Core losses in dual big inductors

[Joerg]

Trying to get 45-50W through a SEPIC, and unfortunatly the output
voltage has to be up to 80V, input up to 30V. Peak primary current is
going to be around 7A because of a wide supply voltage range. Long story
short, a 10uH SRF1280 from Bourns will almost unsolder itself, goes 85C
above ambient. Phsssss ... ouch.

http://www.bourns.com/data/global/pdfs/SRF1280.pdf

We've ordered some with larger inductance since this is all due to core
losses. But of course there comes a point where that won't allow the
peak current we need at the lowest supply voltage.

So ... does anone know a brand or style of inductor that is really low
in core losses? Or bigger SMT footprint ones without being taller?

 

[Lasse Langwadt Christensen]

Trying to get 45-50W through a SEPIC, and unfortunatly the output

voltage has to be up to 80V, input up to 30V. Peak primary current is

going to be around 7A because of a wide supply voltage range. Long story

short, a 10uH SRF1280 from Bourns will almost unsolder itself, goes 85C

above ambient. Phsssss ... ouch.



http://www.bourns.com/data/global/pdfs/SRF1280.pdf



We've ordered some with larger inductance since this is all due to core

losses. But of course there comes a point where that won't allow the

peak current we need at the lowest supply voltage.



So ... does anone know a brand or style of inductor that is really low

in core losses? Or bigger SMT footprint ones without being taller?

looks like coilcraft msd1583 is a bit bigger and only 0.6mm taller

http://www.coilcraft.com/msd1583.cfm

try put in some numbers: http://www.coilcraft.com/apps/loss_coupled/loss_1.cfm

-Lasse

 

[Joerg]

looks like coilcraft msd1583 is a bit bigger and only 0.6mm taller

http://www.coilcraft.com/msd1583.cfm

Thanks, got to order some of those tomorrow. If we can get them into the
layout.

try put in some numbers: http://www.coilcraft.com/apps/loss_coupled/loss_1.cfm

Those simple calcs aren't useful. For example, they do not take voltage
swing across windings into account and that is a major factor in core
loss. I just got 130C temperature rise

 

[Bill Sloman]

Trying to get 45-50W through a SEPIC, and unfortunatly the output
voltage has to be up to 80V, input up to 30V. Peak primary current is
going to be around 7A because of a wide supply voltage range. Long story
short, a 10uH SRF1280 from Bourns will almost unsolder itself, goes 85C
above ambient. Phsssss ... ouch.

http://www.bourns.com/data/global/pdfs/SRF1280.pdf

We've ordered some with larger inductance since this is all due to core
losses. But of course there comes a point where that won't allow the
peak current we need at the lowest supply voltage.

So ... does anyone know a brand or style of inductor that is really low
in core losses? Or bigger SMT footprint ones without being taller?

Nickel/Zinc ferrites have low core losses at high frequencies. You could buy them as RM cores when I last looked, which is quite a while ago. Farnell now stocks just three toroids which I can easily identify as Ni/Zn ferrites, but more specialised suppliers may do better.

This still leaves you with getting the cores wound by that almost extinct species, the low volume coil winder, but I'm sure that if you had to find one you could manage it.

 

[Martin Riddle]

Trying to get 45-50W through a SEPIC, and unfortunatly the output
voltage has to be up to 80V, input up to 30V. Peak primary current is
going to be around 7A because of a wide supply voltage range. Long story
short, a 10uH SRF1280 from Bourns will almost unsolder itself, goes 85C
above ambient. Phsssss ... ouch.

http://www.bourns.com/data/global/pdfs/SRF1280.pdf

We've ordered some with larger inductance since this is all due to core
losses. But of course there comes a point where that won't allow the
peak current we need at the lowest supply voltage.

So ... does anone know a brand or style of inductor that is really low
in core losses? Or bigger SMT footprint ones without being taller?

Isat is 6A, L goes down 30%.

You sure its not copper losses?

I'm a skeptic of your assumtion of core loss

Cheers

 

[Joerg]

Isat is 6A, L goes down 30%.

For the 10uH coil it's 11.2A.

You sure its not copper losses?

Yes, that is also what the Bourns engineer said when I told him the
story and asked for core loss data.

I'm a skeptic of your assumtion of core loss

Tomorrow afternoon we'll know more

 

[Klaus Kragelund]

Trying to get 45-50W through a SEPIC, and unfortunatly the output

voltage has to be up to 80V, input up to 30V. Peak primary current is

going to be around 7A because of a wide supply voltage range. Long story

short, a 10uH SRF1280 from Bourns will almost unsolder itself, goes 85C

above ambient. Phsssss ... ouch.



http://www.bourns.com/data/global/pdfs/SRF1280.pdf



We've ordered some with larger inductance since this is all due to core

losses. But of course there comes a point where that won't allow the

peak current we need at the lowest supply voltage.



So ... does anone know a brand or style of inductor that is really low

in core losses? Or bigger SMT footprint ones without being taller?

