Simple mosfet question

A

Andrew

Jan 1, 1970
0
I remembered a different way to host pictures, so hopefully this will
help!

Here is a link to a picture of the portion of the circuit in trouble
as it is while I've had the trouble with the transistors blowing. As
mentioned, the INPUT goes from 0-4.5V, and the OUTPUT is only tied to
the CLK inputs on two CMOS ICs.

http://picasaweb.google.com/xxragexx/Circuits/photo#5095923033846850770

Now, here is the circuit that did NOT have the trouble and went
through tons of testing. It has C2 where the circuit now does not:

http://picasaweb.google.com/xxragexx/Circuits/photo#5095925344539256034

This circuit shows the potential changes I was considering (adding R2
and keeping C2 from the circuit that worked:

http://picasaweb.google.com/xxragexx/Circuits/photo#5095925348834223346


Hope that works and helps to visualize! Thanks for your help!
 
A

Andrew

Jan 1, 1970
0
You should not bring an unprotected gate lead out. A series R (eg.
10K) from the input and diodes to the supply rails (at the gate) or a
zener at the gate should work for you.

Thanks! As you can tell, I'm a novice, especially in the area of
transistors, sorry for a couple more questions as to the how and why?

I had thought about adding a series resistor at the gate, but I didn't
realize it was necessary with a MOSFET?

I'm not sure how you mean to hook up the diodes or zener? The regular
diodes would be hooked up cathode at the drain (positive rail) and
anode at the gate? And would the zener be hooked up cathode at the
gate and anode at the source (ground)? I'm picturing it that way with
a ~4V zener? And I would only need one or the other? Do either of
the diodes need special ratings to achieve their function?
 
A

Andrew

Jan 1, 1970
0
And would the zener be hooked up cathode at the
gate and anode at the source (ground)? I'm picturing it that way with
a ~4V zener?

Argh! Right after I sent this, I realized the zener would be better
rated > 4.5 and < max Vgs (~ 20V), so a ~12V zener would be good,
right? Not a ~4V which would eliminate the input from ever reaching
the gate!
 
D

default

Jan 1, 1970
0
My transistor is a fairchild (if it matters) and from what I can tell,
with 4.5V Vgs, I don't have to worry about on resistance until ~
400mA. The load resistor for the circuit that is blowing the
transistor is 10k when the transistor is on. With a max voltage of
around 14V, that is 1.4mA Id, way way way under the worry zone as far
as Rds on, right?
Yeah, that is way lower than the worry zone. I withdraw my comment.
Can anyone comment one way or the other on whether a 10nF cap from
gate to source should have any real effect on protecting the
transistor? I keep thinking yes, but it seems you guys don't agree?
Wouldn't it limit potentially damaging spikes to the gate?

Yes it would limit spikes to some extent, providing it is close to the
gate physically, and a low inductance type - ceramic disc or stacked
foil/glass or monolithic ceramic type.

With a single cap on the gate you'd be using the inductance of the
wire up to the gate as an LC low pass filter and that might eliminate
any fast transients present by dropping them across the inductance.

It takes very little energy to punch through a mosfet gate insulator -
A zener or mov will protect against higher energy transients than a
small cap on the gate.

A fast oscilloscope with a high impedance probe will tell you a lot
about what is going on there.
 
D

default

Jan 1, 1970
0
Argh! Right after I sent this, I realized the zener would be better
rated > 4.5 and < max Vgs (~ 20V), so a ~12V zener would be good,
right? Not a ~4V which would eliminate the input from ever reaching
the gate!

Roger that. 4 volts is too low 10-12 ideal.
 
S

Spehro Pefhany

Jan 1, 1970
0
My transistor is a fairchild (if it matters) and from what I can tell,
with 4.5V Vgs, I don't have to worry about on resistance until ~
400mA. The load resistor for the circuit that is blowing the
transistor is 10k when the transistor is on. With a max voltage of
around 14V, that is 1.4mA Id, way way way under the worry zone as far
as Rds on, right?

