Core selection

[Tim Williams]

Question the first: where to find transformer (or inductor) cores?
Mouser and Digikey don't sell them plain. Who's the magnetic
equivalent of Digikey?

Other question: how to select cores? Specifically, I'm looking for
something in the 10kHz (square wave, PWM), 10kVA range, with high
permeability to minimize the number of turns. Saturation should be
fairly high, given the low operating frequency. Ratio 4:1, with about
200Vrms (fundamental component) primary and 50Vrms secondary. The
secondary will be copper tubing, 1/4 or 3/8" dia., so a one-turn
secondary is advantageous.

Given the permeability and saturation, I'm thinking something like
Metglas. I know I need something in the 2-3" range, and I can
calculate things like A_L and A*t(sat) from the properties, but it
would be a whole lot easier if I had both parameters laid out in a
table of standard shapes. I can't really make any estimates on what
size I need if I don't have a standard formula for the geometry, so I
don't even know what inductance and magnetizing current and saturation
I have to look for.

And why do they never specify amp-turns saturation? It's always in
B. I can measure amp turns, I can't measure Teslas. They give A_L by
the core, but not saturation, what the hell?

Tim

 

[Jan Panteltje]

Question the first: where to find transformer (or inductor) cores?
Mouser and Digikey don't sell them plain. Who's the magnetic
equivalent of Digikey?

http://www.kitsandparts.com/

 

[James Arthur]

Question the first: where to find transformer (or inductor) cores?
Mouser and Digikey don't sell them plain. Who's the magnetic
equivalent of Digikey?

Other question: how to select cores? Specifically, I'm looking for
something in the 10kHz (square wave, PWM), 10kVA range, with high
permeability to minimize the number of turns. Saturation should be
fairly high, given the low operating frequency. Ratio 4:1, with about
200Vrms (fundamental component) primary and 50Vrms secondary. The
secondary will be copper tubing, 1/4 or 3/8" dia., so a one-turn
secondary is advantageous.

Given the permeability and saturation, I'm thinking something like
Metglas. I know I need something in the 2-3" range, and I can
calculate things like A_L and A*t(sat) from the properties, but it
would be a whole lot easier if I had both parameters laid out in a
table of standard shapes. I can't really make any estimates on what
size I need if I don't have a standard formula for the geometry, so I
don't even know what inductance and magnetizing current and saturation
I have to look for.

And why do they never specify amp-turns saturation? It's always in
B. I can measure amp turns, I can't measure Teslas. They give A_L by
the core, but not saturation, what the hell?

Tim

https://www.amidoncorp.com/ (ham and hobbyist-friendly)

http://www.cwsbytemark.com/


HTH,
James Arthur

 

[Winfield Hill]

Question the first: where to find transformer (or inductor) cores?
Mouser and Digikey don't sell them plain.  Who's the magnetic
equivalent of Digikey?

I get most of my cores from Eastern Components, they're an
industrial supplier. http://www.eastern-components.com/
I think they will sell small quantities - you can see.

A place that does sell small quantities, but with more limited
selection, is Amidon. http://www.amidon.com/

 

[Tim Williams]

https://www.amidoncorp.com/ (ham and hobbyist-friendly)

At a glance, they don't seem to offer anything suitable. Given the
poor layout, I could be missing it by a mile...

http://www.cwsbytemark.com/

Just as vague a layout, too. Geez, is it traditional for magnetics
companies to have terse, cryptic layouts? And,

I get most of my cores from Eastern Components, they're an
industrial supplier. http://www.eastern-components.com/
I think they will sell small quantities - you can see.

This website doesn't even appear to have a catalog. So, I'm better
off browsing the manufacturers' websites themselves (which appears to
be all the information on Eastern's website), looking for a material
that *I think* is right, and then what? I can't even tell what they
stock. I could call them, but geez, this is 2009, everything is
supposed to be online!

Right now I'm eyeing the F-290-W at CWS, which tells me that it's a
2.90" o.d. toroid (and there's some other stuff in its row),
and...**** all else.

If it's an average u = 2000 up to B = 0.5T, then H = 0.5 / (u_o*2000)
= .5/2000*pi*4e-7 = 199 A/m. The average path length is ((2.9 +
1.53) / 2) * pi = 7.0 inches, or 0.177 m. So evidently, saturation
might be around 199 * 0.177 = 35.2 At. Sounds like a typical ferrite
toroid. If that's the case, then for one turn, I need I_m(max) <
35Apk, or under 25Arms. At 50Vrms, that's about 2 ohms X_L, which at
10kHz is about 32uH. Claimed A_L is 16,280mH/T^2, which is so full of
shit that it's obvious they meant nH/T^2, so it's actually 16.3uH/
T^2. So I'd need 2 turns to get the minimum inductance. But then At
would double, so I need double the inductance, so I actually need
2*sqrt(2) turns to be on the edge of saturation. Call it 4 for some
margin. Or I could stack a pair of these cores with one turn down the
center.

