driving load at high frequency (300-500khz) at high power (200 to 600V)

hi all.
I previously built a mosfet driver which drove a N/P array at 120V at 500Khz to 1Mhz.
The device work well but I need more power.
This is for an ultrasound cleaning driver.

Ideally I need something that can drive 200 minimum volts into a capactive load (10nf) while switching at 300 to 500Khz.

Ive looked into high voltage mosfet driver section at IRF, and there are 600 and 1200V sets sold
But the problem is that, via the data sheet, there is a huge amount of heat generated on the driver which limits the frequency response.
With the 600V chips, it starts at around 200Khz and with the 1200, it starts at 100Khz.
Its so close to being usable but I need a few hundread more Khz.
example, ir2184
http://www.irf.com/product-info/datasheets/data/ir2184.pdf
example at 1200V, ir21141 starts to heat at 100Khz
**broken link removed**

Anyone have any ideas on a solution.
Maybe a way to do it without gate driver limitations?
Or alternative devices or chips?

one person even recommended a vacuum tube as they have high bandwidth and are not limited by power. is this feasible?

thanks

Does the drive voltage need to be a square wave, or would a sine wave be okay? Does the output need to have a DC bias? If a sine wave works, then you could just resonate the capacitive load with an inductor and easily get lots of power with high efficiency (since the FETs can be soft switched).

DC on the ultrasound transducer is probably not desirable.
the ultrasound trandcuer is piezoelectic
it strains when voltage is applied and there is a practical limit to how much strain occurs.
so if too much DC is present, there is less strain that can occur it gets hit with an electrical square wave.
ideally there should be a large amount of strain for each pulse applied to get a good transformation of electrical to mechanical energy.

a square wave is more ideal but a sine wave might work.
i believe it allows the transdcer to create better wavefront to create cavitiational energy

thanks

Driving it as a resonant load would be more power efficient than a square wave drive. Resonant drivers can be made to self-oscillate in order to deal with tolerance in the load properties. The issue with resonant drive is that the actual power and voltage delivered will not be easily regulated, and will depend on the Q of the load.

If you want to use an H bridge for simple square wave drive (no resonance), then the power dissipation of the FETs is mainly determined by P=V^2*C*f. That's just a strict limitation, and doesn't really depend on things like switch resistance or switching time (so long as it actually switches on fully).

im not against trying this out.
can you help me understand this a little better.
an inductor matched to transducer (simialar to capactior) in series?
is there a link or something, or anotehr term you can use (maybe buck converter?) im not getting a lot of google hits on this.
i do remember doing this like this studying EE about a decade ago but am pretty rusty.


its not really an H bridge, but rather just a half bridge that would be used.
with the half bridge, there is resonance involved.
the pulse waveforms which hit the transducer have to be at resonant frequency of the ultrasound transducer.
the transducer in a simple manner is modeled as a capactior.
resonance produces the most mechanical power out of the transducer.
its a pretty dramatic change in output of 300KHz vs 400 with a 400Khz transducer.
It will levitate water around 1" at 400 but not create a ripple at 300.

Hmm, I forgot that the electrical resonance of a piezo is not necessarily the same as the mechanical resonance. That just means you have to tune the electrical resonance to be equal to mechanical resonance. Just measure the complex impedance of the piezo at mechanical resonance, and then cancel the imaginary part with inductance.

I'm not aware of anyone using this method for driving piezo elements. Resonant H bridge/half bridges are common in solid state tesla coils, and by making them self oscillate it's possible to push tremendous amounts of power through the load, and take the switching devices far past their rated limits.

If you use an SRPP output stage, you only need to drive the bottom MOSFET, which could make things easier.

You might also consider something besides MOSFETs for the output, e.g. IGBTs or ESBTs.

(P.S. I'm not disagreeing with the resonant drive idea, just adding a few other thoughts to the pot)

i gave the resonant drive a try
it seems that at 500khz and 3nf load that the inductance needed to make it osciallate from say 100 t 500V is so small that even the copper wires have larger inducance.
ill copy in some screen shots of pspice.

and i did do this on the breadboard though too and didnt have any luck

and IGBTs tend to be too low to use in frequency responce
from what im seeing, most bomb out in the 100khz range