transformer thermals

Iron losses are zero with DC testing, and certainly not 'lower' at all higher frequencies. Core magnetization goes down with frequency, but losses go UP when the secondary doesn't draw current. And, the core shape determines the magnetization profile, at higher frequencies it's unclear whether the entire core shares the induction (so the behavior at 4 kHz ought to be tested... at 4 kHz.

It sounded like the double the heat assumption was from the 50% of losses being in the iron, when the transformer is used with AC. Even if you double the primary resistive loses to account for the secondary, the core is still being left out. That's how I read it.

Could be- I have no references for core loss in small toroidal tranformers handy. It should be fixed though and less than the losses needed to magnetize the core with no load, so yeah, the 20 + 6 looks ok.

As for the PF has no effect statement, I've never really understood the point on the VA in and VA out ratings on some toroidal transformers. Isolation transfomers are usually marked in a more sensible way like "input 120VAC 8.4A, output 120VAC 8.0A". While these are safe use ratings, it still doesn't answer the question about efficiency at no to full load. You can only guess the worst case heat losses.

Are these ratios exact or is there a fudge factor for losses at rated load? Just curious. Retired transformer designer friend seemed to spend a bit of his time trying to get "the facts" from customers to make designs that would actually work under load. The stubborn customers would get full production runs of stuff that met specs, but didn't work for the application.

Was amgis friendly about making samples?

It looks like they could dump more copper on the typical toroidal transfomer. The diameters for large ones get huge, with gigantic empty holes in the middle. The stacking factor, if that's the right term is terrible on most toroids. Not sure why this is though. They seems to scale the wide and thin vs. taller with more copper stuffed in the hole and around the rest of it.

Was the core really more than 3x oversize?

Exact turns ratio. We know the other parameters: wire resistance, mag inductance, leakage inductance, SRF, saturation. We plug all that into the system Spice model.

Just curious. Retired transformer designer friend seemed to spend a

Not free samples, but I didn't ask for that. We placed a PO that included a few first-articles for verification, with the rest shipped on approval.

No, I'm assuming zero core losses, but figure the copper loss will double if I have 10 amps RMS AC into the primary and load the secondary to get there. That's probably close.

Any decent power transformer runs cold at rated voltage and no load. Core losses can be ignored.

Iron is cheaper than copper.

================================

Max heat loss happens at max load, always.

Maybe on on planet Claire where know nothing jerks like you come from.

..... Phil

For windings, it's a 'fill factor'.

If you check the ratio of copper vs air in the inner diameter of a toroid, as the fill approaches a certain percentage, you'll see a diminishing return for agravated difficulty in fabrication.

Any time manual winding methods enter the equation, costs traditionally skyrocket. There may be different considerations in today's off-shore sourcing.

Forty years ago, you would be considered foolish to ship anything with a high density - resulting in relatively local magnetics fab.

RL

Ultimately yes, but the curve is multidimensional and non-linear, especially when the frequency is high. I wonder if there is some FEM simulation evidence actually supporting this rule.

Sure, but it is a local optimisation. Globally one may get 8W of losses instead of 10 by disturbing this balance.

Yes, this is a good idea in low volume.

Best regards, Piotr

We rated the tranny for 100 Hz to 4K, but the box currently works from

Yes, and copper losses go up with frequency, typically, too.

==========

But as here like to consider what physics allows, it is clearly a fact that filling a toroidal with as much copper as possible maximises the VA for a given core.

This IS the practice for all other types of core shape.

Most toroidals are low and flat - again not optimum but liked by many customers. Same goes for R-cores and U cores which result in low height products.

.... Phil

Those last few turns also have the longest length per turn, further diminishing their 'return'.

The 'practice' is to assume a fill factor of less than 80%, to allow for insulation, wire shape and cover. For mains voltage wall thickness, creepage spacing, and bobbin tolerancing, it's even worse.

Toroids depend upon 3xlayer film overlapping weave for reinforced layering and core or outer wrap, though some cores use fitted caps at the expense of efficiency.

Low and flat toroids are a 'style' (at the expense of efficiency) - the most efficient physical ratio being roughly 2.2:1 / OD:H.

High frequency stuff . . . .

RL

I'm intrigued by the matrix transformer concept: two transformers with primaries in series and secondaries in parallel have both 1/4th the i^2*r losses each, and more dissipation surface per watt compared to a single giant lump of copper buried in steel.

Cheers, James

You'd probably gain more than that on the primary at least (especially at 240V) because less of the volume would be used up by the insulation--you'd have more than twice the copper cross-section and the volts per turn would be the same.

Cheers

Phil Hobbs

Utility transformers are very efficient, over 99%. They eliminate surface area from the efficiency tradeoff by oil cooling, sometimes with radiators and fans.

Absent concern about cooling, the matrix thing sort of collapses with the observation that all the windings can share one core.

Flex PCB winding transformers are neat. Power densities are outrageous.

Hey, are these new?

Well sure they can use one core, that's where we started. The matrix transformer gets its advantages from /distributing/ the core and the windings.

- Heat is spread out over more area, and current-sharing reduces i^2*r loss in each transformer-element.

- Leakage inductance is proportional to turns^2, which is reduced by 1/n when you split, e.g., the secondary into n sections on separate cores.

- And of course switching loss is lowered, since leakage inductance energy is lost to the snubber each cycle in most switchers.

You could make GaN switchers run at super speeds.

A further advantage of low leakage inductance claimed is that the primary and secondary currents are in such tight phase, that the primary winding and synchronous rectifier on the output can share the exact same timing.

I was surprised to read recently that transformers' power-handling capabilities aren't core-size limited, but by one's ability to get out the heat (core-loss, and i^2*r).

That's a nice upgrade. Best I've seen 'til now were 500V.

Cheers, James

They don't spec max winding voltage. They are sending me 10 samples on the condition that I blow up a couple.