A closer look at a simulated Negative resistance coil.

[Peterae]

Hi Smudge
I hope it was OK to start this thread in your bench, i wanted to take a closer look at this proposed coil of yours.
I've moved my post from the partnered thread.

Right, now listen up!!  I have just done a FEMM simulation for Itsu's large ferrite rod 200 mm long by 20mm diameter.  I have assumed the ferrite to be like 3F4 with a mu of 900 and a velocity of propagation of 3.88x10⁸m/S.  I have placed on this core two coils each of 100 turns of 1 mm magnet wire, that is 10 layers of 10 turns per layer.  The coils are 10 mm in length and extend out 10 mm from the core surface.  The gap between the cores is 60 mm so their mean separation (center to center) is 70 mm.  That represents a time delay from one coil to the other of 18nS.  My simulation tells me the inductance of each coil is 541uH and the mutual inductance is 183uH.  Thus the coils connected in series opposing should give 715uH and FEMM confirms this when I set that up (see image).  Now using that mutual inductance figure and the equation for induced negative R in my paper R=-omegaL₁₂sin(phi) where phi is the phase delay I get the following results.

frequency   omega   phase   Neg R
1.00E+03   6.28E+03   1.13E-04   2.61E-04
1.00E+04   6.28E+04   1.13E-03   2.61E-02
1.00E+05   6.28E+05   1.13E-02   2.61E+00
1.00E+06   6.28E+06   1.13E-01   2.61E+02
1.00E+07   6.28E+07   1.13E+00   2.09E+04

Considering the coils have a DCR of only 0.176 Ohms this suggests that we should get self oscillation at 100kHz.  I can't believe it is that easy.  Of course the AC resistance of the coils will be higher and I have not taken account of core losses.  Also there could be reflections from the ends of the core coming into play.  But the gap is not optimised, I just used 60 mm as a starting point.  I think this does illustrate the potential this approach has for getting OU.

Smudge
             

Smudge i have 2 ferrite rods from an aerial, not sure what type of ferrite they are but their dimensions are 127mm x 9.525mm, any chance you could tell me what coils i need to wind to try your simulated setup.  

I have 18SWG (1.219mm) & 21 SWG (0.813mm)
Cheers
Peter

[Smudge]

Hi Peter,

I think you need coil formers that you can slide along the rod.  The axial width must be small compared to the rod length, so formers that are less than 1cm wide would be good.  You need a reasonable number of turns, more than 10 but probably less than 100.  Put one coil at the centre of the rod and measure inductance against frequency to find the roll-off point.  Then put two coils, use that roll-off frequency, energise one coil and measure the voltage from the other coil.  See if you can find any phase shift of that induced voltage as you move the coils apart.  If you can get a change of phase against distance you have measured the velocity of propagation which can be used to predict results.  Also the value of that open circuit voltage can allow you to calculate the mutual inductance between the coils.  Now you can play with the coils connected in series bucking mode.  The measured inductance of that series connection allows you to calculate the coupling factor if needed.  Now you are really looking for an anomalous effect that could be masked by losses, so how do we find that.  One way is to add another coil so as to make a transformer, connect the bucking coils to a load and do careful input, output power measurements.  Plot COP against separation distance and against frequency (lots of measurements ) and look for a peak.  If that peak goes above 1 then so much the better

Cyril

[Verpies]

You need a reasonable number of turns, more than 10 but probably less than 100.

I will add my 2 cents' worth of commentary:
Use even number of layers to cancel any component of current that is parallel to the solenoid's axis.
The inductance increases with the square of the number o turns, the resistance increases linearly, thus more turns yield a higher L/R ratio.

Unfortunately, more turns also leads to more interwinding capacitance, that will degrade the high frequency response of the coil and cause LCR oscillations.

@Smudge
How does the interwinding capacitance vary with the wire diameter and number of turns, all other parameters being equal ?

[Peterae]

Thanks Smudge

First i will work on the coil formers & bobbins and report back.
I will use 18 SWG, former for 8 turns per layer with 8 layers for start, so 64 turns each coil.

Thanks verpies

Even number of layers

PS if we were worried about interwinding capacitance could we not wind bifilar and connect like a tesla bifilar coil.

infact could we just use Teslas coil and slide that on the rod, it would only be 1 turn wide   and easily made flat on cardboard.

Cheers
Peter

[Smudge]

I will add my 2 cents' worth of commentary:
Use even number of layers to cancel any component of current that is parallel to the solenoid's axis.
The inductance increases with the square of the number o turns, the resistance increases linearly, thus more turns yield a higher L/R ratio.

