Forums

transformer core flux propagation speed

Started by Jamie M April 13, 2012
Robert Macy wrote:
> > Think more sequential than average. Actually, a domain wall does not > have 'pressure' on it until an adjacent wall flips. Then the field can > build against it. Sort of like dominoes. > > For me to understand what was going on, I used to envision a field of > wheat, blowing in the wind. The wind hits a stock and it tends to bend > over allowing the next stock to 'see' the wind and so on. Some stocks > are stiffer than others so the wave is not so uniform. Importantly, It > makes a great image for picturing wave propagation. Plus, *IF* the > wind changes direction before the field is down, you can start to > envision the standing wave patterns moving across the field, even see > how the stocks in one section are not even going down the right > direction, but the opposite direction, so instead of helping, they are > hindering. Anway, any allegory that helps intuitive understanding has > some value, look at what Tesla came up with after watching ??, which > was an imperfect allegory, too. He came up with the induction motor.
'Stocks'? I think you mean 'Stalks' -- You can't have a sense of humor, if you have no sense.
On Apr 17, 8:55=A0am, "Michael A. Terrell" <mike.terr...@earthlink.net>
wrote:
> Robert Macy wrote: > > > Think more sequential than average. Actually, a domain wall does not > > have 'pressure' on it until an adjacent wall flips. Then the field can > > build against it. =A0Sort of like dominoes. > > > For me to understand what was going on, I used to envision a field of > > wheat, blowing in the wind. The wind hits a stock and it tends to bend > > over allowing the next stock to 'see' the wind and so on. Some stocks > > are stiffer than others so the wave is not so uniform. Importantly, It > > makes a great image for picturing wave propagation. Plus, *IF* the > > wind changes direction before the field is down, you can start to > > envision the standing wave patterns moving across the field, even see > > how the stocks in one section are not even going down the right > > direction, but the opposite direction, so instead of helping, they are > > hindering. =A0Anway, any allegory that helps intuitive understanding ha=
s
> > some value, look at what Tesla came up with after watching ??, which > > was an imperfect allegory, too. He came up with the induction motor. > > =A0 =A0'Stocks'? =A0I think you mean 'Stalks' > > -- > You can't have a sense of humor, if you have no sense.
Another model shot!
boB wrote:
> > On Mon, 16 Apr 2012 19:10:56 -0500, "Tim Williams" > <tmoranwms@gmail.com> wrote: > > >"boB" <K7IQ> wrote in message > >news:paooo7d5m6qsglnavjs3d2q9p9agjrflue@4ax.com... > >> What I mean is.... If you had a super quiet wide range hall effect > >> or other magnetic sensing device in the gap of a (ferrite ?) core, > >> what would be the smallest AC and/or DC signal change you > >> could see ?? I know there is some noise floor in there but > >> it's kind of hard to read some of the lit I've googled. > > > >http://en.wikipedia.org/wiki/Barkhausen_effect > > Looking at the B-H curve on that page, it doesn't look like > particularly hard material, but it might not be magnetically > "to scale" but was just a convenient curve to use for > illustatration. Since I'm set up to do the test in the lab > already, I will have to try it on ferrite at least while > varying the DC current. It might even be worse than > any other noise but I haven't noticed it yet. > > I wonder how much gain is needed ? Maybe a hall effect device > is too noisey to eve hear this Barkhausen effect and that's one > reason the wiki artilce mentions using a coil ? Looks like it could > even sound like somewhat of a zipper noise. > > OK, so I went to the listed youtooob link and they show a guy rotating > a magnet near an antenna loop looking thing. This is kinda confusing > to me cuz I don't see how that is going to excercise through the B-H > loop as shown in the wikipedia article.... And it's a magnet, not a > soft magnetic core. Something doesn't seem quite right there. > > boB > > > > >This isn't directly due to thermal noise (flipping magnetic spins, domain > >fluctuations), which will be much weaker (though possibly noticable around > >the Curie temperature). > > > >Very soft materials (low remenance) should be quieter than those with high > >remenance; blatant example, applying a field to a permanent magnet won't > >cause any substantial flipping until the entire coercive force is applied > >(~1e5 A/m for NdFeB, IIRC), at which point the whole thing changes quite > >rapidly. > > > >Tempted to set up an amplifier and try listening to a core. Should be able > >to get a small ferrite toroid up to Curie with only the soldering iron > >handy. > > > >Tim
