mosfet failure

[andy]

I'm using a BUZ10 n-channel mosfet to switch a 1 ohm electromagnet coil on
and off. It's been working OK for a while, and is now stuck open with a
resistance of 11 ohm between drain and source, and around 50 ohm between
gate and drain or source. Any ideas on what might have caused this? It's
unlikely to be electrostatic damage, because there is a 10k resistor
permanently soldered between gate and source, which was there before is
started malfunctioning.

It has a heatsink on it - an aluminium screw-on one about 4cm square with
4 fins on it. Can't find the C/W rating for this.

 

[andy]

I'm using a BUZ10 n-channel mosfet to switch a 1 ohm electromagnet coil on
and off. It's been working OK for a while, and is now stuck open with a
resistance of 11 ohm between drain and source, and around 50 ohm between
gate and drain or source. Any ideas on what might have caused this? It's
unlikely to be electrostatic damage, because there is a 10k resistor
permanently soldered between gate and source, which was there before is
started malfunctioning.

It has a heatsink on it - an aluminium screw-on one about 4cm square with
4 fins on it. Can't find the C/W rating for this.

forgot to say - the voltage is 12V DC from a lead acid battery. Drops to
about 10-11V when the current is on. I've rebuilt it with a new BUZ10, and
it's working again, but it would be good to have some idea why it faled.

also, I'm wondering if because of this voltage drop, I would be better
running all the electronics in the circuit from a 6V linear regulator with
a decent sized capacitor on the output, and just having the 2N3906 output
driver transistor, the mosfet and the coil running from 12V. This would
also mean I could use 74HC cmos instead of 4000 series.

The existing circuit is like this:

http://skullcap23.tripod.com/daily-water3-bw.png
http://skullcap23.tripod.com/daily-water3-bw.jpg

Is there any reason why this shouldn't work?

 

[andy]

put a fly back diode across the solenoid coil.
the cathode side on the + and anode on the -
side of the coil.

done that already.

 

[John Fields]

---
Sounds like you've punctured the gate oxide. You're only allowed 20V
between the gate and the source or between the gate and the drain, and
who knows how much inductance you have in the wiring, so when you
switch the the thing off, you could easily be stressing the gate oxide
to failure. The cure for this is to connect one 15V Zener from the
gate to the drain, with the cathode connected to the drain, and
another 15V Zener from the gate to the source, with the cathode
connected to the gate. Use 1 watt Zeners and connect them as close as
you can to the FET. Also, it wouldn't hurt to put a nice big
electrolytic across the supply with the + terminal connected to the
cathode of the 1N4001 and the - terminal connected to the FET source.
But, not knowing how you've got this thing wired up, that might be
easier said than done. Why don't you take a picture of what you've
got and post it to your website so we can see what we're dealing with?
---

---
Probably won't matter much once the thing goes to work, since a total
of two seconds ON out of every 86400 won't heat it up much. Also,
fully enhanced, the worst case channel resistance is 70 milliohms, so
even with 12A flowing through it, that's only 10 watts, and that's
only for one second.
---

forgot to say - the voltage is 12V DC from a lead acid battery. Drops to
about 10-11V when the current is on. I've rebuilt it with a new BUZ10, and
it's working again, but it would be good to have some idea why it faled.

also, I'm wondering if because of this voltage drop, I would be better
running all the electronics in the circuit from a 6V linear regulator with
a decent sized capacitor on the output, and just having the 2N3906 output
driver transistor, the mosfet and the coil running from 12V. This would
also mean I could use 74HC cmos instead of 4000 series.

---
That's not a _bad_ idea, but the problem with it is that with the the
2N3906's emitter at 12V and its base incapable of being driven higher
than 6V, it'll always be ON, so you'll need an NPN to drive it. That
means more parts, a higher quiescent current load for the PV array,
and since the regulator will have its own quiescent current
requirements, an even heavier load on the PV. Life's a bitch, ain't
it?^)

 

[Jamie]

put a fly back diode across the solenoid coil.
the cathode side on the + and anode on the -
side of the coil.

