SystemD_2kW, any interest for an open design?

[ChocoHolic]

...also thinking about the emitter switched approach.
The unpleasant thing is that ST starts its portfolio at 1700V and the most types have higher Rdson... Not the right thing for a 200V application.

Hm, discrete solution for emitter switched?
Ugh - component count. Mr. Nitpicker is back
On the other hand, the MosFet will be a low voltage type with low gate charge
and could be driven directly from the IRS20957, without additional totem pole driver.
Did Workhorse say 'tempting'??? ...have to dig deeper...

[phase_accurate]

I still owe you the other modulation variant. Here it is. Instead of Multiplying the fed-back voltage (which is deterimining the hysteresis) in some way or applying clamping somewhere in the voltage divider the hysteresis can be set solely by clamping form the beginning. Simulated performance is almost as good as the version with multiplier. The gain of E89 was set to 0.35. BTW: C314 is there to guarantee convergence of SPICE (also the other variants had this) it is not needed in a real-life circuit.

Regards

Charles

[ChocoHolic]

I have the feeling that the emitter switched approach is coming to a dead end for our application from very fundamental considerations.

- It is hard to find 200V/50A type. Especially when interested in reasonably fast types and high current gain.
- Even when considering to parallel three (ugh!!) of the famous 2SC2922 at 50A we would get a current gain around 20 (blue eyed without considering frequency).
With that we would need to feed 2.5A base current. No fun.
...even if averaging this over music program, we will still end up in something like 500mA. Not promising for a boot strapped floating high side supply.

In case you find 200V-BJTs which can handle 50A in a single device and provide a current gain of >50 at 50A, then it could become a viable option.
Please don't hesitate to post your findings.
The idea is really tempting, except that I am not aware of fitting real world devices.

Overall these considerations fit to the fact that ST is using this technology only at high voltage, where MosFet technology is a pure catastrophy.

[ViennaTom]

...at 200V, there is definitely no need for such strange Emitter switch. MOSFETs are fine and easier to drive..

[ChocoHolic]

Originally Posted by phase_accurate

.... the hysteresis can be set solely by clamping ....

I definitely like your new approach.

[ChocoHolic]

I would definitely love to see a high-current-200V-version of IXKF 40N60SCD1...

[ChocoHolic]

The extreme dv/dt > 10kV/us are IMHO related to the high Qrr and sudden current snap back of the diode as soon as the Qrr is removed.
During removing Qrr at output currents of 50A the current through the MosFet is ramping up to more than 100A (50A for the load and more than 50A to feed the recovery peak of the opposite body diode recovery peak). When Qrr is removed the body diode snaps back and drops current with di/dt of multiple kA/us. Means suddenly there are more than 50A unneeded current stored in the parasitic inductances, which deliver this current to the parasitic capacitances without any delay, resulting in unsaint dv/dt.
I case of low Qrr and soft recovery or mostly capacitive behavior like shottkys, such excessive dv/dt are not likely.

Freewheling
V30200C is back in the race
HFB35HB20 also promising
STTH3002C another candidate

For series shottky I defintely like the
60CTTN015
but it should be easy to find multiple suitable types.

Any comments or real life experiences on this?
Further diode proposals?

[ViennaTom]

For shunting the body diode, you may also try 2 x IPP110N20N3 (and no diodes) instead 1 x 4668 plus diodes.
HFB35HB20 and STTH3002 IMHO can not compete to V30200C.

[ChocoHolic]

Originally Posted by ViennaTom

HFB35HB20 and STTH3002 IMHO can not compete to V30200C.

Agree...

[ChocoHolic]

Originally Posted by ViennaTom

For shunting the body diode, you may also try 2 x IPP110N20N3 (and no diodes) instead 1 x 4668 plus diodes.

IRFP4668 with diodes appears more preferably to me, because it is more forgiving regarding dead time and idle losses.
The shunting method only works with really precise dead time adjustment and all my experiments with that where fitting to theory which tells me that you have to adjust to remarkable cross conduction during idle in order to minimize charge transfer to the body diode in the switching moment at high currents.
It is possible, but again no lunch for free