yes, i built a buck converter, and got 98% efficiency at 60 KHZ using irfz44 mosfets hard switched at 5 amps per pair.
50 volts going in, 24 volts going out. so i lost 10 watts. --it wouldn't have mattered if it was a boost converter working backwards.
one whole watt of power was lost in those two 680uF 200 volt capacitors. the 1uF film caps were required to prevent the mosfets from going into avalanch during the switching transition, and the two toroidal inductors killed the ripple going into the power supply (this was required, as there was half a volt of ripple IIRC on the input to this thing.), but the inductor on the output was not required, because the current ripple was only 2 amps or so.
http://johansense.com/bulk/50_buck1.jpgI would estimate that half the mosfet losses was due to driving them from a signal generator with a 50 ohm output impedance, and then adding 20 more ohms to turn the fets on, as there should have only been 2.2 watts lost per fet at 5 amps (they were in parallel to get 10 amps out of this machine) instead, i was at double that, at 60KHz. this used to not be that bad.
now consider that for a few extra dollars mosfets today are probably one third the resistance, twice the voltage and half the gate charge.
so any common man should be able to throw a proper half bridge driver at two mosfets and get 98% efficiency as either a buck or a boost converter.
heck, consider the stats of the induction heater i'm working on right now:
4x 500 or 600 volt mosfets, i can't recall which, at 70 milliohms each.
20 amps at 400 vdc going into the HBridge, out of the H bridge comes a square wave into a resonant load at 60-140KHZ.
24$ worth of mosfets, switching 8 Kilowatts, (20 amps average turns into 28 amps RMS, 28^2* 70 milliohms is 55 watts) 55 watts per switch lost, but each pair is only on half the time, so it becomes 110 watts lost at 8 KW.
since the body diode is canceled out, and at resonance there is no flyback diode loss, i can get 98.6% efficiency.
but just double the cost and that is halved, not to mention the inductance of the packages probably limit me to 4KW, not to mention the fact that if i bought new, i would need to spend 5$ in film capacitors to hold the DC bus together.
so lets say i run 24$ worth of mosfets at 14 amps rms each, this turns into 13 watts at 50% duty cycle which is 26 watts lost turning 400 volts dc into 100khz ac at 14 amps rms or 10 amps average.
If i spend 20$ on silicon carbide diodes, i could turn this into a hard switched two phase boost or buck converter and get about the same efficiency, as high as 99%. so i'm at about 25$ per kilowatt plus an inductor and capacitors.
heck, the last induction heater i didn't even need to put heatsinks on the mosfets, which may have been why it generated noise sufficient to block out FM stations.. lol. but the point is, the waveform should look like this:
http://johansense.com/induction_heater/6/P1020775.JPGnow, in the case of this waveform, there is 500 nanosecond of dead time between the switches, and its at resonance, which is why the voltage climbs so slowly. (2 volts per nanosecond.
When you get to hard switching, as in the case with buck or boost converters, things get a bit more sketchy.
http://johansense.com/induction_heater/prototype3.1_W.JPGyou'll see that there's an extra 30 volts that shows up across like, one inch of pcb trace between the mosfet and the catch diode, and that's only at 100 volts on the dc bus, what happens when there's 400 and twice the current draw?
all that energy can potentially be absorbed in avalanche with low voltage power circuits, HIGH currents (say running 30 amps at 20% duty cycle through a TO-222) and people wonder why they get warmer than the back of napkin math said it should.
so in the case of boost or buck converters, since we have to hard switch big inductors, we throw more phases at it, and then we add: (
)
a resonant snubber.
these things are really not that hard to design, but they typically add a lot of $$ to the problem.