Barkhausen noise used to be a huge problem in hard disk heads, back when the pole pieces were large enough to have more than one magnetic domain. For the past decade or more, the domain walls have been pinned by the geometry, so it's not such a big worry. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
On 4/17/2012 1:05 PM, Phil Hobbs wrote:
> boB wrote: >> >> On Mon, 16 Apr 2012 19:10:56 -0500, "Tim Williams" >> <tmoranwms@gmail.com> wrote: >> >>> "boB"<K7IQ> wrote in message >>> news:paooo7d5m6qsglnavjs3d2q9p9agjrflue@4ax.com... >>>> What I mean is.... If you had a super quiet wide range hall effect >>>> or other magnetic sensing device in the gap of a (ferrite ?) core, >>>> what would be the smallest AC and/or DC signal change you >>>> could see ?? I know there is some noise floor in there but >>>> it's kind of hard to read some of the lit I've googled. >>> >>> http://en.wikipedia.org/wiki/Barkhausen_effect >> >> Looking at the B-H curve on that page, it doesn't look like >> particularly hard material, but it might not be magnetically >> "to scale" but was just a convenient curve to use for >> illustatration. Since I'm set up to do the test in the lab >> already, I will have to try it on ferrite at least while >> varying the DC current. It might even be worse than >> any other noise but I haven't noticed it yet. >> >> I wonder how much gain is needed ? Maybe a hall effect device >> is too noisey to eve hear this Barkhausen effect and that's one >> reason the wiki artilce mentions using a coil ? Looks like it could >> even sound like somewhat of a zipper noise. >> >> OK, so I went to the listed youtooob link and they show a guy rotating >> a magnet near an antenna loop looking thing. This is kinda confusing >> to me cuz I don't see how that is going to excercise through the B-H >> loop as shown in the wikipedia article.... And it's a magnet, not a >> soft magnetic core. Something doesn't seem quite right there. >> >> boB >> >>> >>> This isn't directly due to thermal noise (flipping magnetic spins, domain >>> fluctuations), which will be much weaker (though possibly noticable around >>> the Curie temperature). >>> >>> Very soft materials (low remenance) should be quieter than those with high >>> remenance; blatant example, applying a field to a permanent magnet won't >>> cause any substantial flipping until the entire coercive force is applied >>> (~1e5 A/m for NdFeB, IIRC), at which point the whole thing changes quite >>> rapidly. >>> >>> Tempted to set up an amplifier and try listening to a core. Should be able >>> to get a small ferrite toroid up to Curie with only the soldering iron >>> handy. >>> >>> Tim > > Barkhausen noise used to be a huge problem in hard disk heads, back when > the pole pieces were large enough to have more than one magnetic > domain. For the past decade or more, the domain walls have been pinned > by the geometry, so it's not such a big worry.
Hi, Does Barkhausen noise stop above a certain frequency, maybe proportional to the magnetic domain sizes? Also for an extreme example, for AC core losses, normally they are thought of as caused by eddy currents, but if the frequency is (extremely!) high enough, maybe the eddy current losses will start reducing, but will the transformer core also start to lose its inductance properties so that it is not useful? Just curious maybe 50 years in the future core losses will not matter! cheers, Jamie
> > Cheers > > Phil Hobbs
Jamie M wrote:
