800+ Watt DC-DC converter ferrite transformer design

On 11/26/2013 9:49 PM, P E Schoen wrote:
> I want to make a DC-DC converter for 48 VDC to about 300 VDC at a
> continuous power of about 800W to drive a 1 to 2 HP three phase
> induction motor.

1HP is ~ 750W.  at 800 watts plus the efficiency if your 3 phase 
inverter, you'll be pushing it for 1 HP let alone 2.

Also be careful of your input impedance.  At 800W with 48V input and 
assuming 80% efficiency (960W) you're looking at 20 amps input current 
for an apparent input impedance of 2.4 ohms @ 48Vin.   You need to be 
sure your 48V source impedance (including cabling) has a total loop to 
be less than 1/4th that value (0.6 ohms)  Even that is critical at 
something closer to 1/8th would be better.  (.3 ohms) to deal with motor 
load dymanics, startup surge etc.



I have some E55 cores and coilformers which should do
> the job, and also some E47/20/15 which might work OK. The E55 is an
> Epcos type N27 and the E47 is probably the same.
>
> http://www.mouser.com/ds/2/136/e_47_20_16-75097.pdf
> http://www.mouser.com/ds/2/136/e_55_28_21-73714.pdf
>
> I found a website that shows a simple procedure to determine the turns
> required for a transformer in a half-bridge topology using three
> capacitors. I plan to use the same circuit but not use the additional
> series capacitor and instead use two 20 uF 100 VAC PP capacitors in
> series across the DC bus and the center to one end of the primary.
>
> http://tahmidmc.blogspot.com/2013/02/ferrite-transformer-turns-calculation_22.html
>
>
> The LTSpice for the basic circuit is on my server:
> http://enginuitysystems.com/pix/48V-320V_DCDC_HalfBridge_2Cap.asc
>
> I made a spreadsheet to automate the selection process and for the E55
> core and 50 kHz it came up with 23 turns of #8 AWG for the primary at 33
> amps and 144 turns of #16 for the secondary at 5.3 amps. It appears to
> have 8 watts of ferrite losses and about 6 watts primary and secondary
> copper losses for total efficiency of about 97%.
>
> Here is the (OpenOffice) spreadsheet. Please have a look and see if it
> is at least close to being accurate:
> http://enginuitysystems.com/files/Ferrite_Transformer.ods
>
> Another question I have is if would be helpful to use heat shrink tape
> to compress the ferrite halves:
> http://www.mcmaster.com/#heat-shrink-tape/=pjzuzt

I would suggest glue.  regular tape would serve to keep the cures 
together for testing but once it is ready for prime time I would glue 
the to core halves together.  You don't want to put glue between the 
core halves.  Just a ring around the outside of the cut.  5 minute epoxy 
or something similar would suffice as long as it adheres to ferrite. 
Make sure you clean them with alcohol before gluing to remove oils and 
films.

>
> And, finally, I think it would be good to use Litz wire for this. I
> found some that seems to be a good deal:
> http://www.ebay.com/itm/370951676185
>
> Thanks,
>
> Paul

Litz is OK but make sure it is the solder through type of insulation. 
Otherwise stripping the ends is a pain.

You can also go multifilar smaller gauge instead of litz to get the 
turns to law down in the bobbin with less wasted space between the turns.

In article <l76ve3$kh4$1@dont-email.me>, paul@peschoen.com says...
> 
> "Tim Williams"  wrote in message news:l74saa$guj$1@dont-email.me...
> 
> > Waveform calculator:
> > http://schmidt-walter.eit.h-da.de/smps_e/smps_e.html
> 
> I found what appears to be an error in the calculation of wire size for the 
> primary. For my 48V 750W 50 kHz transformer it gives a primary of 7 turns of 
> 1.82 mm (2.6 mm^2 or about #13 AWG) but the primary current is about 33 A 
> RMS which should be 11 mm^2 (about #6 AWG) at 3A/mm^2.
> 
> I have updated my calculator for multiple primary strands in parallel, with 
> appropriate skin effect calculation. I used 11 turns of 8 strands of #16 AWG 
> (same as secondary), and the primary losses are now about 6 watts instead of 
> 15. Secondary is 138 turns and 15 watts losses. So the secondary would also 
> benefit from multiple strands.
> 
> http://enginuitysystems.com/files/Ferrite_Transformer.ods
> 
> Paul

