Re: What Driver Voltage to Drive a Mosfet? (SMPS app)

On 31 jul, 19:09, D from BC <myrealaddr...@xxxxxxxxx> wrote:
On Tue, 31 Jul 2007 04:04:58 -0700, Wimpie <wim...@xxxxxxxxx> wrote:
On 31 jul, 02:08, D from BC <myrealaddr...@xxxxxxxxx> wrote:
On Tue, 31 Jul 2007 00:07:53 +0100, Eeyore

<rabbitsfriendsandrelati...@xxxxxxxxxxx> wrote:

D from BC wrote:

My power mosfet has an absolute gate spec of 30VDC.
Most of the Mosfet data*** gate behavior graphs only go to 10V.

Is there a drive voltage that'll provide low switching loss?
Looks like there's a 6V to 30V range to choose from.

As others have said, the trick is to turn it on *FAST*. Age proportion of the
losses are during turn-on and turn-ff as opposed to conduction.

So... what you *should* be asking (given a sensible choice of voltage drive) is
gate CURRENT ! The more amps the better.

Graham

I'm picturing driving a mosfet like this..

+----Rsource---+
| |
| |
Vdriver Cgate
| |
| |
+--------------+

V is from the mosfet driver rail voltage.
Rsource is mostly internal mosfet driver resistance.

Here's my confusing thing...

Given that the mosfet driver can change state fast with currents up to
14A peak and the objective to fast charge/discharge the gate for low
Pds switching loss ...It kinda looks like one should make V (the
mosfet driver rail) as high as possible for the fastest charging.

D from BC

Hi "D from BC"

Basically driving the MOSFET with high positive and negative voltage
will result in faster switching times

Driving your MOSFET with high voltage (for example 20 V) will turn-on
the mosfet faster, but increases the drive losses significantly (so
your driver will consume more power).

The gate of the MOSFET has internal resistance (the gate of standard
SMPS MOSFETs is made of polysilicon instead of metal). Even when you
would use a driver with infinite current capability, the drive current
is limited by the MOSFET's internal gate resistance (and dI/dt is
limited by inductances).

You can get some idea of the internal Rgate by looking into the
simulation model of the device you are going to use.

If you want to speed up turn-off, you may use a negative gate drive.
In that case you can pull more current out of the gate (but with
higher gate drive loss).

Before designing your driver circuitry I would recommend you to first
evaluate the losses for you SMPS topology. Maybe you are spending
money (and power) in an almost perfect drive circuitry, while the
reduction in switching loss is insignificant with respect to other
losses.

Unnecessary fast switching will generate more HF noise, so you have to
spend more components to meet EMC requirements.

You mentioned 600 kHz (and higher). Are your using a (semi) resonant
or a zero-voltage-switching topology?

The driver IC has been (probably) designed for the IXYS MOSFETS. They
have several types that have very low internal gate resistance and can
be used into the HF/VHF frequency range.

Best regards,

Wim
PA3DJS
www.tetech.nl

Hi D from BC,

Hey...that's different.. It's been suggested in this thread to only
charge up the gate to a voltage for needed drain current.
That way there's less to discharge.
Little charges and little discharges can be done fast.

Driving many MOSFET above 8 to 10V, doesn't reduce the Rds_on. So at
that point I agree. Also most non-logic drive MOSFET switching
behavior is specified with a 0 to 10 V gate drive (with certain series
resistance).

Regarding switching. One has to see the delay time apart from the rise/
fall time. When you want to turn of a MOSFET that has its gate on 15V,
this will result in a longer delay time than when the gate was at
10V. But I believe, the fall time is not affected by the higher
initial gate voltage.

But... Every Rseries and Lseries (driver internal,trace and internal
mosfet R&L) limits the current therefore limits the charge rate which
in turn sets the mosfet switching speed.

Right, when for a certain MOSFET the internal gate resistance is 2 Ohm
and you are driving from a 0 to 10 V driver, is has no use to select a
driver that can sink and source 10A.

The slower the switching speed...the more Pds switching loss.

Not always

The time between, for example 10% and 90% (the rise or fall time) of
drain current is important. Besides this, you have to be careful when
you turn on a MOSFET, while a diode is conducting (as in a buck
converter), the MOSFET has to remove the reverse recovery charge of
the diode. When you turn on the MOSFET very fast, with a high gate
voltage, the peak drain current can be very high. Because of the diode
reverse recovery charge, very fast switching does not always lead to
lower switching loss. The diode recovery time was the reason for
asking what type of converter you are designing.

One of my clients blew-up some LM5010 buck converter ICs, just because
of diode recovery, changing to a Schottky rectifier did solve the
problem.

Someday I'll learn how to balance the mosfet driver heat, the mosfet
heat and the EMI.
I guess when I burn my finger on the IXYS (IXDD414) mosfet driver,
it's max'd out. :)
(Assuming it's worthy to get the mosfet driver that hot.)

Picking a mosfet driver supply voltage has become a 3 bears story for
me..
Too big...too small...just right..

About resonance...
I'm not using a resonant design.
The power inductors I'm using do ring a bit and I do have to try to
keep inductor parallel capacitance to a min with proper windings.

These capacitances also increase the switching loss, faster switching,
does not reduce the losses caused by winding capacitance. So look the
complete design, not just switching loss.

Depending on the topology, you may expect difficulties in getting a
low leakage inductance, with low capacitance, and safety barrier.

The energy stored in the leakage inductance, you have to dissipate
also (for example in TVS diodes as snubbers). Resonant topologies are
more forgiving at some points.

I could convert to a resonant design and switch at "dead" times but
I'd like to experiment with non-resonant first.

I can imagine your statement. The control range of a single resonant
converter is limited (with respect to, for example, a fly-back or non-
resonant forward converter).

Ahh....you suspect IXYS drivers best with IXYS mosfets...I'll take a
peak at the IXYS mosfets sometime.

They make MOSFETs in a very special case that you normally see in RF
power devices (the terminations are strips with very low inductance).
They claim switching speeds in the low nano-s range. link:
http://www.ixysrf.com/app/switch_mode.html

D from BC

Best regards,

Wim
PA3DJS
www.tetech.nl



.