TC1410 MOSFET driver problems

[Paul E. Schoen]

I just recently received eight SMT PC boards for a boost converter for 12
VDC nominal to 750 mA into a series high power LED string that may drop 24
to 40 VDC. It uses a Microchip 16F684 with 100 kHz PWM output, a TC1410
driver, and a Fairchild HUF75645 MOSFET, with a 10 uH inductor and a 100V
6A Schottky to drive the LEDs. In my prototype, the TC1410 shorted to Vdd
when I raised the input voltage to about 16 VDC with a poorly filtered
supply, and I figured it probably had voltage surges above its 20V maximum.
I replaced the IC with a 3 transistor driver, and it worked fine ever
since.

When testing the new board, the PWM seemed to work OK for an instant, and
then the TC1410 failed. Several others also failed, sometimes with output
shorted to ground, sometimes to Vdd. I added a 100 ohm resistor to the
output, and it failed again. I added anothe 100 ohm resistor to the Vdd
supply, and it survived, but did not properly drive the MOSFET.

The HUF75645 has a gate capacitance of 3800 pF, which is much higher than
the 500 pF for which the driver is characterized, but its internal
resistance of 16 ohms should have limited the output current to less than 1
ampere. It is rated at 500 mA, so I suspect this is the problem. However,
even a smaller MOSFET with 350-400 pF will also draw the same peak current,
although for a much shorter time. Simulating it in LTspice shows only about
100 mW total gate power for the 3800 pF gate. I have ordered some TC1413
drivers, rated at 3 amps, and characterized for 1500 pF, but they have 2.7
ohm resistance, so they would see pulse current of 6 amps at 16 VDC. The
app notes show direct connection to the MOSFET gate.

I can add 20-30 ohms to the gate and still get reasonable switching times,
while limiting the peak current to 500 mA for the existing device. However,
I'm fairly certain that the one driver failed with 100 ohms. The Vdd supply
is isolated from the main 10-16 VDC supply with a diode, and bypassed with
100 uF and 1 uF. The main supply is bypassed with 0.47 uF. I am very
concerned about reliability, as these boards will be potted in an assembly
that will not be repairable. It is a high power LED diving flashlight that
will have a parts cost of well over $100.

I'm waiting on a response from Microchip, but I am concerned that these
drivers may have reliability problems. I am considering redesign of the
boards to use the simple three-transistor driver that has been working for
a while in my rough prototype. Any suggestions?

Thanks,

Paul

 

[[email protected]]

[...]
Try #2.

PWM is too high. Most of the driver chips fall over beyond 30kHz.
Pushed one chip to 100kHz, ran but was taking 1A while smoking.

 

[[email protected]]

[...]

The test circuit in the data sheet is a 100 kHz square wave. The rise and
fall times of the IC are 25-40 nSec. While the circuit was working, the
driver stayed cold, and my simulation showed less than 100 mW. The
three-transistor driver has rise and fall times of about 200 nSec. The
Microchip App Note shows a 250 kHz example with 16 mW driver power.

I think there is something else wrong. I have not heard back yet from
Microchip. I have also requested samples of some TI parts that have an
output stage consisting of MOSFETs in parallel with bipolars, which
provides better drive through the transition area where the MOSFET actually
switches.

Thanks for your comments.

Paul

Yes. I missed the 100kHz test diagram. Just spiced it myself though
and am seeing roughly 100ma rms gate drive but more interestingly (at
switch off) there's a 10V peak, delayed negative voltage pulse feeding
back to the driver (about 200nS wide). Seems the driver's reverse
limit is 5V.
john

 

[Paul E. Schoen]

[...]

The test circuit in the data sheet is a 100 kHz square wave. The rise
and
fall times of the IC are 25-40 nSec. While the circuit was working, the
driver stayed cold, and my simulation showed less than 100 mW. The
three-transistor driver has rise and fall times of about 200 nSec. The
Microchip App Note shows a 250 kHz example with 16 mW driver power.

