PopularFX
Home Help Search Login Register
Welcome,Guest. Please login or register.
2020-09-24, 08:35:35
News: A feature is available which provides a place all members can chat, either publicly or privately.
There is also a "Shout" feature on each page. Only available to members.

Pages: 1 2 [3] 4 5 6 7 8
Author Topic: A closer look at a simulated Negative resistance coil.  (Read 58701 times)
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Cyril,

On looking at a ferrite catalog, Peterae's test is probably being done with something like Fair-rite material 78 which has an initial mu of 2300 and a roll off at 1 Mhz:

http://www.fair-rite.com/newfair/materials78.htm

I assume that we want the initial mu to be as high as possible, since a higher mu will represent both a longer delay, and less loss. Unfortunately the HF ferrite materials have mus around 40-150. Do you think these tests will work with the typical low-mu HF pulse transformer materials? How important is having the mu high?

orthofield
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
Fair enough so we want a large toroid using 3F4 ferroxcube material.

Equivalent to 3F4 are the following

Material 75G
Epcos N92
MMG F49
TDK PC50
FDK 7H10
Magnetics K
Nicera BM29

TDK have discontinued PC50 in favour of a newer PC95 material
http://www.mhw-intl.com/news/2009/10/tdk-power-ferrite-pc95-verses-pc50/

PC95 is electrically conductive at 6 Ohm per metre and is mn-zn ferrite

I've tried tracking down 3F4 made by other manufacturers and so far not found any big toroid's that are the same material, infact it's beginning to appear this type of ferrite is being discontinued.
« Last Edit: 2015-02-05, 19:11:14 by Peterae »
   
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Peterae,

The 3F4 has an initial mu of around 900. I wonder if this will cause a higher attenuation, or if it really matters? I guess we will see...

orthofield
   
Group: Moderator
Hero Member
*****

Posts: 1408
Hi Cyril,

On looking at a ferrite catalog, Peterae's test is probably being done with something like Fair-rite material 78 which has an initial mu of 2300 and a roll off at 1 Mhz:

http://www.fair-rite.com/newfair/materials78.htm

I assume that we want the initial mu to be as high as possible, since a higher mu will represent both a longer delay, and less loss. Unfortunately the HF ferrite materials have mus around 40-150. Do you think these tests will work with the typical low-mu HF pulse transformer materials? How important is having the mu high?

orthofield

Higher frequency has a larger effect than higher mu since the induced neg R goes with omega^2 while the sin(phi) goes with sqrt(mu).  Peter has just sent me some Farnell data on large toroidal cores and it looks like 3C90 material will do the job.

Smudge
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
OK ordered 1 of these
http://uk.farnell.com/ferroxcube/t102-66-25-3c90/ferrite-core-toroid-3c90/dp/2103392

and some 1 Ohm thick film resistors for the load.

Dam if i ordered 1 hour earlier then they would have been here tomorrow.
   
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Smudge,
Thanks for answering my question.
I remember you saying at one point that the higher the mu, the longer the delay, which makes sense from a magnetic energy storage standpoint.
I did some looking in my book on bubble memories, and found that the velocity of a domain wall in a bulk sample is based on the Landau-Lifshitz-Gilbert equation for the amount of torque the domains of a particular material see with a given magnetization H. The actual equation for a particular material is:

V = Y/a * Lw/pi * deltaH

Where Y is the gyromagnetic ratio of the material involved, a is a damping factor, Lw is the width of the wall, and of course delta H is the impulse field applied. The damping factor is much smaller for films than for bulk materials which is why bubble memories use films.
You can see from this that for a given core with its material and width that the rate of change of the applied magnetic field is the major determiner of how fast the wall gets moving, and thus how short the delay between source and load coils is. Oddly to me, a lower amplitude H field in the source leads to a longer delay since the domain wall is moving at a slower rate
There doesn't seem to be any dependence on mu at all, and the subject is not mentioned in the book. Maybe this is because the energy in the domain wall is really mechanical-- a moving torque so to speak, rather than actually magnetic.

orthocoil
   
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Smudge, all,

Correction, the velocity of the domain wall is based on the absolute value of H, not on delta H.

orthocoil
   

Group: Professor
Hero Member
*****

Posts: 1864
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
verpies the price is large at £95.87 and 3 times the price of the one we bought, but if we can prove the concept works then it could be bought.

