MOSFET driver killed with no particular reason

[[email protected]]

hi to everyone. I mysteriously killed 5 MOSFET half bridge drivers and
wish to see if anyone can help.

I'm making a N-MOSFET H bridge circuit that switch a sine wave output
using unipolar PWM for an inverter project. The gate driver I choose
was IR21834 and I began testing half of the bridge on breadboard
according to the datasheet circuit. Connecting input lines manually to
V+ or GND to produce the output, the lower MOSFET would first work,
but it always happened that somehow after changing the circuit such as
connecting driver input here and there or switching the value of
boostrap capacitor, at a random time the chip would go crazy . The
chip would feel hot touch, followed by increase in current drawn, all
of which signifying an internal short circuit. It seems the hide side
circuit is causing problem, since the low side output would usually
still functioning afterwards.

I believe such short circuit behavior can only be caused by shorting
high side output to ground or to V+, and I swear I didn't, nor did I
connect bootstrap diode and capacitor wrongly (I redid the circuit
several times and same error can't happen over and over again). I
thought it might be chip design problem but as I tried chip from other
manufacture, L6388 from ST, the same happened. Now I'm really
threatened as I don't want to kill any more chips. Although I might
find out the reason myself but that would probably take another 10
chips dedicated to destroy, and this is why I ask for help here.

While not expecting a direct reason to the problem, I am REALLY REALLY
glad if someone who have killed MOSFET driver before (for whatever
reasons) to share his/her story, and also tell me what to avoid when
working with drivers, thousands thanks in advance!

 

[Spehro Pefhany]

hi to everyone. I mysteriously killed 5 MOSFET half bridge drivers and
wish to see if anyone can help.

I'm making a N-MOSFET H bridge circuit that switch a sine wave output
using unipolar PWM for an inverter project. The gate driver I choose
was IR21834 and I began testing half of the bridge on breadboard
according to the datasheet circuit. Connecting input lines manually to
V+ or GND to produce the output, the lower MOSFET would first work,
but it always happened that somehow after changing the circuit such as
connecting driver input here and there or switching the value of
boostrap capacitor, at a random time the chip would go crazy . The
chip would feel hot touch, followed by increase in current drawn, all
of which signifying an internal short circuit. It seems the hide side
circuit is causing problem, since the low side output would usually
still functioning afterwards.

I believe such short circuit behavior can only be caused by shorting
high side output to ground or to V+, and I swear I didn't, nor did I
connect bootstrap diode and capacitor wrongly (I redid the circuit
several times and same error can't happen over and over again). I
thought it might be chip design problem but as I tried chip from other
manufacture, L6388 from ST, the same happened. Now I'm really
threatened as I don't want to kill any more chips. Although I might
find out the reason myself but that would probably take another 10
chips dedicated to destroy, and this is why I ask for help here.

While not expecting a direct reason to the problem, I am REALLY REALLY
glad if someone who have killed MOSFET driver before (for whatever
reasons) to share his/her story, and also tell me what to avoid when
working with drivers, thousands thanks in advance!

Read ALL the application notes carefully. Sounds like latchup
(probably from poor layout).


Best regards,
Spehro Pefhany

 

[[email protected]]

This picture shows what I did:
http://www.geocities.com/w2kwong/MOSFET.jpg
(NOTE: bootstrap diode is built in)

Plus what Spehro said as latchup really fall into my mind, as I have a
"feel" that chip internal logic crashed somehow. As I'm already using
decoupling cap, can you explain a bit more on that thx.

 

[[email protected]]

hi to everyone. I mysteriously killed 5 MOSFET half bridge drivers and
wish to see if anyone can help.
......

While not expecting a direct reason to the problem, I am REALLY REALLY
glad if someone who have killed MOSFET driver before (for whatever
reasons) to share his/her story, and also tell me what to avoid when
working with drivers, thousands thanks in advance!

Avoid drivers - Hrmph!