I think there is something else wrong. I have not heard back yet from
Microchip. I have also requested samples of some TI parts that have an
output stage consisting of MOSFETs in parallel with bipolars, which
provides better drive through the transition area where the MOSFET
actually
switches.

Thanks for your comments.

Paul

Yes. I missed the 100kHz test diagram. Just spiced it myself though
and am seeing roughly 100ma rms gate drive but more interestingly (at
switch off) there's a 10V peak, delayed negative voltage pulse feeding
back to the driver (about 200nS wide). Seems the driver's reverse
limit is 5V.
john

My reply from Microchip:

Problem Resolution:
I suspect the problem is that you are driving too much current. Since we
don't know what the impedance of the mosfet is, we don't offer parts with
built in limiting, but you can do this as you obsereved by adding a
resistor.

You will need a driver with more current output. More current output
translates to more speed. Also, reducing the capacitance of the gate will
also help to decrease time, since there is less capacitor to charge.

To know what currents are being driven, put a scope probe on either side of
the series resistor into the mosfet gate and put the scope in differential
mode. You will then see the spike of current driven by the driver. If it
exceeds the maximum spec, you will need a larger resistor to properly
protect the mosfet.

You can figure out the worst case resistor size with ohms law. For example
16V into ground is a 16V difference, with a 500 mA drive, is 32 ohms. So
your 100 ohms is more than enough protection assuming these cases.
However, its probably overkill for this part. In either case, you said you
were migrating to a 3A part, so 1/6th the resistance should provide
adequate protection, or about 6 ohms.

==============================================================================

I ran a simulation with their TC1413 3 amp driver, and it seems that the
current is limited to 2.5 amps using 3 ohm driver resistance. The gate
resistance of the transistor seems to be only 0.8 ohms. Much depends on the
actual rise and fall times before being lowered by the capacitor as shown
on the test circuit. I used 20 nSec, but if I use 10 nSec the current peak
is 3.5 amps, again over the spec. I have a hard time believing a 10 nSec
transient 20% over specified peak would destroy the device. Also, I am
fairly sure I first put a 100 ohm limiter on the output and it still
failed.

I did not see the negative voltage pulse you describe. I am modeling the
gate very simply with a 9 nF inductance, and the 3800 pF capacitance. I
have a model for the HUF75645, which I will try. I think I will be safe
using the 3 A drivers and a smaller MOSFET with 350 pF gate, but I am leery
of the Microchip parts now, and I may try the TI TPS2817, which has a much
beefier output stage.

If you do see a negative spike that may cause the problem, I'd appreciate a
copy of your LTSpice ASC file.

Thanks,

Paul

 

[Terry Given]

[...]

The test circuit in the data sheet is a 100 kHz square wave. The rise
and
fall times of the IC are 25-40 nSec. While the circuit was working, the
driver stayed cold, and my simulation showed less than 100 mW. The
three-transistor driver has rise and fall times of about 200 nSec. The
Microchip App Note shows a 250 kHz example with 16 mW driver power.

I think there is something else wrong. I have not heard back yet from
Microchip. I have also requested samples of some TI parts that have an
output stage consisting of MOSFETs in parallel with bipolars, which
provides better drive through the transition area where the MOSFET
actually
switches.

Thanks for your comments.

Paul

Yes. I missed the 100kHz test diagram. Just spiced it myself though
and am seeing roughly 100ma rms gate drive but more interestingly (at
switch off) there's a 10V peak, delayed negative voltage pulse feeding
back to the driver (about 200nS wide). Seems the driver's reverse
limit is 5V.
john

My reply from Microchip:

Problem Resolution:
I suspect the problem is that you are driving too much current. Since we
don't know what the impedance of the mosfet is, we don't offer parts with
built in limiting, but you can do this as you obsereved by adding a
resistor.

