MOSFET motor drive

[Peter Gottlieb]

Hi all...

I'm building an electric go-cart for the kids using a 24 Volt motor. I have
an H bridge with each leg being 2 paralleled 8 milohm 50 amp MOSFETs. Each
half bridge is driven by a Linear Technology driver which has lockout to
prevent shoot-through.

After I built it, I decided to test with a 75 Watt 24 Volt bulb and a
current limiting power supply. I plugged my bridge assembly into a test set
I built which sends either a "brake" (both lower sets of mosfets on),
"forward" (one upper and one lower mosfet sets on) or "reverse" (opposite
sets of mosfets on) command.

When I give the forward or reverse command, the bulb lights (opposite cmd
gives reversed polarity as expected) but the LT chip starts heating up big
time. I see that the drive starts going down after a few seconds until some
mosfets are in the linear region - not good (although no problem with just
the bulb, I have them on a big heat sink).

Anyone have any idea what's happening? There aren't that many inputs to
that chip and they all look proper, power is clean, nothing going on on the
ground.

I tried hooking up the motor and it surged like crazy with the power supply
showing overvoltage, then current limit (at 25 amps) all over the place.
That is difficult to probe due to the currents and my not wanting to let it
do that for too long. The mosfets didn't even get mildly warm, but again,
the LT chip started getting hot rapidly. Unplugging the LT chip allowed me
to manually drive the mosfets and they seem just fine with nothing but the
usual capacitive gate load. A 400 MHz scope shows no oscillations anywhere.

The LT chip is on the H bridge assembly and the inputs are isolated.

It's gotta be something simple, but what?

 

[Spehro Pefhany]

Hi all...

I'm building an electric go-cart for the kids using a 24 Volt motor. I have
an H bridge with each leg being 2 paralleled 8 milohm 50 amp MOSFETs. Each
half bridge is driven by a Linear Technology driver which has lockout to
prevent shoot-through.

After I built it, I decided to test with a 75 Watt 24 Volt bulb and a
current limiting power supply. I plugged my bridge assembly into a test set
I built which sends either a "brake" (both lower sets of mosfets on),
"forward" (one upper and one lower mosfet sets on) or "reverse" (opposite
sets of mosfets on) command.

When I give the forward or reverse command, the bulb lights (opposite cmd
gives reversed polarity as expected) but the LT chip starts heating up big
time. I see that the drive starts going down after a few seconds until some
mosfets are in the linear region - not good (although no problem with just
the bulb, I have them on a big heat sink).

Anyone have any idea what's happening? There aren't that many inputs to
that chip and they all look proper, power is clean, nothing going on on the
ground.

You didn't give a part number, and I'm not going to search the LTC
site and make guesses, but if it's a charge pump type high-side driver
you cannot run them at 100% on. Now, there ought to be some kind of
undervoltage lockout that prevents the MOSFETs from ever going linear,
but I'm not familiar with LTC's chips.

I tried hooking up the motor and it surged like crazy with the power supply
showing overvoltage, then current limit (at 25 amps) all over the place.
That is difficult to probe due to the currents and my not wanting to let it
do that for too long. The mosfets didn't even get mildly warm, but again,
the LT chip started getting hot rapidly. Unplugging the LT chip allowed me
to manually drive the mosfets and they seem just fine with nothing but the
usual capacitive gate load. A 400 MHz scope shows no oscillations anywhere.

The LT chip is on the H bridge assembly and the inputs are isolated.

It's gotta be something simple, but what?

Best regards,
Spehro Pefhany

 

[Robert Monsen]

The chip might have some kind of thermal shutdown built in. Are you using a
BIG heatsink?

Regards

 

[Peter Gottlieb]

Not on the chip. The fets are on a huge one though, way larger than
necessary.

 

[Eric Y. Chang]

Peter Gottlieb ([email protected]) wrote:
: Hi all...

: When I give the forward or reverse command, the bulb lights (opposite cmd
: gives reversed polarity as expected) but the LT chip starts heating up big
: time. I see that the drive starts going down after a few seconds until some
....
: I tried hooking up the motor and it surged like crazy with the power supply
: showing overvoltage, then current limit (at 25 amps) all over the place.

Not a good idea if you fail the light bulb test. Why did you proceed?

: That is difficult to probe due to the currents and my not wanting to let it
: do that for too long. The mosfets didn't even get mildly warm, but again,

Does your scope have a storage mode on it?

: the LT chip started getting hot rapidly. Unplugging the LT chip allowed me
: to manually drive the mosfets and they seem just fine with nothing but the
: usual capacitive gate load. A 400 MHz scope shows no oscillations anywhere.

This is key. A lot of problems are due to drain-gate coupled oscillations.
You must have done a test to exclude Miller driver oscillations. That is to
replace the MOSFET's with little capacitors (with resistors in series) and
no B+. If the LT chip gets hot, there's a big problem that you will have to
solve by breaking things down. If not, there is an oscillation that you
must have missed. So what was the result of this test?

 

[Jim Thompson]

Everyone knows this scenario all too well. You are at the airport,
late for your flight, running to the gate, all the while you are on a
critical conference call. Then, all of a sudden, you hear the sound of
death 'beep beep', your cell battery goes dead. You frantically
search for your charger that you have packed away in your bag. You
have lost a crucial call and more importantly your ability to
communicate. Never lose your ability to communicate. Cellboost gives
you that power! Visit shithead.com for more info. "Don't Be Caught
Dead".

