Rediscovering Zaev’s ferro-kessor

[partzman]

Vasik,

Thanks for your reply! 

It is interesting that you did not experience any noticeable gain in the flyback arrangement.  I think there are many factors that would affect the result.

Smudge has written papers on the clockwise BH curve so maybe he would have some comments on your work here.

Regards,
Pm

[Vasik041]

Here updated version of "observing_negative_resistance.pdf"
I can see tiny gain even in LTSpice simulation...

[partzman]

Here updated version of "observing_negative_resistance.pdf"
I can see tiny gain even in LTSpice simulation...

Nice work!    I ran your sim and got a little different results for the dissipation levels as you can see in the attachment.  Still a slight gain of ~1% though.  I'm using version IV with modified trap and alternate solver with all other parms default.

Regards,
Pm

[Vasik041]

Hi Pm,

I use older LTSpice XVII and I set maximum timestep 10ns, otherwise it "optimize" too much.
Here again, we have a dilema, is this gain real or just a simulation error ?

Regards,
Vasik

[partzman]

Hi Vasik,

Yes that is the question especially at these low gain levels.  However, it is worth pursuing IMO!

Regards,
Pm

[partzman]

Vasik,

I have been reading the documentation of your experiments you have posted in this thread and I would highly recommend to all FE researchers to do the same!  You have many interesting results both in simulation as well as bench work.

I too believe that sims can show OU with the proper models and you have worked with mostly non-linear devices.  I would like to ask if you have any idea how to model the MVP or A field?  Same question goes for the E field or S=EXH?  I have a bench device that exhibits a COP of ~1.20 that uses either the MVP or the Poynting flow as the gain mechanism and would love to be able to simulate it but haven't figured out any way to do that at this point in time.

Again, I really appreciate the work you have achieved!

Regards,
Pm   

[Vasik041]

Hi Pm,

Thank you for nice words, I appreciate it
You can just write equations in LTSpice and it will do simulation. Or you could use SciLab (or Mathlab) for generic simulation. If you share more details about you setup, perhaps I get better idea how to simulate it.

BR,
Vasik

[partzman]

Vasik,

OK, attached is a pix of the basic transformer used in my aforementioned bench experiments.  The cores are EC-52 with a center leg gap and the circuit configuration is that the outside coils are the primaries and are connected in a buck mode to drive the center coil secondary.  The secondary Ls is biased by a constant current inductor Lcc with Lcc>>Ls.  The voltage polarity on Ls is such that when the primaries are being charged, the current in Lcc increases and is clamped at the peak of the primary current.  The device gain comes from the differential current in Lcc.

However, if one examines the layout, the MVP of the primaries is in phase with the Ls MVP resulting in a current gain increase in Lcc that is greater than the applied voltage across Lcc using di=E*t/L.  In the pix, the transformer on the right will exhibit a current increase of ~10% while the transformer on the left will exhibit an increase of ~15% or greater and also produces a larger COP.  Therefore my logic says that the additional coil heighth of the primaries and secondaries provides a greater area for the aiding MVP field between the adjacent winds of the primaries and secondary.  I could be wrong of course!

The two core transformer arrangement has reached a COP~1.25 as the best at the moment but I'm sure this can be improved upon over time.

Regards,
Pm

[Vasik041]

Pm,

Here an example of magnetic circuit simulation in LTSpice.
I used it when I was experimenting with MEG-like setups.
mu.lib defines core non-linearity mu(H) and mu(B).
May be it will be useful for your setup too. You can add electrical part of setup to the same model if you like.

I would suggest analyze some simpler differences between setups e.g. coils inductance and parasite capacitance.
There might be some simpler explanation for observed difference.

BR,
Vasik

[partzman]

Vasik,

OK thanks for the files and info.  I have tried to simulate the transformers with gyrator/capacitor modeling but they come out conservative.  IOW, since the cores are operating in a linear mode as far as I can tell, I assume there must be some outside energy that is not accounted for in the sim.  I used G/C modeling because I couldn't seem to model the 3-leg core properly otherwise!  The G/C modeling will allow for non-linear core characteristics and separate electrical characteristics but I'm not that proficient with it yet.

I have run this arrangement with non-linearity on the bench and there was gain but not as good as the linear modes. 

I will continue to investigate to see if the gain is explainable by more conventional means.

Thanks for your help! 

Regards,
Pm

Edit: The nonlinear modes were symmetrical and I am studying your comments on asymmetrical non-linearity as perhaps this may provide additional gain.

[Vasik041]

Pm,

I've been thinking about your setup and it boils down to this simple arrangement.
We here have a current source (power supply + Lcc) connected to variable Ls and load.
Thinking from this perspective you can clearly see what affects performance: I and dLs/dt.
I hope it make sense

BR,
Vasik

[partzman]

Pm,

I've been thinking about your setup and it boils down to this simple arrangement.
We here have a current source (power supply + Lcc) connected to variable Ls and load.
Thinking from this perspective you can clearly see what affects performance: I and dLs/dt.
I hope it make sense

BR,
Vasik

Vasik,

I understand your arrangement but my device is different.  In my previous description, consider that Lcc and Ls are charged in series to a given current level then they are clamped to form a current loop.  The primaries are then ramped up with a current with voltage polarity across Ls such that the Lcc/Ls loop current increases.  Basically there is little to no dLs/dt change as the core is sufficiently gapped to remain in a linear mode.

At the end of a given period, the energy in the primaries is returned to the supply, the current in Lcc is clamped and held, and the current in Ls diminishes to a certain level.  When all the energies are summed in this cycle, there is a gain.   

Regards,
Pm

[Vasik041]

Pm,

I remember you had a thread with description of your device, scope traces etc.
I think it was two years ago. I can't find that thread anymore.
Could you please give a link ?

Thanks,
Vasik

[partzman]

Vasik,

Check your PMs.

Pm

[partzman]

Vasik,

In your pdf "fe_research2" on page 79 your simulated results of your parallel ring MEG design shows an output of ~5.58mJ.  The modulator input appears to be minuscule!!!  Would you care to discuss this?

Regards,
Pm

[Vasik041]

Vasik,

In your pdf "fe_research2" on page 79 your simulated results of your parallel ring MEG design shows an output of ~5.58mJ.  The modulator input appears to be minuscule!!!  Would you care to discuss this?

Regards,
Pm

Hi Pm,

Yes, simulation looks good
It use technique I described to you - there is magnetic circuit and electronic in same model.
As you see, this work from 2012, I spent considerable time researching MEG-like setups.
It is very difficult to make non-standard core configurations with high quality.
My practical results not so good. With setup like this I got only 50% efficiency.
There were some reports about successful experiments e.g. with resonance modulator, but I stopped that line of research.

BR,
Vasik

[partzman]

Hi Pm,

Yes, simulation looks good
It use technique I described to you - there is magnetic circuit and electronic in same model.
As you see, this work from 2012, I spent considerable time researching MEG-like setups.
It is very difficult to make non-standard core configurations with high quality.
My practical results not so good. With setup like this I got only 50% efficiency.
There were some reports about successful experiments e.g. with resonance modulator, but I stopped that line of research.

BR,
Vasik

Hi Vasik,

OK, thanks for the explanation and I too agree that accurate models of core configurations are difficult to say the least.  I have preferred to use the current controlled voltage source models which I see you use as well but when non-linear effects are introduced plus PM bias, accuracy seems to be an issue!

Regards,
Pm