part 1 751 bytes content-type:text/plain; (decoded 7bit)
> What would be the chip-of-choice to driver a bunch (4 or 8)
> N-Mosfets
> from PIC output pins? The Mosfets are of course not 5V-types, they
> need
>>10V at their gates.
If you don't mind a few glue parts consider multiple versions of the
driver cct I posted a day or two ago.
Key questions are,
- are these low side FETs (presumably)?
- Switch how fast?
- What supply voltage for FETs?
- What load?
- What power levels?
'My' (time honoured) circuit will meet most needs well and cheaply.
Parts count per channel - 2 x small NPN, 3 x R, 1 x Z, 1 x D,1 x cap
(optional).
Very amenable to SMT implementation.
See recent post for notes.
I'll post it again - 'only' 3K.
Russell
part 2 3085 bytes content-type:image/gif; (decode) part 3 35 bytes content-type:text/plain; charset="us-ascii" (decoded 7bit)
> > What would be the chip-of-choice to driver a bunch (4 or 8)
> > N-Mosfets
> > from PIC output pins? The Mosfets are of course not 5V-types, they
> > need
> >>10V at their gates.
>
> If you don't mind a few glue parts consider multiple versions of the
> driver cct I posted a day or two ago.
>
> Key questions are,
>
> - are these low side FETs (presumably)?
> - Switch how fast?
> - What supply voltage for FETs?
> - What load?
> - What power levels?
>
> 'My' (time honoured) circuit will meet most needs well and cheaply.
> Parts count per channel - 2 x small NPN, 3 x R, 1 x Z, 1 x D,1 x cap
> (optional).
> Very amenable to SMT implementation.
>
> See recent post for notes.
>
> I'll post it again - 'only' 3K.
>
>
> Russell
>
> from points of easy, cost, performance, safe, definitely use op.
use what?
> - are these low side FETs (presumably)?
yes
> - Switch how fast?
from DC to up to above audio (20~30kHz)
> - What supply voltage for FETs?
>= 12V
> - What load?
1..10A
> - What power levels?
I think the kind of FETs I have in mind (for instance BUZ11) max out at
26A 50V so maybe 5A at 48V = 240 W would be possible.
> 'My' (time honoured) circuit will meet most needs well and cheaply.
Yeah, but I was hoping for an integrated solution, maybe not as cheap as
discrete components but cheaper to manufacture in medium quantities.
I often use the list of parts available at my local supplier as
candidates. Currently looking at L6204. It is much cheaper here than http://www.findchips.com suggests. Intended to drive a bipolar stepper, but it
has 4 independent outputs, so it can be used to drive 4 FETs. Maybe I
can use the enable inputs for a low-power-supply lockout.
Wouter van Ooijen
-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu
>I often use the list of parts available at my local supplier as
>candidates. Currently looking at L6204. It is much cheaper here than
>http://www.findchips.com suggests. Intended to drive a bipolar stepper, but it
>has 4 independent outputs, so it can be used to drive 4 FETs. Maybe I
>can use the enable inputs for a low-power-supply lockout.
>
>Wouter van Ooijen
IRF has some nice chips with LV lockout built-in. If you can provide
that externally, Microchip's parts such as TC1426CPA, which is only 0.68
USD/100
for a dual 1.2A driver in DIP8/SOIC8 is pretty reasonable (34 cents/channel)
Or use logic level MOSFETs and maybe you don't need a driver at all-
(assuming 25mA of gate drive will do, rather than what a real MOSFET
driver does) and can perhaps use the BOR in the PIC (pulldowns and gate
resistors only).
Or, of you do need a driver with logic level MOSFETs, the standard circuit is
just two complimentary transistors in a dual emitter-follower configuration.
The cheapest solution for relatively low performance is to use a CD4049/4050
which will drive about 60mA and do the level translation from 5V to 10V
for less than 2 cents/channel (2500) but watch latchup issues.
> Or use logic level MOSFETs and maybe you don't need a driver at all-
> (assuming 25mA of gate drive will do, rather than what a real MOSFET
> driver does) and can perhaps use the BOR in the PIC (pulldowns and
> gate
> resistors only).
>
> Or, of you do need a driver with logic level MOSFETs, the standard
> circuit is
> just two complimentary transistors in a dual emitter-follower
> configuration.
>
> The cheapest solution for relatively low performance is to use a
> CD4049/4050
> which will drive about 60mA and do the level translation from 5V to
> 10V
> for less than 2 cents/channel (2500) but watch latchup issues.
If you are running at say 20 kHZ MAX frequency then you have say 25 uS
on and off times. If however you have 20 Khz frame rate PWM you may
have pulses in the few uS or faster range.
To switch a FET rapidly you must sweep the appreciable amount of
charge in and out of the gate capacitance rapidly. To do this in much
faster than 1 to 10 uS (so pulses in this region are sharp edged) you
need 100's of mA to amps of drive. The MEAN current is usually 1 - 10
mA range but the peaks are far higher. The cct i provided and any real
FET driver will do this. A PIC pin or typical CMOS gate won't. Look at
the data sheets and do the arithmetic and decide how much current you
need to get adequately fast switching for your application. A CMOS
gate or PIC pin will often be 1 to 2 orders of magnitude slower than a
gate driver.