 Litz wire would be a great choice, expensive as hell for your size 
however :)
 
  We have a resource of copper tape at work and can run that through the 
lacquer process. Works well for high freq apps and high current.
Jamie

On 11/28/2013 2:38 AM, P E Schoen wrote:
> "Tim Williams"  wrote in message news:l74saa$guj$1@dont-email.me...
>
>> Waveform calculator:
>> http://schmidt-walter.eit.h-da.de/smps_e/smps_e.html
>
> I found what appears to be an error in the calculation of wire size for
> the primary. For my 48V 750W 50 kHz transformer it gives a primary of 7
> turns of 1.82 mm (2.6 mm^2 or about #13 AWG) but the primary current is
> about 33 A RMS which should be 11 mm^2 (about #6 AWG) at 3A/mm^2.
>
> I have updated my calculator for multiple primary strands in parallel,
> with appropriate skin effect calculation. I used 11 turns of 8 strands
> of #16 AWG (same as secondary), and the primary losses are now about 6
> watts instead of 15. Secondary is 138 turns and 15 watts losses. So the
> secondary would also benefit from multiple strands.
>
> http://enginuitysystems.com/files/Ferrite_Transformer.ods
>
> Paul
>

  I'm confused by your wire description.
You said,
 >I used 11 turns of 8 strands
 > of #16 AWG (same as secondary)
  Does this mean no Litz wire?
  I have wound transformers and inductors
with as many as four strands paralleled,
that helped reduce heat vs. a single larger wire.
  I never wound any with Litz.

  A 38 Gauge 66 strand litz is the same diameter as
a 16 gauge solid wire.
  I'd like to see the equivalent Rac of 8 solid 16 Gauge vs.
8 litz 38/66. They should be equal size.
              Mikek

  Useful formulas here:
newenglandwire.com/products/litz-and-formed-cables/theory.aspx

On Thu, 28 Nov 2013 12:34:50 -0500, "Maynard A. Philbrook Jr."
<jamie_ka1lpa@charter.net> wrote:

>
> Litz wire would be a great choice, expensive as hell for your size 
>however :)

We (www.fluxeon.com) use Litz wire in our induction heaters.  We also
sell small quantities to the public as a service since it is so hard
to get in small quantities.  We have some huge #4 equivalent stuff if
you need it.  Just go to the web site, to the store and buy what you
need by the foot.

John
John DeArmond
http://www.neon-john.com
http://www.fluxeon.com 
Tellico Plains, Occupied TN
See website for email address

"Neon John"  wrote in message=20
news:ptah99dprcht9km6sehk5u2o7vgg6plika@4ax.com...

> We (www.fluxeon.com) use Litz wire in our induction heaters.  We
> also sell small quantities to the public as a service since it is so
> hard to get in small quantities.  We have some huge #4 equivalent
> stuff if you need it.  Just go to the web site, to the store and buy
> what you need by the foot.

I checked the website and they also offer 7x52/38 which is equivalent to =
#10=20
AWG and rated for about 17 amps at 600 CM/A (3.3 A/mm^2). It's listed at =

$1.13 while the larger size 7x7x52/38 is $8.31. That should be good for=20
17*7=3D 119 amps which is equivalent to #4 at 350 CM/A (which may be too =
much=20
for continuous duty in a transformer core).

The website does not say that the price is per foot. I figure I will =
need=20
about 8 ft for two primaries of 11 turns, which would cost about $9 plus =

shipping, which is a $15 flat rate (which was not stated until =
checkout). So=20
my cost would be $24.

I could use the 7x3x21/40 Litz wire I found on eBay for $20/40 ft with =
free=20
shipping. This is rated for about 7.8 amps, so I would need twice as =
much,=20
or 16 ft. But for $20 I would have enough for two transformers.

