310Vdc 1.8kW isolated supply using ZVS oscillator from 120VAC input.

[davekni]

Documenting the isolated 310Vdc bulk cap supply feeding buck converter of my QCW coil.  Requirements (features) of this circuit:
1) Charge bulk caps (54000uF) from 0V to 310V.
2) 120VAC 15A input.
3) High power factor to get 1.8kW without significantly overloading 15A circuit.
4) Output isolated from line voltage.  Allows grounding Vbus- for convenient QCW scoping at full power.

Used a ZVS oscillator for this design (higher voltage version of typical cheap induction heater circuits).  ZVS oscillators are very simple, but have many potential problems.  I'm still intrigued with these even with bits of added complexity needed to avoid problems.  ZVS oscillator drives a ferrite transformer with coupling ~0.84.  Coupling is low enough to avoid oscillation dropping out independent of load impedance.  Circuit diagram shows actual measured inductances and coupling factors for transformer after construction:



Two primary ZVS power transistors are STP60H65DRF IGBTs.  FETs M2 and M3 (AOD4S60) provide cross-coupling (instead of diodes).  Gate supply voltage is "p19v_sw".  It is switched between 0V and 19V to enable and disable oscillation.  I'd previously avoided gate drive switching in ZVS oscillators due to the need to dump energy from the main series inductor(s) on power input (L3+L4 in this circuit).  However, energy is not all that high.  And clamping voltage across inductor(s) is more efficient than clamping to ground.  D12+D13 provide this clamping.  Some clamping across IGBTs is still necessary to handle wiring and leakage inductance energy.

As has been discussed in several ZVS oscillator threads, ZVS oscillators have trouble starting if main power is applied before gate power.  That makes a problem for gate power control.  Main power is always present.  I found through simulation that splitting the power feed inductor into two and adding a resonant capacitor C7 makes startup possible with gate supply after main power.  This C7/L3 resonance is adjusted to work with main ZVS oscillation frequency range.  Functions in both simulation and now in practice.  Worked fine for two days of local science festival last month (starting 10 times/second during regulation).  However, there is no protection in this circuit in case oscillation fails to start.  Even one failed start will cause current to ramp up until something burns out.  See farther below for my next idea to improve starting.

Edit:  Or perhaps a failed start would become a delayed start when the conducting IGBT desaturates, with Vce increasing enough to turn on opposite IGBT.  If current has not ramped too high, perhaps everything survives.  This would be a good reason to keep Vge around 15V.  There are no high peak currents in normal ZVS oscillator operation (as there are in DRSSTC H-bridges).  Vge above 15V provides only marginal reduction in Vce at normal operating current.

Control circuit is bang-bang (on/off) regulation at 310Vdc.  Minimum off time limits cycling to 10Hz to avoid excess power in clamp diode D13.  During initial bulk cap charging and after each QCW arc, output power is at its maximum 1.8kW and input power factor is reasonably high.  (Haven't measured exactly, probably around 0.9.)  Of course, bang-bang control makes for low power factor during regulation.  Total power is low during regulation, so power factor isn't as important.

No one will want to copy my discrete transistor control implementation.  Posted here anyway for completeness:



My next idea (that works in simulation) is to make a delay timer for each IGBT gate.  If either gate stays high (on) for too long (as when oscillation fails to start), it is forced off by this delay timer.  As with above control circuit, no one is likely to copy my discrete transistor delay.  Below is my simulated but untested circuit:



This one increases input capability to 240VAC at 40A using 1350V IGBTs.  I'm initially building a version of this circuit as higher power drive for my SSTC (changing to DRSSTC).  (Existing drive is 3kW.)  Power board with IGBTs is built now, but not control nor most of MMC.  Control will be times with line voltage zero-crossings, further reducing energy and power in clamp diodes.  (Should have done that in the initial 120VAC supply too.)

Note that this new circuit uses diodes for cross-coupling as with standard low voltage ZVS oscillator circuits.  (Few reasonable FET options at 1350Vds.)  Diodes feed emitter follower gate buffer.  4.1V shift is added between buffer and IGBT gate.  This DC shift keeps Vge low enough even when Vce forward drop gets high.  Plan to increase gate supply voltage a bit to keep Vge high enough when on, to at least 20V or perhaps 24V.  Schematic still labels gate supply as 19V.