If you are running at 20 KHz a FET driver will help greatly and at 20
kHZ PWM frame rate it's mandatory. In production volumes using through
hole or SMT the discrete circuit is liable to be cost competitive in
many applications
>> 'My' (time honoured) circuit will meet most needs well and cheaply.
>
>Yeah, but I was hoping for an integrated solution, maybe not as cheap
>as discrete components but cheaper to manufacture in medium quantities.
Have you looked at the Intersil HIP408x series? Primarily designed for PWM H
Bridge Motor Drive type use, but may be suitable for individual FET drive.
At 02:05 AM 6/4/2005 +1200, you wrote:
>>Or use logic level MOSFETs and maybe you don't need a driver at all-
>>(assuming 25mA of gate drive will do, rather than what a real MOSFET
>>driver does) and can perhaps use the BOR in the PIC (pulldowns and gate
>>resistors only).
>>
>>Or, of you do need a driver with logic level MOSFETs, the standard circuit is
>>just two complimentary transistors in a dual emitter-follower configuration.
>>
>>The cheapest solution for relatively low performance is to use a CD4049/4050
>>which will drive about 60mA and do the level translation from 5V to 10V
>>for less than 2 cents/channel (2500) but watch latchup issues.
>
>If you are running at say 20 kHZ MAX frequency then you have say 25 uS on
>and off times. If however you have 20 Khz frame rate PWM you may have
>pulses in the few uS or faster range.
>
>To switch a FET rapidly you must sweep the appreciable amount of charge in
>and out of the gate capacitance rapidly.
Putting some numbers on it, a drive current of only about 38mA is required to
have switching losses 50% of conduction losses with the specified
MOSFET at 10A/20kHz, assuming the datasheet 40nC gate charge and a
resistive load.
The gate drivers (especially the >10A ones) are great if you need to switch
at 100kHz or 500kHz, they *may* be overkill if your PWM frequency can be
kept relatively low and some switching losses tolerated (in return, the softer
switching means less problems with stray inductance in the layout and less
EMI,
but more heat to get rid of).
> To do this in much faster than 1 to 10 uS (so pulses in this region are
> sharp edged) you need 100's of mA to amps of drive. The MEAN current is
> usually 1 - 10 mA range but the peaks are far higher. The cct i provided
> and any real FET driver will do this. A PIC pin or typical CMOS gate
> won't. Look at the data sheets and do the arithmetic and decide how much
> current you need to get adequately fast switching for your application. A
> CMOS gate or PIC pin will often be 1 to 2 orders of magnitude slower than
> a gate driver.
Oh, surely, but it *may* be fast enough.
> In production volumes using through hole or SMT the discrete circuit is
> liable to be cost competitive in many applications
I agree that the discrete solution is quite viable, especially for a simple
low-side driver where LV lockout can be separately provided.
Don't forget the LV protection- especially for battery-powered stuff- if the
voltage can drop so that there is insufficient gate voltage to drive the MOSFET
fully on and there is no lockout it will surely come to tears sooner or
later.
Thanks for the driver circuit - very useful (and cheap
to build, I think around 0.20 USD).
Questions that I didn't see explicitly answered in the
original notes - the top rail should be at least the
Vgs+.7 for the desired Rds, correct? I assume that
the trigger input to Q6 can be +5. right?
> > What would be the chip-of-choice to driver a bunch
> (4 or 8)
> > N-Mosfets
> > from PIC output pins? The Mosfets are of course
> not 5V-types, they
> > need
> >>10V at their gates.
>
> If you don't mind a few glue parts consider multiple
> versions of the
> driver cct I posted a day or two ago.
>
> Key questions are,
>
> - are these low side FETs (presumably)?
> - Switch how fast?
> - What supply voltage for FETs?
> - What load?
> - What power levels?
>
> 'My' (time honoured) circuit will meet most needs
> well and cheaply.
> Parts count per channel - 2 x small NPN, 3 x R, 1 x
> Z, 1 x D,1 x cap
> (optional).
> Very amenable to SMT implementation.
>
> See recent post for notes.
>
> I'll post it again - 'only' 3K.
>
>
> Russell
Did you measure how fast it can switch FET on and off ?
WBR Dmitry.
> 'My' (time honoured) circuit will meet most needs well and cheaply.
> Parts count per channel - 2 x small NPN, 3 x R, 1 x Z, 1 x D,1 x cap
> (optional).
> Very amenable to SMT implementation.
>
> See recent post for notes.
>
> I'll post it again - 'only' 3K.
> Did you measure how fast it can switch FET on and off ?
Short answer: Around 1 Mbps without too much extra work.
Long answer: Fast enough for my use :-)
ie observe switching waveforms with a scope and ensure transitions are
adequately fast. Measure temperature rise for a given load as an
indication that all is OK. Measure Vdson with scope to ensure it's as
low as expected.
In practice I use this most with 20 or 25 kHz PWM with usually about 6
bits.of resolution so smallest bits are about 1/25,000/2^6 = around 1
uS pulses . With the simple circuit with speed up capacitor across the
base drive resistor it's getting somewhat unhappy at that speed. Extra
work would get it faster but many applications are OK at 1 uS pule
rates. At eg 10 uS pulses it's far happier.
paradise.net.nz>