Death to spammers! Complaints to: [email protected]

...Jim Thompson

 

[Yzordderex]

I assume you are turning off one switch in leg before turning on
another. Need a microsecond or two commutation delay. Never want to
turn on two across at once! I also assume you are running some kind
of anti-parallel diodes across each fet. It is a risky venture to use
intrinsic diodes to commutate current.

Gate driver should mount close to fet. Keep leads short. Gate drive
resistor should be about 500/current, or ballpark 10 ohms or less for
a 50a fet. This is a rule of thumb just to get motor turning, and may
not be final value. Driver should be capable of at least an amp or so
pulse. I've used two 9v batteries in series to run upper IGBT
switches in a pinch. Would need good capacitor to hold up voltage to
get nice gate drive pulses. Yep, good power supply decoupling caps on
each driver chip.

Miller capacitance is between drain and gate. This must be charged
when turning on fet. Current goes into gate to turn on, and out of
gate to turn off. Just think of a capacitor connected directly from
gate to drain and go through brain exercise to see what currents need
to be to drive voltage across cap. Driver circuit must be able to
sink and source these large currents. Look at output of driver chip
with one probe, and directly on gate with other probe. Assume you
have dual trace scope. If AOK you will see small step in gate voltage
where drain voltage is rising or falling. timeframe is 10ns-100ns
plus or minus. Reducing drive resistor will make step smaller and
reduce losses, but tradeoff is perhaps diode noise and driver chip
losses. If you make resistor larger, step gets larger, and eventually
you will see oscillations on gate. Not a good place to run. If you
have more than enough heatsink stay at manufacturers upper
reccomendation for gate drive resistor. Let the fet run a little
hotter and keep the noise down.

 

[Peter Gottlieb]

I assume you are turning off one switch in leg before turning on
another. Need a microsecond or two commutation delay. Never want to
turn on two across at once! I also assume you are running some kind
of anti-parallel diodes across each fet. It is a risky venture to use
intrinsic diodes to commutate current.

The LT driver takes care of the logic to make sure both fets don't go on at
once. I'm not using external diodes, perhaps I should, I was just going to
trust the internal ones. What should I use?

Gate driver should mount close to fet. Keep leads short. Gate drive
resistor should be about 500/current, or ballpark 10 ohms or less for
a 50a fet. This is a rule of thumb just to get motor turning, and may
not be final value. Driver should be capable of at least an amp or so
pulse. I've used two 9v batteries in series to run upper IGBT
switches in a pinch. Would need good capacitor to hold up voltage to
get nice gate drive pulses. Yep, good power supply decoupling caps on
each driver chip.

LT chip is right next to the fets, mounted on a little board on the heat
sink assembly. I'm using 10 ohm resistors. The driver is well decoupled
and good for several amps.

Miller capacitance is between drain and gate. This must be charged
when turning on fet. Current goes into gate to turn on, and out of
gate to turn off. Just think of a capacitor connected directly from
gate to drain and go through brain exercise to see what currents need
to be to drive voltage across cap. Driver circuit must be able to
sink and source these large currents. Look at output of driver chip
with one probe, and directly on gate with other probe. Assume you
have dual trace scope. If AOK you will see small step in gate voltage
where drain voltage is rising or falling. timeframe is 10ns-100ns
plus or minus. Reducing drive resistor will make step smaller and
reduce losses, but tradeoff is perhaps diode noise and driver chip
losses. If you make resistor larger, step gets larger, and eventually
you will see oscillations on gate. Not a good place to run. If you
have more than enough heatsink stay at manufacturers upper
reccomendation for gate drive resistor. Let the fet run a little
hotter and keep the noise down.

Oh, so Miller oscillation is when the gate resistor is too high a value and
so decouples the gate from the driver too much, combined with the feedback
from the output and the gain?

I'll look at the waveforms on the scope. I have one of those monster Tek
7000 series scopes so can presently do 4 channels. Perhaps I will store and
photograph if it looks interesting.

Now I just have to find another chunk of time, perhaps tonight, to get back
into the lab.

 

[Peter Gottlieb]

Did you zap a MOSFET gate? It doesn't always cause a complete failure.
Pull out the chip and make sure there's no current path through the
gates.

I shall double check tonight. I thought I checked, but I may have missed it
as I kept getting interrupted the last time I worked on it. That would
certainly cause this problem.

 

[Paul Burridge]

Did you zap a MOSFET gate? It doesn't always cause a complete failure.
Pull out the chip and make sure there's no current path through the
gates.

Surely he's not using a dual-gate MOSFET??

 

[Peter Gottlieb]

Surely he's not using a dual-gate MOSFET??

I think he meant that I might have one of my 8 fets with a shorted gate.
Easy enough to test. Maybe I'll go down to the lab now and check.

 

[Peter Gottlieb]

All mosfets check good. I'm going to feed the LT chip a proper PWM signal
and see if that is the problem.

 

[Peter Gottlieb]

Gates are fine. Waiting to get back to the lab for further testing.