A different approch to switch mode power supplies

[Downunder35m]

We love our switch mode power supplies and they have come such a long way...
More efficient than a traditional transformer, lighter and by weight often much more powerful.
And down to the most basic level they ARE still just normal transformers.
In case you don't know how they work, here is a dirt basic explanation:
The AV from the mains is rectified to DC.
Then this DC is chopped up to result in a high frequency set of impulses or for the better ones are square wave or even sine wave.
Either way the AC from the mains was basically transformed into high frequency AC that goes into a transformer.
On the output of said transformer you get the required AC to rectify it to the required DC for your device.
Like a reverse inverter if you like....

We do all this chopping because at higher frequencies a transformer is much more efficient than on the usual 50 or 60Hz.
And the variation of this principle are indeed plenty-full.
If we look at the less common ones we can find things like cheap induction heater circuits, step up and down converters, neon sing transformers....

If you and me would design a basic switchmode power supply we would probably base our little transformer on the same principles as any traditional one.
Meaning we match the input voltage and required output voltage through the number of turns and their ratio on the transformer.....
But do we have to do it this way ?

Power supplies for the masses need to be cheap.
And so these power supplies usually operate anywhere between just 15kHz and around 200kHz.
Above that the ferrite material for the core becomes a problem and below that they often cause too much audible noise.
Diodes to rectify high frequency AC are a bit more expensive than standard types if you require a few amps or more.
But why not utilise resonance for a switchmode power supply ?

As with many of my bad ideas I got this one while doing something totally different - playing with an old flyback transformer trying to get the best arc out of it.
You might know these standard circuits with basically just a transistor a 'power' coil and a feedback coil.
I used a circuit that also used a diode and capacitor over the transistor - utilising some of the otherwise wasted energy from the collapsing field.
For some reason I saw a connection here.
What I wrapped around this flyback core basically makes the entire thing a resonant transformer.

[Downunder35m]

After playing around with different capacitor values I got tired of it and reverted to some old school Layden jars to adjuts the capacitance by changing water levels.
The output as measured through my Nickel plated eye sight and based or arc length, colour and thickness was somewhat far from related to my water level adjustments.
I started to note it down like a basic chart and this showed me the output peaks at various water levels.
No, I did not measure the value of the bottles I kept it dirt simple....
I also noticed that the number of turns on the 'power' coil dramatically changed the output - but again not always as expected.
The feedback coil only had to drive the transistor hard, so I left it untouched once this worked good enough to be reliable.

A higher number of turns on the 'power' coil resulted in stronger arcs but in MOST, not all cases also a lower power consumption.
On the other hand a few turns too many and the output would be crappy.
The least possible number of turns that wouldn't blow things up resulted in the longest arcs.
So you could say so far so good and logic works just fine.
That is until I added the diode and bottles to try to find the best capacitance for the three different 'power' coil winding numbers I picked.
Three with the risk of blowing the transistor, 5 and 9 turns.
For all them I could find water levels that resulted in a nicely ionised arc with good power while the power consumption was surprisingly low.
But also levels where the output was more like and angry hissing with very thin arc that had no problem setting paper on fire instantly.
In this configuration though the input power was the highest of all tests.

Clearly, in a brute force configuration with a single transistor, the recovered energy plays a vital role both on the efficiency and the output power levels.
And as clear as this is that this recovered energy going back into the coil(s) results in harmonics on the high voltage output coil.
What IF .....

For a normal switch mode power supply the amount of turns are quite manageable if it is only for demo and not high output.
We can match the capacitor to the power coil to be resonant.
But what if we also match the secondary coil in the same way to the primary ?
Of course we won't use rectified mains for this but something like a 12V DC power supply.....
Logic tells us we won't be able to keep it all in resonance because our load will change and affect things.
On the other hand our resonant primary side should not be affected too badly by this.

In theory, at resonance our poor transistor wouldn't really have any switching loads.
The thing is that we also know that basically a PWM signal with 50% duty cycle is far far ideal to drive a transformer.
But that what our common switchmode power supplies are...
The rest to make this signal 'better' is basically cheating it closer to a sine wave to reduce interference.
And for the better ones we use H-bridge configurations anyway, so why bother?

A good switchmode power supplu soes not get hot under load like a normal transformer would.
It is those transistors that get hot and other passive components.
The actual transformer, if not designed to flimsy barely gets hot.
So we know that despite their higher efficiency of up to 95% there is still room for improvement.
And if this improvement would result in far fewer required parts that could fail.....

