Baxandall converter strange output

[davekni]

2sk941 should be fine.  The two gate-drive FETs can be larger devices, but will waste extra power due to gate charge.  The key for the gate-drive FETs is low threshold voltage.  (BTW, I'm going to experiment with some depletion-mode devices there, with negative threshold voltages.  Ideal would be threshold around 0V, but such isn't produced to my knowledge.)

[dreamth]

2sk941 should be fine.  The two gate-drive FETs can be larger devices, but will waste extra power due to gate charge.  The key for the gate-drive FETs is low threshold voltage.  (BTW, I'm going to experiment with some depletion-mode devices there, with negative threshold voltages.  Ideal would be threshold around 0V, but such isn't produced to my knowledge.)

bss/bsp135 might be cool for really high voltage zvs...but probably bsp125 will work fine...

[davekni]

Yes, bsp125 should work fine.  bsp135 looks quite similar to the parts I ordered to test (DN2470K4-G and CPC3960ZTR).  There is a detail in using depletion-mode parts.  The diode across gate-source in my ZVS schematic for these gate-drive FETs will need to be replaced with 2 or 3 diodes in series to get higher forward voltage drop - enough to turn off the depletion-mode FET.

[dreamth]

I also wonder if a very fast detector schotky(60 VRRM, 15ma) or just regular drop 1n4152 or higher current diodes  would be better than 1n914...i don't really know which parameter is more important for those diodes, being low drop and fast being able to witstand big currents at high speed or having lower leakage at high Vak.

[davekni]

1N4152 or 1N914 are fine - almost any reasonably-fast signal diode is fine.  They conduct current only due to Cgd of the small FETs and the 5k resistors that I just added to aid startup with slow voltage ramp.  (If gate supply comes up before the main supply, the 5k resistors aren't needed.  Depending on gate capacitance of the small and large FETs, sometimes the diodes aren't needed either.  The diodes just prevent over-charging the gates.)

[dreamth]

1N4152 or 1N914 are fine - almost any reasonably-fast signal diode is fine.  They conduct current only due to Cgd of the small FETs and the 5k resistors that I just added to aid startup with slow voltage ramp.  (If gate supply comes up before the main supply, the 5k resistors aren't needed.  Depending on gate capacitance of the small and large FETs, sometimes the diodes aren't needed either.  The diodes just prevent over-charging the gates.)

Thank you! You're really helpful.

[dreamth]

Now that the lower power fet is initiating the oscillation i suppose a great number of parallleled power fet might be used ...Would a classical mazzilli have a problem with paralleled mosfets at startup ?

[davekni]

Most of the cheap Chinese induction heater kits use multiple FETs in parallel.

Even the conventional diode-coupled circuit will start up reliably if the gate pull-up resistors are connected to a separate supply that is turned on before the main supply.

[dreamth]

Why is that your circuit doesn't have any gate resistor?

[davekni]

In the common ZVS oscillator, pull-up resistors charge the gated and diodes to the opposite drains discharge the gates.  In the version I posted here, small FETs both charge and discharge the large FET gates.

[dreamth]

In the common ZVS oscillator, pull-up resistors charge the gated and diodes to the opposite drains discharge the gates.  In the version I posted here, small FETs both charge and discharge the large FET gates.

Then who's discharging the small fet's driver  gate?

[davekni]

The easiest way to understand analog circuits is with a simulator - some variety of SPICE.  I prefer LTSpice, but there are other free simulators available, both online and for download and local execution.

The small FET gates are fixed at 12V.  As a small FET's source is charging a large FET gate, it is simultaneously discharging its Vgs, since source voltage is rising towards 12V.

[dreamth]

The easiest way to understand analog circuits is with a simulator - some variety of SPICE.  I prefer LTSpice, but there are other free simulators available, both online and for download and local execution.

The small FET gates are fixed at 12V.  As a small FET's source is charging a large FET gate, it is simultaneously discharging its Vgs, since source voltage is rising towards 12V.

