Smps Ferrites, Frequency and Loss

[D from BC]

Ugh...I have to change a non-isolated 700khz Cuk converter into an
isolated version.
Safety reasons.
Currently, the power inductors feel like 30C. Nice @~100W transfer.

I suspect, popping in an isolation transformer running at 700khz will
create a humongous drop in efficiency.

Transformer Stresses
Vprimary: 111Vrms (square wave, D~40%)
Iprimary: Positive peak: 0.7Amps, Negative peak: -1.2Amps
f: 700khz
Lp: 300uH and up
N ratio: 1:1
=one big ass hot ferrite transformer?

I'm trying to figure out the 'sweet spot' for the smallest transformer
size for a 20C temp rise.
I'm too newbie at magnetics for smps design. Should I drop f to 100khz
and redesign?

Or get crazy and make f higher and use an air core transformer.
But there's the crappy coupling of ~0.5, emissions and wire losses.


D from BC
British Columbia
Canada.

 

[Paul Mathews]

Ugh...I have to change a non-isolated 700khz Cuk converter into an
isolated version.
Safety reasons.
Currently, the power inductors feel like 30C. Nice @~100W transfer.

I suspect, popping in an isolation transformer running at 700khz will
create a humongous drop in efficiency.

Transformer Stresses
Vprimary: 111Vrms (square wave, D~40%)
Iprimary: Positive peak: 0.7Amps, Negative peak: -1.2Amps
f: 700khz
Lp: 300uH and up
N ratio: 1:1
=one big ass hot ferrite transformer?

I'm trying to figure out the 'sweet spot' for the smallest transformer
size for a 20C temp rise.
I'm too newbie at magnetics for smps design. Should I drop f to 100khz
and redesign?

Or get crazy and make f higher and use an air core transformer.
But there's the crappy coupling of ~0.5, emissions and wire losses.

D from BC
British Columbia
Canada.

I'm guessing you've already looked at Pressman pg. 242.

This may be of interest:
http://scholar.lib.vt.edu/theses/available/etd-4198-143730/unrestricted/Ch1.pdf

Take to investigate intellectual property rights if you pursue the
above approach.

700 kHz not really at the bleeding edge these days, although you'd
definitely be more comfortable an octave or 2 lower. You need to
design for flux levels considerably below saturation, since core
losses will tend to be the determining factor. In other words, the
core ends up being larger than you'd predict by scaling up from a
lower freq. Note, too, that core loss specs are usually stated in
watts/volume unit, so, for a given flux level, a larger core
dissipates more power. It makes an interesting design optimization
exercise, particularly when you factor in skin and proximity effects.
My experience has been that you're better off, unless you have to
achieve very high power density, going down in frequency to the region
where everything just fits in the available volume.
Paul Mathews

 

[D from BC]

I'm guessing you've already looked at Pressman pg. 242.

This may be of interest:
http://scholar.lib.vt.edu/theses/available/etd-4198-143730/unrestricted/Ch1.pdf

Take to investigate intellectual property rights if you pursue the
above approach.

700 kHz not really at the bleeding edge these days, although you'd
definitely be more comfortable an octave or 2 lower. You need to
design for flux levels considerably below saturation, since core
losses will tend to be the determining factor. In other words, the
core ends up being larger than you'd predict by scaling up from a
lower freq. Note, too, that core loss specs are usually stated in
watts/volume unit, so, for a given flux level, a larger core
dissipates more power. It makes an interesting design optimization
exercise, particularly when you factor in skin and proximity effects.
My experience has been that you're better off, unless you have to
achieve very high power density, going down in frequency to the region
where everything just fits in the available volume.
Paul Mathews

Nope never saw that book.. However, I did found it in Google books..
Yup..that's the one. P.242. Fig. 6.18
Perhaps the most insane topology to pick when beginning to learn smps
design...At my rate of understanding, the patents might expire by the
time I figure out the magnetics.

One time I spotted (Epcos) loss specs in terms of kW/m^3. ..Who's
using a cubic meter of core these days.. Why not Megawatts/cubic
kilometer Geez...
Expressed with mW/cm^3 is better when cores are way below 100 cubic
centimeters.