Thanks for the link. It seems like a good source for induction heaters =
if I=20
ever have the need.

Paul=20

"P E Schoen" <paul@peschoen.com> wrote in message 
news:l7bb4h$fmm$1@dont-email.me...
> I checked the website and they also offer 7x52/38 which is equivalent to
> #10 AWG and rated for about 17 amps at 600 CM/A (3.3 A/mm^2). It's
> listed at $1.13 while the larger size 7x7x52/38 is $8.31. That should be
> good for 17*7= 119 amps which is equivalent to #4 at 350 CM/A (which may
> be too much for continuous duty in a transformer core).

Strange...

Note that anything with 7 bundles is 14% useless: you get six around one 
central core, which never moves out from the center and therefore exhibits 
hugely greater resistance than the others.  I've only ever seen 3 and 5x 
bundles from NEWT, but I've seen Chinese stuff that's 7-way before.  Doesn't 
make sense why anyone would make it that way, unless they simply didn't know 
how to do it properly.

Tim

-- 
Seven Transistor Labs
Electrical Engineering Consultation
Website: http://seventransistorlabs.com 

On 11/29/2013 11:10 PM, Tim Williams wrote:
> "P E Schoen" <paul@peschoen.com> wrote in message
> news:l7bb4h$fmm$1@dont-email.me...
>> I checked the website and they also offer 7x52/38 which is equivalent to
>> #10 AWG and rated for about 17 amps at 600 CM/A (3.3 A/mm^2). It's
>> listed at $1.13 while the larger size 7x7x52/38 is $8.31. That should be
>> good for 17*7= 119 amps which is equivalent to #4 at 350 CM/A (which may
>> be too much for continuous duty in a transformer core).
>
> Strange...
>
> Note that anything with 7 bundles is 14% useless: you get six around one
> central core, which never moves out from the center and therefore
> exhibits hugely greater resistance than the others.  I've only ever seen
> 3 and 5x bundles from NEWT, but I've seen Chinese stuff that's 7-way
> before.  Doesn't make sense why anyone would make it that way, unless
> they simply didn't know how to do it properly.
>
> Tim
>

The skin effect explanation for Litz wire is wrong all through.  If it 
were correct, Litz would be lossier than solid, because in solid wire, 
at least the current has a straight shot on the thin outside layer, 
whereas with Litz, all of it spends a lot of time inside the bundle.

The skin effect argument is far from straightforward in the presence of 
other conductors, and especially of ferrite cores.  You can't just take 
the 1-D isolated conductor result and wave it over the design like a 
dead chicken.

The actual benefit is due to reducing eddy current loss in the wire due 
to dB/dt.  Copper tape winding is about equally effective IIRC.

Cheers

Phil Hobbs

"Phil Hobbs"  wrote in message=20
news:8fKdncAw8Lc99QTPnZ2dnUVZ_qadnZ2d@supernews.com...

> The skin effect explanation for Litz wire is wrong all through.  If it =

> were correct, Litz would be lossier than solid, because in solid
> wire, at least the current has a straight shot on the thin outside
> layer, whereas with Litz, all of it spends a lot of time inside the
> bundle.

> The skin effect argument is far from straightforward in the presence
> of other conductors, and especially of ferrite cores.  You can't just
> take the 1-D isolated conductor result and wave it over the design
> like a dead chicken.

> The actual benefit is due to reducing eddy current loss in the wire
> due to dB/dt.  Copper tape winding is about equally effective IIRC.

Very interesting. Is there any data available showing the real =
performance=20
of Litz wire vs single thick strand and multifilar coils? I also wonder=20
about the Lorentz force, which tends to press together parallel =
conductors=20
with current flowing in the same direction, so it seems that the current =
may=20
tend to try to flow more in the center. And there is a lot of =
distributed=20
capacitance among all those isolated conductors, so at high frequency =
there=20
may be some current flow among coils at different potentials.