You will need a driver with more current output. More current output
translates to more speed. Also, reducing the capacitance of the gate will
also help to decrease time, since there is less capacitor to charge.

To know what currents are being driven, put a scope probe on either side of
the series resistor into the mosfet gate and put the scope in differential
mode. You will then see the spike of current driven by the driver. If it
exceeds the maximum spec, you will need a larger resistor to properly
protect the mosfet.

You can figure out the worst case resistor size with ohms law. For example
16V into ground is a 16V difference, with a 500 mA drive, is 32 ohms. So
your 100 ohms is more than enough protection assuming these cases.
However, its probably overkill for this part. In either case, you said you
were migrating to a 3A part, so 1/6th the resistance should provide
adequate protection, or about 6 ohms.

==============================================================================

I ran a simulation with their TC1413 3 amp driver, and it seems that the
current is limited to 2.5 amps using 3 ohm driver resistance. The gate
resistance of the transistor seems to be only 0.8 ohms. Much depends on the
actual rise and fall times before being lowered by the capacitor as shown
on the test circuit. I used 20 nSec, but if I use 10 nSec the current peak
is 3.5 amps, again over the spec. I have a hard time believing a 10 nSec
transient 20% over specified peak would destroy the device. Also, I am
fairly sure I first put a 100 ohm limiter on the output and it still
failed.

I did not see the negative voltage pulse you describe. I am modeling the
gate very simply with a 9 nF inductance, and the 3800 pF capacitance. I
have a model for the HUF75645, which I will try. I think I will be safe
using the 3 A drivers and a smaller MOSFET with 350 pF gate, but I am leery
of the Microchip parts now, and I may try the TI TPS2817, which has a much
beefier output stage.

If you do see a negative spike that may cause the problem, I'd appreciate a
copy of your LTSpice ASC file.

Thanks,

Paul

John has most likely identified the culprit. Bugger spice, measure the
actual circuit. Spice wont tell you your circuit parasitics. and
ultimately you dont care what spice says, you care what your circuit does.

many FET drivers used to get real pissy of their output was pulled below
(or above) the supply. this was usually sneakily "mentioned" in app
notes by schottky diode clamps on the outputs, often sans explanation.
Some better FET driver chips specifically mention reverse poer flow
protection, and that schottky clamps arent needed. Just to make life
fun, tacking them on might not help, if there is too much L in series
with the clamps.

also the 3n8 capacitance probably doesnt include Cmiller. but 100mW is
unlikely to make it a thermal problem (even if it were out by a factor
of 3), your finger test confirms this, so that makes it a transient problem.

I cant see ABSE, but if you want to email me a pic of your layout, I can
have a quick squiz for you (I can read protel and pads directly,
otherwise a PDF, JPG etc).

Cheers
Terry

 

[Paul E. Schoen]

Terry:

Thanks for your reply. I have sent an email to YourName at ieee dot org.
However, it bounced. You may email me directly. Mine is real.

Paul

 

[Terry Given]

Terry:

Thanks for your reply. I have sent an email to YourName at ieee dot org.
However, it bounced. You may email me directly. Mine is real.

Paul

Paul, I sent you an email. you missed the underscore in mine, which
conveniently hides beneath the underline

my_name @ ieee.org

Cheers
Terry

 

[[email protected]]

[...]

I did not see the negative voltage pulse you describe. I am modeling the
gate very simply with a 9 nF inductance, and the 3800 pF capacitance. I
have a model for the HUF75645, which I will try. I think I will be safe
using the 3 A drivers and a smaller MOSFET with 350 pF gate, but I am leery
of the Microchip parts now, and I may try the TI TPS2817, which has a much
beefier output stage.

If you do see a negative spike that may cause the problem, I'd appreciatea
copy of your LTSpice ASC file.

Thanks,

Paul

Rough and dirty, I cobbled basic bits together based on extrapolated
driver details and picked a FET with a big Cgs Cgd. Unknown item was
load capacitance, it's size is important for the transient.

