Flyback transformers, small, high power, where?

[Joerg]

Probably. I wasn't advocating that you go do it -- just mentioning that
I have seen it done. We had some pretty severe hight constraints and a
whole bunch of other conditions (like, we were going to have a bazillion
layers on that board anyway, and the lower voltage) that were amenable to
it.

When the Linear Tech app engineer suggested it the only thing that kept
him from being thrown out of the room was his long history of being
right. And dammit -- he was right!

LTC guys are almost always right

 

[legg]

It can be done but it's expensive. 8-10 layers, board becomes brittle if
skinny, and so on. It does look elegant though.



And probably at a much higher frequency than 100kHz.

Not a flyback, either, I'll bet.

RL

 

[Joerg]

They are used more often these days. For example, the Excelsys series
of modular supplies use planar transformers with mains volage
primaries. A bit pricey, but they have available good "medical"
isolation (4kV) high-rel and low noise options.

Here's the output module with the planar transformer:
http://media.digikey.com/Photos/Excelsys Tech Photos/POWERMOD SERIES.JPG

I have not attempted to determine how many layers are in the board.

You can see the traces, looks like a megeehoitz switcher. Here, we are
talking 100kHz flyback, single-digit Dollars, way different ballgame.

 

[Joerg]

Don't most CRTs have flyback xfrmr? Look fer one in an itty bitty tv
or o-scope. I've gotta tektonix 213 dmm w/ about a 2" CRT. The
flyback hasta be pretty small. (shrug)

Not this small.

 

[Joerg]

It would be easier to isolate only the LV winding, considering that
most of the guts in the area will be hazardous.

Then the whole core becomes hot which would creat some real issues. But
I am not so worried about that, there's the insulating bobbin and we
could pot a lot of things.

If it's a tube, make use of ALL of the available headroom - unless
it's a square tube. (if it were square fitting the board diagonally
would also permit increased hight profile ..... and increased board
real estate)

It's a square. So technically more a "profile" rather than a tube.

Care to reference the standard?

RTCA/DO-160.

 

[legg]

It doesn't follow the usual standards, this is a special hi-rel app. We
need certain breakdwon limits and creepage paths for everythin
line-side. Don't have them in my head right now but more than 0.150" to
chassis, for example. This goes into a narrow tube so that presents a
real estate issue.

The core itself can be on the isolated (LV) side.

It would be easier to isolate only the LV winding, considering that
most of the guts in the area will be hazardous.

If it's a tube, make use of ALL of the available headroom - unless
it's a square tube. (if it were square fitting the board diagonally
would also permit increased hight profile ..... and increased board
real estate)

Care to reference the standard?

RL

 

[Tim Williams]

It isn't a popular method because that way you cannot reasonably get the
number of turns for a 180V peak primary. The only time I have seen
planar transformers is in signal isolation or pulse transformers.

The main trouble is, how many layers can you get and how narrow can you make
the traces? Board fab rules really set the limit on voltage here -- easy to
wind a transformer with 40AWG wire, much harder to print something
equivalent. 180V, at that frequency, on a small core, is probably in the
100t range, which is an awful lot of very fine printing.

I've got a design with 10W planar transformers, which works well. The
outline is much larger than your requirement, voltage is low, turns count
low, and frequency somewhat higher. It's only a two layer board, so the
size could come down a lot.

In the same outline, I think I could push 20W, with a high voltage primary,
given a 6 or 8 layer board. But the frequency would still be high, and
flyback (using only half the B-H curve) would be pushing it (I'd probably do
half bridge for such a circuit, at some expense in coupling/bypass caps)...

Tim

 

[legg]

legg wrote:

Then the whole core becomes hot which would creat some real issues. But
I am not so worried about that, there's the insulating bobbin and we
could pot a lot of things.



It's a square. So technically more a "profile" rather than a tube.



RTCA/DO-160.

That standard doesn't have physical requirements for isolation
barriers. This is usually covered by a different reference in spec or
procurement docs for the aircraft company doing the purchasing.

If purchasing told you .15 hazardous to frame, then you'll have to
count on completely wrapped/encapped sections in an isolator this
small. There is no creepage distance available. Multiply-insulated
wire on the secondary might be an option. I don't know how small the
diameters get, but the film layers will rapidly dominate their volume.

