Why do I keep killing these MOSFETS?

[Mechatrommer]

https://www.onsemi.com/pub/Collateral/LM324-D.PDF (I started using the TI datasheet, but it seemed to contain a technical error/mistake (as regards the output current), so used the onsemi one).

so i guess correctness is to its own context. your linked datasheet is not much different to my reference here (regarding this matter)

There are a number of limitations with the LM324, which might make choosing a better op-amp, a good idea. Such as its limited voltage output swing, compared to more modern parts (e.g. full output rail to rail). They (modern/better/rail-to-rail-output) are still relatively cheap, so why not use them ?

name the rail-rail opamp the same or close price point as LM324 or TL074/072/071, 30V++ supply range. i'll buy it everyday, as i'm clearing up my old stock.

[MK14]

There are a number of limitations with the LM324, which might make choosing a better op-amp, a good idea. Such as its limited voltage output swing, compared to more modern parts (e.g. full output rail to rail). They (modern/better/rail-to-rail-output) are still relatively cheap, so why not use them ?

His power-up negative input will be at 0v as well as potentially his + input depending on pot setting, he also needs rail-rail input support, or at least an opamp which is guaranteed to not seize up under this condition.

I was only trying to start the ball rolling (onto better op-amps), when I mentioned rail-to-rail outputs. The more additional specifications I mentioned, the more likely I would cause someone to disagree with some of it, so I avoided piling on the extra specs.

Even when I choose an op-amp for my own projects, I can get into huge dilemmas as to which one to choose. With some of the specifications (potentially), you have to choose what one, two or three requirements will be met, well. Then compromise on the other specifications.
E.g. Price vs functionality vs ease of use, etc etc.
Sometimes even the lack of availability in some packaging formats, limits ones choices.

[MK14]

https://www.onsemi.com/pub/Collateral/LM324-D.PDF (I started using the TI datasheet, but it seemed to contain a technical error/mistake (as regards the output current), so used the onsemi one).

so i guess correctness is to its own context. your linked datasheet is not much different to my reference here (regarding this matter)

There are a number of limitations with the LM324, which might make choosing a better op-amp, a good idea. Such as its limited voltage output swing, compared to more modern parts (e.g. full output rail to rail). They (modern/better/rail-to-rail-output) are still relatively cheap, so why not use them ?

name the rail-rail opamp the same or close price point as LM324 or TL074/072/071, 30V++ supply range. i'll buy it everyday, as i'm clearing up my old stock.

I tend to agree with you.
If he/she (OP) was doing this for a production run, with tight cost targets and/or does not want to spend a dollar or few, on better op-amp(s). Then I agree with you.

I also have been disappointed/surprised at how pricey the better op-amps can be. For hobbyist or one-offs (or very low volumes) it does not matter that much. But for any kind of serious production volume (and tight cost targets), it is a problem/issue.

Also breadboard friendly DIP packages, seem to be largely forgotten in the later/latest op-amps as well, forcing the use of adapter boards, if breadboarding. Which also probably adversely hits some of the beginners on this forum.

[CaptainNomihodai]

Thanks for all the help everyone. I'll implement these suggestions when I get home from work today.

[Mechatrommer]

For hobbyist or one-offs (or very low volumes) it does not matter that much. But for any kind of serious production volume (and tight cost targets), it is a problem/issue.

if it can be bought next door it doesnt matter, otherwise buying 1 piece from china or america land to here is not practical either even for hobby job. i tend to buy parts at least 10 pieces for each chip as many as i think i'll need to save shipping cost. cost can multiply dramatically for high end chips and is going to be problematic for poor hobby like me who has very limited portion of budget to invest in the hobby. edit: and we hobbiest need to buy excess parts to consider for blowing out during building and testing. blowing out $10 part is not fun.