> > On 4/17/2012 1:05 PM, Phil Hobbs wrote: > > boB wrote: > >> > >> On Mon, 16 Apr 2012 19:10:56 -0500, "Tim Williams" > >> <tmoranwms@gmail.com> wrote: > >> > >>> "boB"<K7IQ> wrote in message > >>> news:paooo7d5m6qsglnavjs3d2q9p9agjrflue@4ax.com... > >>>> What I mean is.... If you had a super quiet wide range hall effect > >>>> or other magnetic sensing device in the gap of a (ferrite ?) core, > >>>> what would be the smallest AC and/or DC signal change you > >>>> could see ?? I know there is some noise floor in there but > >>>> it's kind of hard to read some of the lit I've googled. > >>> > >>> http://en.wikipedia.org/wiki/Barkhausen_effect > >> > >> Looking at the B-H curve on that page, it doesn't look like > >> particularly hard material, but it might not be magnetically > >> "to scale" but was just a convenient curve to use for > >> illustatration. Since I'm set up to do the test in the lab > >> already, I will have to try it on ferrite at least while > >> varying the DC current. It might even be worse than > >> any other noise but I haven't noticed it yet. > >> > >> I wonder how much gain is needed ? Maybe a hall effect device > >> is too noisey to eve hear this Barkhausen effect and that's one > >> reason the wiki artilce mentions using a coil ? Looks like it could > >> even sound like somewhat of a zipper noise. > >> > >> OK, so I went to the listed youtooob link and they show a guy rotating > >> a magnet near an antenna loop looking thing. This is kinda confusing > >> to me cuz I don't see how that is going to excercise through the B-H > >> loop as shown in the wikipedia article.... And it's a magnet, not a > >> soft magnetic core. Something doesn't seem quite right there. > >> > >> boB > >> > >>> > >>> This isn't directly due to thermal noise (flipping magnetic spins, domain > >>> fluctuations), which will be much weaker (though possibly noticable around > >>> the Curie temperature). > >>> > >>> Very soft materials (low remenance) should be quieter than those with high > >>> remenance; blatant example, applying a field to a permanent magnet won't > >>> cause any substantial flipping until the entire coercive force is applied > >>> (~1e5 A/m for NdFeB, IIRC), at which point the whole thing changes quite > >>> rapidly. > >>> > >>> Tempted to set up an amplifier and try listening to a core. Should be able > >>> to get a small ferrite toroid up to Curie with only the soldering iron > >>> handy. > >>> > >>> Tim > > > > Barkhausen noise used to be a huge problem in hard disk heads, back when > > the pole pieces were large enough to have more than one magnetic > > domain. For the past decade or more, the domain walls have been pinned > > by the geometry, so it's not such a big worry. > > Hi, > > Does Barkhausen noise stop above a certain frequency, maybe > proportional to the magnetic domain sizes? Also for an extreme > example, for AC core losses, normally they are thought of as caused by > eddy currents, but if the frequency is (extremely!) high enough, maybe > the eddy current losses will start reducing, but will the transformer > core also start to lose its inductance properties so that it is not > useful? Just curious maybe 50 years in the future core losses will not > matter! >
AFAIK it isn't white, but it goes well up into the megahertz. The spectrum depends on everything including what you had for breakfast. (Like so many ferromagnetic things.) Barkhausen seems to have made a career out of investigating inconvenient effects, so that he got his name associated with a lot of annoyances: Barkhausen noise, Barkhausen oscillations, and the Barkhausen criterion for when your amp becomes an oscillator, or vice versa. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
"Jamie M" <jmorken@shaw.ca> wrote in message 
news:jml88f$a9n$1@speranza.aioe.org...
> Does Barkhausen noise stop above a certain frequency, maybe > proportional to the magnetic domain sizes?
Probably on par with the domain relaxation time, which depends on material; most likely this corresponds to the cutoff frequency as well.
> Also for an extreme > example, for AC core losses, normally they are thought of as caused by > eddy currents, but if the frequency is (extremely!) high enough, maybe > the eddy current losses will start reducing, but will the transformer > core also start to lose its inductance properties so that it is not > useful?