Have you tried higher frequency? (and possibly lower inductance at the same time) Core loss increases with frequency (f^m), but if all other is kept the same, the flux density is lowered and resulting losses are reduced (B^n). n is normally higher than m

You could try reducing the inductance at the same time to reduce RDC losses

Cheers

Klaus

 

[Klaus Kragelund]

Have you tried higher frequency? (and possibly lower inductance at the same time) Core loss increases with frequency (f^m), but if all other is kept the same, the flux density is lowered and resulting losses are reduced (B^n). n is normally higher than m



You could try reducing the inductance at the same time to reduce RDC losses

Download the Epcos tool and try different operating points:

http://www.epcos.com/web/generator/Web/Sections/DesignSupport/Tools/Ferrites/Page,locale=en.html

Cheers

Klaus

 

[Joerg]

Yes, we've tried higher frequency. Then the inductor almost began
unsoldering itself. On the FLIR image it looked like a nuclar meltdown.
But we did order one inductor with less inductance, just in case, to try
running the thing faster.

Download the Epcos tool and try different operating points:

http://www.epcos.com/web/generator/Web/Sections/DesignSupport/Tools/Ferrites/Page,locale=en.html

Thanks, I'll try that. Except I don't know what ferrite the various mfgs
use and they won't tell me.

 

[Joerg]

Might a boost or boost-doubler configuration work? Is it always stepping up?

Unfortunately not. It has to work with several input voltages and the
output must be controllable from 0V to 80V.

The boost-doubler should use the copper better, because there's only one winding
that has to carry the load current. Pure boost is even better because it rides
directly on the input voltage; the first bunch of volts are free.

Yeah, I could really use the "first bunch of volts for free" here,
because bunch would mean 20-30V. But, can't have it.

How tall can you stand? Coilcraft has some nice parts. Maybe you could parallel
two (or more!) if you have area but no height.

About 0.400" would be the limit. Area is really tough, the board looks
like a subway in Tokyo during rush hour. Lasse suggested a bigger one
but if I tell my layouter he is probably going to faint.

I was really disappointed that this core didn't perform better.

I just got a copy of Switching Power Supply Design, 3rd Ed, by Pressman et al,
mostly to check on prior art. It has a lot of weird switcher configs, including
a 2-phase flyback. Which could be a 2-phase boost or boost-doubler.

Strangely, they don't mention either the sepic or the boost-doubler.

That's one reason I don't buy books much anymore. The real tricky stuff
often isn't in there.

 

[Joerg]

John Larkin wrote:

[...]

That's one reason I don't buy books much anymore. The real tricky stuff
often isn't in there.

I tried the boost-doubler, and it works step-up or step-down. As Rob
says, it's a Sepic without the second winding.

Core loss is a wash, but copper is 2:1 better.

Copper is really not problem at all, we could easily live with 200mOhms
here. It's the core loss that literally cooks them.

One problem with ferrite is that it generally doesn't lend itself to be
heatsinked.

 

[Joerg]

John Larkin wrote: [...]


I just got a copy of Switching Power Supply Design, 3rd Ed, by Pressman et al,
mostly to check on prior art. It has a lot of weird switcher configs, including
a 2-phase flyback. Which could be a 2-phase boost or boost-doubler.

Strangely, they don't mention either the sepic or the boost-doubler.

That's one reason I don't buy books much anymore. The real tricky stuff
often isn't in there.
I tried the boost-doubler, and it works step-up or step-down. As Rob
says, it's a Sepic without the second winding.

Core loss is a wash, but copper is 2:1 better.

Copper is really not problem at all, we could easily live with 200mOhms
here. It's the core loss that literally cooks them.

One problem with ferrite is that it generally doesn't lend itself to be
heatsinked.

No? How about a dab of epoxy to a copper pour, with thermal vias to a
ground plane? That's got to reduce theta by some number of times. The
layout needs to suck heat from the lead connections, too. Clever layer
interleaves and such.

Easy enough to try. One copper winding makes a damned fine RTD.

This board is a very dense one with several power converters on it. So I
made it have a good "heat-sucking" core and a cooling strip on the side
which is at the same time the mounting strip. Tons of vias. And this
inductor is located right at the thermal strip for cooling reasons.
Epoxy is one option but I am thinking more about thermally conductive
pliable plastics to sink heat away by pressing agains a metal surface.
Can't remember the name, have to dig it out.

However, first I want to push these losses down. Because burning several
watts in the inductor isn't a good thing in the first place, it dings us
on efficiency. Plus heat sinking has its limits if the enclosure is
small. The whole box will eventually become a hot-plate.

 

[Joerg]

Isat is 6A, L goes down 30%.

You sure its not copper losses?

I'm a skeptic of your assumtion of core loss

Well, now we know it is core loss. We have received the shipment with
higher inductance, which of course comes at the price of higher DC
resistance and higher copper losses. Result -> lower temperatures. We'll
have to see where the sweet spot is and that can only be found out
experimentally because no reliable data is available about core loss.

 

[Joerg]

[...]

This board is a very dense one with several power converters on it. So I
made it have a good "heat-sucking" core and a cooling strip on the side
which is at the same time the mounting strip. Tons of vias. And this
inductor is located right at the thermal strip for cooling reasons.
Epoxy is one option but I am thinking more about thermally conductive
pliable plastics to sink heat away by pressing agains a metal surface.
Can't remember the name, have to dig it out.

