half bridge mosfet switching

[[email protected]]

I have a simple half mosfet bridge. The mosfet i use has crss 350pF and
Coss as 870pF. So the Cds would be 520pF. I have a gating pulse to
these mosfet with deadtime of about 500ns, which more than enough to
prevent shoot through. Suppose i have no load connected to the bridge
and i just apply the gating pulse. The bridge keg should draw current
that is charging the Cds. When i calculated it, it comes out to be 1A.
But when i do simulation and experiment, i see it to be more than 6A. I
am not able to track down the logic for it. Any MOSFET expert please
explain me.
kristo.

 

[Pooh Bear]

I have a simple half mosfet bridge. The mosfet i use has crss 350pF and
Coss as 870pF. So the Cds would be 520pF. I have a gating pulse to
these mosfet with deadtime of about 500ns, which more than enough to
prevent shoot through. Suppose i have no load connected to the bridge
and i just apply the gating pulse. The bridge keg should draw current
that is charging the Cds. When i calculated it, it comes out to be 1A.
But when i do simulation and experiment, i see it to be more than 6A. I
am not able to track down the logic for it. Any MOSFET expert please
explain me.

I don't consider myself to be an *expert* but I happen to have exactly this
on the bench right now. Vbus is 320 V. Similar specs to your example.

No load, it draws virtually nothing ( other than transformer magnetising
current when that's connected )

What switching frequency ? Have you been purely simulating or actually got
a working example ?

Graham

 

[[email protected]]

i am switching at 80khz. i see both in experiment and simulation. still
wondering why?

 

[Artem]

6A at 320v is

320*6
1920 Watt.

You mosfet should be a VERY hot (but only for a short time )

If you mosfet did not burn-out and cool. Than mean that you measuring
current wrong.

 

[Pooh Bear]

i am switching at 80khz. i see both in experiment and simulation. still
wondering why?

If that much curent is being drawn, something must be getting *very* hot
very fast !

So, what's getting hot ? What's your Vbus ?

Graham

 

[Artem]

If that much curent is being drawn, something must be getting *very* hot
very fast !

So, what's getting hot ? What's your Vbus ?

actually if 6A is pulsed current. Transistors could be cool. But
measuring pulsed current is not simply task.

I this that if transistors is cool, it's ok. All right.

 

[Terry Given]

actually if 6A is pulsed current. Transistors could be cool. But
measuring pulsed current is not simply task.

It can be easy. As there is no load, slap a resistor in series with the
+DC bus. If there really is cross-conduction the voltage will collapse.
use a small cap (just enough to charge up the expected FET capacitance)
after the resistor.

I this that if transistors is cool, it's ok. All right.

he might be measuring gate charging current....if he has a low-side
current sense resistor. If so, change the layout so gate current is not
included (route gate driver V- to FET side of current sense resistor).

Cheers
Terry

 

[Artem]

It can be easy. As there is no load, slap a resistor in series with the
+DC bus. If there really is cross-conduction the voltage will collapse.
use a small cap (just enough to charge up the expected FET capacitance)
after the resistor.

Current transformer on the shielded cable may be more useful, even at
DC current.

he might be measuring gate charging current....if he has a low-side
current sense resistor. If so, change the layout so gate current is not
included (route gate driver V- to FET side of current sense resistor).

Gate current 15v / 10Om * 4transistors = 6A.
But it's too simply. And current pulses must be much more shortest.

 

[[email protected]]

hi,
the mosfets is getting hot. i am not measuring gate current. I am
measuring the dain current. my Vbus is 400V. i am using IRFp460 mosfet.

 

[Artem]

the mosfets is getting hot. i am not measuring gate current. I am
measuring the dain current. my Vbus is 400V. i am using IRFp460 mosfet.

MOSFET with radiator or without radiator? If MOSFET getting hot WITH
radiator, this is very bad.

Check common DC current. (Before condencer, after bridge, with
oscilloscope and shunt, rms value).

1. Are you shure that MOSFETs is completely close/open?
2. How you make a gate driver?
3. What about parasitic diode in MOSFET? Is this diode fast enough?
4. Did you use a some kind of snubbers?

 

[[email protected]]

since the current pulse occurs during turning on either mosfet on the
half bridge...
since Vbus=400V, the current pulse is 6A peak.
the approx. average power dissipated = 200*3*1/20=30Watts...1/20 is
current conduction time by the period it repeats.
should the mosfet get hot with this power.

 

[Terry Given]

hi,
the mosfets is getting hot. i am not measuring gate current. I am
measuring the dain current. my Vbus is 400V. i am using IRFp460 mosfet.

If they are getting hot then your measurement is probably real.

What does your gate waveform look like? if it has a broad flat spot at
about Vth then your gatedrive impedance is too high (and is losing the
fight with Cmiller)

What value of Rgate?