So, is that right?

God, it's no wonder people are afraid of inductors, they sure try hard
making them impossible to use. Seems to me there could be big
business simply setting up a website that actually *tells* you
things! Parametric search, even (including A*t saturation)!

Tim

 

[[email protected]]

Question the first: where to find transformer (or inductor) cores?
Mouser and Digikey don't sell them plain.  Who's the magnetic
equivalent of Digikey?

Other question: how to select cores?  Specifically, I'm looking for
something in the 10kHz (square wave, PWM), 10kVA range, with high
permeability to minimize the number of turns.  Saturation should be
fairly high, given the low operating frequency.  Ratio 4:1, with about
200Vrms (fundamental component) primary and 50Vrms secondary.  The
secondary will be copper tubing, 1/4 or 3/8" dia., so a one-turn
secondary is advantageous.

Given the permeability and saturation, I'm thinking something like
Metglas.  I know I need something in the 2-3" range, and I can
calculate things like A_L and A*t(sat) from the properties, but it
would be a whole lot easier if I had both parameters laid out in a
table of standard shapes.  I can't really make any estimates on what
size I need if I don't have a standard formula for the geometry, so I
don't even know what inductance and magnetizing current and saturation
I have to look for.

And why do they never specify amp-turns saturation?  It's always in
B.  I can measure amp turns, I can't measure Teslas.  They give A_L by
the core, but not saturation, what the hell?

Tim

Hello Tim,

You said 50Vrms at 10 kHz, 1 turn. This means you need peak flux = 1.2
mT. So using a ferrite up to B=200mT, you need a cross section of
about 1.2m/0.2 = 0.006 m^2. That is a square with 3" sides. I never
saw ferrite cores with that large cross section.

Probably you have to increase the number of secondary turns and/or use
a low loss material that has higher Bsat (so you can design at higher
peak flux density). This will result in less cross section for the
magnetic path.

Given the power, a complete electrical / thermal design is required
when the duty cycle of this transformer will be high.

Best regards,

Wim
PA3DJS
www.tetech.nl

 

[Hammy]

http://www.kitsandparts.com/

Newark (farnell) has them you can buy in small quantities. There is a
ten to fifteen day delay unless you live in the EU but the standard
$12.00 shipping charge still applies. They have all the hardware
bobbins clips etc as well.

Here the search page or it might take a while to find them.

http://canada.newark.com/jsp/search/results.jsp?N=511+1001332&isGoback=false

 

[Spehro Pefhany]

Newark (farnell) has them you can buy in small quantities. There is a
ten to fifteen day delay unless you live in the EU but the standard
$12.00 shipping charge still applies. They have all the hardware
bobbins clips etc as well.

Here the search page or it might take a while to find them.

http://canada.newark.com/jsp/search/results.jsp?N=511+1001332&isGoback=false

Thanks, Hammy!


Best regards,
Spehro Pefhany

 

[Tim Williams]

Newark (farnell) has them you can buy in small quantities. There is a
ten to fifteen day delay unless you live in the EU but the standard
$12.00 shipping charge still applies. They have all the hardware
bobbins clips etc as well.

Here the search page or it might take a while to find them.

http://canada.newark.com/jsp/search/results.jsp?N=511+1001332&isGobac....

Ah yes, they have some, and I recall Allied has similar products. But
that's EMI stuff, I need bigger. The biggest toroid I see there is
about an inch across- I'd need a lot of those.

Surprisingly, the Fair-Rite website seems to be fairly new. The
scripting runs really slow though. At least it contains information,
for once!

Tim

 

[legg]

Hello Tim,

You said 50Vrms at 10 kHz, 1 turn. This means you need peak flux = 1.2
mT. So using a ferrite up to B=200mT, you need a cross section of
about 1.2m/0.2 = 0.006 m^2. That is a square with 3" sides. I never
saw ferrite cores with that large cross section.

Probably you have to increase the number of secondary turns and/or use
a low loss material that has higher Bsat (so you can design at higher
peak flux density). This will result in less cross section for the
magnetic path.

Given the power, a complete electrical / thermal design is required
when the duty cycle of this transformer will be high.