We can find ZVS or Zero Voltage Switching on a lot of things these days to advertise energy efficiency with out actually explaining how it supposed to do that.
In lame terms it means that whatever switches the load does so while the AV voltage is at zero.
Resulting in no switching losses or transistors getting hot.
We can apply the same for a power supply to drive a transformer.
And we can do so with just a pulsed DC as the input if we can recover the energy from between those pulses - where the field in the transformer collapses.
Funny enough we do just this in modern ignition coils for our cars....
Where to start though ?

[Downunder35m]

A resonant transformer is one thing.
But driving it with pulsed AC while recovering what is lost when the magnetic field collapses ?
How could this even work?
Let's start with this pink elephant in the room...
If we drive a transformer, be it with feedback coil or not, with a pulsed signal than we always have to deal with voltage spikes.
Hence the cheating in our power supplies as they are hard to get rid of or to prevent.
But why not utilise them instead ?

We know a simple diode and capacitor trick gets our primary into resonant regions while recovering otherwise lost/wasted energy.
And a similar approach is used to keep the RF noise down in our power supplies.
On the output side however we use big capacitors and a suppression coil to filter out those spikes and create a more or less smooth DC.
We also often use some big resistor here to smooth things out that only really wastes more energy.
Why not use those spikes and the negative ones we just drain away through a diode ?
Let those positive spikes go into the capacitor and create a better filter on the output.
But get rid of this 'choke' coil behind the transformer and the diode to eliminated those negative spikes - feed them back instead!

Don't let the diode bleed to ground, feed it back right when the next impulse comes to lower what the primary has to provide.
Like you would, well, on something like a good Tesla coil.
With this 'in sync' with our impulse going into the transformer we also create an overlapping magnetic field.
In our normal switchmode power supplies the 'ringing' from transformer coils is a big part of creates those interferences through a wide spectrum of frequencies.
Now try to follow me here as hard is seems:
In a resonant transformers those ringing effects, the unwanted harmonics, would be preferable because they would be harmonic and utilised.
And what we feed in and how feed it in is on the same harmonic, resonant, frequency.
Means all those effects would be WANTED ones that we can use.
What would this theoretical and impossibly ideal power supply pose in challenges for the creator?
And I don't mean in terms of getting the impossible mix of transformer, frequencies, resonance or such tough things right....
Just the problem of how to get those formerly unwanted negative voltage spikes turned into a positive so to say....
Let's assume weed a positive impulse into our primary coil.
When this impulse is gone the collapsing electromagnetic field results in a NEGATIVE Voltage spike.
Hence the need for diode on DC relays to protect our electronics...
If we use a diode to direct this harmful impulse away - what would happen if we could ADD it with the next positive impulse coming in to 'lower' the negative side or ground of our impulse ?
Instead of the 12V from our power supply we probably end up with an impulse of around 150 or more Volt going into our transformer.
Most of it though with very little 'ooomph' so to say.
A lot of Volts but barely any Amps coming back through our diode.
(totally ignoring the problems for our power supply and impulse generator here!)
BUT if we would also have a capacitor able to handle the Voltage and sharp spikes we could bring this Voltage down a lot while storing enough juice for a much higher Amp rating on our addition.
Of course things get rather complicated if you have to deal with two grounds and one primary coil that needs to be fed.
Or do they ?
Does it really matter to the transistor it it switches just our primary coil to ground to make a connection OR if there is a separate 'ground formed by our diode and capacitor above?
I mean, all we really would have to do is to add another transistor that goes on with our impulse to bleed the capacitor to ground of the primary coil with the this impulse.
And a feedback coil, like a simple mosfet driver coil can be used to delay the switching on of our main transistor for the primary coil.
To avoid all this nonsense we could also just use a suitably size inductor to connect our capacitor to ground.....
And suddenly an impossibly complex problem comes down to just two resonant systems working in harmony with still just one switching transistor.....

If it really is THAT simple then why doesn't anyone do it?
Tolerances and because simple gets really complex in resonant systems.
The required capacitors, diodes and precision in terms of acceptable tolerances mean you blow whatever is doing the switching to pieces if anything goes out of spec only a tiny bit.
Simply too expensive to do.
On the other hand there is the thing that no one actually ever tried....
We do it all but we we never do it all together for some reason.
Because we learned that forcing things and having 'acceptable losses' is far easier and cheaper than any other solution.
Today though we start to re-think this approach and try go green with everything to safe the planet....
And while switchmode power supplies ARE the way to go for now, are they really the answer ?