Agreed so more turns means higher Q.  But this is only for a certain wire diameter in which more turns occupy more space.  If you have a coil former that determines the winding area and you fill it with wire (i.e. more turns means finer wire gauge) then R also goes as the square of number of turns, and Q remains the same.

Unfortunately, more turns also leads to more interwinding capacitance, that will degrade the high frequency response of the coil and cause LCR oscillations.

That might not be a bad thing if we find that we need to operate a resonant circuit, provided that the self resonance is lower than the optimum frequency.

@Smudge
How does the interwinding capacitance vary with the wire diameter and number of turns, all other parameters being equal ?

I don't have a formula to hand but I am sure there will be one out there somewhere

Smudge

[Smudge]

If we were worried about interwinding capacitance could we not wind bifilar and connect like a tesla bifilar coil.

infact could we just use Teslas coil and slide that on the rod, it would only be 1 turn wide   and easily made flat on cardboard.

Not only that but you could stack Tesla coils together to make more turns.  I think you need to limit the outer diameter because the outer turns don't couple well to the core, and that will limit the turns for each flat coil.  Rule of thumb, outer turns should not be more than 1 core diameter from the core surface.  It would be interesting to compare your normal coils with stacked Tesla coils occupying the same volume and same turns to see if there is any difference.  But that could come later.

Cyril

[Peterae]

Sure we can try that later,they would be easier to make i think.

I have 1 coil  glued and drying, 8 turns wide, 8 high and slidable, but i have to say it's really hard getting the turns side by side on the last 4  rows, the wire is really thick and hard to lay down manually.
I dropped down to my smaller wire as well, just could not do it with the 1.2mm wire.

[Peterae]

OK 2 coil made just gluing No 2

1 Step

Put one coil at the centre of the rod and measure inductance against frequency to find the roll-off point.

How am i going to do this.
I can measure inductance at 100Hz,1KHz,10KHz using my inductance meter, but not by varying a frequency linearly.

[Smudge]

OK 2 coil made just gluing No 2

1 StepHow am i going to do this.
I can measure inductance at 100Hz,1KHz,10KHz using my inductance meter, but not by varying a frequency linearly.

Resonate it with known capacitor values and see what frequency you get.

[Peterae]

OK 

[Peterae]

Initial Test using LCR Meter with coil fitted middle of ferrite rod

Test @ 120Hz
R = 0.1055 Ohm
L = 453 uH
Z = 0.3603 Ohm

Test @ 1KHz
R = 0.111 Ohm
L = 454.3uH
Z = 2.8571 Ohm

Test @ 10KHz
R = 0.2783 Ohm
L = 453.1uH
Z = 28.474 Ohm

EDIT
I will add data below as i check different capacitors

So
100nF gives max pk-pk at about 21.27KHz 8.16V

OK switched to cap box
330nF gives 13.25KHz @7.92V before roll off
90nF gives 26KHz @ 8.16V
70nF 30KHz @ 8.16V
50nF 37.5KHz @ 8.16V
30nF 45KHz 8.24V
20nF 62KHz @ 8.208V
10nF 87KHz @ 8.32V
5nF 144KHz @ 8.32V

Calculated inductance
330nf 437uH
90nF 417uH
70nf 402uH
50nf 360uH
30nf 417uH
20nf 330uH
10nf 335uH
5nf 244uH

[Smudge]

Well done Peter.  I have put all your measurements onto a plot with a log scale to show the typical roll-off curve for inductance.  The mu will have the same roll-off characteristic.  So your optimum frequency will be between 10 and 20 kHz I reckon.

Smudge

[Smudge]

I have looked at the effect of self-capacitance of the coils and find it doesn't affect the value of negative R.  But it does affect the series L, increasing the impedance until you hit resonance where it is then difficult to drive current hence get any anomalous power.  So verpies is right, you need to minimise self capacitance in the coils.  I have amended my paper to include this, also to discuss optimum arrangements to maximise anomalous power.  It is now clear that ferrite rods are not the best way to go, you need a closed magnetic circuit (of which there are plenty where OU is claimed but they haven't latched on to the delay phenomenon).

Smudge

[Peterae]

Thanks Smudge

So whats my next move, do i abandon the ferrite rods.

I have a huge ferrite ring  of unknown ferrite type, 10cm OD 6.5cm ID & 2cm high

I've just taken off a load of insulation tape so looks a bit sticky so will clean it up.

Do i need the 2 coils slideable on a toroidal core?

[Smudge]

Thanks Smudge

So whats my next move, do i abandon the ferrite rods.

I have a huge ferrite ring  of unknown ferrite type, 10cm OD 6.5cm ID & 2cm high

I've just taken off a load of insulation tape so looks a bit sticky so will clean it up.

Do i need the 2 coils slideable on a toroidal core?