   
Group: Moderator
Hero Member
*****

Posts: 1408
Hi Smudge, all,

Correction, the velocity of the domain wall is based on the absolute value of H, not on delta H.

orthocoil

The reason I treat ferrite like a magnetic dielectric with the velocity related to permeability and permittivity is (a) work done on 3F4 material that indicates this could be so and (b) claims of delays in the tens of nanoseconds that also fit that assumption.  There may be much slower effects for domain wall movement that could on the one hand improve matters but on the other hand since energy is involved it could make matters worse.  Hence the need for experiments.

It may be that the previous measured effects are not due to mu or even domain wall movement at all, maybe the dielectric constant of ferrite is much greater than the figure of about 7 that I use.  My trusty Reference Data for Radio Engineers says that dielectric constants as high as 100,000 have been measured on several ferrites having a small amount if divalent iron in their composition.  So maybe the EM propagation previously seen is due to that, and it represents a weak transmission running ahead of the slow domain wall one.  All I can say at the moment is a large 3F4 toroidal core in a magnetic delay transformer exhibited anomalous effects at 14MHz, and the use of bucking coils would considerably amplify that effect and make it appear at a lower frequency.

Smudge
   
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Mudge,

OK, now I see your reasoning a bit better... Is the 3F4 study on the 'partnered coils' list somewhere?

The guy who did extensive study of the dielectric constant of various ferrites was J. L Snoek, but his book came out in 1947. There is also a paper in my files somewhere on 'anomalous dispersion in ferrites'. I'll do some looking and see if there is more hard data on 'ferridielectric dispersion'. I found a paper on e in NiZn right away-- around 40 and starts dropping off at a maximum of 40 Khz or so. Fig. 4 on pg. 5 shows the inverse relation of resistivity to e for these ferrites.
On the other hand the Brockman patent attached shows that Snoek measured e of MnZn in the 10-4 to 10-5 range, resulting in high core losses:

"In view of the large permeability and the large dielectric constant the wave length of the electromagnetic wave generated in the core becomes of the same order of magnitude as the dimensions of the core in its usual form and thereby standing waves are established in the core".

I'm not sure exactly what was seen in the toroid, but this standing wave could certainly cause an interesting situation. Although he is representing this as a loss, the standing wave is actually adiabatic and transmits energy quite well, although of course energy is extracted from the source so not OU in itself.

If indeed the effect you are seeing is due to a magnetoelectric wave through the ferrite, then the geometry, the size, and the bulk mu and permittivity would factor into the resonances, so only experiment will show what they are in any particular case, without some wave mechanical simming inside a ferrite-- ouch.

The domain wall movement could also also be very fast with short pulses with a high rise time, so there is the possibility of the combination of domain wall and magnetoelectric effects.
 
There is a wealth of literature about this subject, so maybe I should send to your email rather than to the group at large.. does everyone want to get the full ream of papers and patents I can potentially send ? :-)

orthofield



The reason I treat ferrite like a magnetic dielectric with the velocity related to permeability and permittivity is (a) work done on 3F4 material that indicates this could be so and (b) claims of delays in the tens of nanoseconds that also fit that assumption.  There may be much slower effects for domain wall movement that could on the one hand improve matters but on the other hand since energy is involved it could make matters worse.  Hence the need for experiments.