[MK14]

For hobbyist or one-offs (or very low volumes) it does not matter that much. But for any kind of serious production volume (and tight cost targets), it is a problem/issue.

if it can be bought next door it doesnt matter, otherwise buying 1 piece from china or america land to here is not practical either even for hobby job. i tend to buy parts at least 10 pieces for each chip as many as i think i'll need to save shipping cost. cost can multiply dramatically for high end chips and is going to be problematic for poor hobby like me who has very limited portion of budget to invest in the hobby. edit: and we hobbiest need to buy excess parts to consider for blowing out during building and testing. blowing out $10 part is not fun.

I agree that the availability/affordability of components, varies between different hobbyists, how old they are and where they live (Country).  Also how much of the hobbyists wealth, they are willing to commit to their electronics hobby.

Going partly off-topic to reply to the post.
The problem with many of the older, standard op-amps, such as the 741 and many others. Is that the microcontroller (used in many projects), may be running at a rather low voltage, such as 3.3V
Some of the modern, rail to rail input and output op-amps will still work fine at 3.3V, for use around the A2D converter of the MCU.

This makes hobbyists who expect to often do microcontroller projects (not all hobbyists), running at lower supply voltages, needing to stock up on such op-amps. E.g. For more complicated A2D circuitry.

In the OP's case, without analyzing their circuit in too much detail. They need the op-amp to be able to climb above the 10 ohm resistors voltage drop, and have enough extra voltage to drive the (non-logic level) Mosfets. The LM324's may run out of output voltage headroom, as they are not rail to rail output, op-amps.

As discussed in other posts, some of the other advantages of better op-amps, can also help in circuits like the OP's.

But I also concede that if the circuit was better designed (as also already discussed). It would be able to work better, even with a LM324.

My opinion is that the LM324 is still useful in some simple/limited circuit use situations. But as the op-amp circuit becomes more and more demanding/complicated. Better/modern op-amps, become increasingly important.

I'm NOT brave enough to say on a public forum, that a LM324 with a clever enough circuit design CAN perform the OP's circuit function, for fear of being criticized too heavily.

Proof/Example: All I said so far was a rail to rail output, would be good/useful, and I'm already getting lots of flak

[David Hess]

I'm NOT brave enough to say on a public forum, that a LM324 with a clever enough circuit design CAN perform the OP's circuit function, for fear of being criticized too heavily.

I am.  The LM324/LM358 will work fine in this application although CaptainNomihodai does not give any performance requirements.  Performance will be limited by its speed and drive capability and the fact that it is not a precision operational amplifier.  At worst, some frequency compensation will be needed.

1. Its input common mode range includes its negative supply.
2. Its output stage is current limited
3. Its output cannot sink significant current at the negative supply because of a base-emitter junction but the Vgs threshold of the MOSFETs makes this unnecessary.
4. It only suffers from phase reversal if an input is pulled *below* the negative supply.  I do not see that happening here.

Before I knew better, I built a very similar circuit for testing inductors and ended up using an LM318.  *That* is asking for trouble but it worked great.

[bson]

I'd add gate pulldowns so the gate can't drift towards drain enough to exceed max Vgs and kill the MOSFET while the LM324 output is in the crossover region.  (I'd also use an opamp that's biased to be active over its entire output range.)

[David Hess]

I'd add gate pulldowns so the gate can't drift towards drain enough to exceed max Vgs and kill the MOSFET while the LM324 output is in the crossover region.  (I'd also use an opamp that's biased to be active over its entire output range.)

I agree with adding the pulldowns for safety reasons but if the LM324's output stage is in its crossover region, it will not be if the gate tries to drift anywhere.

Take a look at the LM324 schematic and you can see the problem.  Ignore the current limit transistor and there are 3 Vbe junctions in series with the bases of the upper Darlington and lower emitter follower tied directly to together at the class-A output of the previous stage.  So the output can wander about 3 Vbe total or 1.8 volts, so what?