Best regards,

Wim
PA3DJS
www.tetech.nl

Or putting it more generally, for fully reversing AC flux,

Ae = V / (2.f.n.Bpk)

Ae = core cross-section in square meters
V = average volts applied
n = number of turns
Bpk = permitted peak flux determined for core loss limited
or saturation-limited design.

RL

 

[James Arthur]

At a glance, they don't seem to offer anything suitable. Given the
poor layout, I could be missing it by a mile...

They offer a range of E-cores, toroids, in a variety of materials.

I assumed you'd want a mess of ferrite E-cores in mat'l #77--you'll
have to stack 'em to get to 10kVA.

https://www.amidoncorp.com/items/65

Just as vague a layout, too. Geez, is it traditional for magnetics
companies to have terse, cryptic layouts? And,



This website doesn't even appear to have a catalog. So, I'm better
off browsing the manufacturers' websites themselves (which appears to
be all the information on Eastern's website), looking for a material
that *I think* is right, and then what? I can't even tell what they
stock. I could call them, but geez, this is 2009, everything is
supposed to be online!

Right now I'm eyeing the F-290-W at CWS, which tells me that it's a
2.90" o.d. toroid (and there's some other stuff in its row),
and...**** all else.

If it's an average u = 2000 up to B = 0.5T, then H = 0.5 / (u_o*2000)
= .5/2000*pi*4e-7 = 199 A/m. The average path length is ((2.9 +
1.53) / 2) * pi = 7.0 inches, or 0.177 m. So evidently, saturation
might be around 199 * 0.177 = 35.2 At. Sounds like a typical ferrite
toroid. If that's the case, then for one turn, I need I_m(max) <
35Apk, or under 25Arms. At 50Vrms, that's about 2 ohms X_L, which at
10kHz is about 32uH. Claimed A_L is 16,280mH/T^2, which is so full of
shit that it's obvious they meant nH/T^2, so it's actually 16.3uH/
T^2. So I'd need 2 turns to get the minimum inductance. But then At
would double, so I need double the inductance, so I actually need
2*sqrt(2) turns to be on the edge of saturation. Call it 4 for some
margin. Or I could stack a pair of these cores with one turn down the
center.

So, is that right?

God, it's no wonder people are afraid of inductors, they sure try hard
making them impossible to use. Seems to me there could be big
business simply setting up a website that actually *tells* you
things! Parametric search, even (including A*t saturation)!

Tim

http://www.mag-inc.com/software/inductor.asp

Cheers,
James Arthur

 

[legg]

Hello Tim,

You said 50Vrms at 10 kHz, 1 turn. This means you need peak flux = 1.2
mT. So using a ferrite up to B=200mT, you need a cross section of
about 1.2m/0.2 = 0.006 m^2. That is a square with 3" sides. I never
saw ferrite cores with that large cross section.

Probably you have to increase the number of secondary turns and/or use
a low loss material that has higher Bsat (so you can design at higher
peak flux density). This will result in less cross section for the
magnetic path.

Given the power, a complete electrical / thermal design is required
when the duty cycle of this transformer will be high.

Best regards,

Wim
PA3DJS
www.tetech.nl

Or putting it more generally, for fully reversing AC flux,

Ae = V / (2.f.n.Bpk)

Ae = core cross-section in square meters
V = average volts applied
n = number of turns
Bpk = permitted peak flux determined for core loss limited
or saturation-limited design.

RL

 

[Rich Grise]

Question the first: where to find transformer (or inductor) cores?

http://www.google.com/

Good Luck!
Rich

 

[Tim Williams]

They offer a range of E-cores, toroids, in a variety of materials.

I assumed you'd want a mess of ferrite E-cores in mat'l #77--you'll
have to stack 'em to get to 10kVA.

https://www.amidoncorp.com/items/65

#77 is starting to look like the material of choice. Or something
similar, like 75 or 78.

The largest E-core Amidon offers is rated for "about 200W", which
suggests I'd need roughly 50 of them for the 10kW level I'm interested
in.

On an indirectly linked page, I discovered the data:
https://www.amidoncorp.com/specs/2-40.pdf

This says the largest core has a winding window of 2 * 0.593 x 0.375
inch (using an E-E arrangement). A stack of 50 would be 50 * 0.605 =
30" thick, which is certainly possible, but would stick out one side
of my chassis. On the plus side, I would certainly be able to push
all the voltage through one turn. A single piece of 3/8" tubing would
fit without too much trouble, though leakage inductance to the primary
wouldn't be great (though it doesn't need to be). Evidently, A_L
would be 5.3 * 50 = 265uH/T^2, which would be fairly "ideal". But it
seems like an awful lot of overkill, not to mention way too expensive
($312 for 50 E-cores? no thanks).

