P-Channel driver for high speed

[freeman3020]

Hi all

I searched a lot for P-Channel driver for fast speed switching buck converter about 30khz
I found attached driver from MPPT circuit but result very bad in gate , I use 4 irf4905 in parallel

what's the solution

original signal:


result :

 

[dick_freebird]

If your PMOSFET can take the full 25V then a bang-bang power
MOSFET driver would give you the speed you want. But many
FETs cannot stand 25V Vgs.

FET drivers for load switch applications seem like what you
want. Some new ones include charge pumps so they can
drive NMOS instead. These application-oriented drivers
will incorporate the level shifting and gate voltage limiter
circuitry as is cobbled here from discretes (only faster
and better).

Most buck converters don't use PMOSFETs, outside of the
integrated buck chips. If your 30kHz is continuous and the
duty cycle range is sane then totem pole half bridge
drivers with bootstrap high side NMOS drive could be the
best bet.

There are also many controller ICs which support this
but include the PWM regulation as well. Linear Tech (ADI)
has many such, some good up to 100V in.

Don't fixate on the PMOS high side switch if all you need
is some chopping done. It will drive you to less than ideal
choices & outcomes. Only simplicity really argues for it,
and your discrete lashup may not be any better cost-wise.

 

[freeman3020]

Thank you dick_freebird

but if I have pmos on my hand and drivers not available in my market, is it mean its difficult to drive pmos with this frequency by most common component?

 

[dick_freebird]

TIP31 is a fat boy and probably could drive that FET
(turnon, at least) quicked than it does in the 'scope
photos. You might experiment with putting a "speed-up"
capacitor across the 100-ohm emitter resistor.

Turnoff is a different problem and the 470 ohm, BCxxx
transistor network looks way too weak.

You might look at old Unitrode databooks (Silicon General,
before them) for MOSFET driver chip schematics that
could show you how high current drivers were done,
back when NPN bipolar was the only choice. You can
find them on archive.org, if nowhere else (modern data
sheets often are scrubbed of too-informative schematics
but these parts are waayyy old).

Having done a few FET drivers in 40V IC technology I'd
say you could get it done with under 10 transistors, 8
of them switching NPNs and two, power NPNs. A few
50V Schottky diodes might help the low side switch
act right, speed-wise, if you can find low enough Vf.
Then get yourself SPICE models and start in on
optimizing the bias resistors you will see in the schems.

 

[freeman3020]

I have and idea ,

if I used negative voltage regulator , link 7912 , then I used normal totem pole driver , but how to input base of transistor? with negative voltage?

 

[CataM]

Attached is a discrete driver for the big PMOSes you have (4 in parallel). I have simulated it for a similar big FET as yours, specifically the IRF7210 (it has higher total gate charge than the one you use).

Also, the layout is very important. Keep all gate traces and gate drive areas small. Add decoupling capacitors.

 

[freeman3020]

Attached is a discrete driver for the big PMOSes you have (4 in parallel). I have simulated it for a similar big FET as yours, specifically the IRF7210 (it has higher total gate charge than the one you use).

Also, the layout is very important. Keep all gate traces and gate drive areas small. Add decoupling capacitors.

sorry , but the file damaged

 

[CataM]

O.K. Attached again. You surely noticed it is an ".asc" file, so you need to install LTSpice.

 

[mtwieg]

The original circuit schematic looks ok, I've used similar circuits in the past with success. Though I wouldn't recommend using a power BJT like the TIP31C.

Your waveform suggests something is not connected correctly, or a component is damaged...

 

[FvM]

Your waveform suggests something is not connected correctly, or a component is damaged...

I wonder which node voltage is shown in the waveform?

 

[zarzisur]

I have used a P-Mosfet driver for 20khz which is very simple.