Push-Pull driver <<1ohm Rdson?

[Joerg]

Hello Folks,

Looking for a staunch chip to drive transformers in the 30-150kHz range.
The most powerful (and available) one I found is the MIC4451 but it's
1ohm Rdson. Way too much.

There are bigger ones but they usually have a charge pump and the
efficiency is pretty poor when operated at less than 3A:
http://rocky.digikey.com/weblib/Infineon/Web Data/TDA21201prelimDS.pdf

What I need is n/p channel push-pull output (not two n-channels), no
charge pump because the clock sometimes stops, and ideally 100-200mohm
Rdson. External FETs aren't so hot, too much cross conduction. I was
hoping there'd be n/p synchronous buck converters. But nope, all with
charge pumps.

Any ideas?

 

[Jon Slaughter]

Hello Folks,

Looking for a staunch chip to drive transformers in the 30-150kHz range.
The most powerful (and available) one I found is the MIC4451 but it's 1ohm
Rdson. Way too much.

There are bigger ones but they usually have a charge pump and the
efficiency is pretty poor when operated at less than 3A:
http://rocky.digikey.com/weblib/Infineon/Web Data/TDA21201prelimDS.pdf

What I need is n/p channel push-pull output (not two n-channels), no
charge pump because the clock sometimes stops, and ideally 100-200mohm
Rdson. External FETs aren't so hot, too much cross conduction. I was
hoping there'd be n/p synchronous buck converters. But nope, all with
charge pumps.

Any ideas?

--

FS---FDD8424H---Dual N & P Channel half-bridge 40V@20A 54mOhm.pdf

 

[Mike Monett]

Hello Folks,

Looking for a staunch chip to drive transformers in the 30-150kHz range.
The most powerful (and available) one I found is the MIC4451 but it's
1ohm Rdson. Way too much.

There are bigger ones but they usually have a charge pump and the
efficiency is pretty poor when operated at less than 3A:
http://rocky.digikey.com/weblib/Infineon/Web Data/TDA21201prelimDS.pdf

What I need is n/p channel push-pull output (not two n-channels), no
charge pump because the clock sometimes stops, and ideally 100-200mohm
Rdson. External FETs aren't so hot, too much cross conduction. I was
hoping there'd be n/p synchronous buck converters. But nope, all with
charge pumps.

Any ideas?

If you had no luck on a driver and wanted to use a charge pump for the
widest choice of output transistors, could you turn the bottom FET on when
the clock stops?

This would keep the charge pump capacitor primed and ready to go on the
first positive clock. The same thing would work for a full h-bridge. It
might take a bit of glue logic to get things started again in the correct
phase but it shouldn't be too difficult.

Regards,

Mike Monett

 

[[email protected]]

Hello Folks,

Looking for a staunch chip to drive transformers in the 30-150kHz range.
The most powerful (and available) one I found is the MIC4451 but it's
1ohm Rdson. Way too much.

There are bigger ones but they usually have a charge pump and the
efficiency is pretty poor when operated at less than 3A:http://rocky.digikey.com/weblib/Infineon/Web Data/TDA21201prelimDS.pdf

What I need is n/p channel push-pull output (not two n-channels), no
charge pump because the clock sometimes stops, and ideally 100-200mohm
Rdson. External FETs aren't so hot, too much cross conduction. I was
hoping there'd be n/p synchronous buck converters. But nope, all with
charge pumps.

Any ideas?

Hi Joerg

Why is a charge pump not acceptable? Are you concerned about the time
it takes to build up the gate charge.

What do you mean when you say "External FETs have too much cross
conduction?

Dick

 

[Joerg]

FS---FDD8424H---Dual N & P Channel half-bridge 40V@20A 54mOhm.pdf

Thanks, Jon. Although they conduct already quite well between 2-3V Vgs
so there'll still be considerable cross conduction. I am operating at
12V. Maybe I'll place zeners in the gate drive to burn off some drive level.

 

[Joerg]

If you had no luck on a driver and wanted to use a charge pump for the
widest choice of output transistors, could you turn the bottom FET on when
the clock stops?

Yes, that would be no problem.

This would keep the charge pump capacitor primed and ready to go on the
first positive clock. The same thing would work for a full h-bridge. It
might take a bit of glue logic to get things started again in the correct
phase but it shouldn't be too difficult.

I was thinking about glue logic with some "poor man's" one-shots in
there. Certainly an option but I wanted to avoid it, keeping complexity
down.