RTCA also doesn't have 240V sources in any of it's four categories. I
don't have access to the newer (2001) section (16) that covers
harmonics, so this may have changed too. Were you just going to use
the extra headroom to ride through the 180V surges?

Doing a rough calculation for toroids that might limbo under the 6mm
headroom, it looks like you'll be wanting peak flux densities at
100KHz, in a complete energy transfer part, that exceed ferrite
saturation, just to fit wire in. That's going to cook, no matter what
material is used, or core shape.

If you went for incomplete energy transfer with a gapped amorphous
core ($$$) the gap/turn ratio can come close to 1mm/turn, with peak to
average current at ~3:1 at low line, but deltaB will still be a
killer. I don't think the PFC chip will go for incomplete energy
transfer - looks like a critical conduction job to me and they all
want CET.

There might be parts with better volumetric efficiency, but not that
much better.

RL

 

[Joerg]

That standard doesn't have physical requirements for isolation
barriers. This is usually covered by a different reference in spec or
procurement docs for the aircraft company doing the purchasing.

It does specify pressure loss performance and then Paschen's law gets you.

If purchasing told you .15 hazardous to frame, then you'll have to
count on completely wrapped/encapped sections in an isolator this
small. There is no creepage distance available. Multiply-insulated
wire on the secondary might be an option. I don't know how small the
diameters get, but the film layers will rapidly dominate their volume.

There are many ways to maintain the insulation, for example split
bobbin. That's not really my concern because stuff can be potted up,
size is.

RTCA also doesn't have 240V sources in any of it's four categories. I
don't have access to the newer (2001) section (16) that covers
harmonics, so this may have changed too. Were you just going to use
the extra headroom to ride through the 180V surges?

Yes, of course no 240V but we need headroom for the surges, and at the
DC peaks of those.

Doing a rough calculation for toroids that might limbo under the 6mm
headroom, it looks like you'll be wanting peak flux densities at
100KHz, in a complete energy transfer part, that exceed ferrite
saturation, just to fit wire in. That's going to cook, no matter what
material is used, or core shape.

My preference would be E-core and some newer materials start showing up
as E-cores, just not in the small sizes we need. It's like going into a
sporting goods store and all running shoes are 11 and up. No problem for
me but my wife would not like that.

If you went for incomplete energy transfer with a gapped amorphous
core ($$$) the gap/turn ratio can come close to 1mm/turn, with peak to
average current at ~3:1 at low line, but deltaB will still be a
killer. I don't think the PFC chip will go for incomplete energy
transfer - looks like a critical conduction job to me and they all
want CET.

There might be parts with better volumetric efficiency, but not that
much better.

Cost is not the prime parameter here as long as it isn't outrageous. And
that's what I am finding, the millisecond you want a custom size ferrite
and bobbin they charge through the nose. But, starting to look overseas
now. It could be like with taylored suits, something that is really only
affordable in Asia. To say it politely, right now I am a bit non-plussed
regarding the willingness of domestic companies to go out of their way
and make it happen. This would be "the" chance for one of those to get
into a cutting edge market.

If push comes to shove we could use a chip that does flyback-PFC at
several hundred kHz. But I haven't found such a beast yet, probably
because that usually result in an EMC nightmare.

 

[legg]

legg wrote:

There are many ways to maintain the insulation, for example split
bobbin. That's not really my concern because stuff can be potted up,
size is.

Well, I sort of admire the relaxed attitude, but it would probably
drive me crazy, if I was on the team. I like to see goals that are
demonstrably possible, even if just on paper, for some good reasons.

Yes, of course no 240V but we need headroom for the surges, and at the
DC peaks of those.

Suggest you nail down actual requirements, if you know you're possibly
testing the physical limits in the hardware being designed to do the
job.

My preference would be E-core and some newer materials start showing up
as E-cores, just not in the small sizes we need. It's like going into a
sporting goods store and all running shoes are 11 and up. No problem for
me but my wife would not like that.

It doesn't matter what core shape you use, unless the materials
exhibut remarkably lower core loss than the current leader
- ie ferrite.

Reports and specs on other materials talk about lower loss, but this
is a term used relative to other powder formulations only.

