Models for RF transformers with extreme ratios?

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

On Apr 29, 11:31 am, Joerg wrote: ...

... What impedances?

Try searching for current-sense applications (where the primary is a single turn).

A reasonable starting point for a broadband transformer is a primary inductance whose reactance is the same as the primary side impedance. So if the bottom end is 2MHz and the primary is designed for 1 ohm, you'd want about 80nH or a bit more. An FT114-61 should give you about that with a single turn. But 100 turns gives you 800uH which will resonate with only 2pF at 4MHz. Properly loaded (10k ohms resistive) it should be reasonably damped. A Spice model will tell you pretty nicely what the response will be for any reasonable assumed coefficient of coupling; for example, assuming 80nH:800uH with k=0.9 (which seems like it ought to be pretty easy), and 2pF||10k secondary load, driven from a 1 ohm source, you get a 3dB bandwidth from about

1MHz to 12MHz, with no apparent resonance effects. k=0.8 and Cload=4pF only cuts the top end to about 6MHz. (The bottom is determined mainly by the reactance. Get to low enough coupling and the effective turns ratio drops, but at k=0.8, the mid-band is only a fraction of a dB below the "theoretical" 40dB voltage stepup.)

About what you assumed, around an ohm at roughly rectified AC level. That one ohm will be a whole 'nother story but I'll get it there. Somehow :-)

Thanks, Tom. I am currently at around 300-400nH primary but any capacitance on the stepped up side is killing things. The secondary in those cases is always largish because it needs to withstand a lot of breakdown voltage. The only feasible method is to wind the packet on top of the primary or use compartmentalized bobbins and make sure the other layer maintains enough clearance. The latter not so much for HV breakdown but to avoid stray capacitance to ground.

Probably the best avenue is to get a suitable large core and make one. I was hoping there was some example data from core/bobbin manufacturers and such but the usual suspects didn't have anything.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

A bandwidth of around an octave implies _some_ resonance.

You'll have a huge juggling job between leakage inductance and primary inductance.

The more loss you can stand the better chance you'll have of making it work. I'd feel a strong sense of accomplishment if I got 50% of my input power coming out of my secondary.

I recall reading in some ARRL publication or another (the one on transmission line transformers, I think) that to achieve extreme ratios you can often do better using two stages -- in your case perhaps three?

4:1 * 5:1 * 5:1 = 100:1.

Disclaimer: never done it, I wasn't there, it's not my fault, etc.

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

Oh, yeah, and I forgot to ask: how much power? ;-) My comments came from a low-power context, though they tranlate. The reactance:impedance thing should stay the same, assuming you avoid core nonlinearity. It actually came as a bit of a surprise to me how low a winding reactance is when you get to the low frequency cutoff. The relatively simple model I suggested has worked well for me: get the coupling coefficient up to extend the high end. The model matches several RF transformers I've measured. Up till resonances and other capacitive effects get to you, you can generally extend the response of a transformer by driving it with a lower source impedance and loading it with a higher impedance. I have a 1:1 audio transformer that, when driven with a low impedance and loaded with about 2k ohms, is flat within +/-0.1dB from 0.6Hz to 109kHz, but quite a bit worse if driven from 600/loaded with 600, and it shows resonant peaking at the high end if loaded too lightly.

Suggest you go for a core with modest permeability, probably around

100 (depending on path area and length), so you can drop the inductance down some from where you are. Harry D. seems to know a lot about this sort of thing; maybe he'll have some ideas.

Cheers, Tom

Eric Tart Red "Arbeitsbush fur den HF Techniker"

The very first chapter is about the RF transformers and the simulated line transformers, their equvalents and the compensation.

But the ratio of 100:1 in one stage doesn't seem reasonable; the sensible way would be breaking this into the series/parallel connection of the transformers.

This type of the impeadance matching can be done by a bandpass filter type of network (also using many stages); I would do it that way.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Not at liberty to tell ;-)

Yes, it's a compromise to push the upper end a bit. On line transformers it's to save cost on the copper. So when I need really low standby power I often use a 230V transformer at 120V.

Yes, a really low drive impedance is key. I'll just whip up a few and measure them on the bench. The data I can find at manufacturers is mostly based on usage at the impedance they are marketed for.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

On extreme ratios you just can't avoid it.

Yes, and you can never have half turns like at some restaurants.

Unfortunately that won't work in this case. It's usually only ok if you have an active stage inbetween.

Maybe you could come visit at the hospital if it blows up in my face ;-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hmm... how's the efficiency of that approach?

We have a design where there's a four-pin power connector, with two of the pins being 120-240VAC (nominally) and the other two being jumpered or not to configure a relay to "configure" a transformer to be either in series or parallel. Originally the idea was to keep the voltage at the transformer's secondary the same at 120V vs. 240V as you'd expect since the Vicor module that the secondary feed didn't have a 2:1 input voltage range. At some point, though, I found a different power module (thanks to Terry Given) that had very wide input range and suggested we get rid of the relay and proprietary 4-pin power connector (just go back to the regular IEC ones), etc., but there was an objection that running a 240VAC transformer at 120VAC would be "very inefficient." That didn't seem right to me -- if anything it seems as though it's probably a skosh more efficient at 120VAC since you're not pushing it anywhere near saturation -- but I don't have a strong enough background in power transformer design to rigorously debate the issue.

