MOSFETs are getting better, year by year

[John Larkin]

On Wed, 19 Jun 2019 20:23:06 +0000 (UTC), Uwe Bonnes
<bon@hertz.ikp.physik.tu-darmstadt.de> wrote:

John Larkin <jjlarkin@highland_snip_technology.com> wrote:
That's kinda private. But regular cmos logic chips can drive low-side
GaN fets. ACT maybe. All three sections of NL37WZ16US in parallel is
awesome.

The high-side driver in a GaN totem pole is harder.


Texas LMG1020 is also fast option.

Uwe

The Tiny part is about as fast and 1/10 the price. It doesn't drive as
hard, but it's fine for the smaller EPC fets.


--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[John Larkin]

On 19 Jun 2019 14:53:24 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

John Larkin wrote...

On 18 Jun 2019, boB <boB@K7IQ.com> wrote:
On 18 Jun 2019, John Larkin wrote:

We are having good luck with the EPC BGA GaN parts as
super fast switches. Pain to assemble, but electrically
great. Unlike SiC, they are happy with 4 or 5 volts
of gate drive.

https://www.dropbox.com/s/k1k4i1r2nhuhnyk/T577_40V_Pulse.jpg?raw=1
(40 volt supply, 50r out, loaded to make 20 volts)
Gotta find a use for that circuit.


Would be Interested to see the drive circuit for that.

That's kinda private. ...

I suspect John makes discrete driving circuits.

His 1kV pulser is slightly faster than mine, and
we're both using the same SiC parts. I think he's
seen my circuit, RIS-764Gb, but I haven't seen his.

I could show it to you privately, but we want to keep it secret as
long as we can. We only plan to sell it to OEMs under NDA.

I'm using 1.0 ohm from half of a UCC21520, with NPN /
PNP SOT-89 emitter-followers. driving +15V through
4.7 ohms and -3.5V through 1.5 ohms, to SiC gate.
It's a pretty serious gate drive, 6 watts of power,
can go to 5MHz rep rate. Interleave for 10MHz.

I'm driving SiC fets with GaN fets, which are in turn driven by the
Tiny logic buffers, which are themselves driven by various things.

I blew up an impressive number of parts getting this to work. Learned
a lot.


--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[Winfield Hill]

John Larkin wrote...

I'm driving SiC fets with GaN fets, which are in turn
driven by the Tiny logic buffers, which are themselves
driven by various things.

I blew up an impressive number of parts getting this
to work. Learned a lot.

OK, John, thanks, I'm certainly not going to go to that
much trouble to get an extra one or two ns of speed. I
can appreciate, a little, how much work it is. What I
found was that one blown part could take out an entire
set of associated parts, and I really missed having my
technician, ready to take over and replace the parts.

My scheme was to get everything working and then turn up
the repetition rate, until failure. After it was clear
I couldn't make 10MHz, I came up with the interleaved
5MHz approach.


--
Thanks,
- Win

 

[John Larkin]

On 19 Jun 2019 17:20:34 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

John Larkin wrote...

I'm driving SiC fets with GaN fets, which are in turn
driven by the Tiny logic buffers, which are themselves
driven by various things.

I blew up an impressive number of parts getting this
to work. Learned a lot.

OK, John, thanks, I'm certainly not going to go to that
much trouble to get an extra one or two ns of speed. I
can appreciate, a little, how much work it is. What I
found was that one blown part could take out an entire
set of associated parts, and I really missed having my
technician, ready to take over and replace the parts.

One advantage to the BGA GaN fets is that I can't even think about
soldering them myself. I send them to my magicians downstairs.

For production, we're putting the GaNs on little surface-mount
castellated (mouse bite) baby boards, which become throwaway parts.

https://www.dropbox.com/s/jmhyros7en61tnd/T852B_Glob_2.jpg?dl=0

Those little fets are really fragile, like tiny glass bricks. We
glob-top them to protect them and maybe improve the thermals a little.
Once they are globbed, they are impossible to rework, so we don't want
them on our main board.


--

John Larkin Highland Technology, Inc

lunatic fringe electronics