 

[John Fields]

Sounds like you've punctured the gate oxide. You're only allowed 20V
between the gate and the source or between the gate and the drain,

 

[andy]

---
Sounds like you've punctured the gate oxide. You're only allowed 20V
between the gate and the source or between the gate and the drain, and
who knows how much inductance you have in the wiring, so when you
switch the the thing off, you could easily be stressing the gate oxide
to failure. The cure for this is to connect one 15V Zener from the
gate to the drain, with the cathode connected to the drain, and
another 15V Zener from the gate to the source, with the cathode
connected to the gate. Use 1 watt Zeners and connect them as close as
you can to the FET. Also, it wouldn't hurt to put a nice big
electrolytic across the supply with the + terminal connected to the
cathode of the 1N4001 and the - terminal connected to the FET source.
But, not knowing how you've got this thing wired up, that might be
easier said than done. Why don't you take a picture of what you've
got and post it to your website so we can see what we're dealing with?
---

http://skullcap23.tripod.com/p1010003-1024.jpg
http://skullcap23.tripod.com/p1010004-1024.jpg
http://skullcap23.tripod.com/p1010005-1024.jpg

These show the battery, terminal block, protection diode, and the piece of
veroboard I've built the mosfet onto. The thin wires from the terminal
block will go to the board with the control electronics on.

---
That's not a _bad_ idea, but the problem with it is that with the the
2N3906's emitter at 12V and its base incapable of being driven higher
than 6V, it'll always be ON, so you'll need an NPN to drive it. That
means more parts, a higher quiescent current load for the PV array,
and since the regulator will have its own quiescent current
requirements, an even heavier load on the PV.

I was thinking of making a low-quiescent-current regulator using a couple
of transistors and a zener diode.

Life's a bitch, ain't
it?^)

sometimes, yes.

 

[John Fields]

http://skullcap23.tripod.com/p1010003-1024.jpg
http://skullcap23.tripod.com/p1010004-1024.jpg
http://skullcap23.tripod.com/p1010005-1024.jpg

These show the battery, terminal block, protection diode, and the piece of
veroboard I've built the mosfet onto. The thin wires from the terminal
block will go to the board with the control electronics on.

---
YEOWWWW!!! That's just plain scary!

For starters,

1. The MOSFET should be mounted in the channel between the fins on the
heat sink, NOT sandwiched between the heatsink and the Veroboard.

2. There's no need to use #10 wire. #16, with a resistance of about 4
milliohms per foot will give you a drop of 50 millivolts per foot with
12 amps flowing through it, which your solenoid will never miss if you
keep the wiring SHORT.

3. The clamp diode needs to be mounted on the board with the MOSFET,
the whole thing needs to be cut down substantially, and the solenoid
needs to be connected to the Veroboard with SHORT wires, whether you
use that terminal block or solder it directly to the board.

4. You need to learn how to solder. Tin the wires and the Veroboard
before you solder the wires down, and either use a hotter iron or (if
the one you have has adequate wattage) hold the one you've got on the
joint until the solder flows properly.

5. You should keep the battery, the solenoid, and the Veroboard as
close to each other as possible. It's not very important to keep the
control circuitry close to the rest of the circuitry, (matter of fact,
I think farther would be better than closer) but you should connect an
electrolytic cap in parallel with a ceramic cap where the battery
leads connect to the board with the control circuitry on it. Also,
I'd run the supply leads directly from the battery to the control
board instead of letting them pass through the other board in order to
eliminate ground loop problems.

 

[John Fields]

oh, I thought I had done a neat job of it apart from the messy soldering.

---
Well... ;^)
---

ok. This will make it harder to solder the solenoid cable to the mosfet
leads though (less of an end to turn over and solder the cable to)

---
You're not supposed to solder the cable to the MOSFET leads, you're
supposed to solder the MOSFET leads to the Veroboard traces and then
solder the solenoid cable to the traces the MOSFET is soldered to.
That way, when you fry a MOSFET, you don't have to unsolder
_everything_ just to replace the MOSFET.
---

I know it's a bit heavy - it's just what they had at the shop.


That I don't get - why does it matter to have short wires?

---
The longer the wire the greater the inductance, and since E = LdI/dt,
the greater the inductance the higher the amplitude of the voltage
transient generated when the MOSFET switches off will be. Those long
wires are also antennas, and the EMI pulse radiated by the wire will
be stronger for long wires than short.
---

Fairy snuff.


Again, not sure why it matters to have short wires? The solenoid will be
about 3 feet above ground level on top of a water tank, which doesn't seem
like a good place to mount the box with the electronics. I was thinking of
having all the electronics including the mosfet in a plastic box under
cover with the battery next to it, the solar panel nearby, and wires
running up from this to the solenoid.