(...So we stuck with the original design...)

---Joel

As usual it depends :-)

A larger primary inductance reduces core losses, it thus becomes less warm and under light loads it can be a good thing. At higher currents copper losses would cause a penalty.

Then main advantage is that the core won't saturate if, say, a unit is running on generator power and an impatient maintenance guy gooses the throttle a bit until the old spark plugs that should have been replaced last year burn themselves cleaner and it stops misfiring.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Hi Joerg,

So if you need, e.g., 250W through the transformer (with either 120V or 240V input), it seems that you'd really want to buy, e.g., a 500W 240V transformer, yes? ...which of course will be bigger and spendier...

Ah, good point.

Not quite that bad but yes, you need to provide a bigger one. For high power you'd be better off buying or designing a good 90VAC-260VAC switcher.

I've seen a lot of grief in that area :-(

A lot of engineers believe all that can be permanently muffled by some MOV here and there. However, those work like a bank account and one fine day ... KABLOUIE.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Oh, ye of little faith. I have some FT82-67 toroid cores. I wound one with 100 turns of #29AWG solid enameled copper wire, and put one turn of #14AWG through as a primary. I loaded the secondary with 10k ohms in series with 51.1 ohms, with the receiver port of an HP8753E across that 51.1 ohms. The primary is in parallel with 1 ohm, and 50 ohms goes from that off to the source port of the analyzer. The response below 20MHz is what I consider to be very close to what my model predicts; I get a peak response at 10.68MHz, with bandwidths

-0.5dB 7.28-12.71MHz, -1.0dB 5.72-14.24MHz, -2.0dB 3.48-16.69MHz and

-3dB 2.89-18.76MHz. I would expect that Joerg, at his much higher power levels, should still have relatively little trouble getting an octave bandwidth in a single transformer (to -1dB, anyway) at the somewhat lower frequencies he's dealing with. He admittedly will have to be careful to avoid actual power loss and to minimize stray capacitance.

Cheers, Tom

Oh yeah, agreed. In this particular design it was actually an isolation transformer: Until you got to the secondary and past some more filtering/isolation, you couldn't use semiconductors... the transformer was supposed to help make the entire widget HEMP resistant.

I expect that a mag amp/vacuum tube-based switcher would meet the HEMP requirements, but there wasn't the time for that. :-) It would be entertaining to build one that way... Pressman's switching power supply book discusses one he did using a mag amp for the controller, but still using BJTs as the switches. Although using, e.g., nuvistors instead might reduce the efficiency so much that it'd no longer look that attractive vs. a linear.

We have a few spark tubes to help as well? :-)

---Joel

Hi Joel,

like Joerg said, it depends. A lot of transformers really beat the crap out of the steel, which saves turns, so losses are almost entirely due to the core - Microwave Oven Transformers are a drastic example of that. In such a case halving the voltage but doubling the current can easily result in greatly reduced losses, even though copper loss quadrupled.

My 3kVA light dimmer design couldnt use the original autotransformer, which had almost 1.8T peak flux density (nope that aint a typo). I cranked that down to a more reasonable 1.5T, but of course core "memory" caused me grief, so it now runs at 80mT.

That nasty little autotransformer got filthy hot, and even at 16A the copper loss was small (ICR the actual number). I'm astounded that anyone would design such a pig. funny thing is, the 800mT replacement is 2x larger and costs 1.4x more :)

Cheers Terry

Please give some more information about why you can never have half turns. I remember overheating a transformer that used a half turn. Never tried half turns again. But what is the reason the half turn gets hot? Mike

With a toroid core, it should be obvious why you can have only integral numbers of turns. In an E-I core, or a pot core with openings on both sides, you can have a wire exit a different place than it entered. The loop then closes around one of the outside "legs" of the core. Note that this is equivalent to a full number of turns around the center post, and one turn around the outer post, with the two connected in series. IF the magnetics are balanced, the field in the outer leg will be half the field in the center leg. But this happens only if there is no current in the turn around the outer leg. Note that the "half turn" is not strongly coupled to the rest of the turns, and as a result adds a lot of leakage inductance. I don't see why the "half" turn itself should get hot, but if it diverts the magnetic field into the other leg in such a way that it significantly increases the core loss in that leg, it could lead to excess power loss in the transformer.

Cheers, Tom

Which got me to thinking: you can keep the magnetics in the two outer legs balanced (that is, the rate of change of flux per unit time) if you put a turn around each and put those two turns in parallel. However, each will see half the flux that's in the center leg, so will contribute half a turn's voltage...this could be an interesting way to get a high step-up ratio with fewer secondary turns: only 50 turns instead of 100, to get a 1:100. That could be an advantage in keeping the parasitic capacitance on the secondary at bay, though the effective capacitance is generally a very weak function of the actual number of turns--and for modest permeability cores, significantly lowers the pri:sec coupling as compared with having the windings co- axial.

Here is a patent on the subject:

--
Regards,

John Popelish