More or less. If you model core losses entirely as eddy currents (ignoring hysteresis or material properties), then the cutoff frequency corresponds to the skin depth in the material: when the skin is shallower than the core thickness, less core is utilized and the effective permeability drops. Skin effect, in turn, occurs when the current flow produces a field opposing the applied field. If you can reduce the conductor dimensions (stacks of iron sheets, or powder), or use a higher resistivity material (by adding silicon to the steel, or using a ferrite), the cutoff frequency rises again. Permeability has the same frequency response / amplitude / phase interplay that any other filter does, so it should be no surprise that, if normal (real) permeability results in an inductive core, a lossy (phase shifted, complex or imaginary) permeability results in a resistive core, and that resistive phase shift necessarily occurs over a frequency range where permeability is dropping with rising frequency. Indeed, all core materials I've seen look almost entirely resistive at or past some frequency. Some graphs: http://www.ferroxcube.com/prod/assets/3c90.pdf 3C90 is a typical MnZn ferrite used for generic power applications. Essentially equivalent to Fair-Rite #77, and I think Epcos/TDK N87 and Magnetics Inc. type P, though I should check those. The graph of complex permeability vs. frequency shows that total permeability is essentially constant and real up to 1MHz or so, where the imaginary component is getting noticable. (Typically, you can operate a ferrite near saturation up to frequencies where mu'' ~ mu' / 100, in this case about 200kHz; at higher frequencies, losses are higher and you have to reduce flux density to keep it from overheating). If you recall |mu| = sqrt((mu')^2 + (mu'')^2), i.e. the vector sum of real and imaginary components, you might notice that, at the point the curves intersect, the value of each is about 70% the flat range (about 1/sqrt(2)). This means, at the intersection, the value of |mu| doesn't actually dip. As a result, |mu| actually remains fairly constant up to almost 4MHz. The implication is, the magnetizing impedance of an inductor or transformer remains essentially inductive from DC to 4MHz, though the phase shifts substantially past 1MHz. Beyond 4MHz, the impedance remains flat or drops further, because |mu| also drops. In fact, the roll-off can be calculated from the graph. The tail drops at a rate of about 2 decades in mu for one decade in F, so the impedance Z = j*2*pi*F*mu drops off inversely with frequency past this point. This is a second order (-20dB/decade) drop, which is physically significant: if it were simply resistive, it would be 10dB/decade instead. Ferrite beads graph this more directly. See for example: http://search.murata.co.jp/Ceramy/image/img/PDF/ENG/L0110S0100BLM31P.pdf Page 2 shows typical |Z| vs. F, and the components vs. F for each individual part on the following page. Now, these are components with metal buried in the ferrite, so the effective terminal capacitance is significant as well; one would certainly expect the impedance to drop at very high frequencies, and indeed it does. It turns out ferrite has a high dielectric constant (which, because of the resistivity, is obviously a complex number as well), so this is significant even for very small ferrite chips. Another material of note, Fair-Rite #43 (I'd guess at some equivalents, but it's actually kind of funky so I won't go out on a limb here), which is typically used for high frequency transformers and ferrite beads. http://www.fair-rite.com/newfair/materials43.htm You can see the "resistive band" is much wider, roughly 2MHz and up. It'll still make a good transformer at 20MHz, but it will be very lossy. Something else pertinent to draw from these graphs: mu vs. temp is provided. You can see there's a sort of "stochastic resonance" approaching Curie point, which means the domains are getting easier to align -- more thermal energy makes more flips, which reduces the number of spins available to use (saturation flux density is reduced -- see the B-H curves at different temperatures), but they're easier to use. FYI, "stochastic resonance" is a big phrase for, if a system has a dead band or hysteresis, adding a low-level stimulus (be it a coherent waveform or random noise) actually improves the noise floor of the system. For example, dithering and averaging an ADC can yield more bits accuracy. Strictly, "stochastic" only applies to systems where random noise is used. Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms
Robert Macy wrote:
> > On Apr 17, 8:55 am, "Michael A. Terrell" <mike.terr...@earthlink.net> > wrote: > > Robert Macy wrote: > > > > > Think more sequential than average. Actually, a domain wall does not > > > have 'pressure' on it until an adjacent wall flips. Then the field can > > > build against it. Sort of like dominoes. > > > > > For me to understand what was going on, I used to envision a field of > > > wheat, blowing in the wind. The wind hits a stock and it tends to bend > > > over allowing the next stock to 'see' the wind and so on. Some stocks > > > are stiffer than others so the wave is not so uniform. Importantly, It > > > makes a great image for picturing wave propagation. Plus, *IF* the > > > wind changes direction before the field is down, you can start to > > > envision the standing wave patterns moving across the field, even see > > > how the stocks in one section are not even going down the right > > > direction, but the opposite direction, so instead of helping, they are > > > hindering. Anway, any allegory that helps intuitive understanding has > > > some value, look at what Tesla came up with after watching ??, which > > > was an imperfect allegory, too. He came up with the induction motor. > > > > 'Stocks'? I think you mean 'Stalks' > > Another model shot!
Was she a stunning beauty, and what camera did you use? ;-) -- You can't have a sense of humor, if you have no sense.