Where does cross sectional area fit into this, anyway? Isn't that
absorbed into A_L? So, as long as I am given A_L, I can calculate
inductance and saturation at will? And saturation only involves path
length, right? -- by amperes per meter, they mean *A/m*, not A.m/m^2
(like how resistivity is actually ohm.m^2/m)?

Ok, so, this is Usenet, right? If I've made an error, surely there
would have been fifty people in the first hour telling me what an
idiot I am -- since this has not happened, I can only assume my
calculations are correct??? Then why do I calculate that a moderately
sized toroid (like the FT-290-W) will suffice, whereas others have
suggested that I need something approximately as thick as my ankle?

Tim

 

[[email protected]]

#77 is starting to look like the material of choice.  Or something
similar, like 75 or 78.

The largest E-core Amidon offers is rated for "about 200W", which
suggests I'd need roughly 50 of them for the 10kW level I'm interested
in.

On an indirectly linked page, I discovered the data:https://www.amidoncorp.com/specs/2-40.pdf

This says the largest core has a winding window of 2 * 0.593 x 0.375
inch (using an E-E arrangement).  A stack of 50 would be 50 * 0.605 =
30" thick, which is certainly possible, but would stick out one side
of my chassis.  On the plus side, I would certainly be able to push
all the voltage through one turn.  A single piece of 3/8" tubing would
fit without too much trouble, though leakage inductance to the primary
wouldn't be great (though it doesn't need to be).  Evidently, A_L
would be 5.3 * 50 = 265uH/T^2, which would be fairly "ideal".  But it
seems like an awful lot of overkill, not to mention way too expensive
($312 for 50 E-cores?  no thanks).

Where does cross sectional area fit into this, anyway?  Isn't that
absorbed into A_L?  So, as long as I am given A_L, I can calculate
inductance and saturation at will?  And saturation only involves path
length, right? -- by amperes per meter, they mean *A/m*, not A.m/m^2
(like how resistivity is actually ohm.m^2/m)?

Ok, so, this is Usenet, right?  If I've made an error, surely there
would have been fifty people in the first hour telling me what an
idiot I am -- since this has not happened, I can only assume my
calculations are correct???  Then why do I calculate that a moderately
sized toroid (like the FT-290-W) will suffice, whereas others have
suggested that I need something approximately as thick as my ankle?

Tim

Hello Tim,

Regarding the Usenet. Be happy that there are also polite people that
don't start to roar immediately. In addition it can also be that
others are not competent enough to tell you whether you are right or
not, or don't have the time to do the complete math. Magnetics is
mysterious in the eyes of many people.

I belong to the people that think that for a 1Turn secondary you need
a large cross section to avoid core saturation. Designing transformers
is balancing between core losses, copper losses, availability of
materials, money, etc.

Best regards,

Wim
PA3DJS
www.tetech.nl
please remove first three letters of alphabet in case of PM.

 

[Fred Bartoli]

Rich Grise a écrit :

http://www.google.com/

Good Luck!
Rich

At least you didn't suggested a PIC!

 

[Tim Williams]

Rich Grise a écrit :



At least you didn't suggested a PIC!

Damn, I like that. You'll, or no wait, Jan will have to send me
some PICs. Then I can put them across this transformer's secondary
and see what happens. >:-D

Tim

 

[Nico Coesel]

#77 is starting to look like the material of choice. Or something
similar, like 75 or 78.

The largest E-core Amidon offers is rated for "about 200W", which
suggests I'd need roughly 50 of them for the 10kW level I'm interested
in.

On an indirectly linked page, I discovered the data:
https://www.amidoncorp.com/specs/2-40.pdf

This says the largest core has a winding window of 2 * 0.593 x 0.375
inch (using an E-E arrangement). A stack of 50 would be 50 * 0.605 =3D
30" thick, which is certainly possible, but would stick out one side
of my chassis. On the plus side, I would certainly be able to push
all the voltage through one turn. A single piece of 3/8" tubing would
fit without too much trouble, though leakage inductance to the primary
wouldn't be great (though it doesn't need to be). Evidently, A_L
would be 5.3 * 50 =3D 265uH/T^2, which would be fairly "ideal". But it
seems like an awful lot of overkill, not to mention way too expensive
($312 for 50 E-cores? no thanks).