Laird and Bergquist make thermal transfer stuff, limp and sticky and
about the look and feel of used chewing gum. Thetas are in the 1-5
w/m-k sort of range. You could put a patch of that on the bottom of an
inductor and squish it down before reflow.

Indeed, didn't know Laird had it:

http://www.digikey.com/product-detail/en/A16104-20/926-1141-ND/2445451

Kinda pricey though.

We just Gerbered our 2-cubic-inch, 5 kilowatt laser driver. That
wasn't too bad thermally, because the expected pulse rate is low. It
does use the schmitt-trigger boost-doubler converter to charge the
caps, with a single tiny Coilcraft inductor. So I have a SOT23 fet in
the boost power supply and a TO247 in the laser driver stage.

I just got the first bare boards on a 300 amp driver, where thermals
matter. These ROHS gold boards are gorgeous.

https://dl.dropboxusercontent.com/u/53724080/PCBs/D100_Bare_Bottom.JPG

Nice. Are you going to solder it lead-free? I try to avoid that wherever
I can.

The D2PAK power fets are on the opposite side. Thermal vias and
interleaved planes transport heat to the big gold pad, which can be
bolted to a heatsink block between the board and a baseplate.

That's just about how I did it on our board, except that we use bigger
mounting screws because ours goes into rough environments.

 

[Bill Sloman]

Thanks, I'll try that. Except I don't know what ferrite the various mfgs
use and they won't tell me.

Another argument for finding your own ferrite core and winding your own coils - or in this day and age, buying a core pair intended for use with printed windings, and getting your own windings printed, though probably not on the same circuit board as you use for the rest of the circuit, since getting a decent copper fill factor depends on putting very thick layers of copper on very thin substrates in a multilayer board.

 

[Tim Williams]

A noble idea, but planar magnetics take up an awful lot of board area.

I suppose if one were to crank up the board layers, components could be
placed on the outer surfaces while routing the windings through inner
layers between. Keep low voltages away (current sense and the like).
Might be tolerable that way, but still, lots of area, and copper losses
are intrinsically worse due to the poorer fill factor.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://seventransistorlabs.com

Thanks, I'll try that. Except I don't know what ferrite the various mfgs
use and they won't tell me.

Another argument for finding your own ferrite core and winding your own
coils - or in this day and age, buying a core pair intended for use with
printed windings, and getting your own windings printed, though probably
not on the same circuit board as you use for the rest of the circuit,
since getting a decent copper fill factor depends on putting very thick
layers of copper on very thin substrates in a multilayer board.

 

[Bill Sloman]

A noble idea, but planar magnetics take up an awful lot of board area.

Which is why I explicitly rejected putting the winding on the same board asthe rest of the circuit.

I suppose if one were to crank up the board layers, components could be
placed on the outer surfaces while routing the windings through inner
layers between. Keep low voltages away (current sense and the like).
Might be tolerable that way, but still, lots of area, and copper losses
are intrinsically worse due to the poorer fill factor.

The fill factor for round copper wires isn't 100% - closer to 70% IIRR - and some board manufacturers can put down copper layers that are thicker thanthan board material, so you can probably get pretty close to round wire copper fill factors, and - in theory - might even be able to do a bit better.But it would take a specialist printed circuit board manufacturer to get you there.

 

[[email protected]]

Trying to get 45-50W through a SEPIC, and unfortunatly the output

voltage has to be up to 80V, input up to 30V. Peak primary current is

going to be around 7A because of a wide supply voltage range. Long story

short, a 10uH SRF1280 from Bourns will almost unsolder itself, goes 85C

above ambient. Phsssss ... ouch.



http://www.bourns.com/data/global/pdfs/SRF1280.pdf



We've ordered some with larger inductance since this is all due to core

losses. But of course there comes a point where that won't allow the

peak current we need at the lowest supply voltage.



So ... does anone know a brand or style of inductor that is really low

in core losses? Or bigger SMT footprint ones without being taller?



--

Regards, Joerg



http://www.analogconsultants.com/

Try the Murata A series

http://www.digikey.com/product-detail/en/60A103C/811-2450-2-ND/3178904

The graph shows about 15C above ambient for 7A and 10uH.

 

[Joerg]

Try the Murata A series

http://www.digikey.com/product-detail/en/60A103C/811-2450-2-ND/3178904

I can't, I need a dual inductor.

The graph shows about 15C above ambient for 7A and 10uH.

They all show that kind of stuff but it is deceiving. It does not take
into account core losses and those are the lion's share in many switcher
designs.

 

[Joerg]

Does anyone actually specify core loss? Or a Spice model that
actually works in transient analysis? All I find model AC (small
signal) lossiness only.

Not really. There is a lot of hand-waving, inaccurate information,
missing data. With inductors it usually already ends at the telephone,
at the point where you ask which core material they use. Almost like
asking Coca Cola about the secret sauces and magic potions in their
concentrate.

That secrecy is the reason why I never really looked into simulation
software for core loss. If you don't know where they get the cores it
wouldn't make much sense.