Hello everyone,
I want to design 2000V high voltage switch for pulse generator.
It will run around 50~100kHz and duty will be variable (%1 to %50). 3300V switches are not cheap and not good for high frequency. Current will be around 3~5A.
So I want to try series mosfet drive, 2 series 1700V SiC is very good for me. I found some schematics, I simulate it but I wonder do you have some suggestion for me?
Some techniques:
Hmm
Those are both cascade options… with some benefits and trade offs.
This is reminding me of a paper I saw years ago. I think the author used a stack of SiC JFETs connected much like your first figure, and a small (like 40V) Si MOSFET to control the source of the bottom JFET.
On topology:
You’ll need a diode for the return side. Not sure if a useful 3300V SiC diode will be available.
How about moving to something like a Neutral Point Clamped (NPC) converter (see
https://www.vincotech.com/products/by-topology/more-about-npc.html) which effectively splits the job into two half bridges? Could help with control too; you move from a 2 level (0, 2kV) to a 3-level (0, 1kV, 2kV) output which means the volt second product your inductor is filtering is halved, and therefore the required inductor value is halved.
Hello again,
JFET looks good but 1700-2000V JFET is hard to find.
I want to use same topology on Flyback Converter too. So I have to use mosfet in series
Voltage sharing isn't perfect, if the bottom MOSFET goes too fast you will get some avalanche.
Those designs are a bit old in the tooth,
supercascodes are the new hotness. They tend to have avalanche diodes, so if something avalanches it's at least not the SIC FETs.
UF3N170400B7S seems relatively affordable and available ... for the moment ...
I tried 3300V Mosfet, result is not good.
IGBT is not good for 50-100kHz.
I knew UF3N170400B7S from UnitedSiC, little expensive and only one manufacturer but i like supercascodes.
If you have another idea, it will be good for me.
G3R350MT12D looks nice too.
I think you can go faster with GDTs than supercascodes though. For low frequencies and extreme duty cycles you will just need a gate FET for turn off.
I'm using SiC since 2016. So I know them very well but for 2-2.5kV, they are not cheap.
Supercascode looks very good. But it's not easy to implement. But i will try
Ignoring supply for a moment, the Coolmos 800/950v devices look like a nice budget alternative. Very small internal gate resistance and good gate max voltage. If you could drive them hard with synchronized +/-15V drivers they should be able to go fast
I generally don't like gate transformers. They have to have very small leakage inductance, small primary/secondary cap ext.
It's diffucult to control around %1/%99 duty.
I'm searching solution without gate transformers. Otherwise I can use isolated mosfet driver
It can be more easy
If you want to go fast, there's always optically isolated gate drivers using isolated power supplies.
It's hard to see how a cascade will ever be as fast as a stack of synchronous driven MOSFETs.
You are right.
But think about one PSU (like flyback with UC38xx for chip crisis
) that has working input voltage is around 1500-2000V. So it's hard to make startup. For this series mosfet looks good solution. Otherwise I have to find some voltage source for mosfet driver to startup the PSU.
Alternatively, take a look at the offerings from Behlke (
https://www.behlke.com/ ). Finished modules doing what you want.
Its just not worth building an iso DC-DC.
Probably, but I like the concept of PCB transformers (ie. coils on either side of the core, some ferrite plate on top and bottom).
Watch out for specmanship on gate drivers and DC-DC converters. The working voltage is the number you need, not the peak withstand.
Which one actually specs kV+ working voltages? I clicked through a couple and none did.
Some of the reinforced Infineon isolated gate drivers with integrated transformer do spec kV working voltages though.
I think I didn't explain my problem fully.
If I want to build "cheap" pulse generator or Flyback PSU, I don't want to use isolated gate drivers and modules. Otherwise I wouldn't have posted my problem here...
I used lots of DCDC converters for gate drives and modules for PSU. But i want to say it again, they are not cheap!
If I want to use DCDC converters in PSU, I said before, there will be startup problem.
I don't want to use gate transformers because, good gate transformers not easy to find and they have some issues too.
Its just not worth building an iso DC-DC.
Probably, but I like the concept of PCB transformers (ie. coils on either side of the core, some ferrite plate on top and bottom).
Watch out for specmanship on gate drivers and DC-DC converters. The working voltage is the number you need, not the peak withstand.
Which one actually specs kV+ working voltages? I clicked through a couple and none did.
Some of the reinforced Infineon isolated gate drivers with integrated transformer do spec kV working voltages though.
The Murata MJG6 14mm series are the best candidates I know in terms of affordable non-specialized components. Creepage/clearance is 14mm, test voltage is 10.2 kV and they specify "Continuous barrier withstand voltage (non-safety)" as 3000 V DC.
As far as i remember all specialized mosfet gate driver ics deliver 1 A or more gate drive capability, to rapidly charge/discharge gate capacitance (e.g. 1 nF). To derive these current spikes from the output side using a capacitive voltage divider will result in a capacitive load of some nF onto the output, e.g. 2x 4.7 nF in the case of two mosfets. Don't know if that will support "fast" switching. Turn off time of the upper mosfet will depend on load current.
Regards, Dieter
What is the load that you will be connecting to the high voltage pulser. I have use very similar circuits that you posted here. Trying google search of Dr Jake Baker. His thesis was on pulser designs. I designed a high voltage pulser for space that will be driving a capacitor load so not much power. I am switching a delta voltage of 1500V in 33ns. I do use a gate transformer just to drive the bottom stack mosfet. The transform has to have very low parasitic capacitance otherwise you will have common mode noise that will affect the performance and even have crossconduction. My email is rbola35618@aol.com