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

Tue May 16 11:36:31 EDT 2017

Chris,

at that low frequency, indeed you can make the individual amps for 100 
ohm and put them in parallel, or for 25 ohm and put them in series. 
There should be no phasing problems, at such a low frequency. But you 
have to make sure that both amplifiers have extremely similar responses, 
in terms of gain curve mainly.

The advantage of using a combiner is that it isolates one amp from the 
other, so the system is highly tolerant to differences between the 
individual amps, down to the point where one amp module can completely 
fail, and the other will continue limping along. Without a combiner, the 
failure of one module would probably cause the demise of the other too.

I have been giving some thought to your transformer. First, I suggest 
NOT buying that giant toroid. Bigger isn't always better. A huge core 
has a lot of ferrite, that causes losses! That huge toroid has an 
enormous space for winding, which you would never take advantage of. A 
smaller core in a better design can provide far better performance. So, 
let's start optimizing this design:

The first step is turning it into an autotransformer. The schematic you 
linked shows a conventional transformer, with separate primary and 
secondary, and both of them having one end grounded. That's very 
inefficient! There is no reason at all to use separate primary and 
secondary windings, if anyway they will be connected together! So, for a 
start I would change this into an autotransformer: 7 turns total, with a 
tap at 5 turns - or multiples of that, depending on the core used. This 
eliminates 36% of the wire, and strongly improves coupling, without any 
ill effect at all. Also transformer action is now required for only 571W 
instead of the full 2000W, which allows using a far smaller core, and 
far less total copper.

In that autotransformer, the current coming from your amplifier enters 
the transformer through the tap, and splits in two parts: 5/7ths of the 
current flows "up" through the 2 turns and into the load, while 2/7ths 
flows "down" through the 5 turns and returns via ground. The voltage 
applied to the 5 turns induces 2/5ths of that voltage in the 2 turns. So 
the output voltage is 7/5ths of the input voltage, while the output 
current is 5/7ths of the input current, the transformer works at only 
2/7ths of the total power (571W), and at the output you still get the 
full 2000W.

There is an old adage: Engineering is a combination of material and 
brains. The more you use of one, the less you need of the other.

Okay. Now lets try to come up with a good transformer for those 571W. I 
will write here as I attempt to design it, so you can learn how to do it.

A good core shape is an RM or a pot core. They have bobbins (easy to 
wind), round center legs (even more easy to wind), a much shorter path 
length than a toroid, and they are available in suitable materials. The 
catch is that they aren't very large. So, let's take the largest RM core 
offered by FairRite, and see very dumbly how it works out.

This core is available both in the 95 and the 98 materials. They are 
quite similar, but I like 95 better because of its flatter loss versus 
temperature curve. So, the chosen core would be the 6295420121.

First let's find out how many turns we need. 2000W on 50 ohm is 316V. 
This core has a cross sectional area of 1.95cm². Its volume is 14.36cm³. 
How much power can we make it dissipate? That's a decision one has to 
take. I would say, 2W is fine for continuous use, some more is 
acceptable for intermittent use. So, at 2W to be on the safe side, 
139mW/cm³ loss is acceptable. Looking at the material loss chart given 
by Fair-Rite, an acceptable maximum flux density at 136kHz seems to be 
0.11T.

Now we can use equation 4 on my page

http://ludens.cl/Electron/Magnet.html

to calculate the required number of turns:

316V /4.44 / .000195m² / 136000Hz / 0.11T = 24.4 turns

That's the minimal requirement. Since we need multiples of 7 turns, 
let's use 28. So the recipe is 28 turns total, with a tap at 20 turns.

We don't need to make the 28 turns of the same wire, since the top 8 
turns carry more than twice as much current as the lower 20 turns. To 
evenly distribute losses, it's better to distribute copper cross-section 
according to actual current flow.

Also it's hard or impossible to wind very thick, stiff wire on such a 
bobbin, and on top of that thick wire suffers badly from skin effect. It 
follows that you should wind this transformer with several strands of 
thinner wire. That invites using a single size of wire, but using more 
strands for the 8 turn winding than for the 20 turns.

For best coupling it would also be optimal to interleave primary and 
secondary layers. In this case you could first wind a layer with 10 
turns, then one layer with 8 turns, then a third layer with 10 turns, 
and interconnect the three layers properly so that the 8 turn layer ends 
up at one end of the other 20 turns. This scheme is still reasonably 
easy to do, but doesn't allow us to use optimal copper cross sections 
for each winding... Anyway, let's try a modification of it:

The winding space on the bobbin is roughly 18*8mm. We can start from 
enamelled wire of roughly 0.7mm diameter (AWG #22), and wind 10 turns 
with two strands side-by-side. That's a total width of 14mm, which 
should fit in the 18mm bobbin space despite some slight kinks and 
imprecisions. Try to keep the winding centered, leaving some empty space 
at each side, because this reduces the risk of flashover between layers.

Then wind two or three layers of Mylar or Kapton tape, cut just a tad 
wider than the bobbin, so that it seals well against the bobbin sides.

Then comes the 8 turn winding. For simplicity let's use the same wire, 
but 4 strands instead of 2. Wind 4 turns, with the 4 strands nicely 
side-by-side. That will use up most of the bobbin width. Then wind one 
or two layers of Mylar or Kapton tape, threading the four wires through 
it, and then wind the other 4 turns. So this is a double-layer winding, 
with both ends coming out of the bobbin on the same side.

Now wind another two or three turns of Kapton or Mylar tape, and then 
wind the topping layer of 10 turns of 2 strands of wire. Finish with 
another few layers of tape.

The whole thing should be only around 5mm tall, fitting comfortably in 
that bobbin.

Now the windings have to be interconnected. A lot of wires will be 
sticking out of the bobbin... The two 10-turn windings have one end on 
each side of the bobbin, while the 8-turn winding has both ends on the 
same side. First thing is to take one end of one 10-turn winding, and 
the opposite end of the other 10 turn winding, and join them. This can 
be done on top of the Mylar tape, at a place of the bobbin that will end 
up in one of the core's openings. Make the connection nice and short.

Then the still free end of a 10-turn winding on the same side of the 
bobbin where both 8-turn ends come out, has to be joined to the CORRECT 
end of the 8 turn winding. The correct one is the ENDING, not the 
BEGINNING, assuming that you wound everything in the same direction. So, 
join that, fit the core (you can tape it together for now, later glue or 
clamp it), and the transformer is ready for testing.

I would expect this to work pretty well, although it's not entirely 
optimized. We could have used more strands of a thinner wire, and 
interleave 3 primary with 2 secondary layers, for example. Anyway it's 
MUCH better than winding separate primary and secondary on a stack of 
large toroids, let alone a giant and non-optimally shaped toroid!

Manfred

========================
Visit my hobby homepage!
http://ludens.cl
========================