High side driving N-channel Mosfet with PWM

[Phil in Melbourne]

So I would have thought that this would be a simple problem but the more
information I get the more I get confused......


I want to PWM a whole lot of LEDs. The current consumption of the bank of
LEDs is in the order of 15 Amps, at 12V.

For reasons I won't go into, my only option is to use a high side mosfet.

I have chosen an N-channel MOSFET (IRL3803) mainly because of its huge
current capability and very low on resistance of 7 milli ohms.

I am hoping to use PWM at a reasonably high frequency, say 20khz or higher,
and I want to use an efficient MOSFET driver IC that is designed for the
job.

I understand that to switch an N channel mosfet on the high side of the
supply line, I have to drive the gate with 10V **more** than the supply
voltage. This is correct yes?

This is where the question comes in. I am looking at the option of using the
IRS2001 to switch the high side fet. But looking at the datasheet, I am
confused. How can the IRS2001 generate a voltage greater than VCC without
actually being connected to the high side voltage????? There is no talk of a
charge pump in the description and in any case it does not know what
potential the fet is connected to..

Maybe it is not suitable for my application as the IRS2001 seems to be used
for switch mode power supplies? I can't make sense of the first diagram on
the datasheet.

here is a link to the datasheet:

http://www.irf.com/product-info/datasheets/data/irs2001spbf.pdf



OK so if the above IC is not suitable, I have been looking at the MAX621.

http://datasheets.maxim-ic.com/en/ds/MAX620-MAX621.pdf

This one looks to be more straight forward. But I'm not sure how this IC
will go with 20Khz switching. There are many **low side** mosfet drivers
that are in the 2 to 7 amp range to ensure that the mosfets are hard on or
hard off within a few nS. This one doesn't look like it will do anywhere
near that switching current.


Any other ideas for a high side N channel driver IC that packs a punch?

 

[[email protected]]

So I would have thought that this would be a simple problem but the more
information I get the more I get confused......

I want to PWM a whole lot of LEDs. The current consumption of the bank of
LEDs is in the order of 15 Amps, at 12V.

For reasons I won't go into, my only option is to use a high side mosfet.

I have chosen an N-channel MOSFET (IRL3803) mainly because of its huge
current capability and very low on resistance of 7 milli ohms.

I am hoping to use PWM at a reasonably high frequency, say 20khz or higher,
and I want to use an efficient MOSFET driver IC that is designed for the
job.

Why so high? If it's just for brightness control, you don't need that
surely?

 

[legg]

This is where the question comes in. I am looking at the option of using the
IRS2001 to switch the high side fet. But looking at the datasheet, I am
confused. How can the IRS2001 generate a voltage greater than VCC without
actually being connected to the high side voltage????? There is no talk of a
charge pump in the description and in any case it does not know what
potential the fet is connected to..

Maybe it is not suitable for my application as the IRS2001 seems to be used
for switch mode power supplies? I can't make sense of the first diagram on
the datasheet.

here is a link to the datasheet:

http://www.irf.com/product-info/datasheets/data/irs2001spbf.pdf

For your application, VCC will connect to "up to 200V" in the
application schematic, on page #1. The drive supply VB is
conventionally bootstrapped using the diode in the same drawing.

RL

 

[Phil Allison]

"Phil in Melbourne"

I want to PWM a whole lot of LEDs. The current consumption of the bank of
LEDs is in the order of 15 Amps, at 12V.

For reasons I won't go into, my only option is to use a high side mosfet.

** How very dubious ....

I have chosen an N-channel MOSFET (IRL3803) mainly because of its huge
current capability and very low on resistance of 7 milli ohms.

I am hoping to use PWM at a reasonably high frequency, say 20khz or
higher,

** What the bloody hell for ?????

That is just as idiotic and dubious as your first requirement.

and I want to use an efficient MOSFET driver IC that is designed for the
job.

** YOU want FAR too many weird sounding things YOU cannot JUSTIFY !!

So THAT makes YOU another fuckwit, time wasting, PITA fucking TROLL.


Try proving me wrong, anytime - pal.



...... Phil

 

[Mark]

Why so high? If it's just for brightness control, you don't need that
surely?

if you strobe LEDs too slow, then it causes a strobe effect if there
is any movement... if you scan your eye past some LED rear car lights
you can get the strobe effect and it is very annoying... even if the
LED is not moving, your eye is always moving

I don't know how fast you need to pulse to avoid this but a few
hundred Hz is too slow. 5 kHz might be OK I don't know...

At 5 kHz you might get audio interference so that is a good reason to
go to 20 kHz.

Mark

 

[Jamie]

if you strobe LEDs too slow, then it causes a strobe effect if there
is any movement... if you scan your eye past some LED rear car lights
you can get the strobe effect and it is very annoying... even if the
LED is not moving, your eye is always moving

I don't know how fast you need to pulse to avoid this but a few
hundred Hz is too slow. 5 kHz might be OK I don't know...

At 5 kHz you might get audio interference so that is a good reason to
go to 20 kHz.

