Synchronous rectification implementable for high voltage/High power (llc converter)?

[micktosin]

I am looking to implement synchronous rectification instead of diode in my LLC resonant converter, but most research papers results favors Schottky diodes at higher voltage due to the trivial loss of Vf, couple with the extra effort and expense required for synchronous rectification(SN). The problem is my output is 400Amps and 200Volts and so far finding the right Schottky is a challenge. Are there any papers out there or research that demonstrate an implementation of SN to achieve high efficiency, if not what are the other options you guys can suggest to achieve higher efficiency compared to normal diode.

 

[jiripolivka]

I am no expert on power rectifiers. But to my knowledge, synchronous rectifiers to replace Schottky diodes make the only sense in reducing the 0.3...0.6 V forward voltage drop. In high-voltage rectifiers this is negligible.
It is not impossible and at high currents like 400 A heat-sinking may be reduced, for the (quite high) cost of control electronics for the stack of paralleled MOSFETs that can carry 200 V in reverse.

 

[FvM]

It's basically possible with MOSFET, but when choosing MOSFET for reasonable chip area utilization, the voltage drop won't be considerably lower than with self-steered rectifiers.

 

[micktosin]

Thanks. I will have to give it a try to compare their results myself. Even an increase of 4-10% efficiency is enough to justify their use considering I am paralleling multiples modules of converters.

 

[mtwieg]

At such high voltages, using a standard silicon rectifier as opposed to a silicon schottky is probably okay (high voltage schottky diodes lose most of their nice characteristics anyways). If you really want improved efficiency a SiC schottky is probably the best option.

 

[Orson Cart]

At 200V there is no advantage with mosfet rectifying as the internal diode is slower than an U-fast 300V diode, and could well cause extra losses in your mosfets... we have some experience with this....

 

[micktosin]

At 200V there is no advantage with mosfet rectifying as the internal diode is slower than an U-fast 300V diode, and could well cause extra losses in your mosfets... we have some experience with this....

That's certainly disappointing, I was hoping synchronous rectification would be the savior rather than using conventional diodes. By the way, do you have any published paper on it, because I now have to do a literature survey to convince my lecturer for my final year project.

 

[KlausST]

Hi,

with three paralleled schottky diodes like SKR130/12 you have a voltagedrop of about 1.2V at 400A.
This means about 480W of power loss.

To improve this you need MOSFETS with lower than 3mOhm on resistance. But with that high voltage rating it´s hard to find..

The only way i can imagine is to use rectifing diodes combined with a FET. You can swith the FET on if current through diode is more than 100A and switch it of when current drops below 50A. Or similar values.
It depends on your switching frequency if this works...

good luck.

Klaus

 

[FvM]

SKR130/12 is a 1200 V standard rectifier without reverse recovery specification. I don't think that you'll use it for a mid-frequency converter.

 

[Anna Conda]

If you slow the switching freq down to 20 - 50kHz say, and put some generous snubbers across each mosfet, then you may be able to parallel enough mosfets to halve or lower the cond losses.
At 400A, 200V it would be sensible to divide up the transformers anyway.

You will need a 250v fet as a minimum, and some serious low ESR caps right at the rectifers (lots of) as the AC current will be around 90 amps rms. 90^2 x 0.001 ohm = 8.1 watt. If you can get 0.001 ohm (ESR) of cap(s).

As to the mosfets, lets try the FDA69N25 ($3.50) 250V, 69A, 41m-ohm, to get the volts down to say 0.2 volt the max I per fet is 5A, so 80+80 fets needed, do-able but messy and a lot of gate drive needed.
Watts in the rectifier would now be 400x.2 = 80W, so pretty good heatsinking still needed (40W ave + 40W ave)

The internal diode speed is 210nS @ 100A/uS dIf/dt - so pretty slow and rising with temp, some hefty snubbing needed across each fet to limit spikes to 10V or so.

Regards, Anna.

 

[Orson Cart]

A very good concise analysis.

 

[KlausST]

Hi,

You will need a 250v fet as a minimum

I think you need fets with a rated Vds of at least twice the output voltage. This is 400V better more than 450V.

Klaus

 

[FvM]

I think you need fets with a rated Vds of at least twice the output voltage. This is 400V better more than 450V.

The 250 V calculation is assuming a H-bridge. It can be implemented without excessive overvoltage. 250 or 300 V transistor rating is a reasonable assumtion in my view.

I think the most disillusioning point is the huge transistor oversizing that's be required to achieve on-state voltage levels below regular silicon rectifiers. The switching behaviour is probably less problematic for a resonant converter with moderate commutation currents.

 

[Orson Cart]

FvM is correct, 250V Fet for a full bridge rectifer - single o/p winding which more fully utilises the transformer, the centre tapped arrangement shown at the start of the post would require 400-500V fets - the beauty of LLC is that the rectifying devices see only the o/p volts (or 2x for the CT) due to the low ESR filter caps right at the rectifers.
For a current doubler output you would have 2 diodes and two chokes (200A ave each) but there will now be volt spikes on the diode turn off which will need to be snubbed, and the PIV now has to be >400V as for the C tap.

The large amount of C across the rectifying mosfets (because of so many of them) will mean that any sharp transitions imposed on the o/p (due to current limit action, sudden short of the o/p etc) will likely cause resonant overshoot and mosfet killing spikes unless properly snubbed for these events (diodes and zeners across each fet can be used to clip overvolt spikes also - and zeners to the gate if done properly).

Also excessive leakage inductance in an LLC converter gives rise to hard switching of the diodes at full power, and this will cause volt spikes on the slow mosfets if not snubbed.

 

[KlausST]

Hi,

The op shows one picture. There is a half bridge for the transformers primary. This not the point of discussion.

The point of discussion is rectifying the output. Here a see a center tapped transformer winding and a half bridge rectifier.
The shown diode's voltage rating surely is twice the output voltage, and so must be the fets paralleled to the shown diodes..
Btw: A full bridge rectifier usually has twice the power loss than a half bridge rectifier.. but the op wants less power loss.


It seems i have a lack of information.... ;-)

Klaus

 

[FvM]

It seems i have a lack of information....

I don't think so. The point is that if synchronous rectification at this voltage level would be implemented, H-bridge is the only reasonable topology. You are right in so far that the double voltage drop of the bridge topology biases the trade-off even more towards passive rectifiers.

Another reason why H-bridge topology comes into sight is that similar converters are often designed for bidirectional operation. A H-bridge is the "natural" solution in this case.

 

[Anna Conda]

Sorry, I wasn't extremely clear in my post, it is easier to get fets with lower on drop for a full wave rectifier, e.g. 250V ones, if you go only to 0.4V each fet bank you still need 120 fets to do this in a full bridge and the losses will be 0.8 x 400 = 320W, at full power.

Using a push-pull (CT) output rectifier and 500V fets at say 0.1 ohm each, you need 200 + 200 fets to achieve this (0.2V on, and 80W total), 100 + 100 of 500V fets will give 0.4V and 160W in total (still 200 moderately expensive fets).

240 x 240V fets will give 0.274V + 0.274V = .548V * 400A = 220W, so more losses but much cheaper fets...

Regards, Anna.

p.s. correction: just found N-channel 500 V, 0.035 Ω, 68 A, MDmesh™ II Power MOSFET
in a TO-247 package, at about $10 ea, thus 70 + 70 of these give 0.2V @ 400A (80W total)
i.e. 1.4W per fet average, stand up on pcb with no h-sink and a wee fan to blow over them, 2 lots of 8 x 9 array
approx 160mm x 100mm each array, fan (92mm) positioned above, heavy due to 140 fets, but effective.