That coil looks just right so I suggest abandoning your ferrite rod.  Don't bother with sliding coils, just wind some in diametrically opposite positions as per my sketch.  With one coil not connected do the measurements you have just done (on the rod) on the other coil and we'll find out what ferrite it is.  It might even be the 3F4 ferrite.  Then go ahead and make a transformer.  Perhaps you could cut down your ferrite rod to fit inside the big toroid across a diameter and use this for the primary coil.  Then you could have something that will blow up your RF source  .  Wistiti already did this with his arrangement.  I think that if you have a load resistance that is too low it cannot absorb the anomalous energy so it gets fed back to the input.

Smudge

[Peterae]

ok it's slidable a few CM either side of where it's located

dropped the turns to 32

LCR Tests

100Hz
R=0.09 Ohm
Z=2.1016 Ohm
L=3.34mH
q=23.0445

1KHz
R=0.1801 Ohm
Z=20.892 Ohm
L=3.3mH
Q=117.76

10KHz
R=0.7247 Ohm
Z=205.25 Ohm
L=3.265mH
Q=291

[Smudge]

ok it's slidable a few CM either side of where it's located

dropped the turns to 32

LCR Tests

100Hz
R=0.09 Ohm
Z=2.1016 Ohm
L=3.34mH
q=23.0445

1KHz
R=0.1801 Ohm
Z=20.892 Ohm
L=3.3mH
Q=117.76

10KHz
R=0.7247 Ohm
Z=205.25 Ohm
L=3.265mH
Q=291

That looks like a mu of over 2000 so it won't be 3F4.  But no matter it will still be good because that high mu means a smaller propagation velocity.  When you have done the capacitor bank measurements we'll find the roll-off and the optimum frequency.  Then you will really be sailing .

Smudge

[Peterae]

Sounds good  

5nF 113KHz @ 14.18V
10nF 76KHz @ 14V
20nF 31.5KHz @ 14.16V
40nF 15.82KHz 14.24V
80nF 13.92KHz 13.84V
100nF 9.58KHz
330nF 4.64KHz 13.73V

I will process the data and add it here.  
Calculated Inductance
113KHz 397uH
76KHz 439uH
31.5KHz 1.277mH
15.82KHz 2.531mH
13.92KHz 1.635mH
9.58KHz 2.761mH
4.64KHz 3.567mH

Looks like i need more data points, was in a rush tonight, hopefully get more time tomorrow night.  
and it also looks like i have a duff reading at 13.92KHz

I just went back to try and home in again and got different results
It's really hard to find the roll off spot only a very small voltage drop at first over quiet a large span of frequency
6nf 88KHz
7nF 57KHz
8nF 46KHz
9nf 39KHz
10nF 48KHz
11nf 41KHz
12nF 50.4KHz

Not sure why 12nF is so different maybe one of my caps is way off in the box, they are high voltage caps 1500V so maybe +/- 20%
I've run out of time to try and work out whats going on right now, Smudge see if you can make sense of any of the data, if not i will try again tomorrow, it's only 2 degrees right now in my shed and earlier it was toasty warm 16 deg C maybe that's changing something.

Maybe a better way is to measure the capcaitance instead of relying on the printed value.

[Smudge]

I don't want to discourage any experimentations, but I looked at the paper smuge posted, about the negative resistance, and I would say it's a bit more complex than that, since coils have inter winding capacitance, resistance, etc.

However, the math deriving the negative resistance does look right assuming a phase "delay" in the mutual coupling of ideal inductors. But that's the big IF, is there really a phase shift?   Yes there is always propagation delay, between objects spaced apart, since energy travels with the speed of light, but does that produce extra energy?  No I don't believe so.

So what's the problem?  

Well it appears at face value to be a case of mixing apples with oranges, i.e. Mixing time domain analysis with sinusoidal steady state analysis.   Talking about a phase shift implies steady state, meaning the transients have died down, and we're exciting the circuit with a pure sinusoidal signal.

EM

Hi EM,

Sine waves are not steady state and you can handle sine waves in either the frequency domain or the time domain.  For a time delay of t the phase delay is omega*t.  So I don't understand where you are coming from.  And there is evidence of the time or phase delay along a core, it has been measured.  And it is considerably less than c.  I have no doubt that Peter will measure it on his big toroid when he gets that far.

Smudge  

[Smudge]

I've run out of time to try and work out whats going on right now, Smudge see if you can make sense of any of the data, if not i will try again tomorrow, it's only 2 degrees right now in my shed and earlier it was toasty warm 16 deg C maybe that's changing something.

Maybe a better way is to measure the capcaitance instead of relying on the printed value.

