The following two posts are copied over from OU.com. I'm including it just to keep easy reference and it may - hopefully - explain some of my own eccentric thinking related to this.
kindest as ever,
Guys, it's on my mind and I hope you'll get this.
The fact is that the spike that is generated from BEMF is well known. According to mainstream it's the energy that is stored in the circuit materials as a result of the input of energy from a supply source. Our circuits are designed that this voltage can induce a current flow with a counter clockwise path through the Zener diode. What we found was that the sum of the energy that was returned from the circuit and the energy dissipated at the resistor also exceeded the amount of energy first supplied by the source. In effect, the counter clockwise path through the battery supply also served to recharge it - thereby taking our co-efficient of performance to a value greater than 1. This was variously hotly contended - denied or replicated. Take your pick. They all resulted and they generated varying degrees of indifference - denial and abuse. Most attempts at explaining this simple event was cloaked and hidden in utter obscurity. And, predictably, notwithstanding some scholarly presentation of the results - the entire event faded from view.
Then despite some serious allegations against my rights to progress this at all - I was invited to develop this in a CPUT laboratory. The thinking was this. We'd start with a conventional typical element - resistor and then systematically vary the resistor to realise the optimised value of the inductance - then thinking that this would need to be increased. So. We would start off with minimal inductance and then work upwards. The surprise was to find that this element actually gave better results than any we'd previously tested. The trick, apparently, was to reduce and not increase the inductance. That was the first surprise. We could generate a significant 100 degrees C with - surprise, surprise, a zero discharge of energy from the supply.
Also interesting was that this result was NOT frequency dependent. But, invariably, the circuit would generate it's own resonating frequency - at a little over 50% on - which, in line with previous findings - also overrode the duty cycle setting. The difference was this. Before we had a result that invariably 'cost' the battery - albeit only a small fraction. This time there was clear evidence that the battery was now the happy beneficiary of more energy than was first measured to be supplied.
However, there was still no clear evidence of what exactly was going on. Also apparent was that while the technology was scalable - at approximately a 20 degree rise for every battery added - there was an upper limit determined by the amperage that the zener could manage. So. The next test was to up the ante by putting those MOSFETs in parallel. I went for the full monty - at about 30 amps - thinking that this would still keep the battery voltage in line with the DSO's voltage tolerances. That was when I recorded our 'first surprise' in my blog. What was immediately apparent was there was an antiphase relationship of voltage on the source and ground rail - that spoke volumes. When the drain voltage peaked - the source voltage was at it's lowest. And when the drain voltage 'troughed' the source voltage was at its highest. In effect, the returning energy trumped the output - every time - and all the way through each cycle. Also. The resonance - that was always restricted to a long spike and some ringing - now 'flattened out' and for a brief period gave a resonating waveform where there was clear early indications of absolute re-inforcement at each phase and stage. But also apparent was that this resonance actually only occured when the signal at the gate defaulted to negative. In effect - it was a negative triggering - and that's where the benefit had been hiding.
Now. If you think about it. Under usual circumstances, the initial spike that then generates the back EMF - occurs at the last moment that the switch applies a positive signal at the gate and when the circuit is, effectively closed. But this then rings flat and out and does not appear again until the next cycle. What was now evident was that the discharge of current from this spike has only one path to discharge - through a Zener diode that can barely tolerate 6 amps. So what does it do if there is more than 6 amps worth of potential difference in that voltage spike? It can only discharge this as heat over the sundry components including that poor punished diode. Now. With a wider path established through those MOSFET's and their zeners - all in parallel - then the current flow is enabled to the full potential of that voltage spike.
But. And here's the thing. The value of that stored potential difference - that was first established by the flow of current from the battery supply - is now developed on the circuit from the collapsed voltage at the spike. This generates a positive potential difference or a potential 'clockwise' flow of current from the circuit material. And there is no resitriction to the flow of energy from the battery as, now, the signal at the gate is negative. A negative will not repel a positive. And THERE is the benefit. Both negative and positive voltages now have a path to discharge their voltage - in either a clockwise or an anticlockwise direction to an extent or at a value that is commensurate - not with the initial discharge from the supply - but with the potentials in the circuit material itself. What is intriguing is this. There are two entirely different voltages - resulting in two entirely different current flow paths - and they never vary - the one from the other - in their periodicity. It's as steady as a heart beat. And always - I have never seen this vary - there is more energy in the anti clockwise direction than the clockwise. And - in either direction - the beneficiary is to the heat on the resistor and to the retained charged condition of the battery supply.
And this is how subtle is the tuning. It can be tuned to retain the 'off' time to that period that is precisely as long as is required to ensure that the advantage is to the battery. Then. One can increase the offset so that current is actually discharged from the battery during that short 'on' time. The spike is then HUGE. And the subsequent ringing - or resonance rather - triggered during the 'off' time is also then correspondingly increased. Now we have a condition where the energy dissipated at the resistor is greater than is allowed under conventional power delivery. It's acting with all the advantage of a booster converter. And yet there is no discharge from the supply. That's where this technology goes from adequate to super efficient.
Which brings me back to the point of this post with apologies for its length. I feel I'm testing all kinds of tolerances here. That oscillation - that resonance - is adjustable to whatever value is required to ensure that there is a zero discharge from the supply. And the beauty of those wonderful DSOs is that they enable that fine tuning. The math function does an immediate calculation of the instantaneous voltage. And when this crosses into the negative value then one knows that the best tuning has been reached. I am not sure how this can be managed without that sophisticated measuring instrument. It's possibly going to be problematic.
But try it out for yourselves. If and where you use one MOSFET try two - or more. And set the gate signal to negative to enable that closed path condition in both directions. You will see the benefit for yourselves. It's so much more reliable than our previous tests. And the results are also that much more conclusive. I am well aware of the fact that my presentations are usually met with a parade of those who know better and see some need to prove this wrong. It is not wrong. And it's too important to second guess.