It may be that the previous measured effects are not due to mu or even domain wall movement at all, maybe the dielectric constant of ferrite is much greater than the figure of about 7 that I use.  My trusty Reference Data for Radio Engineers says that dielectric constants as high as 100,000 have been measured on several ferrites having a small amount if divalent iron in their composition.  So maybe the EM propagation previously seen is due to that, and it represents a weak transmission running ahead of the slow domain wall one.  All I can say at the moment is a large 3F4 toroidal core in a magnetic delay transformer exhibited anomalous effects at 14MHz, and the use of bucking coils would considerably amplify that effect and make it appear at a lower frequency.

Smudge
[/quote]
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
OK after a missed delivery i have wound 1 coil of 32 turns on the new toroid.


Data Below
32 Turns of same wire.
LCR Test @ 100Hz
L = 4.45mH
Z = 2.79 Ohm
R = 0.0998 Ohm
Q = 28

LCR Test @ 10KHz
L = 4.366mH
Z = 274.41 Ohm
R = 0.785 Ohm
Q = 351

100pf 9mhz
205pf 6.7mhz
500pf 2.4mhz
1000pf 1.6mhz
2000pf 960khz
5000pf 403khz
10nf 200khz
20nf 88khz
50nf 36khz
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
So my calculated Inductance L is as follows for the above measurements

100pf 9MHz L= 3.1272uH
205pf 6.7MHz L = 2.7526uH
500pf 2.4MHz L =  8.7952uH
1000pf 1.6MHz L =  9.8946uH
2000pf 960kHz L = 13.743uH
5000pf 403kHz L = 31.193uH
10nf 200kHz L = 63.326uH
20nf 88kHz L = 163.55uH
50nf 36kHz L = 390.9uH

So it looks to me that we need to operate at no more than 400KHz and maybe 100KHz
PS the graph does not look right to me? Inductance seems to fall off to fast for this core, although is a lot better than my previous core.

OR
Maybe i need to do more measurements with large capacitance values, if that's the case then maybe we can only go in the 10's of KHz  ???
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
OK Added more data points
Not sure what to make of this, my DC Inductance is way higher than my previous core but the rolloff seems to be still low between 10 & 40KHz?

100pf 9MHz 3.13uH
205pf 6.7MHz 2.75uH
500pf 2.4MHz 8.8uH
1000pf 1.6MHz 9.9uH
2000pf 960kHz 13.7uH
5000pf 403kHz 31uH
10nf 200kHz 63uH
20nf 88kHz 163uH
50nf 36kHz 390uH
100nf 13KHz 1.49mH
150nF 7.44kHz 3.05mH
200nF 6.31kHz 3.18mH
330nf 4.66kHz 3.53mH
430nf 3.9kHz 3.8mH

Anyway i will continue.

Next Step is to wind a second coil and feed the primary waveform into a digital monostable to square it up and same for secondary coil and then i will be able to scope the phase shift if it all goes to plan.
   
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Peterae,

I like your methodical approach. It gets results in the end.

The 2C90 core has a much higher DC mu than the other core, and this shows up as expected.
But the sharp drop off in calc. L at way less than 40 Khz doesn't conform to the datasheet at all, where complex mu falls off at around a Mhz.
This is really just a guess, but magnetic losses do drop off more rapidly in this material between 25 and 100 Khz at some given B. So maybe this has an effect on your L caclulations?

I found this interesting article by Jean-Louis Naudin about time delays in magnetic cores.

http://jnaudin.free.fr/dlenz/DLE22en.htm


He demonstrates a sizable magnetic domain delay in a laminated transformer I core.
He drove one end of a laminated power transformer I core with an air coil and then used a search coil, and hall probe to measure the magnetic delay. At a location of 22 mm down his laminated core, the actual magnetic field was 180 degrees out of phase.
This is not at all what Cyril is talking about. He is talking about an electromagnetic resonance, rather than a motional resonance (basically domains sloshing back and forth in a core).
But these domain movement resonances do show up pretty readily, so you could expect to see them in separated coils. It seems like in some ways that ou would be easy once you have a delay between pri and sec of a transformer, no matter what the cause. One could simply load the moving coil in his test with a given load, resulting a load phase lag, and then move the coil up and down until the phase lagged flux is adding to the drive flux instead of opposing it, at the location of the drive coil.

orthocoil

   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
Hello ortho

Thanks for the Naudin link  O0 & help

Back to the bench, i have 2 coils now and 2 primary coils each with 6 turns in bucking mode i will post data below this post a little later.