Solve both problem if they are that with one resistor from the gate to ground.  If the MOSFET's Vgs threshold is about 5 volts, then a 1k resistor between the gate and source will sink 5 milliamps protecting the MOSFET and forcing the LM324 output into class A operation.  This also adds 5 milliamps of error on the output so use a larger resistor to ground if this is not acceptable or something more complicated.

[MrAl]

Thanks for the reply, but I'm not quite sure I follow you. I get what you're saying, that I can't assume the MOSFETS to be identical, but I'm not sure how to align them with a resistor (I'm still working through that linked doc, but some of it's a bit over my head I think). Anyway, I was trying to account for mismatch by driving them with different opamp outputs. I now realize that this didn't matter since I had the inverting inputs tied together, and have modified my design accordingly. Is this what you were talking about?
That all being said, does mismatch even matter that much here, given what the MOSFETS I'm using are supposed to be able to handle? They're rated for 50A continuous and 300W power dissipation. Like I said, the circuit is maxing out at 3.3A with a 36V power source, so that's about 110W dissipation from the MOSFETS (once you account for what's dissipated from the power resistors).

Hello,

Based on your reply in reply #2 your feedback arrangement is not that good.
One op amp depends on the other op amp which then depends on the first op amp again which then again depends on the second op amp..etc.  That is both unnecessary and harder to analyze intuitively.

To fix this is really super easy.  Just use the pot reference for one op amp ONLY, and use the output of one transistor as the reference for the other op amp, which then will be forced to output the same voltage within a tiny percent.  The circuit is shown below.  In this new circuit, the second op amp depends on the first circuit output but that's the end of the analysis, and it must output the same voltage across the second set of load resistors because it's reference is the first output and it's feedback is from the second set of load resistors where it has to regulate the voltage.  In short it is a tracking regulator.

Note this does not say anything else about the circuit, such as the incredibly high power dissipation in the two mosfets at some operating points.

[T3sl4co1l]

Simple fixes:

1. 10mA minimum is a good point to start reasoning from.  20 or 40mA typical would also be fine.  It's not like it's critical.  Certainly nothing to get pedantic about.  (I say this, and I'm one of the biggest pedants here!)

Current is also limited by dV/dt, since I = C * dV/dt.  10mA into 10nF is 1V/us, a bit slower than the LM324 is, so the resistor will be useful (i.e., it will have a worst-case voltage drop near half the supply voltage, so it's doing something, and it's not too big, not too small).

If the gate were, say, 1nF or less, it wouldn't be a big deal, because the LM324 can't move that fast in the first place.  A 100 ohm resistor would be fine: mainly to avoid possible oscillation.  (External compensation would still be recommended.)

2. The LM324 will not blow up.  It might get warm if it breaks into full oscillation.  Even in a short, it's rated for "continuous" duty.  You have to work to destroy one of these (or, as the phrase goes, "it takes a special kind of stupid"... ).

3. The transistor won't blow up.  15V is well within its rating.

That said, a G-S zener diode (preferably a large one, like P6KE15.0A or SMAJ15A) will help keep G-S voltage within limits, even under surge conditions.

That is, consider what happens when you connect a large 50 or 100V capacitor to the load terminals: the voltage rises within nanoseconds, limited only by series inductance and circuit damping.  The peak current, during that transient, may be many amperes, and a fraction of it will be seen by the LM324 because the transistor has so much capacitance.

Designing the circuit around transistor failure is also handy; in that case, you might make the series gate resistor, and source/shunt resistor(s) fusible types, and add a protection diode to the LM324 output pin, so it doesn't get damaged by transistor failure (i.e., suppose it fails D-G shorted, S open).

4. If low supply current consumption is not necessary, then a pull-down resistor will stabilize the LM324's crossover distortion under quiescent conditions.  Step changes in one direction will still be messy, but it will eventually settle out of it.