On Apr 17, 9:52=A0pm, "Tim Williams" <tmoran...@gmail.com> wrote:
> "Jamie M" <jmor...@shaw.ca> wrote in message > > news:jml88f$a9n$1@speranza.aioe.org... > > > Does Barkhausen noise stop above a certain frequency, maybe > > proportional to the magnetic domain sizes? > > Probably on par with the domain relaxation time, which depends on materia=
l;
> most likely this corresponds to the cutoff frequency as well. > > > Also for an extreme > > example, for AC core losses, normally they are thought of as caused by > > eddy currents, but if the frequency is (extremely!) high enough, maybe > > the eddy current losses will start reducing, but will the transformer > > core also start to lose its inductance properties so that it is not > > useful? > > More or less. =A0If you model core losses entirely as eddy currents (igno=
ring
> hysteresis or material properties), then the cutoff frequency corresponds=
to
> the skin depth in the material: when the skin is shallower than the core > thickness, less core is utilized and the effective permeability drops. =
=A0Skin
> effect, in turn, occurs when the current flow produces a field opposing t=
he
> applied field. =A0If you can reduce the conductor dimensions (stacks of i=
ron
> sheets, or powder), or use a higher resistivity material (by adding silic=
on
> to the steel, or using a ferrite), the cutoff frequency rises again. > > Permeability has the same frequency response / amplitude / phase interpla=
y
> that any other filter does, so it should be no surprise that, if normal > (real) permeability results in an inductive core, a lossy (phase shifted, > complex or imaginary) permeability results in a resistive core, and that > resistive phase shift necessarily occurs over a frequency range where > permeability is dropping with rising frequency. =A0Indeed, all core mater=
ials
> I've seen look almost entirely resistive at or past some frequency. > > Some graphs:http://www.ferroxcube.com/prod/assets/3c90.pdf > 3C90 is a typical MnZn ferrite used for generic power applications. > Essentially equivalent to Fair-Rite #77, and I think Epcos/TDK N87 and > Magnetics Inc. type P, though I should check those. =A0The graph of compl=
ex
> permeability vs. frequency shows that total permeability is essentially > constant and real up to 1MHz or so, where the imaginary component is gett=
ing
> noticable. =A0(Typically, you can operate a ferrite near saturation up to > frequencies where mu'' ~ mu' / 100, in this case about 200kHz; at higher > frequencies, losses are higher and you have to reduce flux density to kee=
p
> it from overheating). > > If you recall |mu| =3D sqrt((mu')^2 + (mu'')^2), i.e. the vector sum of r=
eal
> and imaginary components, you might notice that, at the point the curves > intersect, the value of each is about 70% the flat range (about 1/sqrt(2)=
).
> This means, at the intersection, the value of |mu| doesn't actually dip. =
=A0As
> a result, |mu| actually remains fairly constant up to almost 4MHz. =A0The > implication is, the magnetizing impedance of an inductor or transformer > remains essentially inductive from DC to 4MHz, though the phase shifts > substantially past 1MHz. =A0Beyond 4MHz, the impedance remains flat or dr=
ops
> further, because |mu| also drops. > > In fact, the roll-off can be calculated from the graph. =A0The tail drops=
at a
> rate of about 2 decades in mu for one decade in F, so the impedance Z =3D > j*2*pi*F*mu drops off inversely with frequency past this point. =A0This i=
s a
> second order (-20dB/decade) drop, which is physically significant: if it > were simply resistive, it would be 10dB/decade instead. > > Ferrite beads graph this more directly. =A0See for example:http://search.=
murata.co.jp/Ceramy/image/img/PDF/ENG/L0110S0100BLM31P.pdf
> Page 2 shows typical |Z| vs. F, and the components vs. F for each individ=
ual
> part on the following page. =A0Now, these are components with metal burie=
d in
> the ferrite, so the effective terminal capacitance is significant as well=
;
> one would certainly expect the impedance to drop at very high frequencies=
,
> and indeed it does. =A0It turns out ferrite has a high dielectric constan=
t
> (which, because of the resistivity, is obviously a complex number as well=
),
> so this is significant even for very small ferrite chips. > > Another material of note, Fair-Rite #43 (I'd guess at some equivalents, b=
ut
> it's actually kind of funky so I won't go out on a limb here), which is > typically used for high frequency transformers and ferrite beads.http://w=
ww.fair-rite.com/newfair/materials43.htm
> You can see the "resistive band" is much wider, roughly 2MHz and up. =A0I=
t'll
> still make a good transformer at 20MHz, but it will be very lossy. > > Something else pertinent to draw from these graphs: mu vs. temp is provid=
ed.