Next i will try for the phase delay and then check the mutual coupling.

Seconday windings
35 Turns each as previously detailed above
Connected series bucking
LCR @ 10KHz

L=378uH
R=0.2915R
Z=23.8R

Then
Primary on top 6 Turns each
Individual coil LCR Tests @ 10Khz
R=0.0462R
L=174.66uH
Q=241
Z=10.5R

Then
Primary LCR Test in series bucking mode @ 10KHz
L=15uH
R=0.0488R
Q=19.4
Z=0.94R

Cheers
Peter
« Last Edit: 2015-02-14, 18:15:04 by Peterae »
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
OK the digital monostable did not work out, i think the amplitude varies slightly and makes any phase shift thats detected as false, i am triggering a schmitt at a certain voltage not zero crossing point  :-\

So onto the mutual Inductance.
So i placed a 1 Ohm resistor in series with one of the main 32T coils and scoped across the resistor (Green Scope Chan)
and then scoped the other 32T coil (Yellow Chan)
I did this at 12.8KHz see photo curvsvol1

I did a scan up in frequency and found that about 1.2MHz the sig gen current would change phase with the second coil's voltage see photo
curvsvol2 and i have a link to a video of me adjusting the sig gen either side of this phase change.
https://www.youtube.com/watch?v=VwP4Sn8gwsw&feature=youtu.be
   
Group: Moderator
Hero Member
*****

Posts: 1408

I found this interesting article by Jean-Louis Naudin about time delays in magnetic cores.

http://jnaudin.free.fr/dlenz/DLE22en.htm


He demonstrates a sizable magnetic domain delay in a laminated transformer I core.
He drove one end of a laminated power transformer I core with an air coil and then used a search coil, and hall probe to measure the magnetic delay. At a location of 22 mm down his laminated core, the actual magnetic field was 180 degrees out of phase.


I can't accept that much delay occurring in transformer steel so I did a quick FEMM. simulation.  JLN's hall probe is not measuring the flux in the core, it is measuring the flux outside the core, in fact the flux lines leaving the core normal to the surface.  You can see in the simulation that there is a 180 degree change of that surface flux, but it has nothing to do with propagation delay.  The red line is where FEMM plotted the normal component of B and you can see it starts off negative close to the coil then crosses over to positive some distance from the coil.

Have just got back from a weekend away so will look at Peter's results next.

Smudge
   
Group: partzmans Private Group
Full Member
*

Posts: 201
Hi Smudge,

Hmm, yes I agree with you about JLN's test. Most of the flux travels at right angles to the laminate and so it is more of a flux motional effect (interesting in itself!) rather than a domain movement. 
It does show that there are at least two and maybe three mechanisms that can show up as a delay.

orthocoil
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
OK i placed a 1 Ohm NI Resistor in series with my primary and scoped across the resistor and coil, i had to invert the coil probe wave because of the common ground probe connections.

The waveforms can be seen at 41KHz in snap vipri Green Chan across 1 Ohm & Yellow across bucking primary.

Next the 1 Ohm load resistor across the 32T bucking coils as load see snap sec1ohm

PS don't worry Smudge if this does not workout to be OU, it fun to do and worthy of experimentation either way  O0

So my scope computes RMS values for I & V
If i multiply these 2 together then i have a power input of 2.1225mW

As i just have a load resistance then presumably i can rake the RMS value of the voltage across the 1 Ohm and do V^2/R

Which gives me an output power off 144uW

I have seen a lot larger amplitude on my load resistor and this happens when i switch scales on my Sig Gen, i will investigate this as i do not believe there is an increase in power input but only a constantly changing frequency until it stabilizes, and as it seems to stabilize the amplitude also falls.