This, by the way, is another handy feature of "single supply" amps: they are well behaved on VEE-referenced resistive loads.  In this case, it keeps the high-side output transistor active, so it behaves like a simple emitter follower.  Only when the output needs to change rapidly, will the low-side transistor be turned

Tim

[MrAl]

Simple fixes:

1. 10mA minimum is a good point to start reasoning from.  20 or 40mA typical would also be fine.  It's not like it's critical.  Certainly nothing to get pedantic about.  (I say this, and I'm one of the biggest pedants here!)

Current is also limited by dV/dt, since I = C * dV/dt.  10mA into 10nF is 1V/us, a bit slower than the LM324 is, so the resistor will be useful (i.e., it will have a worst-case voltage drop near half the supply voltage, so it's doing something, and it's not too big, not too small).

If the gate were, say, 1nF or less, it wouldn't be a big deal, because the LM324 can't move that fast in the first place.  A 100 ohm resistor would be fine: mainly to avoid possible oscillation.  (External compensation would still be recommended.)

2. The LM324 will not blow up.  It might get warm if it breaks into full oscillation.  Even in a short, it's rated for "continuous" duty.  You have to work to destroy one of these (or, as the phrase goes, "it takes a special kind of stupid"... ).

3. The transistor won't blow up.  15V is well within its rating.

That said, a G-S zener diode (preferably a large one, like P6KE15.0A or SMAJ15A) will help keep G-S voltage within limits, even under surge conditions.

That is, consider what happens when you connect a large 50 or 100V capacitor to the load terminals: the voltage rises within nanoseconds, limited only by series inductance and circuit damping.  The peak current, during that transient, may be many amperes, and a fraction of it will be seen by the LM324 because the transistor has so much capacitance.

Designing the circuit around transistor failure is also handy; in that case, you might make the series gate resistor, and source/shunt resistor(s) fusible types, and add a protection diode to the LM324 output pin, so it doesn't get damaged by transistor failure (i.e., suppose it fails D-G shorted, S open).

4. If low supply current consumption is not necessary, then a pull-down resistor will stabilize the LM324's crossover distortion under quiescent conditions.  Step changes in one direction will still be messy, but it will eventually settle out of it.

This, by the way, is another handy feature of "single supply" amps: they are well behaved on VEE-referenced resistive loads.  In this case, it keeps the high-side output transistor active, so it behaves like a simple emitter follower.  Only when the output needs to change rapidly, will the low-side transistor be turned

Tim

Hi,

See the new circuit in the post just before yours.  It's a good way to do it as long as the mosfets can handle the power.

[T3sl4co1l]

See the new circuit in the post just before yours.  It's a good way to do it as long as the mosfets can handle the power.

Yes, this works, though it's rather peculiar.  A simpler shortcut is simply tying both +in's to the pot wiper: same reference, but not passed through a feedback loop.

Tim

[Audioguru]

The circuit is simply an opamp driving a Mosfet that is a source follower. There are two of these circuits, cross-coupled but they do not oscillate, instead one latches on and the other latches off.
The one that is latched on dissipates about 72.5W in the Mosfet if the supply is the shown 50V so it overheats and is destroyed.

The Mosfet can dissipate 300W when its case is held at or below 25 degrees C with liquid nitrogen or something else extremely cold.

I show the voltages when power is applied and only one Mosfet becomes turned on. There is no input signal to switch the circuit so that the other Mosfet turns on and it turns off the one that was already turned on.

[CaptainNomihodai]

Here is the circuit updated based on suggestions I've received here. The Zeners are 14V because I have a bunch of them (and that's higher than the opamp will output anyway). R1 and R2 don't have values right now since I need to go through some of my parts bins to find low value resistors that can handle a fair amount of power (I'm currently paralleling a bunch of big power resistors to get about 2.5R, but it's pretty hideous and I'd like a lower value), and then the value of R3 depends on that.
Since that 300W dissipation figure is apparently fictitious, I guess I'll have to parallel some more FETS in order to run this circuit up to 10A like I was hoping...
I haven't actually built this new one yet, I'll update once I do.

