Simulation of the Muller Dynamo

[poynt99]

Here is my attempt at simulating Romero's output wave form.

Currently using simple linear coupling, so the "Bias" magnets are not useable. The output across both generator coils connected as shown (series-canceling) is zero volts, i.e. perfect cancellation.

I've also tried a basic test with the core marked in the top left; with similar results so far.

The simulation is based on a motor (VROTOR) driving the rotor in a sine wave mode. L1 and L4 are the top and bottom poles respectively of the rotor magnet, and the top and bottom generator coils are self-evident. The output is +/-5Vp if either L2 or L5 connection is inverted, i.e. in series-adding.

L3 and L6 may be used later (with a magnetic core) to emulate the bias magnets placed on the generator cores.

This is a starting point only. Any suggestions how this circuit may be modified to achieve that output wave form?

Regards,
.99

[EMdevices]

very nice, you should be able to gain some basic insights, but for a more serious simulation of this dynamo I think a different software is more appropriate. 

I see you're trying to implement the biasing action.    Are the inductors you're using the right inductor model that uses saturation?

Maybe you can implement voltage pulses having Gaussian functional shapes, for the rotor magnets.

EM

[poynt99]

very nice, you should be able to gain some basic insights, but for a more serious simulation of this dynamo I think a different software is more appropriate.  

I see you're trying to implement the biasing action.    Are the inductors you're using the right inductor model that uses saturation?

Maybe you can implement voltage pulses having Gaussian functional shapes, for the rotor magnets.

EM

I guess I should have been more specific; it's an attempt at simulating the electrical (and to a degree, the magnetic) characteristics of the Romero dynamo using SPICE.

As I mentioned, the bias magnets can be used when a magnetic core model is used. I am currently using a linear (air) core, and thus the bias is unusable at the moment. It is there if/when I switch back to a magnetic core model again.

.99

[poynt99]

PSpice files.

[teslaalset]

PSpice files.

Thanks Poynt,

It's running here on Cadence Pspice 16.3 meanwhile.
Can you give me a few short instructions how you composed those transformers (what library and what components did you use) and how I can change the core materials (I have the 'magnetic.lib' already included in the project) ??
I am building up some experience but this is where I am stuck.

[poynt99]

teslaalset,

The setup of coils in this simulation is such that there are two groups of 3 coils; Top Group=L1, L2, L3 and Bottom Group=L4, L5, L6. Each group is coupled together via their own "K" element. In the case where you want to try air-core coupling, use the K_linear component found in the "ANALOG" library. If you want to try some of the Ferroxcube cores, delete the K_linear component, and add any one of the Ferroxcube cores located in the "magnetic" library.

Once you place a number of separate inductors and a "K" coupling on your diagram, you open up (double-click) the "K" device and add the inductor reference designators (i.e. L1, L2, etc.) to the list.

When using the K_linear coupling device, the value for each inductor is chosen as a value in Henries. The coupling value can be selected anything from say 0.1 to 1.0. Choosing a value of "1" is unachievable in practice, but it emulates a perfect linear core with 100% coupling, and no saturation. For standard realistic air-core coupling, I usually choose a K coupling factor between 0.5 and 0.7.

When using any one of the non-linear cores in the "magnetic" library, you must replace the inductance value of each separate inductor with a "no. of turns", for eg. "300". I usually leave the K coupling factor at a value of 1 for these "real" cores.

The core lines were simply drawn in using the line tool.

Hope that provided the answers you were interested in.

Regards,
.99

[teslaalset]

teslaalset,

The setup of coils in this simulation is such that there are two groups of 3 coils; Top Group=L1, L2, L3 and Bottom Group=L4, L5, L6. Each group is coupled together via their own "K" element. In the case where you want to try air-core coupling, use the K_linear component found in the "ANALOG" library. If you want to try some of the Ferroxcube cores, delete the K_linear component, and add any one of the Ferroxcube cores located in the "magnetic" library.

Once you place a number of separate inductors and a "K" coupling on your diagram, you open up (double-click) the "K" device and add the inductor reference designators (i.e. L1, L2, etc.) to the list.

When using the K_linear coupling device, the value for each inductor is chosen as a value in Henries. The coupling value can be selected anything from say 0.1 to 1.0. Choosing a value of "1" is unachievable in practice, but it emulates a perfect linear core with 100% coupling, and no saturation. For standard realistic air-core coupling, I usually choose a K coupling factor between 0.5 and 0.7.