I have to confess  i have been a bit lazy on te sim part as i have no relevant experience with simulating transformers.I am reading everything i can on the subject rigt now, but i can't make the sim work when adding the two transformers k coupling although the unloaded converter without transformer secondaries works really well.One core is TDK PC40 with 1.5mm gap and about 5uH inductance, the other is E+E , 0.2mm gap with about 580uH leakage inductance when primary shorted, which is 1/3 of the total 1.5mH measured inductance of te secondary with primary in air.There must be smth interesting i don't know  that the primary winding of the E +E flat  transformer should be 20uH with those 2x 5uH in series yet it measures 60mH with secondary in the air so the difference of 40uH might be lekage series inductance of primary?  Another interesting measurement is on the PC core where the primary is 2x4 turns, the secondary is only 2 turns yet the secondary measures twice the inductance of each primary which is counterintuitive...
I tried simulation with loaded transformers to see the dissipated power on 2sk941, but i can't make it work for now, not even ideally...more sim knowledge is needed on my side...
Today i'm gonna try 3 IRFZ46n in parallel driven by one 2sk941. Hopefully i'll scrap the need for a heatsink this way.
P.S.i saw that i cannot upload .asc or rar files...asc files should be top priority as this is an electronics forum...

[dreamth]

and the size of the transformers

[dreamth]

loading a single transformer there's no waveform anymore...
until i add some parasitics to input inductor and the output looks as it should.
also the driver power dissipation looks pretty good to me at half the load althoug at full load some oscillations show up but i am really far from a true simulation of te real ting anyway.Now i'm more confident.

[dreamth]

Now here's smth i really am worried about: the power dissipation on fet drivers increases proportionally  with the number of paralleled fets probably because of the gate capacitance of power transistors so some series resistors might need to be added there just to make the circuit more linear although i tried that already and the dissipated power on drivers is increasing instead of decreasing...just another counterintuitive thing, but it probably interferes with the converter efficiency...
 Yet compared with original Mazilli your circuit simulates  much better in terms of thermal efficiency.I'll see later how the real thing behaves.

[Twospoons]

With these types of symmetrical circuits the transient sim can sometimes get stuck in perfect balance - since all the parts are mathematically identical. It helps to deliberately unbalance the circuit (like the real world) to get things started. I find adding a few pF to one of the power fet gates is usually enough.
Oh, and V1 should be a single step from 0 to 12V, not a fixed 12V.

And always include the series resistance of any inductors, even just a token milliohm - that helps with convergence.

[dreamth]

And always include the series resistance of any inductors, even just a token milliohm - that helps with convergence.

Indeed , i couldn't make even one transformer work untill i added some 0.2 ohm to the input inductor, then naturally added simillar series resistance to other inductors too, also added some 5...10pf winding capacitance too as i had some convergence rules already that didn't help enough.I still need to find out how to properly sim two transformers in parallel and actually learn to sim true transformers.Just saw some CUK material on TDK PC40, but its a bit too theoretical for me at the moment:
https://core.ac.uk/download/pdf/4876634.pdf

[Twospoons]

So long as you include winding resistances, sims with transformers in parallel should work fine.  If you don't then the simulator will be unable to determine the inductor currents in its initial DC operating point calcs.
The same sort of thing happens if you have capacitors in series - you end up with a floating node of indeterminate voltage.

Specifying initial conditions for critical parts is another way to sort this out (setting initial current in L or voltage in C)

[dreamth]

I don't think i know how to specify the initial current for inductors but for now its clear that  until now i couldn't  make ltspice accept  two transformer primaries in parallel even with series resistance specified....
As i already tried my circuit with Baxandall converter , i will tell that i opted for a different version as having two transformers in parallel is raising the resonant frequency and ask me for bigger parallel capacitor if i don't want the power to go too low.PC 40 core is specified up to 250khz but honestly the power got lower when going from 40khz to 80 khz maybe due to semiconductor  inherent losses.
So i went a different road: I paralleled  a secondary winding from the pc 40 core with the primary of the second high voltage transformer and the coupling losses are covered by the lower resonant frequency while a dead short on the high voltage side will be much more forgiving to the converter's transistors .I already tested that for a few times and it works very well.You have some taste of my circuit in the photo.