The eddy current loss reduction makes sense, and the Litz wire may be=20
analogous to thin insulated laminations in transformers and motors =
designed=20
for higher frequencies.

I think I will order some of the Litz wire, as it seems that it will =
help=20
reduce the power dissipation and hence temperature, and it may be easier =
to=20
wind. But my first prototype will probably be with just multifilar =
windings.=20
In fact, I made a smaller transformer with an inner secondary winding of =

about 145 turns of #26, and an outer primary winding with 8 turns of 2=20
parallel #17 and 2 parallel #18. I just reused this wire that was on a=20
computer power supply transformer good for about 400 watts.

I tried to test it using my HP 3312A function generator, but it does not =

drive the primary very well, although it seems to do OK at 300 kHz. I =
will=20
need to rig up a proper drive circuit using the push-pull 2 capacitor=20
topology.

Thanks for the information.

Paul=20

"Phil Hobbs" <pcdhSpamMeSenseless@electrooptical.net> wrote in message 
news:8fKdncAw8Lc99QTPnZ2dnUVZ_qadnZ2d@supernews.com...
> The skin effect explanation for Litz wire is wrong all through.  If it 
> were correct, Litz would be lossier than solid, because in solid wire, at 
> least the current has a straight shot on the thin outside layer, whereas 
> with Litz, all of it spends a lot of time inside the bundle.

Lossier, per resistive length, factored by current-carrying cross sectional 
area?  (Making some sort of estimation of current density and resistivity in 
the areas where current does flow...)  Of course, it ends up better in the 
end, because you can use much finer wire, which gives much more perimeter, 
and thus more cross sectional area for current to flow in, even though it's 
choked up much worse from being forced into to a constant average current 
density.

Compared to single strands in free space, even of much larger diameter than 
the individual strands, the stuff is lossier.  If you look at Rac/Rdc for 
decreasing strand diameters, the single free strand might level off at, I 
forget, 28AWG or something, at say 100kHz, whereas in a big Litz cable (say, 
a thousand strands), it keeps going until 36 or 38AWG, and even then, the 
total resistance for equivalent area is larger (in addition to the increased 
length due to the weave).  It's like making copper a better resistor (or 
alternately, a worse inductor).

The nice part is, you get to carry more total current, in an only slightly 
larger volume, which is significantly smaller than the volume required of a 
single massive strand.  That is to say, at high frequencies, a large solid 
conductor is O(N), while fine conductors are O(N^2).  Litz has a smaller 
constant multiplier on that Big-Oh than a single fine strand, but 
appropriately chosen, it scales independently of frequency (as diameter 
squared), something a solid conductor doesn't (it's perimeter limited).  One 
of those things that "shouldn't work" by certain physical principles, but 
when considered holistically by the engineer, works great. ;-)

But anyway, in a seven-conductor construction, the always-central bundle is 
completely surrounded by fields from the other six, and so has much more 
eddy current losses, or higher Rac, or stronger proximity effect, or thinner 
skin depth, or however one likes to say it (they're all aspects of the same 
phenomenon, after all).

> The skin effect argument is far from straightforward in the presence of 
> other conductors, and especially of ferrite cores.  You can't just take 
> the 1-D isolated conductor result and wave it over the design like a dead 
> chicken.

Yes.  Proximity effect is all over the stuff, which is why the strands have 
to be so much finer than the free space skin depth would suggest.  Even 10 
strands of 28AWG will be noticeably higher in resistance than 1/10th of a 
single strand.  100 or 1000 strands need strands finer and finer still.  The 
scaling between number of strands and required decrease in strand diameter 
is of course "far from straightforward", for the same reason.

> The actual benefit is due to reducing eddy current loss in the wire due to 
> dB/dt.  Copper tape winding is about equally effective IIRC.

I haven't seen any analyses of tape, but I've seen it used here and there. 
Trouble is, the field around a conductor 'wants' to be round, and forcing it 
to wrap around a foil conductor is somewhat counterproductive.  It 
necessarily must penetrate the conductor, particularly along the edges.  The 
conductor must be thin enough to allow this; a thick conductor will shield 
its self-induction, and you get standard skin effect along the edges (within 
a constant factor).

