[Top] [All Lists]

Re: [Amps] Real time tests to see if an RF transformer is saturating?

Subject: Re: [Amps] Real time tests to see if an RF transformer is saturating?
From: Manfred Mornhinweg <>
Date: Tue, 13 Jun 2017 15:37:37 +0000
List-post: <>

Thanks  for  clarifying  that,  I  have  the  20 turn winding with the
grounded  end coming out on the side of the core with no terminal pins
and joining the 8 turn winding on the side with the pins.  I can't see
it will affect things though.

I think you have it correct, but I'm not sure. As long as it works, it's
fine, though! :-)

Whilst I have your magic ear maybe you
can  advise  on  the next heating issue please? With two amps combined
the  centre  toroidal  core  and winding get pretty hot if I use OPERA
with  its  32 minute transmission. The two end cores and windings stay
cool.  I  wonder  if you can suggest a better core or whatever please?

I can try.

In that lowpass filter, the center inductor has twice the inductance and
works at twice the voltage but the same current as the two end
inductors. Yet the designer in his infinite wisdom made all three on
identical cores. Thus the center inductor has 1.41 times the turns as
each of the end inductors.

I don't remember your absolute power level anymore. Assuming 2kW, the
voltage on the load is 316V, and that's also the voltage on each end
inductor, while the center one has 632V on it. This is at the design
frequency of 136kHz. The voltages get higher at higher frequencies,
which might be important in case the inductors came out a little low in
inductance and thus are working above the filter design frequency. But
let's assume this is not the case.

The end inductors have 52 turns each, so they work at 6.1 volts per
turn. Instead the center inductor has 73 turns, and given twice the
voltage runs at 8.7 V/turn.

Considering the effective cross sectional area of the T-225A-2 core,
this causes a flux density of 37mT for the end inductors, and a core
loss of 9.7W in each. Instead the center inductor works with a flux
density of 52mT, and a core loss of 20.8W. No wonder it gets hot, but I
don't swallow that the end inductors stay cool! Surely not as hot as the
center one, but at nearly 10W core loss they should get pretty warm too.

So much for the cores. The wire is #14, is working at 6.3A, the end
inductors have around 3.8 meters of it, while the center one has around
5.3m of it. The AC resistance at 136kHz is 0.08Ω for the end inductors,
resulting in 3.1W of wire loss, while the center coil's AC resistance is
0.11Ω, resulting in 4.4W wire loss.

Clearly core heating is most of the problem, rather than wire heating.

Much larger toroids won't fit in that box, but it looks like a stack of
two of those cores would fit. Using two cores, and 52 turns, you would
get the correct inductance, all cores would work at identical flux
density, and the wire length on that stack would be roughly 5.7m. As a
result the center inductor would have a core loss of 19.4W and a wire
loss of 4.6W, so the total loss would be almost identical as with a
single core! The only thing causing some temperature reduction would be
the greater surface area available. It doesn't look like a good solution.

Let's try another material. Generally when a core material is too lossy,
you have to go to a lower permeability one. This would be material 6.
It's not available in T-225A size, but T225-6 cores exist. Stacking two
of these is just a tad taller than a single T-225A. So let's try that.

The slightly lower permability is almost compensated by the slightly
taller stack, so that you would need 75 turns. The core loss comes out
as 16.8W, and the wire loss as 4.6W. So there is a small reduction in
heating, but not enough to be worthwhile.

There aren't many iron powder materials available in such big toroids,
so I don't see much more to do there. Your last option without changing
the box might be using two of those toroids, individually wound with 52
turns each, connected in series. That has the same core loss as stacking
the two cores, has higher copper loss than the other solution, but gives
you the full dissipation surface of both cores. And with some luck it
might fit in that box.

Let's try our luck with a compact ferrite core instead of a powdered
iron toroid. For simplicity, I will try with the same core you used for
your autotransformer, and I will assume that you have tested it enough
to confirm that at 2kW for all the time you need, it doesn't overheat,
but does get quite warm, which would indictae that the design is about

Ferrite has a much higher permeability than powdered iron, and that
calls for a different technique to build inductors. You absolutely need
a split core (not a toroid), and you basically calculate turns number so
as to get acceptable flux density, and then you introduce an air gap to
reduce the huge resultig inductance to the value you need. This air gap
also gives enough stability against temperature-induced variations in
the permeability, and aging.

