Maker Pro
Maker Pro

gate drive transformer vs isolated supply fet driver

J

Jamie Morken

Jan 1, 1970
0
Hi,

I am trying to decide what is the best way to drive a 600V Hbridge,
which requires fast (optimally 10ns or shorter) switching.

The two options I see are either a gate drive transformer, 1 primary,
and two secondaries for driving each leg of the 600V Hbridge. Will the
leakage inductance of the gate drive transformer may cause excessive
gate ringing at this fast switching speed?

The second option is an isolated 18Volt supply with an optoisolated
high current fetdriver IC. The required common mode resistance of the
optoisolator will have to be 60kV/us at 10ns switching speed for
600Volts I think, the fastest opto I've seen so far is 15kV/us
(HCPL-3120) I guess there are faster ones though?

Which of these two options would work best at this high voltage, fast
switching?

cheers,
Jamie
 
D

D from BC

Jan 1, 1970
0
Hi,

I am trying to decide what is the best way to drive a 600V Hbridge,
which requires fast (optimally 10ns or shorter) switching.

The two options I see are either a gate drive transformer, 1 primary,
and two secondaries for driving each leg of the 600V Hbridge. Will the
leakage inductance of the gate drive transformer may cause excessive
gate ringing at this fast switching speed?

The second option is an isolated 18Volt supply with an optoisolated
high current fetdriver IC. The required common mode resistance of the
optoisolator will have to be 60kV/us at 10ns switching speed for
600Volts I think, the fastest opto I've seen so far is 15kV/us
(HCPL-3120) I guess there are faster ones though?

Which of these two options would work best at this high voltage, fast
switching?

cheers,
Jamie

IIRC..
GMR isolators 110Mbps
RF isolators 150Mbps
Opto lost the race.

If I understand this right...
Why not use a mosfet driver IC that has a bootstrap circuit for the
high side?
Example:
http://cache.national.com/ds/LM/LM2724A.pdf


D from BC
British Columbia
Canada.
 
J

Joerg

Jan 1, 1970
0
Jamie said:
Hi,

I am trying to decide what is the best way to drive a 600V Hbridge,
which requires fast (optimally 10ns or shorter) switching.

The two options I see are either a gate drive transformer, 1 primary,
and two secondaries for driving each leg of the 600V Hbridge. Will the
leakage inductance of the gate drive transformer may cause excessive
gate ringing at this fast switching speed?

The second option is an isolated 18Volt supply with an optoisolated
high current fetdriver IC. The required common mode resistance of the
optoisolator will have to be 60kV/us at 10ns switching speed for
600Volts I think, the fastest opto I've seen so far is 15kV/us
(HCPL-3120) I guess there are faster ones though?

Which of these two options would work best at this high voltage, fast
switching?

I'd first look at the transformer method. I never had a problem sending
rise and fall times under 10nsec across. In fact, sometimes the whole
pulse wasn't much longer than that.

Two things to mind: You need to be or become familiar with multi-filar
winding techniques. Also, make sure any cross conduction issues are
taken into account.

If you can find a suitable transformer off the shelf and it's only one
secondary consider using two of them.

As for optocouplers they usually don't produce much oomph and most are a
bit sluggish for my taste. I'd imagine that you have to swing a
substantial gate capacitance around.
 
J

Joerg

Jan 1, 1970
0
D said:
IIRC..
GMR isolators 110Mbps
RF isolators 150Mbps
Opto lost the race.

If I understand this right...
Why not use a mosfet driver IC that has a bootstrap circuit for the
high side?
Example:
http://cache.national.com/ds/LM/LM2724A.pdf

If you use it in a 600V application .... WHADDABANG!

AFAIK they aren't rated above 30V or so.
 
J

Jamie Morken

Jan 1, 1970
0
Joerg said:
Hi,


I'd first look at the transformer method. I never had a problem sending
rise and fall times under 10nsec across. In fact, sometimes the whole
pulse wasn't much longer than that.

Two things to mind: You need to be or become familiar with multi-filar
winding techniques. Also, make sure any cross conduction issues are
taken into account.

