Reducing size of isolated power supplies

[5PNdcc0Jety0]

Hello world,

I've been recently musing about how is it practically possible to make isolated 220V to 20V AC-DC power supplies without bulky passives.

The most obvious one is the size of isolation transformer, followed by safety capacitors.

Can one make a power supply for a < 100W application using a smaller transformer if instead of flyback topology, you drive the transformer with high frequency AC, and rectify after it?

Is it practical to forego the transformer altogether, and instead use something like piezo transformer, or capacitive isolation?

[David Hess]

The usual solution is to operate at higher frequency so that the passive parts including the transformer can be smaller.

Piezoelectric transformers exist but are commonly used only where very high isolation is required.  I do not think they have very high power density.

[5PNdcc0Jety0]

It's clear that they do get smaller with frequencies, but ones going to simple flybacks are already much smaller today that say 20 years ago.

Would the size of passives needed to do AC-AC conversion to drive a very high frequency transformer be small enough to not to offset the gain from a smaller transformer size?

[David Hess]

It's clear that they do get smaller with frequencies, but ones going to simple flybacks are already much smaller today that say 20 years ago.

The advantage of high frequency operation applies just as much to flyback designs as the others.

Would the size of passives needed to do AC-AC conversion to drive a very high frequency transformer be small enough to not to offset the gain from a smaller transformer size?

I am not sure what you mean.  The advantage of higher frequency applies to the passives and the transformer up until the point where losses increase which require physically larger parts to dissipate increased heating.

[ocset]

if you want 100W and small, then its more difficult because you need PFC if you are in some countries.

Also, SMPS for offline connection are usually with switching frequency <150khz as thats where the EMC limits start.
I mean, the smallest way to do it would probably be to use a Integrated transformer....like a transformer in a PCB.
But they are quite expensive.
Otherwise, i reckon the smallest way for you would be to use a power.com innoswitch, which gives a small solution because the isolation is in the control chip which sits nicely under the transformer.

[penfold]

...
I've been recently musing about how is it practically possible to make isolated 220V to 20V AC-DC power supplies without bulky passives.
...
Is it practical to forego the transformer altogether, and instead use something like piezo transformer, or capacitive isolation?

Do you perhaps some real application with the target numbers you can put to this, or is it a general musing?

One factor that jumps to mind is heat. Picking up an arbitrary converter design (including passives and everything), you'll likely be able to start playing Tetris with the parts and squeeze the envelope in which they fit quite a bit smaller, but in doing so, you lose a lot of the surface area for heat to escape at the same time as increasing the W/m^3. So from that viewpoint, the reduction in the size you could gain by removing (or using sub-optimal) EMI and filtering caps, might not necesarily translate into an overall reduction in size once you've done the thermal calculations. And you may end up having to design in a bit of free space to give yourself more area over which to dissipate if you can't otherwise extract the heat.

[5PNdcc0Jety0]

...
I've been recently musing about how is it practically possible to make isolated 220V to 20V AC-DC power supplies without bulky passives.
...
Is it practical to forego the transformer altogether, and instead use something like piezo transformer, or capacitive isolation?

Do you perhaps some real application with the target numbers you can put to this, or is it a general musing?

One factor that jumps to mind is heat. Picking up an arbitrary converter design (including passives and everything), you'll likely be able to start playing Tetris with the parts and squeeze the envelope in which they fit quite a bit smaller, but in doing so, you lose a lot of the surface area for heat to escape at the same time as increasing the W/m^3. So from that viewpoint, the reduction in the size you could gain by removing (or using sub-optimal) EMI and filtering caps, might not necesarily translate into an overall reduction in size once you've done the thermal calculations. And you may end up having to design in a bit of free space to give yourself more area over which to dissipate if you can't otherwise extract the heat.

Just a common consumer AC-DC in under 100W. Which today basically means USB-PD range of 5V-20V.

[Berni]

Going to high frequency doesn't just help the size of the transformer but also all the other passives. Also all typologies benefit from high frequency because this makes the minimum inductance of the transformer smaller, this means you can use less turns of wire and smaller cores to handle the magnetic fields.

Usually what is more of a problem is heat. Going higher in frequency increases the switching losses, this heat has a harder time getting out when everything is tightly packed together, as well as the device being smaller gives it less surface area to dissipate the heat. So at some level of miniaturization it might become necessary to use a fan to keep it from running too hot.

[Siwastaja]

The days of just increasing frequency to reduce passive size are over long ago (two decades, maybe).

Increasing frequency by itself lowers efficiency and does that in a way that is nontrivial to compensate for.

The key is to concentrate your efforts in improving the efficiency. Then you can, at the same time, increase frequency and reduce the size. These goals work against each other so it's not easy.

Efficiency improvement have other obvious benefits as well - energy bill savings, CO2 emissions from energy usage. Traditionally, these have not been in the interests of the supply manufacturer, and still aren't, but miniaturization is finally driving efficiency higher as well. In many cases, it has also happened that heatsinking (cost of metal, if nothing else!), packaging, shipment volume, retail package size etc. are dominating costs compared to silicon or R&D. Large R&D volumes help. So a modern supply is often highly efficient, over 90% is nothing special now.

[riccardo.pittini]

High frequency doesn't always mean small. The design issue is rather an optimization problem.

High frequency:
- Smaller passives (flyback is not optimal for HF)
- For real HF you need a soft switching topology
- More difficult to achieve high efficiency
- Magnetics design is more difficult and depending on the frequency it might requires litz wires (and above 2-3MHz also Litz don't help... so you have to go back to solid copper)

To get a small power supply as a rule of thumb you need to work on the following topics:
- topology for application
- you cannot take an individual component and optimize it to get a small power supply
- optimize the whole system (not only trafo, but mosfets, capacitance, cooling system, EMC/EMI filter, thermal management

Some of the best design that you can find are Vicor designs and research papers from ETH Zurich (Kolar power electronics group).

[tszaboo]

You can replace the regular FETs with something like GaN so you have larger frequency. And there are technologies like planar transformers, that are smaller, and operate at higher frequencies. Usually higher quality components can reduce the size, there are stacked MLCCs, Polimer capacitors with extremely small ESR, FETs with "smarter" packages, like DirectFET.
The physical construction is something else. Having 2-3 PCBs allows you to use larger area for components, and use any given 3D space more efficiently.