Inverting vs NonInverting Mosfet Driver

[D from BC]

I have some fuzzy recollection of reading to the effect that
noninverting mosfet drivers are more prone to instability due to layout
effects.

Confirm?

Just doing a quick sim in my head:
With poor layout, when a noninverting mosfet driver turns on (sourcing),
the signal ground pops up due to the mosfet gate capacitance (iirc
called ground bounce?) then the driver sees a valid (actually invalid)
'0' threshold to turn the mosfet drive off.. The driver oscillates.

Is it best to use a noninverting mosfet driver in a smps design?

 

[Joerg]

I have some fuzzy recollection of reading to the effect that
noninverting mosfet drivers are more prone to instability due to layout
effects.

Confirm?

Not affirmative

Just doing a quick sim in my head:
With poor layout, when a noninverting mosfet driver turns on (sourcing),
the signal ground pops up due to the mosfet gate capacitance (iirc
called ground bounce?) then the driver sees a valid (actually invalid)
'0' threshold to turn the mosfet drive off.. The driver oscillates.

Is it best to use a noninverting mosfet driver in a smps design?

I have used both, depending on what was needed or sometimes what Digikey
had in stock. Never a problem with either. Of course I never design
anything without a full ground plane. Also, most modern FET drivers have
input hysteresis so it's kind of tough to get them to oscillate without
deliberate and serious feedback. Which I sometimes did, in order to use
them as poor man's switch mode controllers and that trick only works
with the inverting kind.

 

[legg]

I have some fuzzy recollection of reading to the effect that
noninverting mosfet drivers are more prone to instability due to layout
effects.

Confirm?

Just doing a quick sim in my head:
With poor layout, when a noninverting mosfet driver turns on (sourcing),
the signal ground pops up due to the mosfet gate capacitance (iirc
called ground bounce?) then the driver sees a valid (actually invalid)
'0' threshold to turn the mosfet drive off.. The driver oscillates.

Is it best to use a noninverting mosfet driver in a smps design?

You have to test drive any prospective part, by any mfr, in-circuit.
Duals and non-inverters are more prone to misbehaviour. Low voltage
logic level inputs are a mistake to be avoided, wherever possible,
even with a 'ground plane'. Some parts are even sensitive to output
disturbances, never mind ground bounce on the input, regardless of
sales blurbs or specsmanship.

For non-inversion, bypass Micrel MIC4424 parts, if you want to avoid
grey hair. Similar lower-powered Maxim parts MAX4427A or Micrel TC4427
seemed OK, although I recall a lack of internal UVLO, which required
vigilance.

RL

 

[D from BC]

You have to test drive any prospective part, by any mfr, in-circuit.
Duals and non-inverters are more prone to misbehaviour. Low voltage

Ahhhh..

logic level inputs are a mistake to be avoided, wherever possible,
even with a 'ground plane'. Some parts are even sensitive to output
disturbances, never mind ground bounce on the input, regardless of
sales blurbs or specsmanship.

For non-inversion, bypass Micrel MIC4424 parts, if you want to avoid
grey hair. Similar lower-powered Maxim parts MAX4427A or Micrel TC4427
seemed OK, although I recall a lack of internal UVLO, which required
vigilance.

RL

INteresting. Thanks

I'm using a MIC4452 non-inverting.

 

[Joerg]

You have to test drive any prospective part, by any mfr, in-circuit.
Duals and non-inverters are more prone to misbehaviour. Low voltage
logic level inputs are a mistake to be avoided, wherever possible,
even with a 'ground plane'. Some parts are even sensitive to output
disturbances, never mind ground bounce on the input, regardless of
sales blurbs or specsmanship.

For non-inversion, bypass Micrel MIC4424 parts, if you want to avoid
grey hair. Similar lower-powered Maxim parts MAX4427A or Micrel TC4427
seemed OK, although I recall a lack of internal UVLO, which required
vigilance.