In the dimension limits stated; the use of bobbins over film
insulation will just make this worse.

Cost is not the prime parameter here as long as it isn't outrageous. And
that's what I am finding, the millisecond you want a custom size ferrite
and bobbin they charge through the nose. But, starting to look overseas
now. It could be like with taylored suits, something that is really only
affordable in Asia. To say it politely, right now I am a bit non-plussed
regarding the willingness of domestic companies to go out of their way
and make it happen. This would be "the" chance for one of those to get
into a cutting edge market.

If push comes to shove we could use a chip that does flyback-PFC at
several hundred kHz. But I haven't found such a beast yet, probably
because that usually result in an EMC nightmare.

For that matter, any converter chip designed for off-line critical
conduction mode could potentially be usable, whether intended for PFC
application or not. It won't cut down on real estate, unless you can
find an 8-pin part that has also migrated to leadless smd, that
requires less external parts to do the job.

About 100KHz operating frequency: In general, the simple CCM control
is constant on-time, variable frequency, with a limited upper
frequency (maximum power). Controllers likely try to limbo under
150KHz in order to avoid the CISPR/FCC pinch, making controller
tolerances and transformer inductance values a critical factor in
power transfer. That's going to be a headache in practice, without
margins - tough to provide in limited physical/thermal circumstances.

RTCA has it's own EMC requirements, with conducted emissions also
speced above 150khZ. It also has conducted susceptibility requirements
that go down to 10khZ, at current levels that may be significant in
lower powered devices.

Although I hesitate to advise actual use of power-integrated parts,
you may seem some interestink applications of 'linkswitch' parts, at
your targeted power level, at Power Integrations. LNK405-409(EL).

RL

 

[legg]

No worries, I've done things like that before. Sometimes the envelope
needs to be pushed. My motivation goes up tremendously when people
around me say "It can't be done"

Well, if you were not prepared to explain how it can be done, you'd
have run into that a lot and will likely continue to do so. It gets
kind of lame after a while.

The kicker is when you're told it's a requirement; no-one has a clue
how it's supposed to be achieved; your own paper points to a brick
wall.....and you're the one with the job.

I prefer the incremental approach, where simple things you've found
out, most likely by accident, or from working in left field, are
introduced to save money, or improve performance - and there's a
concensus by those expert in their own area of responsibility that it
will fly. It's seldom just your own ass on the line.

For example, this is how an "impossible" cardiac patient interface came
about that was deemed impossible without fiberoptics in the 90's. The
sales guys for the super-expensive fiberoptics were miffed, of course.
Passed UL and IEC601 right away but went beyond by being
defibrillator-proof (I did not consider these med standards safe enough
back then and would not design any other way).



They are as stated in the first post. Basically 120VAC in, LV out, up to
20W, about 100kHz, max 0.500" wide and 0.250" high.

Must have mis-read the voltage elsewhere. It's the 'up to' that
probably needs redefinition, soonest.

Ferrite is ok if someone would make us a "fat" core that's long but no
wider than 0.500" and 0.250" high, plus matching bobbin. Without
breaking the bank, that is.

Well, it's not OK, because the ideal ferrite part, fitting the space
limit, won't do it.

Why?

Bobbin wall thicknesses and minimum dimensional tolerances exceed
achievable film thickness, due to plastic fab restrictions. Anything
that uses volume, unecessarilly, will cut down on power throughput and
result in higher temp rises in the same total volume.

The chip size doesn't matter so much as the transformer takes the lion's
share.

Have you got something besides this real estate to convey the losses
to the tube walls?

But ... it needs a modulator for the current control in there in
order to do the PFC. Ideally with sim files, that's what I like so much
about LTC.

This ain't necessarilly so. No disrespect to LTC.

L6562
FAN7528
MC33260

I've used those a bucks but they are a stretch at 20W even in flyback
configuration. Also, doing PFC with one of these isn't exactly easy.

Never the less, you might find the reading informative. First google
search return on LNK405.

RL

 

[legg]

I suppose one could always think in terms of self-supporting coils
wound on removable cores (wax? aluminium to be etched out with sodium
hydroxide?) with self-bonding wire.