Where does cross sectional area fit into this, anyway? Isn't that
absorbed into A_L? So, as long as I am given A_L, I can calculate
inductance and saturation at will? And saturation only involves path
length, right? -- by amperes per meter, they mean *A/m*, not A.m/m^2
(like how resistivity is actually ohm.m^2/m)?

Ok, so, this is Usenet, right? If I've made an error, surely there
would have been fifty people in the first hour telling me what an
idiot I am -- since this has not happened, I can only assume my

Perhaps nobody has experience in this field? I tried the same thing
you are doing for low power stuff. As someone already typed before:
there are a huge amount of units which need to be converted and each
vendor seems to have their own specification method. Its a world of
pain.

That leads to the question: why do you want to wind your own
transformer? It might be much easier to have a transformer wound which
meets your specs than failing a couple of times. At the power levels
you are talking about, a failure is likely to cause collateral damage.

If you are serious about winding your own transformer I think you
might need to buy some cores first put some windings on them and
verify your calculations. Beware that the method of winding also
influences the behaviour of the transformer.

 

[Tim Williams]

That leads to the question: why do you want to wind your own

transformer? It might be much easier to have a transformer wound which
meets your specs than failing a couple of times. At the power levels
you are talking about, a failure is likely to cause collateral damage.

I'm not too concerned about that (which, you're right to observe,
seems kind of odd given the ten kilowattedness of my ambition). I've
done this plenty of times before- junk ferrites are plentiful, but
small. The biggest cores I have are probably worth 2kVA together (and
that at 100kHz). Testing saturation and inductivity is easy, making
transformers is easy (especially when they have only ten turns). I
really just need the core to do it.

Getting a transformer wound feels like a huge waste of resources,
seeing as I just need the core (whichever size it has to be). I don't
need someone else to design it (assuming I get the correct core), and
I don't need someone else to wind it -- do winders even do 1/4 or 3/8"
copper tubing? -- I can do all that myself, no need to quadruple the
price of this project on services. Consider, BTW, that I am of the
age group where "money is short and time is free". ;-)

If you are serious about winding your own transformer I think you
might need to buy some cores first put some windings on them and
verify your calculations. Beware that the method of winding also
influences the behaviour of the transformer.

That should also not be a problem, about five turns of copper strap
around the toroid with one pipe down the center would be pretty fine I
think. If I need several turns secondary, I can either wrestle the
pipe through, or cut it into segments and fix it back together with
fittings (compression or flare, I'm thinking). The extra connections
won't do any favors for conductivity, but oh well.

Whereas, I suppose if I bunched all the primary turns to one side of
the toroid -- which is tempting, stiff as that copper strap is -- that
side would be more prone to saturation, which might produce
unfavorable results (sooner saturation, higher LL, etc.?).

Tim

 

[James Arthur]

#77 is starting to look like the material of choice. Or something
similar, like 75 or 78.

The largest E-core Amidon offers is rated for "about 200W", which
suggests I'd need roughly 50 of them for the 10kW level I'm interested
in.

On an indirectly linked page, I discovered the data:
https://www.amidoncorp.com/specs/2-40.pdf

This says the largest core has a winding window of 2 * 0.593 x 0.375
inch (using an E-E arrangement). A stack of 50 would be 50 * 0.605 =
30" thick, which is certainly possible, but would stick out one side
of my chassis. On the plus side, I would certainly be able to push
all the voltage through one turn. A single piece of 3/8" tubing would
fit without too much trouble, though leakage inductance to the primary
wouldn't be great (though it doesn't need to be). Evidently, A_L
would be 5.3 * 50 = 265uH/T^2, which would be fairly "ideal". But it
seems like an awful lot of overkill, not to mention way too expensive
($312 for 50 E-cores? no thanks).

Where does cross sectional area fit into this, anyway? Isn't that
absorbed into A_L? So, as long as I am given A_L, I can calculate
inductance and saturation at will? And saturation only involves path
length, right? -- by amperes per meter, they mean *A/m*, not A.m/m^2
(like how resistivity is actually ohm.m^2/m)?

Ok, so, this is Usenet, right? If I've made an error, surely there
would have been fifty people in the first hour telling me what an
idiot I am -- since this has not happened, I can only assume my
calculations are correct??? Then why do I calculate that a moderately
sized toroid (like the FT-290-W) will suffice, whereas others have
suggested that I need something approximately as thick as my ankle?

Tim

No it's not safe to conclude your calcs are right--I didn't
check them.

Here's Terry Given's take on a similar app, complete with
worked out examples and wisdom of the ages:

http://groups.google.com/group/sci....hread/74808f35486301d9/6ae3c6dec198792a?hl=en

HTH,
James Arthur