> You can see there's a sort of "stochastic resonance" approaching Curie > point, which means the domains are getting easier to align -- more therma=
l
> energy makes more flips, which reduces the number of spins available to u=
se
> (saturation flux density is reduced -- see the B-H curves at different > temperatures), but they're easier to use. > > FYI, "stochastic resonance" is a big phrase for, if a system has a dead b=
and
> or hysteresis, adding a low-level stimulus (be it a coherent waveform or > random noise) actually improves the noise floor of the system. =A0For exa=
mple,
> dithering and averaging an ADC can yield more bits accuracy. =A0Strictly, > "stochastic" only applies to systems where random noise is used. > > Tim > > -- > Deep Friar: a very philosophical monk. > Website:http://webpages.charter.net/dawill/tmoranwms
On Apr 17, 9:52=A0pm, "Tim Williams" <tmoran...@gmail.com> wrote:
> "Jamie M" <jmor...@shaw.ca> wrote in message > > news:jml88f$a9n$1@speranza.aioe.org... > > > Does Barkhausen noise stop above a certain frequency, maybe > > proportional to the magnetic domain sizes? > > Probably on par with the domain relaxation time, which depends on materia=
l;
> most likely this corresponds to the cutoff frequency as well. > > > Also for an extreme > > example, for AC core losses, normally they are thought of as caused by > > eddy currents, but if the frequency is (extremely!) high enough, maybe > > the eddy current losses will start reducing, but will the transformer > > core also start to lose its inductance properties so that it is not > > useful? > > More or less. =A0If you model core losses entirely as eddy currents (igno=
ring
> hysteresis or material properties), then the cutoff frequency corresponds=
to
> the skin depth in the material: when the skin is shallower than the core > thickness, less core is utilized and the effective permeability drops. =
=A0Skin
> effect, in turn, occurs when the current flow produces a field opposing t=
he
> applied field. =A0If you can reduce the conductor dimensions (stacks of i=
ron
> sheets, or powder), or use a higher resistivity material (by adding silic=
on
> to the steel, or using a ferrite), the cutoff frequency rises again. > > Permeability has the same frequency response / amplitude / phase interpla=
y
> that any other filter does, so it should be no surprise that, if normal > (real) permeability results in an inductive core, a lossy (phase shifted, > complex or imaginary) permeability results in a resistive core, and that > resistive phase shift necessarily occurs over a frequency range where > permeability is dropping with rising frequency. =A0Indeed, all core mater=
ials
> I've seen look almost entirely resistive at or past some frequency. > > Some graphs:http://www.ferroxcube.com/prod/assets/3c90.pdf > 3C90 is a typical MnZn ferrite used for generic power applications. > Essentially equivalent to Fair-Rite #77, and I think Epcos/TDK N87 and > Magnetics Inc. type P, though I should check those. =A0The graph of compl=
ex
> permeability vs. frequency shows that total permeability is essentially > constant and real up to 1MHz or so, where the imaginary component is gett=
ing
> noticable. =A0(Typically, you can operate a ferrite near saturation up to > frequencies where mu'' ~ mu' / 100, in this case about 200kHz; at higher > frequencies, losses are higher and you have to reduce flux density to kee=
p
> it from overheating). > > If you recall |mu| =3D sqrt((mu')^2 + (mu'')^2), i.e. the vector sum of r=
eal
> and imaginary components, you might notice that, at the point the curves > intersect, the value of each is about 70% the flat range (about 1/sqrt(2)=
).