 

[Joerg]

This is a nice part:

http://www.national.com/pf/LM/LMD18201.html

It has a charge pump, but it also has an oscillator to keep it pumped
even when the input parks at one level for a long time. I've used it
for driving microsteppers, works fine.

Thanks, John. I had looked at that one and the Rdson is a bit highish,
600mohm max. I've also had some issues with National motor drivers with
obsolescence. Thing is, a lot of my stuff remains in production well
over 10 years.

I wish there was a big brother of the LMD18201 somewhere, 1/3rd of its
Rdson or so. I am running this at 12V supply and cost is not a real
concern for this part of the design. IOW if it was $10 a pop it would be ok.

 

[Mike Monett]

Jon Slaughter wrote:
[...]

FS---FDD8424H---Dual N & P Channel half-bridge 40V@20A 54mOhm.pdf

Thanks, Jon. Although they conduct already quite well between 2-3V
Vgs so there'll still be considerable cross conduction. I am
operating at 12V. Maybe I'll place zeners in the gate drive to
burn off some drive level.

You could also put a small inductor in the positive leg to limit the
max current while both FETs are on.

Say the cross conduction current was a rectangular pulse, and you
wanted to limit the current spike to 1A from your 12V supply. If the
conduction overlap was 40ns, a rough calculation give an inductance
of

L = E * dt / di
= 12 * 40e-9 / 1
= 480nH

If you knew the actual pulse shape, a simulation in LTspice would
give a more accurate value. But it would be lower than the above
calculation.

A value this small would have little or no effect on the operation
at 100KHz. But limiting the current spike would greatly reduce EMI
to the rest of the circuit.

Best Regards,

Mike Monett

 

[Joerg]

You could certainly make it out of parts... a couple of LM5112's, or a
dual gate driver chip, a couple of non-overlap parts, and two fets.

Yeah, if I don't find anything I'll do that this afternoon. Some logic
and external FETs. I wanted to avoid it but I guess it's the usual, I
might be alone in the marketplace for this stuff so the mfgs don't
bother. Kind of surprising because motor-PWM sometimes needs this stuff.

But I bet one of the fiercer fet gate driver chips would just work...
maybe parallel 2 or 4 sections.

Well, that's just the point, I haven't found any fiercer ones. And none
where I could parallel on the same chip and get to <250mohm. When I
talked to Micrel about paralleling some MIC4451 the engineer said the
delay differences may not work out well unless I make sure they are all
from the same batch. The latter is totally frowned upon by clients these
days.

A back of the envelope calc showed if I paralleled four and one would
really exhibit the 30nsec worst case delay veer versus the other three
that one would quickly mutate into a puff of smoke. Phssst ... *BANG*

Take a look at the LM5112... it's an interesting part. They cost us
$1.30 in small quantities. Maybe one or three of them could drive your
load.

Nah, I did look at that in detail after you mentioned it last year. The
high side is too wimpy, only the BJT in there has enough oomph and that
leaves too much headroom. The MIC4451 is much better. But not good
enough here. Also comes in a cheaper 9A peak edition.

 

[Joerg]

Hello Folks,

Looking for a staunch chip to drive transformers in the 30-150kHz range.
The most powerful (and available) one I found is the MIC4451 but it's
1ohm Rdson. Way too much.

There are bigger ones but they usually have a charge pump and the
efficiency is pretty poor when operated at less than 3A:
http://rocky.digikey.com/weblib/Infineon/Web Data/TDA21201prelimDS.pdf

What I need is n/p channel push-pull output (not two n-channels), no
charge pump because the clock sometimes stops, and ideally 100-200mohm
Rdson. External FETs aren't so hot, too much cross conduction. I was
hoping there'd be n/p synchronous buck converters. But nope, all with
charge pumps.

Any ideas?

Ok, guys, don't bother with this one anymore. I just wrapped it up with
discrete parts. As usual

 

[Mike Monett]

[...]

A back of the envelope calc showed if I paralleled four and one
would really exhibit the 30nsec worst case delay veer versus the
other three that one would quickly mutate into a puff of smoke.

Phssst. *BANG*

That's exactly what a small inductor in the top FET will prevent.
Check it out. You might be surprised how effective it can be.

You really have to get off this Phssst. *BANG* habit. It makes too
much noise, and everyone gets nervous. Plus it stinks up the place

Best Regards,

Mike Monett

 

[Joerg]

[...]