The result is, eddy currents flow along the edges.  This manifests as skin 
effect.  Except, because we're talking about a somewhat two-dimensional 
conductor, it's really edge effect, and instead of bulk resistivity, 
thickness can be factored into the area resistivity, which edge penetration 
then takes as a factor.  For a finite thickness, edge penetration is deeper 
than the free space skin depth (which is the limit at infinite thickness, 
i.e., an infinite slab), but I don't know by how much.

If depth is inverse with thickness (a crude but not unreasonable guess, 
taking the area resistivity approximation as a suggestion), then one would 
need a conductor of thickness t = d^2/w, for width w and skin depth d.  (If 
t = w, you have a square conductor of dimension d, which is in the right 
order of magnitude.)  Unfortunately, copper at 100kHz is already only a few 
mils, so you need truely microscopic foils to actually achieve full 
utilization across the width of an average bobbin.  That stinks.

Proximity effect still applies, so while you're doing this, you can't just, 
say, wrap ten turns of foil primary, a layer of tape, then ten turns of 
secondary; the innermost facing turns will burn up from all the congestion.

Tape does at least suggest itself nicely for transmission line approaches: 
if the turns are similar, just layer primary and secondary together, with 
tape between, like the plates of a capacitor.  Except with a core in the 
middle.  Isolation capacitance won't be great, but leakage inductance will 
be teensy.  The image currents from primary and secondary will tend to flow 
along the faces as well as the edges, because it looks more like a parallel 
plate transmission line than an isolated foil conductor; that helps 
efficiency a lot.

You can of course apply the Litz trick to foil, but you don't have any free 
lunch; the geometry reduction is still required whether putting together a 
bunch of strips or strands.  Ten strips woven together will have less 
resistance than a single strip of the same width and thickness, but higher 
than 1/10th of an isolated strip that size.  I know of at least one company 
that claims to have some sort of foil technology that reduces Rac like Litz, 
presumably doing some kind of weave.  Tempting to buy a bigass custom part 
from them just to take it apart and look, see how they put the stuff 
together.  I can't imagine it's all that easy to make, considering there are 
only two US companies making the round stuff as is.

Tim

-- 
Seven Transistor Labs
Electrical Engineering Consultation
Website: http://seventransistorlabs.com 

"P E Schoen" <paul@peschoen.com> wrote in message 
news:l7ckot$107$1@dont-email.me...
> The eddy current loss reduction makes sense, and the Litz wire may be
> analogous to thin insulated laminations in transformers and motors
> designed for higher frequencies.

The analogy is no accident -- consider the laminations, which carry an axial 
field (i.e., parallel to the plates) and a transverse eddy current (looping 
around the perimeter; in essence, breaking up and stretching that perimeter 
by sawing the core into laminations increases the perimeter's resistance, 
decreasing eddy currents).  Current flowing down a wire is axial, with a 
transverse magnetic field -- it's a 90 degree analogy, but the same right 
hand rule is at work, generating phase shift, loss and shielding effects.

Presumably, braided steel cable would make excellent "mag-Litz", but in the 
same way it's difficult to make a connection to Litz wire (a soldered lug 
sucks all that evenly-distributed current onto its surface..), it's rather 
difficult to make a solid loop (with little airgap) of steel wires.  (One 
would hope to trace a given strand through the cable and somehow weld its 
ends together to eliminate airgap, doing this for the entire cable...)

Wire core toroids do exist, and work.  The coils can be wound around the 
core, as a traditional (ring core) toroid, or the coils can be ring shaped 
(as in a pot core construction), and the magnetic core wound around that 
toroidally.  They are very rarely seen... mostly as science projects I 
guess?  Does anyone know if anyone actually produced transformers with this 
method, perhaps very early models?

Tim

-- 
Seven Transistor Labs
Electrical Engineering Consultation
Website: http://seventransistorlabs.com