Yes this is what I was wondering about, I guess the multi-filar winding
is to minimize the leakage inductance as much as possible to prevent
gate ringing?
If you can find a suitable transformer off the shelf and it's only one
secondary consider using two of them.

For controlling all quadrants of a fullbridge using gate drive
transformers I think these are the options:

1. 1 transformer with 4 isolated secondaries (one per fullbridge
quadrant)

2. 1 transformer with 3 isolated secondaries, with one centertapped
for the lowside fets, and the other two coils for the high side fets

3. 2 transformers, one with a centertapped secondary, and another with
dual secondaries.

As for optocouplers they usually don't produce much oomph and most are a
bit sluggish for my taste. I'd imagine that you have to swing a
substantial gate capacitance around.


Yes for the isolated supply+optocoupler option I was planning on using
the optocoupler to drive a floating fet driver, ie. TC4452 12Amp fet
driver.

cheers,
Jamie
 
J

Joerg

Jan 1, 1970
0
Jamie said:
Yes this is what I was wondering about, I guess the multi-filar winding
is to minimize the leakage inductance as much as possible to prevent
gate ringing?

It's to make sure the bandwidth is high enough so you get your <10nsec
transitions. And yes, it'll curb ringing at the same time. I'd start
with toroids such as FT43-xx from Fair-Rite. The xx is for the diameter
which you'd have to pick so you can comfortably get the wires wound.

The material depends a bit on max on-time, assuming it's no more than a
few microseconds.
For controlling all quadrants of a fullbridge using gate drive
transformers I think these are the options:

1. 1 transformer with 4 isolated secondaries (one per fullbridge
quadrant)

2. 1 transformer with 3 isolated secondaries, with one centertapped
for the lowside fets, and the other two coils for the high side fets

3. 2 transformers, one with a centertapped secondary, and another with
dual secondaries.

Well, whatever you do, chart the turn-on and turn-off times on a piece
of grid paper first. You don't want to have any overlap such as having
both sides of one leg conducting at the same time.
Yes for the isolated supply+optocoupler option I was planning on using
the optocoupler to drive a floating fet driver, ie. TC4452 12Amp fet
driver.

Then you'd need a supply on that side as well.
 
P

Paul Mathews

Jan 1, 1970
0
Yes this is what I was wondering about, I guess the multi-filar winding
is to minimize the leakage inductance as much as possible to prevent
gate ringing?




For controlling all quadrants of a fullbridge using gate drive
transformers I think these are the options:

1. 1 transformer with 4 isolated secondaries (one per fullbridge
quadrant)

2. 1 transformer with 3 isolated secondaries, with one centertapped
for the lowside fets, and the other two coils for the high side fets

3. 2 transformers, one with a centertapped secondary, and another with
dual secondaries.




Yes for the isolated supply+optocoupler option I was planning on using
the optocoupler to drive a floating fet driver, ie. TC4452 12Amp fet
driver.

cheers,
Jamie



- Hide quoted text -

- Show quoted text -- Hide quoted text -

- Show quoted text -

Some additional tips:
1) Gate drive transformers are often used in conjunction with a PNP
transistor on the secondary side to speed up off transitions. There's
a nice TI/Unitrode App Note called something like "Gate Drive
Techniques" that covers this and other bits.
2) If you're using a current sense resistor in any of your MOSFET
drain circuits, transformer drive gives you the opportunity to drive
the gate-source without having the voltage drop across the current
sense resistor subtract from gate drive. Many app note schematics miss
this point.
3) Beware of flux-walking in full-bridge circuits. This was discussed
in another thread recently. Consider using a series capacitor in the
primary circuit to absolutely prevent this possibility. Do not believe
anyone who says that "current mode control prevents flux walking". It
does, but only some of the time.
Paul Mathews
 
D

D from BC

Jan 1, 1970
0
Yep, a bit sluggish, won't come anywhere close to his 10nsec. But it
sure is an interesting part.

I'm not sure why these chips are not faster.
It's only takes 1.5amps to charge 1000pF in 10nS.
Perhaps the OP can find a faster chip or a high speed version based on
the topology.


D from BC
British Columbia
Canada.
 
Top