<snip>

The HV winding could be wound as a preformed subassembly, possibly
using simple VPI of a polyester resin. To allow for bake-out, I think
you'd be looking at removable teflon, or teflon-coated forms.

As there's probably no room for interleaving, the same could be done
for the secondary, in a side-by side arrangement, separated by a
suitable foil layer extending through the part, unbroken in the
immediate area of the coils.

Core material is a stumper. If ferrite dust composites can be made,
then maybe same thing would be possible with amorphous crystaline
dust. Basically already do that with iron dust and other materials.

If the amorphous HF losses were less than an order of magnitude
greater than ferrite, at the temperature limit, it might be possible
to cojole a permeability out of one mix so as to approach that needed.
I think the loss/volume relationship of amorphous base material is
actually worse than that, though.

There are physical limits to the amount of filler you can force into
organic binders that have a suitable temperature index and
flammability rating. I'm not sure about the insulation characteristics
that would result, either.

Anyways, if the coils and their terminations were imbedded in a
suitable block of the mix, using HPInjection molding, you'd be getting
close to practical limits for the volume available. It'd still get
hot, but I've seen SMD magnetics, with >70degree rises, paraded in the
literature without apparently raising an eyebrow.

 

[UNotBrightEnough]

That MOQ is ok (after sampling), but their flyback stuff is huge, all
regular ferrite.

You wanting a 20W flyback transformer but wanting it in a small form
factor is the big joke here. How many HV designs have you done? Try one
negative 10W supply and one positive 10W supply and tie their returns
together. So two flybacks to achieve the power density you are after.
Two smaller form factor devices would fit your profile and likely allow
for off the shelf sourcing.

Something to think about.

 

[Joerg]

Well, if you were not prepared to explain how it can be done, you'd
have run into that a lot and will likely continue to do so. It gets
kind of lame after a while.

For me those are actually the most fun jobs. When it gets tough.

The kicker is when you're told it's a requirement; no-one has a clue
how it's supposed to be achieved; your own paper points to a brick
wall.....and you're the one with the job.

I prefer the incremental approach, where simple things you've found
out, most likely by accident, or from working in left field, are
introduced to save money, or improve performance - and there's a
concensus by those expert in their own area of responsibility that it
will fly. It's seldom just your own ass on the line.

Lots of cutting edge projects can't be done incrementally at all, one
must start completely fresh. Here is one example:

http://www.analogconsultants.com/papers/pcc_1.pdf

Yes, there are five ICs in that thing. And 64 sensors. This was
definitely deemed impossible in the ultrasound world, and then we
started producing hundreds of these per day. They save lives.

Must have mis-read the voltage elsewhere. It's the 'up to' that
probably needs redefinition, soonest.

Well, that's flexible. Of course you always want to be able to regulate
out surges but if the magnetics don't have enough oomph one may have to
cut conversion.

Well, it's not OK, because the ideal ferrite part, fitting the space
limit, won't do it.

Why wouldn't it?

Bobbin wall thicknesses and minimum dimensional tolerances exceed
achievable film thickness, due to plastic fab restrictions. Anything
that uses volume, unecessarilly, will cut down on power throughput and
result in higher temp rises in the same total volume.

I know but nowadays one can make rather thin-walled bobbins. Almost like
film.

Have you got something besides this real estate to convey the losses
to the tube walls?

One can always thermally couple to the walls, that wouldn't be a
problem. Not great, but man's gotta do what man's gotta do, as John
Wayne said

This ain't necessarilly so. No disrespect to LTC.

L6562

Thanks. It's a regular boost-mode chip though. Maybe it can be used as
flyback, should be possible. LTC doesn't have to offer much in that
area, AFAICT only up to 300kHz but at least that's a factor of three
versus they flyback-PFC chips. And the voltage control in those isn't
too great, would have to roll my own anyhow.

So yeah, that may be an idea. But first I want to find out what it takes
to get a xfmr that can be run under 150kHz.

FAN7528
MC33260

These two have been obsoleted.

Never the less, you might find the reading informative. First google
search return on LNK405.

App note DER-286 for the LNK419 is interesting. If their web site just
wasn't such a mess. The only way to quickly get to their datasheets is
via Digikey.