D from BC
British Columbia
Canada.

 

[legg]

One time I spotted (Epcos) loss specs in terms of kW/m^3. ..Who's
using a cubic meter of core these days.. Why not Megawatts/cubic
kilometer Geez...
Expressed with mW/cm^3 is better when cores are way below 100 cubic
centimeters.

Kw/m^3 = 1E6mW/ (1E2cm)^3 = mW/cm^3

RL

 

[D from BC]

Kw/m^3 = 1E6mW/ (1E2cm)^3 = mW/cm^3

RL

How about Terawatts/cubic kilometer..

1TW/km^3 = 1E15mW/1E15 cm^3 = mW/cm^3


D from BC
British Columbia
Canada.

 

[MooseFET]

How about Terawatts/cubic kilometer..

1TW/km^3 = 1E15mW/1E15 cm^3 = mW/cm^3

No this doesn't work. To make it scientistic you need to make all the
numbers in non-multiplied units. Thus I suggest that the only
reasonable units would be Watts per cubic light second.

 

[Eeyore]

Ugh...I have to change a non-isolated 700khz Cuk converter into an
isolated version.
Safety reasons.
Currently, the power inductors feel like 30C. Nice @~100W transfer.

I suspect, popping in an isolation transformer running at 700khz will
create a humongous drop in efficiency.

Transformer Stresses
Vprimary: 111Vrms (square wave, D~40%)
Iprimary: Positive peak: 0.7Amps, Negative peak: -1.2Amps
f: 700khz
Lp: 300uH and up
N ratio: 1:1
=one big ass hot ferrite transformer?

I'm trying to figure out the 'sweet spot' for the smallest transformer
size for a 20C temp rise.
I'm too newbie at magnetics for smps design. Should I drop f to 100khz
and redesign?

Or get crazy and make f higher and use an air core transformer.
But there's the crappy coupling of ~0.5, emissions and wire losses.

Check out Epcos's Ferrite Magnetic Designer tool. It'll do a lot of the
work for you.

Graham

 

[legg]

How about Terawatts/cubic kilometer..

1TW/km^3 = 1E15mW/1E15 cm^3 = mW/cm^3

Just so long as you can recognize what you're looking at.

RL

 

[Spehro Pefhany]

Just so long as you can recognize what you're looking at.

RL

Most can recognize TWatts
Best regards,
Spehro Pefhany

 

[D from BC]

Check out Epcos's Ferrite Magnetic Designer tool. It'll do a lot of the
work for you.

Graham

I recently tried the tool.
However, I don't know yet what Epcos distributor has a wide variety of
stock and can sell me a few cores.


D from BC
British Columbia
Canada.

 

[Frithiof Andreas Jensen]

I recently tried the tool.
However, I don't know yet what Epcos distributor has a wide variety of
stock and can sell me a few cores.

The magic words is an Evaluation Kit. Often the salesperson will send you
one for free - Coilcraft did for me recently ;-)

 

[John Fields]

Most can recognize TWatts

 

[legg]

Ugh...I have to change a non-isolated 700khz Cuk converter into an
isolated version.
Safety reasons.
Currently, the power inductors feel like 30C. Nice @~100W transfer.

I suspect, popping in an isolation transformer running at 700khz will
create a humongous drop in efficiency.

Transformer Stresses
Vprimary: 111Vrms (square wave, D~40%)
Iprimary: Positive peak: 0.7Amps, Negative peak: -1.2Amps
f: 700khz
Lp: 300uH and up
N ratio: 1:1
=one big ass hot ferrite transformer?

It might not be as bad as you think.

The isolation transformer in a Cuk has to be AC-coupled (input and
output chokes are in series - switch shows up across the primary of
the isolation transformer), and doesn't do any energy storage, You're
working with total reversal around zero.

This gives you a double flux swing for the same loss, as loss is
supposedly computed as being a product of peak flux, rather than pk-pk
swing.

It's a formula for a smaller transformer than might immediately be
expected, anyways.