---
A better plan, IMO, would be to mount the battery, the solenoid, and
the board with the MOSFET, the flyback diode, and protective Zener
(the one from the gate to the source) all together in a box at the top
of the tank with the solar panel nearby, and then to run a twisted,
shielded pair of wires from the battery and the MOSFET gate to the
controller, located in a sealed box somewhere under cover, with
battery+ on one conductor, the MOSFET gate on the other, and battery-
and circuit ground connected to the shield.

 

[andy]

oh, I thought I had done a neat job of it apart from the messy soldering.

For starters,

1. The MOSFET should be mounted in the channel between the fins on the
heat sink, NOT sandwiched between the heatsink and the Veroboard.

ok. This will make it harder to solder the solenoid cable to the mosfet
leads though (less of an end to turn over and solder the cable to)

2. There's no need to use #10 wire. #16, with a resistance of about 4
milliohms per foot will give you a drop of 50 millivolts per foot with
12 amps flowing through it, which your solenoid will never miss if you
keep the wiring SHORT.

I know it's a bit heavy - it's just what they had at the shop.

3. The clamp diode needs to be mounted on the board with the MOSFET,
the whole thing needs to be cut down substantially, and the solenoid
needs to be connected to the Veroboard with SHORT wires, whether you
use that terminal block or solder it directly to the board.

That I don't get - why does it matter to have short wires?

4. You need to learn how to solder. Tin the wires and the Veroboard
before you solder the wires down, and either use a hotter iron or (if
the one you have has adequate wattage) hold the one you've got on the
joint until the solder flows properly.

Fairy snuff.

5. You should keep the battery, the solenoid, and the Veroboard as
close to each other as possible. It's not very important to keep the
control circuitry close to the rest of the circuitry, (matter of fact,
I think farther would be better than closer) but you should connect an
electrolytic cap in parallel with a ceramic cap where the battery
leads connect to the board with the control circuitry on it. Also,
I'd run the supply leads directly from the battery to the control
board instead of letting them pass through the other board in order to
eliminate ground loop problems.

Again, not sure why it matters to have short wires? The solenoid will be
about 3 feet above ground level on top of a water tank, which doesn't seem
like a good place to mount the box with the electronics. I was thinking of
having all the electronics including the mosfet in a plastic box under
cover with the battery next to it, the solar panel nearby, and wires
running up from this to the solenoid.

 

[andy]

---
You're not supposed to solder the cable to the MOSFET leads, you're
supposed to solder the MOSFET leads to the Veroboard traces and then
solder the solenoid cable to the traces the MOSFET is soldered to.
That way, when you fry a MOSFET, you don't have to unsolder
_everything_ just to replace the MOSFET.
---

I was thinking that for a high current load, the veroboard traces wouldn't
be thick enough to carry it. And the cable is too thick to put through one
of the veroboard holes.

---
The longer the wire the greater the inductance, and since E = LdI/dt,
the greater the inductance the higher the amplitude of the voltage
transient generated when the MOSFET switches off will be. Those long
wires are also antennas, and the EMI pulse radiated by the wire will be
stronger for long wires than short. ---

OK. I'm surprised it would matter so much.

---
A better plan, IMO, would be to mount the battery, the solenoid, and the
board with the MOSFET, the flyback diode, and protective Zener (the one
from the gate to the source) all together in a box at the top of the
tank with the solar panel nearby, and then to run a twisted, shielded
pair of wires from the battery and the MOSFET gate to the controller,
located in a sealed box somewhere under cover, with battery+ on one
conductor, the MOSFET gate on the other, and battery- and circuit ground
connected to the shield.
---

I'll think about it - thanks.

 

[John Fields]

I was thinking that for a high current load, the veroboard traces wouldn't
be thick enough to carry it. And the cable is too thick to put through one
of the veroboard holes.

---
Sure they would [be thick enough]. The only thing which would make
the traces fail is if they melted or vaporized with 12 amps going
through them, and even if they were only 1/2 oz. copper, with a nice
coating of solder they should be fine. Remember too, there's only
going to current flowing for about a second.
---

 

[Dr Engelbert Buxbaum]

I'm using a BUZ10 n-channel mosfet to switch a 1 ohm electromagnet coil on
and off. It's been working OK for a while, and is now stuck open

Was there a diode across the coil to deal with spike voltages? If not,
those probably killed the FET. You need something like that:

o + Ub
|
|--------
| |
| ) ____
|( /\ 1N4148
coil | ) / \
|( ----
| ) |
|---------
|
to FET