When using any one of the non-linear cores in the "magnetic" library, you must replace the inductance value of each separate inductor with a "no. of turns", for eg. "300". I usually leave the K coupling factor at a value of 1 for these "real" cores.

The core lines were simply drawn in using the line tool.

Hope that provided the answers you were interested in.

Regards,
.99

Poynt,

Thanks a lot for your instructions. They are pretty clear.
I was indeed a bit confused by the tranformer core lines in your drawing.
Now I understand how they were created.

Just a brief check:
In the model you posted 3C85 materials are applied.
From what I see this is a non-linear type, right?
(since I see current waves that are non-sinusoidal)
So, the coil values in your posted model represent nr. windings, if I understood correctly.

[poynt99]

Just a brief check:
In the model you posted 3C85 materials are applied.
From what I see this is a non-linear type, right?

Yes, and yes. Anything other than "K_Linear" will be a model of a real non-linear core.

(since I see current waves that are non-sinusoidal)

Look closely at the green probe on the schematic; it is a current probe. It is non-sinusoidal because the current is being distorted by the FWBR. Try the "V" probe or "+" and "-" (differential) probe.

So, the coil values in your posted model represent nr. windings, if I understood correctly.

Yes. I am using for example, 5000 (5k) windings for the magnets, and 300 for the gen coils. I used the same ratio when I had the K_Linear core there, i.e. 16mH and 1mH.

You may notice that L2 (LTopCoil) is inverted with respect to orientation. Single-click that inductor to highlight it, then press the "V" keyboard key. The inductor will flip vertically (for V). Now run the sim and you will see only low level "noise" for either the current or voltage, due to perfect canceling.

.99

[teslaalset]

Look closely at the green probe on the schematic; it is a current probe. It is non-sinusoidal because the current is being distorted by the FWBR. Try the "V" probe or "+" and "-" (differential) probe.

You may notice that L2 (LTopCoil) is inverted with respect to orientation. Single-click that inductor to highlight it, then press the "V" keyboard key. The inductor will flip vertically (for V). Now run the sim and you will see only low level "noise" for either the current or voltage, due to perfect canceling.

I had a look at the primary current meanwhile, so this where I saw the distorted current waveform.
Of course, you're right on the secondary, diodes cause distortion

It's pretty handy to have this model, although it's not so easy to generate a comparable flux, the signal generators are pretty limited.
Would be good to have a possibility of an arbitrary waveform generator in Pspice, I can get the flux waveform out of my Ansys simultions and then it would be possible to signal process that into a dedicated input signal generator.

[poynt99]

It's pretty handy to have this model, although it's not so easy to generate a comparable flux, the signal generators are pretty limited.
Would be good to have a possibility of an arbitrary waveform generator in Pspice, I can get the flux waveform out of my Ansys simultions and then it would be possible to signal process that into a dedicated input signal generator.

There are many options for sources in PSpice.

If you can export the flux wave form into a table, we can import that into a PSpice stimulus. 

From what I've seen, a typical wave form looks like a sinusoid, but with the beginning and ends gently tapered (when in-between magnets). I was going to try and create this with the stimulus editor, but if you can export the values, that might be easier and more accurate. Let me know.

.99

[teslaalset]

If you can export the flux wave form into a table, we can import that into a PSpice stimulus. 

From what I've seen, a typical wave form looks like a sinusoid, but with the beginning and ends gently tapered (when in-between magnets). I was going to try and create this with the stimulus editor, but if you can export the values, that might be easier and more accurate. Let me know.

If you can export the flux wave form into a table, we can import that into a PSpice stimulus. 

Poynt,
Attached an example in excel format.
This run simulated an RPM of 3000, so one period is 2.5 ms per magnet.
There is some offset in the signal due to the bias magnet.

Due to the distances between the rotor magnets, it's not pure sinus, so basically you are right in using a pure sinus, assuming rotor distances would be a bit smaller.

[poynt99]

Thanks teslaalset,

Could you please repost that as a .xls ? I'm still using Office 2002.

Thanks,
.99

[teslaalset]

Could you please repost that as a .xls ? I'm still using Office 2002.

Here's a CSV version that should work well with MS office 2002.

[poynt99]

teslaalset,

Is this what you extracted from your model?

I'm wondering if this looks like what we want?

.99