Can anyone comment one way or the other on whether a 10nF cap from
gate to source should have any real effect on protecting the
transistor? I keep thinking yes, but it seems you guys don't agree?
Yes.

Wouldn't it limit potentially damaging spikes to the gate?

In a likely real situation it will reduce them, the reduction will
depend on the nature of the source. It may not be sufficient. Human
body ESD model is around 100pF so with a 20V gate rating (?) you could
have 2000V ESD without exceeding the 20V. But a bit (<100uA) of 60Hz
leakage and you could exceed the 20V.

You should not bring an unprotected gate lead out. A series R (eg.
10K) from the input and diodes to the supply rails (at the gate) or a
zener at the gate should work for you.


Best regards,
Spehro Pefhany
 
A

Andrew

Jan 1, 1970
0
Yes it would limit spikes to some extent, providing it is close to the
gate physically, and a low inductance type - ceramic disc or stacked
foil/glass or monolithic ceramic type.

Yes, it was a ceramic cap, and yes it was close to the gate. Sounds
like the types of transients that are killing this new circuit were
the type that such a cap could handle.

At any rate, I'm going to try the resistor/zener combo for the next
revision. The more I think about it, the more it makes sense, but I
hope I'm not all wrong about why and how it hooks up. A 12V zener,
cathode at the gate and anode at the source, with a 10k resistor
between input and gate is the way to hook it up, and should protect
the transistor gate very well, right? Any transients or spikes above
12V should be dumped directly to the source (ground). Schematic below
for this change:

http://picasaweb.google.com/xxragexx/Circuits/photo#5095948657621738786

Thanks again guys, I really love having the ability to talk to some
people smarter than me about this stuff. Unfortunately, even as
novice as I am, I'm about the highest technical level EE here.
 
A

Andrew

Jan 1, 1970
0
And the power rating of the zener should be very insignificant with a
10k resistor there, right? 200mW should be plenty I'm guessing? Or
is there another reason why I should look at a higher rating?
 
S

Spehro Pefhany

Jan 1, 1970
0
Argh! Right after I sent this, I realized the zener would be better
rated > 4.5 and < max Vgs (~ 20V), so a ~12V zener would be good,
right? Not a ~4V which would eliminate the input from ever reaching
the gate!

Yes, a 12V zener usually is about perfect.

--[10K]---x---------gate
|
\----\
/\
-- source
| |
GND GND

If the gate connection could be disconnected at any time, and
especially if that could cause harm, it's often a good idea to put
something like 100K from the input to ground to ensure things don't
float around.


Best regards,
Spehro Pefhany
 
D

default

Jan 1, 1970
0
And the power rating of the zener should be very insignificant with a
10k resistor there, right? 200mW should be plenty I'm guessing? Or
is there another reason why I should look at a higher rating?

Well with a 10K you'd be charging the gate capacitance slowly - you'd
want to look at the time constant and see if the 10K plus your gate
capacitance will have a negative impact on the rise time. Unless I
botched the arithmetic (double check it) - the shift in phase with 470
pf would be about 5 milliseconds so your 1-2 millisecond pulse would
probably be greatly attenuated (there and gone before the transistor
ever turned on)

Most folks would probably opt for a lot lower resistance - but it is a
trade off. You improve the resistance to a gate spike at the expense
of spending more time in the linear region (not a concern with your
load) and shift the phase of the signal and delay it (turns on later
and slower).

If the 470 cap stays in the circuit, how about a 100 ohm series
resistance plus the zener? That way there'd still be some signal
there to turn on the mosfet.

Yeah, the zener is only there to clamp the input and probably won't
see anything that will harm it - certainly not through a series
resistor. (short of a lightening strike)

Anode goes to the source, cathode goes to the gate. Resistor is in
series with the signal to the gate and zener is on the gate side of
that resistor.

Another thing about the circuit in the schematic - there's no resistor
from gate to source to drain the gate voltage. If the signal source
is doing that - no problem.