But the main show stopper for the LNK4xx series is the switching
frequency. Much lower than for the LTC parts, <70kHz.

 

[Joerg]

legg wrote:

[...]

Core material is a stumper. If ferrite dust composites can be made,
then maybe same thing would be possible with amorphous crystaline
dust. Basically already do that with iron dust and other materials.

If the amorphous HF losses were less than an order of magnitude
greater than ferrite, at the temperature limit, it might be possible
to cojole a permeability out of one mix so as to approach that needed.
I think the loss/volume relationship of amorphous base material is
actually worse than that, though.

This is exactly the reason for my post, to see if there's any new
miracle material. Looking at High-Flux and stuff, maybe there is.

There are physical limits to the amount of filler you can force into
organic binders that have a suitable temperature index and
flammability rating. I'm not sure about the insulation characteristics
that would result, either.

Got to be careful with organic binder. There have been disintegration
issues in the past, after a few years of service.

Anyways, if the coils and their terminations were imbedded in a
suitable block of the mix, using HPInjection molding, you'd be getting
close to practical limits for the volume available. It'd still get
hot, but I've seen SMD magnetics, with >70degree rises, paraded in the
literature without apparently raising an eyebrow.

Do you remember which literature that was in? Most of what I found was
in magazines such as "Power Electronics Technology", like this paper:

http://powerelectronics.com/passive...ite-inductor-cores-pol-converters/index2.html

 

[Joerg]

[...]

App note DER-286 for the LNK419 is interesting. If their web site just
wasn't such a mess. The only way to quickly get to their datasheets is
via Digikey.

But the main show stopper for the LNK4xx series is the switching
frequency. Much lower than for the LTC parts, <70kHz.

OK. I get it. Where's the hidden camera?

This one I didn't understand

<scratching head>

 

[legg]

For me those are actually the most fun jobs. When it gets tough.



Lots of cutting edge projects can't be done incrementally at all, one
must start completely fresh. Here is one example:

http://www.analogconsultants.com/papers/pcc_1.pdf

Yes, there are five ICs in that thing. And 64 sensors. This was
definitely deemed impossible in the ultrasound world, and then we
started producing hundreds of these per day. They save lives.



Well, that's flexible. Of course you always want to be able to regulate
out surges but if the magnetics don't have enough oomph one may have to
cut conversion.




Why wouldn't it?




I know but nowadays one can make rather thin-walled bobbins. Almost like
film.



One can always thermally couple to the walls, that wouldn't be a
problem. Not great, but man's gotta do what man's gotta do, as John
Wayne said




Thanks. It's a regular boost-mode chip though. Maybe it can be used as
flyback, should be possible. LTC doesn't have to offer much in that
area, AFAICT only up to 300kHz but at least that's a factor of three
versus they flyback-PFC chips. And the voltage control in those isn't
too great, would have to roll my own anyhow.

So yeah, that may be an idea. But first I want to find out what it takes
to get a xfmr that can be run under 150kHz.



These two have been obsoleted.



App note DER-286 for the LNK419 is interesting. If their web site just
wasn't such a mess. The only way to quickly get to their datasheets is
via Digikey.

But the main show stopper for the LNK4xx series is the switching
frequency. Much lower than for the LTC parts, <70kHz.

OK. I get it. Where's the hidden camera?

RL

 

[Fred Abse]

How many HV designs have you done?

ROTFLMAO!

 

[Fred Abse]

amorphous crystaline

Oxymoron?

 

[legg]

legg wrote:

[...]

Got to be careful with organic binder. There have been disintegration
issues in the past, after a few years of service.

GF Nylon is glass in an organic binder, as are most bobbin materials.
Magnetic composite aging is more an issue because they're self-heating
with generally positive tempco of loss. And they can all be
misused....

Do you remember which literature that was in? Most of what I found was
in magazines such as "Power Electronics Technology", like this paper:

You may be in luck here, though links to such are not something I
bother to save, if in electronic format. These generally expire too
soon to be of any lasting use.

Following one of your own links yesterday, resulted in a pop-up for
one example of magnetics rises of this order:

http://www.ti.com.cn/cn/lit/ug/sluu866/sluu866.pdf

They're not uncommon, but you have to have your eyes open, before an
eybrow is likely to also rise.

RL