We can base the turns calculation on your existing autotransformer. You
have 28 turns there, for 316V. The center coil in your LP filter works
at twice that voltage, so we need twice the turns. As simple as that.
This will result in the same amount of core loss as the autotransformer
now has.

The Fair-Rite 6295420121 core has an AL value of 7.5µH for one turn. So,
56 turns will give you a whopping 23520µH! But you only want 115µH. That
means that you need to reduce the effective permeability by a factor of
around 205. Given that the material permability is 3000, that means a
new effective permeability of roughly 14.6. Since 3000 is so high
compared to 14.6, we can simplify the calculation by considering that
the ferrite has no reluctance, and that all of the reluctance will be in
the air gap.

This means that if the air gap had the same cross section as the ferrite
core, the gap would need to be the core's path length divided by 14.6.
With the core having a path length of 73.8mm, this means an airgap
length of 5mm. Since separating the core halves introduces a gap both in
the center leg and in the outside return legs, this calls for separating
the core halves by 2.5mm.

But what will happen in this case is that the flux bulges out in the
gaps, using a greater cross sectional area than the ferrite, and as a
result the inductance will end up larger than desired. You need to
further lenghten the gap to achieve the correct inductance. The only
practical way to determine the required gap size is to connect the wound
coil to an inductance meter of any kind, and adjust the gap size to get
the correct inductance. You might end up with a core half separation
much larger than 2.5mm, possibly reaching the point where the core
halves barely enter the bobbin!

If you wind that coil on that bobbin, with wire so thick that the bobbin
is filled up completely, the resulting inductance in air, without a
core, would already be almost half of the 115µH you need! This
illustrates how little you will need to insert the core halves.

This raises the idea of using an air-core, multilayer coil instead. The
problem is that such a coil isn't self-shielding, and needs to be
mounted at a fair distance from the metal surfaces of the box. And your
box is too small for that. The core halves will deliver a fair amount of
self-shielding, even if only slightly inserted.

If my calculation is correct, you could use wire roughly 1.3 to 1.4mm
thick (AWG #16), or rather a bundle of thinner wires totalling the same
diameter, so you can wind it... I calculate roughly 0.12 ohm AC
resistance for that wire, so at 6.3A you would get roughly 4.8W loss in
the wire, plus the 2W core loss assumed for calculation last time. This
represents a considerable improvement over the inductor you have now -
but before calling this a success, you have to try it. There are too
many details where I might have made mistakes.

If you try it, you will need to find a way to fix the core halves in
exactly the right position, holding the air gap constant. For testing
it's probably OK to tape small pieces of cardboard to the cores, like
struts, but a good definitive system might be using non-corrosive
silicone caulk, forming 4 beads of it that join the core halves and also
bind to the coil surface. This is stable over time, heat resistant, and
the silicone's elasticity allows fine-tuning the inductance value by
using any kind of clamp with a screw that presses down on the core.
There you have my new invention: A compression variable inductor, a
great counterpart to the well-known mica compression trimmer! :-)

If you prefer not to try this ferrite kludge, at least you can improve
the cooling of your powdered iron core. Those toroids are mounted in a
way that blocks air flow through their holes, and also impedes free air
flow on their sides! Get rid of that, and make some mounting that allows
free air flow.

And then go and buy a can of flat black spray paint, and liberally paint
the coils and the box (inside and the outside) flat black. That will
drastically improve heat transfer by radiation! I do that even with the
innards of my computers, so that I don't need noise fans! Your very
shiny boxes are formidable heat reflectors!

The  LPF  is detailed here, and they are in individual sealed die cast
alloy boxes,

The photo shows sheet metal boxes, not die cast. But that's irrelevant
in this case.

so  air  flow is none existent, but I suspect even if I
mill two openings in the centre box and put mesh over the apertures it
will  still be marginal. A fan sound like a kludge. Ideas very welcome

I would take my drill press and drill a few hundred holes in those
boxes. In addition to painting everything flat black, except the contact

And after that, stop worrying, because iron powder toroids and enameled
wire can both take some heat.    The schematic is in the link.


Visit my hobby homepage!

Amps mailing list
<Prev in Thread] Current Thread [Next in Thread>