"high voltage" MOSFET

since you are hand soldering, I would go for a TO-220 part. lots of parts to choose from, many probably fine. Look at the tradeoff between Rds_on an d Ciss. You might need small Ciss since you are switching pretty fast; the n keep an eye on temperature if you choose a part with higher Rds_on. It go es without saying if you are switching at 2MHz keep the leads short.

If you can keep it cool, IRFP4668 (in TO-247AC package). One of my favorites, but influenced by the fact I have plenty on-hand. :)

Note that it sports a rather hefty gate-charge, so be prepared to drive it hard to get that low Rds(on) quickly. Typically .008 to .010 Ohms.

So to hopefully helpfully clarify my needs further the application is like an interleaved synchronous-buck-type converter but I'm using it to generate a power sine wave at a couple MHz. There is a goodly amount of dead-time for the switches. The gate driver is an Analog Devices ADuM4120.

Just testing a few MOSFETs from the LTSpice library with fairly high Vds. IRFH5207, Vds 75 volts - blah! Way too slow, these will shoot-thru and smoke themselves instantly. no good for these speeds.

Infineon BSZ900N15NS3G, Vds 150 volts - yes, this is my ballpark. Looks like will drive my load and stay cool under a couple 100mW dissipation no problem. That laptop-motherboard style package seems a bit inconvenient for a copper-clad-board test item, though.

Looking thru the LTSpice lib it looks like there's a set of Infineon parts that have Vds in the range of 100-200 volts, gate charges of about

5-20 nC and Rds-ons in the range of about 0.2-0.5 ohms that, at least in the sim, look like they'll work great at this speed with my Analog Devices 2A isolated gate driver. The package is "PG-TSDSON-8" or "laptop motherboard" as I call it which is a bit inconvenient for hand soldering I think.

Even with a hot-air station once you slap it down to a sheet of copper-clad they're pretty hard to get off I expect if you blow it. Maybe tack their drain pad to risers of copper-clad and let the source and gate pins hang off the side.

Obviously cuz the ~0.2 ish Rds looks ok we're not talking huge output power requirements, here. Trading that off for lower gate charge seems the right way to go.

I think "serveral MHzand Ampere" contradicts and easy to handle package. Did you have a look at the bare die GAN mosfets, like those from EPC?

--
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de 

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt 
--------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------

I think "serveral MHz and Ampere" and easy to handle package contradict. Did you have a look at the bare die GAN mosfets, like those from EPC?

--
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de 

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt 
--------- Tel. 06151 1623569 ------- Fax. 06151 1623305 ---------

You should avoid the temptation to go big, with low Rds(on). Instead go for low Qg and low Coss. Power dissipation terms are f Qg Vg, f Coss Vd^2, and i^2 Rds(on). It'd be good to keep Coss under 50pF. Looking in my giant MOSFET spreadsheet, an FQD5N20L looks good, Qg = 5nC with 5V gate drive, Coss = 40pF, and Rds = 1-ohm, oops, a little high. Or maybe FDD7N20TM, Rds = 0.58 ohms, Coss = 45pF.

--
 Thanks, 
    - Win

Or maybe GaN. Many times better numbers. Easy to drive.

--

John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

Most of the GaN and SiC parts in my spreadsheet are aimed at higher-power applications. They start with higher voltage ratings (600 or 1200V), and actually have higher Qg and Coss values than conventional low- voltage FETs. But then, my spreadsheet is certainly incomplete.

--
 Thanks, 
    - Win

EPC has some dynamite 100 and 200 volt parts in that current range. They are cheap and have amazing performance, the downside being the horrible BGA packages.

--

John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

Yes that sounds the ballpark of what I need. It's driving about a 100 ohm load thru an LC with the N-fets sources hooked to four different rails like an interleaved buck, with bootstrapping on the top FET driver to drive the gate of that one high enough. one node switches e.g.

+25/-25 and the other e.g. +12/-12 gates driven in sequence with some dead-time.

Need to go fairly fast but not ridiculously fast; the 2A ADuM4120 seems up to the task with FETs with gate charges in that range. with the standard MOSFET gate-circuit of diode, small resistor in parallel in line and pull-down at the gate the waveforms look very clean in the sim.

Yes, the ADuM4120 is an amazing part. It's 150kV/us slew rating is one of the best specs out there. That would be 150V/ns, or a 4ns risetime with a 600V pulse. But its 2pF in-to-out capacitance implies i = C dV/dt = 300mA i/o current during the 4ns swing. Worrisome. Better stick to slower edges.

--
 Thanks, 
    - Win

Yes, good to far beyond 2 MHz. For example, EPC2012 has gate charge < 2nC at 5V, so P_drive = f Q Vgs = 0.1W at 10MHz, and with Coss = 73pF, P_loss = f Coss V^2 = 7W.

--
 Thanks, 
    - Win

What's stunning is the Crss of 0.4 pF. That makes all sorts of dumb-looking circuits work.

I've got a GaN totem-pole making 10 volt pulses with rise/fall below

250 ps. From dirt-cheap parts. The circuit works to at least 80 volts, but slows down a bit. The pulses do look a little ratty at that speed, so I'll lowpass them to make prettier edges around 1 ns.

I've tested the EPC GaN parts. They die at just about 2x rated voltage.

I kinda miss the Avago phemts, but the EPC GaNs are really better.

--

John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

Switching that fast at 100V then I hope you are doing ZVS and ZCS swithing

Otherwise you need super junction part or GAN

Cheers

Klaus

snipped-for-privacy@gmail.com wrote in news: snipped-for-privacy@googlegroups.com:

Isn't anything at those levels GAN nowadays?

It's a pulse generator. Totem pole, 50 ohm DPAK resistor, output connector. And maybe the cosmetics lowpass filter.

I'm using the EPC GaN fets.

--

John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"