With mosfets (particularly when you are using larger ones with lots of
gate capacity) you apply a voltage to the gate, and remove the
connection and the mosfet stays on for minutes while the gate charge
slowly bleeds off through dirt on the circuit board or through the
air.

Helps to know how the thing turns off - if you just got lucky this
time and have enough conductive dirt or humidity to turn it off . . .
you'd want to know that so when the weather dries or a new board is
less conductive it would still work.

If, for instance, your signal source is a open collector pulled high
with a resistor - that's no problem because the collector will be low
most of the time and give the gate a path to discharge back into the
transistor driving it.
 
A

Andrew

Jan 1, 1970
0
Well with a 10K you'd be charging the gate capacitance slowly - you'd
want to look at the time constant and see if the 10K plus your gate
capacitance will have a negative impact on the rise time. Unless I
botched the arithmetic (double check it) - the shift in phase with 470
pf would be about 5 milliseconds so your 1-2 millisecond pulse would
probably be greatly attenuated (there and gone before the transistor
ever turned on)

Wouldn't it be ~5 microseconds instead of milliseconds?
Most folks would probably opt for a lot lower resistance - but it is a
trade off. You improve the resistance to a gate spike at the expense
of spending more time in the linear region (not a concern with your
load) and shift the phase of the signal and delay it (turns on later
and slower).

I think it was 5usec which would be no problem, but I have no reason
to use higher resistance that I can think of. I'll lower it down to
the 100 to 200 ohm region, I don't see a problem with that either way.
If the 470 cap stays in the circuit, how about a 100 ohm series
resistance plus the zener? That way there'd still be some signal
there to turn on the mosfet.

I may end up removing the 470pF cap as it was originally there to
filter a noisy signal that should no longer be noisy due to better PCB
layout/design. But I will design as if it is staying.
Yeah, the zener is only there to clamp the input and probably won't
see anything that will harm it - certainly not through a series
resistor. (short of a lightening strike)

That's what I was thinking. I found a nice 12V zener @ 200mW that I'm
going to use.
Another thing about the circuit in the schematic - there's no resistor
from gate to source to drain the gate voltage. If the signal source
is doing that - no problem.

Because I'm not sure about the design of the servo motor driver, I
will probably add a 100k resistor from gate to source. Even if it IS
an open collector pulled high, this wouldn't hurt anything, would it?
With mosfets (particularly when you are using larger ones with lots of
gate capacity) you apply a voltage to the gate, and remove the
connection and the mosfet stays on for minutes while the gate charge
slowly bleeds off through dirt on the circuit board or through the
air.

Helps to know how the thing turns off - if you just got lucky this
time and have enough conductive dirt or humidity to turn it off . . .
you'd want to know that so when the weather dries or a new board is
less conductive it would still work.

If, for instance, your signal source is a open collector pulled high
with a resistor - that's no problem because the collector will be low
most of the time and give the gate a path to discharge back into the
transistor driving it.

I hadn't considered all of this! Thanks for the info!
 
D

default

Jan 1, 1970
0
Wouldn't it be ~5 microseconds instead of milliseconds?
Yes. milliseconds didn't look right to me either. Made the same
mistake twice apparently.
I think it was 5usec which would be no problem, but I have no reason
to use higher resistance that I can think of. I'll lower it down to
the 100 to 200 ohm region, I don't see a problem with that either way.


I may end up removing the 470pF cap as it was originally there to
filter a noisy signal that should no longer be noisy due to better PCB
layout/design. But I will design as if it is staying.


That's what I was thinking. I found a nice 12V zener @ 200mW that I'm
going to use.


Because I'm not sure about the design of the servo motor driver, I
will probably add a 100k resistor from gate to source. Even if it IS
an open collector pulled high, this wouldn't hurt anything, would it?
Probably not. Last time I took a receiver apart it had a 4.7K pull up
for the servo drive.
 
Top