Maybe we need to change the load resistor value to match any possible negative resistance.
   
Group: Moderator
Hero Member
*****

Posts: 1408
It looks like Peter's toroidal core is not the 3C90 ferrite he expected.  From the measured inductance values I have calculated the effective mu and get the chart shown below.  The LF mu is not far off the 3C90 spec, but the mu falls off rapidly above 10KHz.  The 3C90 spec shows it flat to about 1Mhz.  I think a complaint to Farnell would be justified.

Smudge

Edited to include comparison chart
« Last Edit: 2015-02-16, 11:06:11 by Smudge »
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
Seems strange that the first 5 reading i took are good, these were done after the other wrong readings, i wonder if there is anything i could have done to get duff readings, so will re do the tests in a bit and post below before complaining  O0
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
Are there any rules for finding the roll off, i am adjusting the Sig gen for max amplitude and then looking for when it starts decreasing in amplitude, i have a choice to use the frequency immediately it starts decreasing which is very small at first but then the decrease accelerates, i could register a point that the decrease really starts speeding up, at the moment i am looking for the smallest fall in amplitude and using this frequency value.

Can i ask how you are producing your graphs  ???
   

Group: Administrator
Hero Member
*****

Posts: 3756


Buy me some coffee
OK done the measurements again, looks like there are a few false slight drops before the big fall off, not had a chance to look at the new data yet but here it is, hopefully this may be better C.C


500pf 1.75MHz Calc L = 16.542uH
1000pf 1MHz L = 25.33uH
2000pf 735kHz L = 23.444uH
5000pf 312kHz L = 51.056uH
10nf 97kHz L = 269.21uH
20nf 56kHz L = 403.86uH
50nf 20kHz L = 1.2665mH
100nf 12kHz L =  1.759mH
150nF 10.4kHz L = 1.5613mH
200nF 7.8kHz L = 2.0817mH
330nf 5.6kHz L = 2.4477mH
430nf 4.6kHz L = 2.9781mH

EDIT OK this still does not look good, is there anything i could be doing wrong, maybe i should not use a cap box with flying leads to the coil?

Why is my LCR meter saying 4.43mH @ 10KHz and the above is saying 1.56mH @ 10KHz?
   
Group: Moderator
Hero Member
*****

Posts: 1408
Are there any rules for finding the roll off, i am adjusting the Sig gen for max amplitude and then looking for when it starts decreasing in amplitude, i have a choice to use the frequency immediately it starts decreasing which is very small at first but then the decrease accelerates, i could register a point that the decrease really starts speeding up, at the moment i am looking for the smallest fall in amplitude and using this frequency value.

Can i ask how you are producing your graphs  ???

I assumed you tuned the sig gen to the resonant point which would be a peak in the amplitude.  I calculate the relative mu from the inductance formula L=N^2*munought*mu*A/l where A is the core cross section and l is the core length.  These core dimensions are given in the data sheet in mm^2 and mm so you have to convert those to m^2 and m.  I do this in an Excel spread sheet and use the chart function which allows you to use logarithmic scales.

How are you connecting the sig gen to the LC circuit?  One way is to use a series LC and connect the sig gen across the series circuit, then look for maximum amplitude across the L or the C depending on which one connects to ground.  Alternatively use a parallel LC but loosely couple the sig gen via a one turn loop.  Tune for peak amplitude across the circuit.  If you are not looking for the peak this could be why the results are wrong.  Edit, and don't forget to include the scope probe capacitance if it becomes significant.

Smudge
   
Pages: 1 2 [3] 4 5 6 7 8
« previous next »


 

Home Help Search Login Register
Theme © PopularFX | Based on PFX Ideas! | Scripts from iScript4u 2020-09-24, 08:35:35