3 Isolated AC Outputs from 1 AC Source

I missed it too. Wasn't in the original post. How about a toroidal core with 4 windings on it.

Tam

I'd still try these even at 32 KHz. Otherwise a custom transformer it is.

I wonder where the 32 KHz comes from. 100 Hz would work well with a 60 Hz transformer, and going to 128 KHz or so would mean a lot fewer turns on a home made transformer. BTW, I once measured the frequency response of a 60 Hz 12 V transformer. Minimum loss was at about 400 Hz. Can't remember the high frequency 3 db point, but my guess is somewhere around 2 KHz. I tried one as an audio transformer; usable, but with a mellow sound,

Tam

That seems about right. Powerstat variable transformers are rated for operation from 50 to 1000 Hz. They are just specially wound toroids. They are probably better than cheap EI power transformers, with better quality steel and thinner laminations.

Paul

"Tam/WB2TT" <

** All depends on the arrangement of the windings.

A small, bobbin wound tranny with primary and secondary side by side is usable only to about 1 kHz before leakage inductance dominates the output. However, if the secondary and primary are wound on top of each other, the frequency range extends to 10 kHz or more.

A toroidal mains tranny will operate well out to 50kHz.

...... Phil

100V audio line transformers might work for you (after passing through one of these a 32Khz square wave will be significantly rounded), these step down so you you'll need to pair them.

I have no idea if the provide sufficient isolation for your application.

--

Bye.
   Jasen

[starting at 4.5V with 35 kHz modulation, need up to 25 mA output]

Lots of folk have mentioned that '60 Hz' power transformers won't work; that isn't quite right, they work but they heat up and waste power. The ferrite cores like in a switchmode power supply (and this design IS a SMPS by any reckoning) are a more suitable magnetic material.

Custom-winding your transformer is the best way to accomplish this, and it isn't terribly hard. Amidon, Stant, Ferroxcube, TDK all supply the parts. You might use a voltage-doubler rectifier, or tripler, if the turns ratio is inconveniently high.

Don't wind the wire too tight, and put a final tape wrap on the completed windings so they don't come loose. Use a spool with separate sections for the input and the three output windings, to minimize capacitance between them. Regular lacquered magnet wire is good for 500V or so, I kind of prefer Beldsol type because a hot soldering iron burns through the insulation (no need to strip the varnish).

For input drive, a center-tapped winding with your program voltage on the center tap and alternate grounding of the two legs is a good approach. The outputs can be rectified with a blocking capacitor and two diodes, clamp one to GND and the other to a filter capacitor- this is a 'voltage doubler' because the output is equal to the peak/peak excursion instead of the peak-to-common voltage. The 'GND' is, of course, only the reference for the filter capacitor and this ONE section, the other two outputs have isolated 'GND' points, none is really Earth ground.

Many SMPS designs close the loop, by having feedback from one of multiple outputs, and you can add a fifth winding if that is desirable.

Not knowing the situation, it MIGHT be a good solution to make a simple one-output design and just build three of 'em after debugging the first.

** Groper V Groper Alert !

** Most 60 Hz transformers are not able to pass 32 kHz without severe attenuation - due to LEAKAGE INDUCTANCE between the primary and secondary. Toroidal ones are an exception since the two windings are so intimately coupled.

It has nothing to do with the use of iron ( really steel) laminations because at such high frequencies the ( massively oversize) core is barely being magnetised at all.

....... Phil

the leakage can easily be tuned out.

Iron is suprisingly awful at high frequencies. a few tens of mT can generate significant losses at tens of kHz. A typical steel transformer runs at about 1.5T (peak) at 50Hz, so at 50kHz thats about 1.5mT - vanishingly small indeed. But try it, and watch it get HOT.

In the early 90's I did a lot of work with chokes for the output of AC drives, which switched at a mere 2kHz - 8kHz. We started with pre-purchased 3-phase iron core chokes, and they got RED HOT. Even when operating a machine at no load (IOW very little current). Removing the core showed this was almost entirely due to magnetic losses (winding AC resistance was much greater than DC resistance of course, but at very low currents that only accounted for a small fraction (< 10% IIRC) of the observed temperature rise.

part of the problem is the steel isnt really GOSS in an E-I/E-E core, as for about half the magnetic path length the flux is going at right angles to the grain orientation (if any), thereby maximising loss. There are methods of making cores using GOSS that avoid this problem eg fold a strip into a U, then stack multiple U's on top of each other - Voila, all flux flows with the grain.

OR tape-wind a toroid with GOSS.....

Then when you look at the skin depth, you see that the effective core area is a LOT less than a ruler indicates (which is, after all, the reason individial lams are insulated), hence the flux density is higher than the initial calculations suggest.

Interestingly enough, at very high frequencies even Ferrite needs to be laminated - although I have not yet seen such a beast, papers detailing them exist.

1mil metglass was pretty good, although appallingly expensive.

Cheers Terry

"Terry Given"

** The word " tuned " implies at a single frequency - so I guess you mean " tuned " by a series cap that resonates with the leakage inductance at that frequency.

Something to keep in mind for single frequency apps.

But the OP requires the tranny to pass a 32 kHz square wave.

....... Phil

true. with a bit of work you can extend this to more than one frequency.

Or, bifilar wind a EE/EI tranny.

Or, use the leakage inductance and make a quasi- or multi-resonant converter (way, way beyond the OP) that uses the leakage for soft switching. a phase-shifted converter would work well here.

with a push-pull primary, its not much work to bung in a DC choke, add a resonating cap across the primary and make a so-called royer converter, with sinusoidal output (Bill, this is your cue ;). then the tuning really does work.

technically he wants isolated, stepped up DC....but thats just me being pedantic.

Cheers Terry