[Mechatrommer]

if you use zener, i think diode to 15V is redundant. put the zener/diode as close as possible to the gate and drain pins to avoid inductance downstream. i suggest pulldown resistor on opamp +ve input since if the trimpot got damaged and wiper disconnect from resistive element, opamp +ve input will be floating and go whereever it like to. if it goes to gnd its ok, if it goes up then its not ok. a weak >100K (>10X the trimpot impedance) pulldown resistor should be ok. if you are me, i'll connect R6 and R7 to GND, not the high side of power resistor, value should be also greater than 10X the series R4,R5. adding capacitance C1 and C2 like that will only worsen the mosfet turn on and off criteria, i dont like that unless i have a strong driver.

[T3sl4co1l]

Almost -- move R6/R7 to the right, in series between C1/ZD1, C2/ZD2.  And move C1, C2 to before R4, R5.

Also, for increased power dissipation on the cheap, connect a resistor in series with the load.  If the load is always 50V and 10A, then a 5.0 ohm resistor will draw a constant current from a constant 50V supply; all the transistors need to do is be turned on, dissipating almost no power themselves.   Any smaller amount of resistance increases the voltage compliance range (i.e., how low the voltage can go, while still being able to draw up to 10A), while still reducing the power dissipation in the transistors.

Tim

[jdraughn]

I skimmed the first few replies, but didn't see where anyone mentioned the pinouts of the mosfets. You show 2 going to gate, but when I googled the mosfets part number, it goes G,D,S, pin one being gate. Looks like your blowing them by putting 50v into the gate.

[CaptainNomihodai]

I skimmed the first few replies, but didn't see where anyone mentioned the pinouts of the mosfets. You show 2 going to gate, but when I googled the mosfets part number, it goes G,D,S, pin one being gate. Looks like your blowing them by putting 50v into the gate.

The pin numbers on the schematic are wrong (I guess the KiCAD library is wrong?). I have them wired according to the datasheet. Based on what people have said, I'm guessing that what kept killing the FETs was inductive kick on the gate due to the long wire between the gate and the pin and/or having my feedback lines crossed (which I never intended to do -- having them crossed on the schematic was just a mistake -- but they were plugged in and out of the breadboard enough times that it could very well be that they were crossed at least once).

[CaptainNomihodai]

New schematic. R9 and R10 are just to take some of the stress off of the FETS. It may be a while before I actually build these modifications since I need to find the proper resistors.

[T3sl4co1l]

Perfect!

[CaptainNomihodai]

Something occurs to me that makes me wonder if this circuit will work the way I intended, even with all these corrections....
My intent here is that the FETs operate in the linear region, acting essentially as variable resistors and dissipating a lot of power. However, the linear region is when Vgs > Vt & Vds < (Vgs -Vt), correct? The problem I'm seeing is that there is no way that Vds < (Vgs - Vt), i.e., my supply is 50V, if I have, say 1R between source and ground and another 1R between power and drain, just to lessen the burden on the FETs a bit, that still leaves, at 5A per FET, a Vds of 40V forcing it into saturation, and things only get worse if I try to run at lower currents, since the voltage drop across the resistors will be less.
Now that I've typed the above, I'm even more confused because my understanding of MOSFETs (which is obviously wrong, apparently) now dictates that a FET shouldn't be able to stay in saturation without an insanely huge current. My reasoning is that Rdson is generally close to zero, but saturation requires a relatively large Vds, so you would need a large current to create that voltage drop.
I know I'm wrong, but where?

[MK14]

One option is to put the circuit into (often free) simulators, such as LTSpice. Then you can play around with the circuit fairly quickly and try out stuff like that and see how it behaves. You can then learn more about how and why it works (or not), the way it does.
Simulators are not necessarily 100% accurate, but it should give you an approximate idea on how the circuit will work out, WITHOUT blowing up Fets and things, while you are learning/experimenting etc.

http://www.linear.com/designtools/software/#LTspice