LT circuit is ...

Version 4
SHEET 1 1136 680
WIRE 816 -80 720 -80
WIRE 720 -64 720 -80
WIRE 816 -48 816 -80
WIRE 720 32 720 16
WIRE 816 96 816 32
WIRE 864 96 816 96
WIRE 960 96 928 96
WIRE 1024 96 960 96
WIRE 816 112 816 96
WIRE 960 176 960 96
WIRE 1024 176 1024 96
WIRE 528 192 512 192
WIRE 768 192 608 192
WIRE 448 208 352 208
WIRE 352 224 352 208
WIRE 816 304 816 208
WIRE 960 304 960 240
WIRE 1024 304 1024 256
FLAG 352 304 0
FLAG 816 304 0
FLAG 720 32 0
FLAG 1024 304 0
FLAG 960 304 0
SYMBOL voltage 352 208 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
WINDOW 3 22 94 Left 0
SYMATTR InstName V1
SYMATTR Value PULSE(0 15 0 10n 10n 5u 10u)
SYMBOL Digital\\buf 448 144 R0
WINDOW 3 -71 -8 Left 0
WINDOW 123 -71 -29 Left 0
SYMATTR Value vhigh=15 vlow=0 trise=200n tfall=100n
SYMATTR InstName A1
SYMATTR Value2 td=100n
SYMBOL nmos 768 112 R0
SYMATTR InstName M1
SYMATTR Value FDR4420A
SYMBOL ind 800 -64 R0
SYMATTR InstName L1
SYMATTR Value 10µ
SYMBOL voltage 720 -80 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value 15
SYMBOL schottky 864 112 R270
WINDOW 0 32 32 VTop 0
WINDOW 3 0 32 VBottom 0
SYMATTR InstName D1
SYMATTR Value MBR745
SYMATTR Description Diode
SYMATTR Type diode
SYMBOL res 1008 160 R0
SYMATTR InstName R1
SYMATTR Value 60
SYMBOL cap 944 176 R0
SYMATTR InstName C1
SYMATTR Value 10n
SYMBOL res 624 176 R90
WINDOW 0 0 56 VBottom 0
WINDOW 3 32 56 VTop 0
SYMATTR InstName R3
SYMATTR Value 15
TEXT 632 304 Left 0 !.tran 1000u
TEXT 960 80 Left 0 ;Led stuff

 

[Paul E. Schoen]

[...]

I did not see the negative voltage pulse you describe. I am modeling the
gate very simply with a 9 nF inductance, and the 3800 pF capacitance. I
have a model for the HUF75645, which I will try. I think I will be safe
using the 3 A drivers and a smaller MOSFET with 350 pF gate, but I am
leery
of the Microchip parts now, and I may try the TI TPS2817, which has a
much
beefier output stage.

If you do see a negative spike that may cause the problem, I'd appreciate
a
copy of your LTSpice ASC file.

Thanks,

Paul

Rough and dirty, I cobbled basic bits together based on extrapolated
driver details and picked a FET with a big Cgs Cgd. Unknown item was
load capacitance, it's size is important for the transient.

LT circuit is ...
[snip]
========================================================================

I tried the simulation, and saw the transient. However, it was caused
mostly by the small value of load capacitor. I have a 0.47 uF in parallel
with 47 uF, which is needed to absorb the inductive energy without
excessive voltage. The 30 volt MOSFET was breaking down, as was the 45 volt
Schottky. I am using a 100 V transistor and diode. For simulation I used
STB120NF10 and MBR20100CT.

It is possible that the problem could have been caused by the opening of
the 47 uF capacitor in my circuit (it was installed backwards). The 0.47 uF
should have been OK, but 47 nF is not enough.

The rise and fall times of your driver are probably 5 times slower than the
TC1410. The drive current in your simulation peaks at less than 500 mA.

Thanks for your interest. I look forward to hearing from Terry, and perhaps
a second reply from Microchip.

Paul