My first guess is that the driver is way too fast for it's own good so you get
ground bounce and probably a latchup. Or the chip oscillates. Both problems are
due to the layout and decoupling - if there is an application note on how
capacitors should be placed and wires routed then I.M.E. this should be followed
without fail - because this will be the only way that *beeep* chip ever worked
(once, in the lab).

The second guess is that the MOSFET is oscillating, overvolting and killing the
gate then blowing the driver. This is sort-of a layout thing but not always.
Usually I stick a resistor between 10 and 47 Ohms in series with the gate to
preempt any oscillation.

PS:
A discrete bipolar driver curcuit is not more expensive, it is much more
tolerant of abuse and layout and the circuit itself is more robust too. Them
sales people like to boast the biggest slew rates on their powerpoint slides -
while most of the real power electronics applications do not need kV/nsec at the
driver level it at all.

 

[brushhead]

This picture shows what I did:
http://www.geocities.com/w2kwong/MOSFET.jpg
(NOTE: bootstrap diode is built in)

Plus what Spehro said as latchup really fall into my mind, as I have a
"feel" that chip internal logic crashed somehow. As I'm already using
decoupling cap, can you explain a bit more on that thx.

Hello,
What kind of de-coupling cap are you using? You need something with a
very low impedance, in order to supply the current spike for the gate of
the MOSFET.

The other thing is, what is your PWM frequency? You might be buggering
the chip because it's power will be similar to f*c*v^2, as a direct
reflection.

Cheers,
Rob.

 

[[email protected]]

Hello,
What kind of de-coupling cap are you using? You need something with a
very low impedance, in order to supply the current spike for the gate of
the MOSFET.

The other thing is, what is your PWM frequency? You might be buggering
the chip because it's power will be similar to f*c*v^2, as a direct
reflection.

Cheers,
Rob.

I'm using electrolytic plus a green cap for decoupling.
A fairchild application note http://www.fairchildsemi.com/an/AB/AB-9.pdf
suggests the gate resistor to be 4.7ohms. So for 10 to 50 ohms of Rg
suggested by frithiof, would it be too large that may slow down rise
and fall time by too much?

Also it sounds good idea to buy discrete high and low side drivers as
it debugging isolated devices will be easier. Thankyou

 

[brushhead]

I'm using electrolytic plus a green cap for decoupling.
A fairchild application note http://www.fairchildsemi.com/an/AB/AB-9.pdf
suggests the gate resistor to be 4.7ohms. So for 10 to 50 ohms of Rg
suggested by frithiof, would it be too large that may slow down rise
and fall time by too much?

Also it sounds good idea to buy discrete high and low side drivers as
it debugging isolated devices will be easier. Thankyou

I would have though that that would be OK. Increasing the value of tge
gate resistor will increase the switching losses in the FET, but will
reduce the peak current required from the driver; simple Ohm's law that one.

I would not have though that adding discreets will help you though.
Herpahs understanding the current in the FET would be more beneficial.
Have you scoped what the gate drive waveform looks like yet? That might
tell you all you need to know. Also what does the D-S voltage look
like. Is there a lot of overshoot because of a bad layout? Is the
voltage source (CD?) properly de-coupled?

There are loads of things to look at, but one thing my experiences have
taught me is that gate drive is everything to a circuit. You may be
ringing the gate-source capacitance like hell which is causing enourmous
losses int he FET and then killing it.

Cheers,

Rob.

 

[Anthony Fremont]

I'm using electrolytic plus a green cap for decoupling.
A fairchild application note
http://www.fairchildsemi.com/an/AB/AB-9.pdf suggests the gate
resistor to be 4.7ohms. So for 10 to 50 ohms of Rg suggested by
frithiof, would it be too large that may slow down rise and fall time
by too much?

Also it sounds good idea to buy discrete high and low side drivers as
it debugging isolated devices will be easier. Thankyou

I saw the other thread in s.e.misc. If you would have cross posted this
(instead of multiposting) it would have been easier to keep it all straight.


At any rate, is that 100uF cap your Cboot? If so, shouldn't it be more
like .1uF (100nF)?

