CoolMos - ZVT SMPS - theory question

[Petter]

Hi,

In a resonant mode (ZVT) SMPS main stage:

I am under the impression that the faster you are able to switch off the switches, the more likely that you will get zero crossing.

My thinking is that if using advanced devices such as Infineon CoolMos (which switch so fast you easily get resonance), this is just a positive thing and you should drive these as hard as possible. The net positive result of this approach would be to enable ZVT switching at lower power levels than would otherwise be possible (with slower devices)

Then again, perhaps I am wrong and one needs to limit the switching speed of these evice using the recommended circuitry found in Infineon application notes ....

Your thoughts on this would be most welcome.

Petter

[jan.didden]

Petter,

I'm not sure I follow you, but generally a lot of the energy waste is caused by the switching, ie when the switching is too slow, there is a time when there is both voltage across the switch and current through the switch, so this give dissipation = loss of energy. In that respect, the faster the switching, the better. "Driving hard" can cause losses in the driver circuit if you drive the switch harder than necessaary to get the fastest switching.

Zero crossing switching refers to the point in time whenm you do the switching, if you do that when (preferably) both voltage and current are zero, you will cause the least noise and spikes on the output.

Is that what you had in mind?

Jan Didden

[Petter]

Not really

1. ZVT SMPS relies on energy stored in leakage inductance in primary to be larger than that of the stray capacitance of the switch to enable zero voltage switching and thus low switching loss and EMI. The less the load, the more leakage inductance (or additional series inductance) you will need to fulfill this requirement. Thus your switching loss will be maximal at zero load. Goal is to extend ZVT range down as low as possible for a 70V SMPS output.

2. Postulate: A slow switch will waste L energy and reduce chance of successful zero crossing at low power output.

3. Infineon CoolMos typically are recommended to have gate slow-down circuitry. The question is whether this is necessary or even benficial in full bridge ZVT mode. From my understanding of the theory of ZVT it is beneficial to have as fast a switch as possible -- both "on" and "off" -- particularly "off", and thus gate slow-down circuitry of any kind would be harmful.

Is it ?????
Anything I need to know to optimize the gate drive?

Petter

[Circlotron]

CoolMos devices are generally very good but they have very poor body diode reverse recovery. When the incoming device is ZV switched it's body diode is forced to conduct momentarily and when the gate drive comes on a moment later it sets up a situation so that when the main forward current begins it flows through the normal source-drain path and tends not to recover the body diode very much because the source-drain shunts current away from it. When the fet switches off, the stress on the body diode because of the *full* reverse recovery current tapering off to fewer and fewer cells causes second breakdown of just *one* body diode cell (of millions) and then it's all over. The body diodes of some fets are optimised for ZVS and they have very low recovered charge nC and fast recovery. Unsuitable fets pop very noisily when used in ZVS mode.

[Petter]

Thanks Circlotron!

Can I also infer that in a ZVT scenario, one should drive the gates of the FET's as hard as possible (and not worry about oscillations resulting from this since we want to induce oscillations anyway) to improve likelyhood of zero crossing as well as reduce conduction losses?

Petter

[BeanZ]

No! You are not trying to induce oscillations. It is a called a resonant supply because of the transfer of inductive and capacitive energy. If you are not controlling the turn on of the FETs in a controlled manner the power supply will not work properly and generate EMI and waste heat. Gate drive control is essential for any switching power supply. International Rectifier has some ZVS optimized power FETs. The number is something like IRFPS40N50L. This is a 500V part and should allow for good performance when driven from an offline PFC preregulator at about 400V or less.

[Circlotron]

With ZVS, gate voltage rise time isn't a big deal as long as it is up before the initial reverse drain current decreases to zero and changes to the positive direction. For turn-off, drive the gate good and hard down to ground. I have seen some setups that drive the gate an equal voltage below ground and this does improve turnoff speed but keep your eyes open for other effects. Dont know what. Gate voltage ringing is not really a good thing. What are you making?

[Petter]

I am considering making a ZVT full-bridge PWM PSU of several hundred watts for audio power amps. However, I am probably going to revert back to simple 2 transistor forward converter as it will probably be too complicated to start with the ZVT design. A standard front-end also needs to be mated, but I am doing the power stage first.

Petter

[Circlotron]

A standard single fet forward converter has a lot going for it. Having gone the full circle in these things, I have come back to liking them. With a 400v dc supply buss, 2 parallel 900v 1.6 ohm fets and ETD49 core running at 100kHz you can get 400 watts easily with no fan cooling. Gate drive is referenced to earth which is always a nice thing. Single fet switchers when done properly virtually never blow up if they fail. When you have a pair of fets across the buss in a half or full bridge, one misfire and KA-BOOM!

[e96mlo]

I have heard that you have to use power factor correction in a PSU larger than 75 W. Otherwise it wouldn't be approved or legal or similar. I don't know if this is an international rule or if it is just here in Sweden.

Has anyone heard about this?

/Marcus