A back of the envelope calc showed if I paralleled four and one
would really exhibit the 30nsec worst case delay veer versus the
other three that one would quickly mutate into a puff of smoke.

Phssst. *BANG*

That's exactly what a small inductor in the top FET will prevent.
Check it out. You might be surprised how effective it can be.

You really have to get off this Phssst. *BANG* habit. It makes too
much noise, and everyone gets nervous. Plus it stinks up the place

I tried the inductor and wasn't too enthused. With a small damper
resistor inparallel it was kind of ok but still cost efficiency. But
it's done, I just lashed up the usual discrete concoction that provides
proper dead time and all that. Now it's moving on to the dreaded
packaging design. As much fun as eating pea soup and I don't like pea
soup. Ok, with some Johnsonville Brats in there I'll eat it.

 

[Mike Monett]

I tried the inductor and wasn't too enthused. With a small damper
resistor inparallel it was kind of ok but still cost efficiency. But
it's done, I just lashed up the usual discrete concoction that provides
proper dead time and all that. Now it's moving on to the dreaded
packaging design. As much fun as eating pea soup and I don't like pea
soup. Ok, with some Johnsonville Brats in there I'll eat it.

Joerg, I'm surprised it affected the efficiency enough to make a
difference. Did you simulate it or try it on the bench? And did you use the
smallest inductor needed to limit the current? Did it somehow drastically
increase the turnoff times?

With the FETs that were mentioned, the overlap is only 40 or 50 ns or so.
At 100KHz, it occurs every 5uS. That is only 1% of the duty cycle, so it's
not clear how that can cause a significant loss of efficiency. If you have
the time, I'd like to learn more how you did the test.

Xnews chopped the rest of your post, and I can't figure out how to make it
put it back. So I'll have to trust that most people know you

Best Regards,

Mike Monett

 

[Mike Monett]

Joerg, I'm surprised it affected the efficiency enough to make a
difference. Did you simulate it or try it on the bench? And did
you use the smallest inductor needed to limit the current? Did it
somehow drastically increase the turnoff times?

I did a small simulation in LTspice. The SPICE model for the
FDD8424H is available from Fairchild, but it is for PSPice. I didn't
want to take the time to make a model for LTspice, so I used IRF530
and IRF7204 for complimentary MOSFETs. I adjusted the rise and fall
time of the gate drive to give about 50ns cross-conduction.

I increased the series inductor between the MOSFETs to 3uH and used
a 47 ohm damping resistor.

The switching losses end up heating the damping resistor, so the
integral of the power dissipated gives the total loss. For this
simulation, the result is 52.114mW.

If the power delivered to the load is 3 watts, using a series
inductor to minimize shoot-through adds 52.114e-3 / 3 = 1.73% to the
total power dissipation. This is a rather small amount for the
simplicity and reliability gained.

Any circuit changes to reduce the power loss in switching will
probably cost additional power, so the overall gain might be small
or negative.

The conclusion is a small series inductor can be a viable option to
minimize shoot-through and reduce circuit complexity.

The LTSPICE ASC file is below, followed by the PLT file. The .tran
analysis string is set to 1uS to show the switching waveforms.
Increase it to 1ms to calculate the power in R2.