RL

 

[legg]

On Wed, 05 Mar 2008 15:37:31 +0000, Eeyore


I recently tried the tool.
However, I don't know yet what Epcos distributor has a wide variety of
stock and can sell me a few cores.

Once you've got the tool, you can juggle roughly 'equivalent'
materials into your work. You don't want a design that requires a
'designer' ferrite material, anyways.

I don't recommend it (the tool) by the way. Too many non-obvious
assumptions are inherrent in the GUI.

RL

 

[legg]

Most can recognize TWatts
Best regards,
Spehro Pefhany

Better than being recognized as one.

RL

 

[Paul Mathews]

I recently tried the tool.
However, I don't know yet what Epcos distributor has a wide variety of
stock and can sell me a few cores.

D from BC
British Columbia
Canada.- Hide quoted text -

- Show quoted text -

If it's something I have, I'll send you a few for postage. I don't
know Canada situation, but I've had the best results stateside working
with Eastern Components. Other USA names of stocking distributors
include Allstar and Lodestone Pacific. Epcos is pretty slow to provide
samples, in my experience. However, their sample kits are second to
none. Possibly useful factoid: Bobbins can be harder to obtain than
cores, and coated toroids don't need bobbins.
Paul Mathews

 

[D from BC]

Better than being recognized as one.

RL

If Tera and Giga got married, what would the kids be called?
I'm working on a punch line.


D from BC
British Columbia
Canada.

 

[D from BC]

It might not be as bad as you think.

The isolation transformer in a Cuk has to be AC-coupled (input and
output chokes are in series - switch shows up across the primary of
the isolation transformer), and doesn't do any energy storage, You're
working with total reversal around zero.

This gives you a double flux swing for the same loss, as loss is
supposedly computed as being a product of peak flux, rather than pk-pk
swing.

It's a formula for a smaller transformer than might immediately be
expected, anyways.

RL

Yup..that's something I've been thinking about. The Cuk transformer is
not stuck in one quadrant of the B-H characteristic.

For example Material F (Amidon sells it.)
http://magneticsinc.com/pdf/2006_Ferrite_Catalog/2006_Materials.pdf

On the core loss graph is has:
"These curves are determined from AC data"
Great!...That's what I want..
The cores have been tested by sine wave drive with no offset.
(It's square wave in smps app so..some calc error.)

By the looks of it...I can get ~300mW/cm^3 if the AC flux density is
at 300 gauss (coincidence). f= 700khz.

If I got this right...
Using faradays law, this is around 13 turns on a 1cm^2 cross section.
(Erms=4.44*B*Ac*N*f*10E-8)

But I don't know the math between loss density,toroid size and temp
rise.

I'd hate to see how much damage thermal runaway to the Curie point can
do..


D from BC
British Columbia
Canada.

 

[D from BC]

If it's something I have, I'll send you a few for postage. I don't
know Canada situation, but I've had the best results stateside working
with Eastern Components. Other USA names of stocking distributors
include Allstar and Lodestone Pacific. Epcos is pretty slow to provide
samples, in my experience. However, their sample kits are second to
none. Possibly useful factoid: Bobbins can be harder to obtain than
cores, and coated toroids don't need bobbins.
Paul Mathews

Thanks... I'll keep it in mind.
I'll check out those distributors.. I haven't been to those sites yet.
I still have to poke at the math and specs for awhile until I figure
out what's ok..
Some shopping and comparing too.
I might pick the N49 material from Epcos. But don't know yet..


D from BC
British Columbia
Canada.

 

[legg]

But I don't know the math between loss density,toroid size and temp
rise.

I'd hate to see how much damage thermal runaway to the Curie point can
do..

Peak density at f gives core loss in mw/cm^3 to be picked off of the
mfr's core loss chart.

Temperature rise is 1 degree C per milliwatt per cm^2 surface area for
the shape of transformer chosen. (+/- 20%). Don't forget copper loss.

Ferrites with high resistivity, used almost without exception for
power conversion above 50KHz, also exhibit currie temperatures
significantly in excess of most commodity safety isolation systems.

If you see a currie temperature below 150degC, you've picked a signal
transformer or EMC (intentionally lossy) material.

RL