> This means, at the intersection, the value of |mu| doesn't actually dip. =
=A0As
> a result, |mu| actually remains fairly constant up to almost 4MHz. =A0The > implication is, the magnetizing impedance of an inductor or transformer > remains essentially inductive from DC to 4MHz, though the phase shifts > substantially past 1MHz. =A0Beyond 4MHz, the impedance remains flat or dr=
ops
> further, because |mu| also drops. > > In fact, the roll-off can be calculated from the graph. =A0The tail drops=
at a
> rate of about 2 decades in mu for one decade in F, so the impedance Z =3D > j*2*pi*F*mu drops off inversely with frequency past this point. =A0This i=
s a
> second order (-20dB/decade) drop, which is physically significant: if it > were simply resistive, it would be 10dB/decade instead. > > Ferrite beads graph this more directly. =A0See for example:http://search.=
murata.co.jp/Ceramy/image/img/PDF/ENG/L0110S0100BLM31P.pdf
> Page 2 shows typical |Z| vs. F, and the components vs. F for each individ=
ual
> part on the following page. =A0Now, these are components with metal burie=
d in
> the ferrite, so the effective terminal capacitance is significant as well=
;
> one would certainly expect the impedance to drop at very high frequencies=
,
> and indeed it does. =A0It turns out ferrite has a high dielectric constan=
t
> (which, because of the resistivity, is obviously a complex number as well=
),
> so this is significant even for very small ferrite chips. > > Another material of note, Fair-Rite #43 (I'd guess at some equivalents, b=
ut
> it's actually kind of funky so I won't go out on a limb here), which is > typically used for high frequency transformers and ferrite beads.http://w=
ww.fair-rite.com/newfair/materials43.htm
> You can see the "resistive band" is much wider, roughly 2MHz and up. =A0I=
t'll
> still make a good transformer at 20MHz, but it will be very lossy. > > Something else pertinent to draw from these graphs: mu vs. temp is provid=
ed.
> You can see there's a sort of "stochastic resonance" approaching Curie > point, which means the domains are getting easier to align -- more therma=
l
> energy makes more flips, which reduces the number of spins available to u=
se
> (saturation flux density is reduced -- see the B-H curves at different > temperatures), but they're easier to use. > > FYI, "stochastic resonance" is a big phrase for, if a system has a dead b=
and
> or hysteresis, adding a low-level stimulus (be it a coherent waveform or > random noise) actually improves the noise floor of the system. =A0For exa=
mple,
> dithering and averaging an ADC can yield more bits accuracy. =A0Strictly, > "stochastic" only applies to systems where random noise is used. > > Tim > > -- > Deep Friar: a very philosophical monk. > Website:http://webpages.charter.net/dawill/tmoranwms
trigger happy google! Anyway, after working with metglas in the micron ranges, I've come to the conclusion that many magnetic material data sheets are STILL showing macro-effects. Refer to the 3C90 data sheet. The data sheet gives information that is important to you the way you use the core material but does NOT reflect the true basic nature of magnetic materials. What has happened is that from reading data sheets, engineers have gotten a mindset of how magnetism works but that mindset is misleading both as one starts to scale smaller AND if one tries to use that mindset's understanding to solve some kind of performance issue. Instead of really solving a problem, you end up optimizing a 'weak' solution. It's just that when you see losses and permeability roll off for a material, much of those losses and much of that rolloff can be attributed to conductivity and eddy currents minimizing the fields destroying effective permeability, NOT the true nature of the material, but rather its gross effectc because of how it's used. Again, true these parameters are important to the designer, BECAUSE that is the component they're working with, but do not really reflect the nature of the basic material. For example, every one seems to accept that high permeability material rolls off above some low frequency, like 1MHz, NOT TRUE. the EFFECTIVE permeability rolls off above 1MHz because of the way the material has been configured into the form you're using it! *IF* you can gain the luxury of restructuring how you use the material you will see most materials have extremely high permeability above 100MHz Sadly, at around 1-2GHz magnetic material is gone due to the moment of inertia of the magnetic molecule.
