can save a mosfet that's in severe service. With the optocoupler acting as a sort of "fuse" this shouldn't really be necessary, but who knows until the circuit is built and tested and has had a few failures.

This generator coil works to speed up the rotor under short circuit.

http://www.youtube.com/watch?v=YpKZw15A41Y

Cheers

Hi Gyula, Yes it has the part from the motor core inside a square laminated one with curved ends. I can check that no problem but I'll need to be a bit  careful so I get a proper result, I'll have to leave it running while I remove the mount with the coil on it, might take a while and I'm about to go do stuff so I might not get it done for a while I think the rpm will be slightly higher without the core there but the input power probably will too also be higher because of the design of the motor, the pulse width will not change but the input power is related to the rpm, so without the core if the motor does spin faster it might consume more energy,  however it didn't seem to in the video, I didn't look that closely though, I do it and post all is shared.  I had that in mind.  See the input current in the video, that is the current out of the battery at 12.5 volts or so.

Umm there is 15.6 uF across the coil though just to be honest.   

I'll do the test it sounds like a good idea.

Cheers