Mark

32 frames/Sec I believe is the minimum needed to prevent strobe flash in
the human eye.. or some figure like that.



http://webpages.charter.net/jamie_5"

 

[swanny]

You're going about this all wrong, no one makes a dimmable LED supply
like this, the main reason being your excessive ripple current resulting
in unnecessary heating and exponential reduction in LED lifetime.
Another good reason for not doing it this way is to avoid the fairly
huge current spikes resulting from what will be a hefty equivalent load
capacitance producing enough EMI to be detected in another galaxy. And
you want to go to all this trouble to produce a crap design because it's
the only thing you understand. Do some background reading:
http://powerelectronics.com/mag/710PET24.pdf and put more effort into
your planning.

Somebody stuffed up Fig. 2, didn't they.

 

[terryc]

I want to PWM a whole lot of LEDs. The current consumption of the bank of
LEDs is in the order of 15 Amps, at 12V.

Can you explain why you can not just break this down into multiple units
all centrally driven? say 15 units drawing 1amp each, or 30 units drawing
500mA?

 

[Phil in Melbourne]

Can you explain why you can not just break this down into multiple units
all centrally driven? say 15 units drawing 1amp each, or 30 units drawing
500mA?

It would be less convenient, but that is possible, and may end up being the
way I go.
There are motor controllers on the market that PWM high current motors. I
thought it would be reasonably trivial to do seeing as MOSFETS come in high
current versions, and seeing that LEDs don't have as much inductance as
motors do.

I will go for the MAX621 and suck it and see. If the fet overheats, I will
decrease the PWM frequency.

If it still overheats, I will break up the supply into groups of 3A loads.

No big deal either way.

 

[terryc]

It would be less convenient, but that is possible, and may end up being the
way I go.

I was thinking that would avoid the all eggs in one basket. if one unit
fails, it only takes out that fraction.

The other point is the frequency and duty cycle might be such that instead
of having to deal with an intermittent 15Amps, you could be dealing with a
consistent 5amps say or less.

 

[Tim Williams]

32 frames/Sec I believe is the minimum needed to prevent strobe flash in
the human eye.. or some figure like that.

Depends on the nature of the "flash". Film looks okay because it moves
quickly and smoothly between frames (at 24FPS). A light scene moving to a
light scene without dark inbetween is a lot more pleasant than a really
bright scene flashed only every so often. PWM LEDs will do that, because
they're very firmly either ON or OFF, within microseconds I mean. You want
to minimize the apparent movement of a PWM cycle with respect to the eye's
movement, which might be...

Pffft... well lemme see here... I can glance 60 degrees in 0.1 second
pretty easily, which is easily 10' travel at a distance of 8'. An average
LED is 1/4" across, and you want to keep the "distance" between cycles
below a diameter or so. The surface rate is around 100 FPS, or 1200 IPS,
or 4800 LEDs/s. So the minimum duration is around 1/4800 LEDs/s = 1/4800Hz
frequency. Which hey, that's pretty damn close to the earlier estimate in
this thread of about 5kHz minimum. There ya go! Of course, this frequency
can be lower for large LEDs, diffuse arrays, etc., or should be higher for
really close, fast moving stuff (uh, passing pedestrians a few feet away at
60MPH?!?). And if nothing else, you can add a choke filter and recoil
diode (aka buck converter) to smooth out the pulses and get perfectly level
light output. Hey, you could use an iron-core choke to justify the
switching loss savings of 400 Hz operation. :^)

Tim

 

[Tim Williams]

I have chosen an N-channel MOSFET (IRL3803) mainly because of its huge
current capability and very low on resistance of 7 milli ohms.

You can get P-channel MOSFETs in similar ratings. Maybe not quite so small
Rds(on), but certainly in the amperage you need at that voltage. There are
plenty of lower voltage P-channels (whereas the supply kind of dries up
over 200V, sadly). Stick a zener on the gate to protect it from nasty
automotive supply voltages and you can practically drive it from a naked
2N3904 collector sinking to ground.

Tim

 

[Rich Grise]

For reasons I won't go into, my only option is to use a high side
mosfet.

Aw, c'mon! Why not? This could turn out more interesting than the rest
of the thread!

Thanks,
Rich

 

[Rich Grise]

Fine for a stationary source you are staring at. But all
you have to do is blink and you will see a trail of dots
(especially if the LED is very much brighter than its
surroundings) at 32 flashes per second, and a lot faster,
too. If you haven't seen the effect, I suggest you try it.

It shows up on scrolling marquees very dramatically - when it's
scrolling, the letters look tilted, but I believe that's an artifact
of the strobing. If you move your eyes to follow the letters, they
square up.

Cheers!
Rich

 

[nospam]

Aw, c'mon! Why not?

Probably because it is some stupid contraption to go on a car with only the
chassis available as a -ve earth. Probably blow up on the first load dump
it sees anyway .

--

 

[krw]

32 frames/Sec I believe is the minimum needed to prevent strobe flash in
the human eye.. or some figure like that.

Not at all. Movies use 24fps, but open the shutter two or three
times per frame to eliminate the flashing. TV interlaces 30Hz
frames with 60Hz fields for similar reasons.

 

[boB]

"Phil in Melbourne"

Another option would be to use a pulse transformer, and maybe some
caps and diodes to drive the FET from a low side driver.

** How very dubious ....




** What the bloody hell for ?????

That is just as idiotic and dubious as your first requirement.




** YOU want FAR too many weird sounding things YOU cannot JUSTIFY !!

So THAT makes YOU another fuckwit, time wasting, PITA fucking TROLL.


Try proving me wrong, anytime - pal.



..... Phil

So it looks like this one's a bit over your head, huh ?

bob