On 04/18/2012 12:52 AM, Tim Williams wrote:
> "Jamie M" <jmorken@shaw.ca> wrote in message > news:jml88f$a9n$1@speranza.aioe.org... >> Does Barkhausen noise stop above a certain frequency, maybe >> proportional to the magnetic domain sizes? > > Probably on par with the domain relaxation time, which depends on > material; most likely this corresponds to the cutoff frequency as well. > >> Also for an extreme >> example, for AC core losses, normally they are thought of as caused by >> eddy currents, but if the frequency is (extremely!) high enough, maybe >> the eddy current losses will start reducing, but will the transformer >> core also start to lose its inductance properties so that it is not >> useful? > > More or less. If you model core losses entirely as eddy currents > (ignoring hysteresis or material properties), then the cutoff frequency > corresponds to the skin depth in the material: when the skin is > shallower than the core thickness, less core is utilized and the > effective permeability drops. Skin effect, in turn, occurs when the > current flow produces a field opposing the applied field. If you can > reduce the conductor dimensions (stacks of iron sheets, or powder), or > use a higher resistivity material (by adding silicon to the steel, or > using a ferrite), the cutoff frequency rises again. > > Permeability has the same frequency response / amplitude / phase > interplay that any other filter does, so it should be no surprise that, > if normal (real) permeability results in an inductive core, a lossy > (phase shifted, complex or imaginary) permeability results in a > resistive core, and that resistive phase shift necessarily occurs over a > frequency range where permeability is dropping with rising frequency. > Indeed, all core materials I've seen look almost entirely resistive at > or past some frequency. > > Some graphs: > http://www.ferroxcube.com/prod/assets/3c90.pdf > 3C90 is a typical MnZn ferrite used for generic power applications. > Essentially equivalent to Fair-Rite #77, and I think Epcos/TDK N87 and > Magnetics Inc. type P, though I should check those. The graph of complex > permeability vs. frequency shows that total permeability is essentially > constant and real up to 1MHz or so, where the imaginary component is > getting noticable. (Typically, you can operate a ferrite near saturation > up to frequencies where mu'' ~ mu' / 100, in this case about 200kHz; at > higher frequencies, losses are higher and you have to reduce flux > density to keep it from overheating). > > If you recall |mu| = sqrt((mu')^2 + (mu'')^2), i.e. the vector sum of > real and imaginary components, you might notice that, at the point the > curves intersect, the value of each is about 70% the flat range (about > 1/sqrt(2)). This means, at the intersection, the value of |mu| doesn't > actually dip. As a result, |mu| actually remains fairly constant up to > almost 4MHz. The implication is, the magnetizing impedance of an > inductor or transformer remains essentially inductive from DC to 4MHz, > though the phase shifts substantially past 1MHz. Beyond 4MHz, the > impedance remains flat or drops further, because |mu| also drops. > > In fact, the roll-off can be calculated from the graph. The tail drops > at a rate of about 2 decades in mu for one decade in F, so the impedance > Z = j*2*pi*F*mu drops off inversely with frequency past this point. This > is a second order (-20dB/decade) drop, which is physically significant: > if it were simply resistive, it would be 10dB/decade instead. > > Ferrite beads graph this more directly. See for example: > http://search.murata.co.jp/Ceramy/image/img/PDF/ENG/L0110S0100BLM31P.pdf > Page 2 shows typical |Z| vs. F, and the components vs. F for each > individual part on the following page. Now, these are components with > metal buried in the ferrite, so the effective terminal capacitance is > significant as well; one would certainly expect the impedance to drop at > very high frequencies, and indeed it does. It turns out ferrite has a > high dielectric constant (which, because of the resistivity, is > obviously a complex number as well), so this is significant even for > very small ferrite chips. > > Another material of note, Fair-Rite #43 (I'd guess at some equivalents, > but it's actually kind of funky so I won't go out on a limb here), which > is typically used for high frequency transformers and ferrite beads. > http://www.fair-rite.com/newfair/materials43.htm > You can see the "resistive band" is much wider, roughly 2MHz and up. > It'll still make a good transformer at 20MHz, but it will be very lossy. > > Something else pertinent to draw from these graphs: mu vs. temp is > provided. You can see there's a sort of "stochastic resonance" > approaching Curie point, which means the domains are getting easier to > align -- more thermal energy makes more flips, which reduces the number > of spins available to use (saturation flux density is reduced -- see the > B-H curves at different temperatures), but they're easier to use. > > FYI, "stochastic resonance" is a big phrase for, if a system has a dead > band or hysteresis, adding a low-level stimulus (be it a coherent > waveform or random noise) actually improves the noise floor of the > system. For example, dithering and averaging an ADC can yield more bits > accuracy. Strictly, "stochastic" only applies to systems where random > noise is used. > > Tim >
No, no, Tim, you don't understand science at all. ;) Terms like "stochastic resonance" are beautifully designed to help folks like these http://www.santafe.edu/research/publications/sfi-bibliography/detail/?id=126 extract $$ from the feds. What's more scientific than that? Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net