Best Regards,

Mike Monett

Version 4
SHEET 1 948 800
WIRE 224 16 48 16
WIRE 528 16 224 16
WIRE 48 32 48 16
WIRE 224 32 224 16
WIRE 528 32 528 16
WIRE 176 48 144 48
WIRE 480 48 448 48
WIRE 48 128 48 112
WIRE 224 144 224 128
WIRE 256 144 224 144
WIRE 320 144 256 144
WIRE 528 144 528 128
WIRE 560 144 528 144
WIRE 608 144 560 144
WIRE 688 144 608 144
WIRE 320 160 320 144
WIRE 608 208 608 144
WIRE 688 208 688 144
WIRE 256 256 224 256
WIRE 320 256 320 240
WIRE 320 256 256 256
WIRE 224 272 224 256
WIRE 96 352 48 352
WIRE 144 352 144 48
WIRE 144 352 96 352
WIRE 176 352 144 352
WIRE 48 368 48 352
WIRE 560 368 528 368
WIRE 608 368 608 288
WIRE 608 368 560 368
WIRE 688 368 688 288
WIRE 688 368 608 368
WIRE 224 384 224 368
WIRE 528 384 528 368
WIRE 48 464 48 448
WIRE 144 464 144 352
WIRE 448 464 448 48
WIRE 448 464 144 464
WIRE 480 464 448 464
WIRE 528 496 528 480
FLAG 96 352 M1G
FLAG 48 128 0
FLAG 256 256 M1D
FLAG 224 384 0
FLAG 48 464 0
FLAG 256 144 M2S
FLAG 560 368 M3D
FLAG 528 496 0
FLAG 560 144 M4S
SYMBOL Nmos 176 272 R0
SYMATTR InstName M1
SYMATTR Value IRF530
SYMBOL voltage 48 352 R0
WINDOW 123 24 134 Left 0
WINDOW 3 -128 159 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value PULSE(0 12 0 100n 100n 5u 10u)
SYMATTR InstName V2
SYMBOL Voltage 48 16 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value 12V
SYMBOL Pmos 176 128 M180
SYMATTR InstName M2
SYMATTR Value IRF7204
SYMBOL res 304 144 R0
SYMATTR InstName R1
SYMATTR Value 1
SYMBOL Nmos 480 384 R0
SYMATTR InstName M3
SYMATTR Value IRF530
SYMBOL Pmos 480 128 M180
SYMATTR InstName M4
SYMATTR Value IRF7204
SYMBOL ind 672 192 R0
SYMATTR InstName L1
SYMATTR Value 3µ
SYMBOL res 592 192 R0
SYMATTR InstName R2
SYMATTR Value 47
TEXT 40 -40 Left 0 ;'Complimentary PWM Switch Series Inductor R2=52.114mW
TEXT 200 -8 Left 0 !.tran 0 1u 0 10n

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[Transient Analysis]
{
Npanes: 2
{
traces: 3 {34603011,0,"I(L1)"} {34603012,0,"Id(M3)"} {524293,0,"-
Is(M4)"}
X: ('m',1,0,0.0001,0.001)
Y[0]: ('m',0,-0.9,0.1,0.5)
Y[1]: (' ',1,1e+308,0.3,-1e+308)
Amps: ('m',0,0,0,-0.9,0.1,0.5)
Log: 0 0 0
GridStyle: 1
},
{
traces: 2 {34603010,0,"I(R1)"} {524294,1,"V(M4S,M3D)*I(R2)"}
X: ('m',1,0,0.0001,0.001)
Y[0]: (' ',0,-1,1,11)
Y[1]: (' ',1,0,0.3,3.6)
Amps: (' ',0,0,0,-1,1,11)
Units: "W" (' ',0,0,1,0,0.3,3.6)
Log: 0 0 0
GridStyle: 1
}
}

 

[Mike Monett]

Oops - typo. M2S and M4S are mislabeled. They should be M2D and M4D. This
has no effect on the result.

Best Regards,

Mike Monett

 

[Joerg]

I did a small simulation in LTspice. The SPICE model for the
FDD8424H is available from Fairchild, but it is for PSPice. I didn't
want to take the time to make a model for LTspice, so I used IRF530
and IRF7204 for complimentary MOSFETs. I adjusted the rise and fall
time of the gate drive to give about 50ns cross-conduction.

I increased the series inductor between the MOSFETs to 3uH and used
a 47 ohm damping resistor.

The switching losses end up heating the damping resistor, so the
integral of the power dissipated gives the total loss. For this
simulation, the result is 52.114mW.

If the power delivered to the load is 3 watts, using a series
inductor to minimize shoot-through adds 52.114e-3 / 3 = 1.73% to the
total power dissipation. This is a rather small amount for the
simplicity and reliability gained.

Any circuit changes to reduce the power loss in switching will
probably cost additional power, so the overall gain might be small
or negative.

That's just the thing. Whatever you do, in the end you just move the
dissipation from one part to another.

The conclusion is a small series inductor can be a viable option to
minimize shoot-through and reduce circuit complexity.

Yep, so I just did the usual, a concoction of Schmitts, resistors and
diodes. No more cross conduction

The LTSPICE ASC file is below, followed by the PLT file. The .tran
analysis string is set to 1uS to show the switching waveforms.
Increase it to 1ms to calculate the power in R2.

Thanks, Mike, but I am done with that part of the design now. I was just
hoping there was a push-pull driver with much less than 1ohm so I could
drive the transformer directly. Such drivers are usually
process-controlled so their internal cross conduction is minimized.
Something you can't do with tolerance-prone external parts where things
such as Vth stray a lot.