Never had a problem with Micrel MOSFET drivers. What caused the gray hair?

 

[legg]

Never had a problem with Micrel MOSFET drivers. What caused the gray hair?

Basically, I had opposing mosfet outputs turning on by themselves on
the low side of a full bridge driver. The duration of conduction could
be reduced but not eliminated with agressive supply decoupling. The
input of the offender scoped slightly negative during the entire drive
fault period, following a positive glitche of 100nS duration,
possibly generated by it's partner. No other pin-compatible part
behaved this way, in the same physical position.

In it's 'representative schematic', the 4424 input is depicted as
analog, with some kind of current hysterisis introduced to the signal,
at the receiver's output, which is just plain barmy, IMHO.

RL

 

[Joerg]

Basically, I had opposing mosfet outputs turning on by themselves on
the low side of a full bridge driver. The duration of conduction could
be reduced but not eliminated with agressive supply decoupling. The
input of the offender scoped slightly negative during the entire drive
fault period, following a positive glitche of 100nS duration,
possibly generated by it's partner. No other pin-compatible part
behaved this way, in the same physical position.

That almost sounds like a damaged chip. I've mainly used the 4421 but
AFAIK they are all the same architecture. Ok, they aren't really
shoot-through protected but they ran nice and cool at a few hundred kHz.
They do need a really stiff supply with two planes and good X7R caps,
else all hell can break loose. It also does if you hang too big a gate
capacitance onto it, which I guess is why they also make 6A, 9A and 12A
devices.

In it's 'representative schematic', the 4424 input is depicted as
analog, with some kind of current hysterisis introduced to the signal,
at the receiver's output, which is just plain barmy, IMHO.

It actually works, a few hundred mV hysteresis. Not barmy

 

[legg]

That almost sounds like a damaged chip. I've mainly used the 4421 but
AFAIK they are all the same architecture. Ok, they aren't really
shoot-through protected but they ran nice and cool at a few hundred kHz.
They do need a really stiff supply with two planes and good X7R caps,
else all hell can break loose. It also does if you hang too big a gate
capacitance onto it, which I guess is why they also make 6A, 9A and 12A
devices.

This was not a damaged chip - substitutions and iterations in
documentatiomn of the fault established this. All outputs were only
required to drive the gate limiting resistor, which dominated gate
current control in an assisted switching circuit, where the big fet in
question had it's active capacitances discharged in advance by an
external switch.

I suspect it might have been the reverse transfer current hitting the
output, before it was required to be active high, that scrambled
adjacent internal logic of it's partner, somehow, but that is just
speculation. I wasn't prepared to rip apart the entire topology in
order to investigate further, with functional substitutes on-hand.

It actually works, a few hundred mV hysteresis. Not barmy

It obviously doesn't work well enough to be specified in the part's
datasheet, as such. The concept should have been buried on that basis
alone.

RL

 

[Joerg]

This was not a damaged chip - substitutions and iterations in
documentatiomn of the fault established this. All outputs were only
required to drive the gate limiting resistor, which dominated gate
current control in an assisted switching circuit, where the big fet in
question had it's active capacitances discharged in advance by an
external switch.

I suspect it might have been the reverse transfer current hitting the
output, before it was required to be active high, that scrambled
adjacent internal logic of it's partner, somehow, but that is just
speculation. I wasn't prepared to rip apart the entire topology in
order to investigate further, with functional substitutes on-hand.

Sorry to hear that, it's really strange. I have used Micrel drivers a
lot and they always delivered. Typically sans gate resistor because I
like to drive FETs with gusto where permitted.

It obviously doesn't work well enough to be specified in the part's
datasheet, as such. The concept should have been buried on that basis
alone.

Well, at least they state in in the text:

http://www.micrel.com/_PDF/mic4423.pdf

Quote "Following the input stage is a buffer stage which provides
~400mV of hysteresis for the input, ..."