 

[James Arthur]

<[email protected]> wrote in message

A discrete bipolar driver curcuit is not more expensive, it is much more
tolerant of abuse and layout and the circuit itself is more robust too. Them
sales people like to boast the biggest slew rates on their powerpoint slides -
while most of the real power electronics applications do not need kV/nsec at the
driver level it at all.

I'm curious Frithiof, what sort of bipolar driver do you prefer?
Something like this?:

.. Vcc
.. ---
.. |
.. |
.. |/
.. +----|
.. | |>.
.. | |
.. >-+--R1--+--R2--> to FET
.. | |
.. | |<'
.. +----|
.. |\
.. |
.. |
.. ===
.. GND

I may be driving some big -- okay, 'medium' -- FETs, would need gate
drivers, and the cross-conduction thing is otherwise an annoyance.

With regards,
James Arthur

 

[Mark]

Connecting input lines manually to

V+ or GND to produce the output, the lower MOSFET would first work,
but it always happened that somehow after changing the circuit such as
connecting driver input here and there or switching the value of
boostrap capacitor, at a random time the chip would go crazy .

Are you saying that you manually moved the driver input wires by hand
while the power was on?

Mark

 

[qrk]

This picture shows what I did:
http://www.geocities.com/w2kwong/MOSFET.jpg
(NOTE: bootstrap diode is built in)

Plus what Spehro said as latchup really fall into my mind, as I have a
"feel" that chip internal logic crashed somehow. As I'm already using
decoupling cap, can you explain a bit more on that thx.

Wow, you really mean breadboard. Superstrips are unsuitable for these
types of circuits due to the fast switching speeds. The long
interconnection wires can have enough inductance to produce voltage
spikes that exceed the ratings of your devices. This sort of layout
frequently causes gate punch-through on the FETs.

I notice you have a lack of bypass capacitors. You need good quality
bypass at the driver chip and the FETs. These capacitors need to have
shortest possible leads to reduce the lead inductance. You can get
circuits like your present breadboard layout to work if you bypass
properly. Don't expect to drive much of a load with this setup.

This sort of circuit could benefit using much better breadboard layout
practices. Use perf board and copper tape for your ground and power.
Use SMT bypass capacitors in the range of 100nF to 1uF around the
driver and the FETs. You also need bulk capacitance from an
electrolytic capacitor (use a low esr type for switching power
supplies). Your bootstrap capacitor is generally in the range of
100nF.

Try make your interconnections as short as possible. That could help.

 

[[email protected]]

I saw the other thread in s.e.misc. If you would have cross posted this
(instead of multiposting) it would have been easier to keep it all straight.


At any rate, is that 100uF cap your Cboot? If so, shouldn't it be more
like .1uF (100nF)?

Sorry I'm newbie of using google group, how might I do cross post
instead?

 

[[email protected]]

Are you saying that you manually moved the driver input wires by hand
while the power was on?

Mark

YES!!!

 

[[email protected]]

Sorry I'm newbie of using google group, how might I do cross post
instead?

Please ignore this question, I just found out how, and sorry for my
previous multiposting

 

[brushhead]

Please ignore this question, I just found out how, and sorry for my
previous multiposting

Well it's how you learn. I don't cross post, simply for avoiding
encurring the wrath of the Usenet community.

Cheers,

Rob.

 

[Rich Grise]

Well it's how you learn. I don't cross post, simply for avoiding
encurring the wrath of the Usenet community.

Crossposting is better than multi-posting:
http://www.cybernothing.org/faqs/net-abuse-faq.html#2.3

Cheers!
Rich

 

[legg]

hi to everyone. I mysteriously killed 5 MOSFET half bridge drivers and
wish to see if anyone can help.

I'm making a N-MOSFET H bridge circuit that switch a sine wave output
using unipolar PWM for an inverter project. The gate driver I choose
was IR21834 and I began testing half of the bridge on breadboard

A plug-in breadboard?

Dont't use these for power control circuits.

RL