Those measurements are enough to find the roll-off characteristic, see chart.  The fitted curve shows the typical roll off expected, and it is obvious this is not a HF ferrite.  So you need to experiment in the frequency range between say 3 and 10KHz.  Perhaps settle on 10KHz for easy inductance measurements.

I would wind an identical coil on the other side of the toroid then do some some measurements.  Of most importance is the time or phase delay so I would apply a sine wave voltage to one coil and measure the voltage on the other coil, then attempt to establish the phase between the two voltages, e.g by timing between the zero cross overs.  Unfortunately the delay could be less then 0.2 degrees so might not be measurable.  Then I would establish the mutual coupling between coils by measuring the current in the input coil (1) and the voltage from the output coil (2).  The mutual inductance L₁₂ is dPhi₂/di₁ which for linear materials becomes Phi₂/i₁.  Phi₂ can be got from Ph1₂=V₂/(omega*N) where N is the number of turns.  Next I would connect the coils in series bucking and measure the inductance of that series combination.  You will know whether they are bucking because if they are not the inductance will be 4 times that of a single coil, whereas in bucking mode the inductance will be less than that of a single coil.

Then I would wind a few turns of primary coils over the secondary coils, connect them in series bucking, put a load across the secondary and see what happens when you apply an input.  Should work as a transformer but hopefully an OU transformer.

Cyril

[Verpies]

I don't want to discourage any experimentations, but I looked at the paper smuge posted, about the negative resistance, and I would say it's a bit more complex than that, since coils have inter winding capacitance, resistance, etc.

He accounted for the inter-winding capacitance in the new version of his paper.

[Smudge]

I just want to point out that my formula for the induced negative resistance R=-2*omega*L₁₂*sin(phi) where phi is a small phase angle can use the small angle approximation sin(phi)=phi (with phi in radians of course).  Then since phi=omega *t_D where t_D is the time delay for magnetic propagation between the coils you end up with R=-2*omega²*t_D*L₁₂.  Note the omega².  Even small time delays can be significant.

Smudge

edited to add the missing 2 factor

[Verpies]

Can you contrast the electromagnetic propagation delay in a coaxial cable vs. the magnetic propagation delay in a magnetic circuit and contrast the differences in their energy transfer characteristics?

[Smudge]

Can you contrast the electromagnetic propagation delay in a coaxial cable vs. the magnetic propagation delay in a magnetic circuit and contrast the differences in their energy transfer characteristics?

The velocity of propagation in a coaxial cable is determined by the dielectric and is given by v_p=c/sqrt(e₀*K) where e₀ is free space permittivity and K the dielectric constant.  And evidence point towards the propagation along a magnetic rod being v_p=c/sqrt(mu₀*mu_R).  So in that respect they are similar.  However with regard to attenuation one should really compare a magnetic rod with a dielectric waveguide where electric or magnetic flux leakage occurs, but note leakage by itself is not necessarily a loss in energy transfer, it's only a redistribution of energy stored (transformers can work quite well with flux leakage present, it merely appears as a loss-less inductor in the equivalent circuit).  As regards energy loss when energy is transferred from coil to coil we have to take account of the known core losses just as we would in a normal transformer.  I have tried to do this by using the real and imaginary parts of the permeability tensor, and this loss only becomes significant when you get near the cut-off frequency.  But I cannot relate it to an equivalent coaxial cable.  In one of my papers I do give a lumped constant version of the magnetic delay line where the components are all magnetic domain ones, and show how you can use the permeability tensor to obtain values for those components.  That allows you to use classical transmission line theory to solve that network, but you have to be aware that the magnetic values do not have the same meaning as electric ones.  For instant mmf is magnetic "voltage" but it doesn't have the dimension of voltage, its actual dimension is current.  Magnetic flux is magnetic "current" but its dimension is Webers.  Thus the magnetic "attenuation constant" in that delay line does not have the dimensions of power attenuation, we merely use electric analogue to solve the magnetic equations.  As I see it loss ocurs in the magnetic core material and loss occurs by EM radiation.  At the frequencies we would use I think we can discount radiation losses, but I could be wrong there.  In order to match real measurements to the theory in a magnetic delay transformer I had to include some radiation loss, but when practical steps were taken to screen off any radiation it didn't do any good.  So the jury is out.  What I like about this bucking coil theory is that it does not predict a small effect, but we will have to wait and see on that one.

Smudge

[Matt Watts]

This may be partially relevant:
https://www.youtube.com/watch?v=AP0aTogfmxU

Notice how the coils are assembled where intermediate test points can be examined.  Clearly you can see the increasing phase shift as he moves further down the coil.

My feeling is the coil itself IS the transmission line here.  I would have to think the electrical delay is also happening with the magnetic flux, mostly inside the tube.