Halve the switching frequency. that ought to halve the temperature rise.

then change the dead time to, say, 2us.

If the FETs stop getting hot, its a cross-conduction problem. Beware low
Vth and high Vgate - it takes a small fraction of a time constant to
charge from zero up to Vth, but many time constants to discharge from
Vgate down to Vth, so if Rg is high then 500ns may not be enough.
Reducing Rgate reduces the time constant, and hence the required dead time.

If the FETs still get hot then its not cross-conduction but rather
switching loss, which can also be caused by gatedrive.


Also, FETs have positive thermal feedback. For a sufficiently high
Rtheta a FET can suffer thermal run away, and because of the positive
feedback (Rds gets higher with T, and usually I is constant so I^2Rds
increases with T) doubling the heatsink Rtheta can more than double dT.

Cheers
Terry

 

[Terry Given]

Terry Given wrote:
[snip]

If the FETs stop getting hot, its a cross-conduction problem. Beware low
Vth and high Vgate - it takes a small fraction of a time constant to
charge from zero up to Vth, but many time constants to discharge from
Vgate down to Vth, so if Rg is high then 500ns may not be enough.
Reducing Rgate reduces the time constant, and hence the required dead time.

Oh yeah, with ultra-low Vth FETs and a really bad gatedrive, its
possible they dont ever turn off. I saw that once with a logic gate
(0.4V output low) driving a complementary emitter follower (add a Vbe)
driving a Supertex FET with Vth = 1V when cold. Vth dropped to about
0.8V when hot, at which point the damn thing wouldnt turn off.....

but I doubt thats your problem.

Cheers
Terry

 

[[email protected]]

hi ,
intially i thought so much current will not flow through drain of the
top mosfet so i did not put radiators. what i thought was:
since maximum dv/dt across mosfet is 400/100*10-9 and cap of mosfet is
maximum of 500pF. so worst case current would be 2A. is there any wrong
in my fundamental calculations?

yes i checked the mosfet are compeletly open/close. i use ir2110 for
driver with gate resistance of 25ohms. a slight increase in current is
seen when gate resistance is increased. i assume mosfet diode to be
fast enough as i see in datasheet the turn on/off delay and rise/fall
time are in terms of tens/hundrends of nanoseconds.
i have not used any snubber.

 

[Artem]

IRFp640 is not a "logic-level" device.

 

[[email protected]]

terry, if they do not turn off, then i must get shoot through with
which dc link voltage will be collapsing..but the voltage does not
collapse. current is there only during the turn ON of the mosfets.

 

[Artem]

intially i thought so much current will not flow through drain of the
top mosfet so i did not put radiators. what i thought was:
since maximum dv/dt across mosfet is 400/100*10-9 and cap of mosfet is
maximum of 500pF. so worst case current would be 2A. is there any wrong
in my fundamental calculations?

In my experience IRFp640 and IRF720 without load and radiator was cool.

yes i checked the mosfet are compeletly open/close. i use ir2110 for
driver with gate resistance of 25ohms. a slight increase in current is
seen when gate resistance is increased. i assume mosfet diode to be
fast enough as i see in datasheet the turn on/off delay and rise/fall
time are in terms of tens/hundrends of nanoseconds.

-------------
trr Reverse Recovery Time 680ns - max
-------------
Try to increase death-time to 2000ns (at least for test).

You can connect 2 ultra-fast diodes for each transistors:
one in serial between source and "-" and another between "-" and drain.
By this way you can "disconnect" internal diode.

i have not used any snubber.

Maybe it's bad idea.

 

[Terry Given]

IRFp640 is not a "logic-level" device.

I couldnt be bothered to look, so used the weasel words "but I doubt
that is your problem"

 

[Terry Given]

terry, if they do not turn off, then i must get shoot through with
which dc link voltage will be collapsing..but the voltage does not
collapse. current is there only during the turn ON of the mosfets.

you will only see that problem when using insanely low threshold FETs,
at elevated temperatures, with a lousy gatedrive

 

[Terry Given]

hi ,
intially i thought so much current will not flow through drain of the
top mosfet so i did not put radiators. what i thought was:
since maximum dv/dt across mosfet is 400/100*10-9 and cap of mosfet is
maximum of 500pF. so worst case current would be 2A. is there any wrong
in my fundamental calculations?

yes i checked the mosfet are compeletly open/close. i use ir2110 for
driver with gate resistance of 25ohms. a slight increase in current is
seen when gate resistance is increased. i assume mosfet diode to be
fast enough as i see in datasheet the turn on/off delay and rise/fall
time are in terms of tens/hundrends of nanoseconds.
i have not used any snubber.

thats a fairly high gate resistor, the IR2110 has a fair bit as well.
Try a diode across Rg, anode to FET, so turn OFF is a lot faster than
turn ON.