My POWER MOSFETS end up in smoke!!! Any idea?

[borescopeit]

My POWER MOSFETS end up in smoke!!! Any idea?

Here is my setup:
I run a dry cell capable of consuming 350 amps, feed it with 14.2VDC from 220AMP Delco heavy duty alternator through #4 OFC wires that handle around 70 amp (this is the amps the cell runs at). Will upgrade my wires to #1 welding cables (will carry through up to 150 amps).

My KOH e-lyte is at 21%.
POWER MOSFET is ON Semiconductor NTB125N02R capable of driving 125A 24VDC @ 25C.

When I run it at 100% duty @ around 20KHz (it is that fast to handle up to 5 MHz) everything is just fine, being cooled by 12V 2" fan stays at 35C, but when I lower its duty cycle it immediately heats up to 102C and if I do not shut it off on time it ends up in a garbage bin.

Does anybody experienced the same sh1t?
Should I try lower switching frequency while lowering its duty cycle? I use microcontroller to drive the FET so I can preset any frequencies. Best freq. I found to be for my cell is around 3.6 & 23 Khz.

Also, I am playing with alternating two different freqs. every X seconds.

Any advice is appreciated!

[bobsbbq]

My POWER MOSFETS end up in smoke!!! Any idea?

Here is my setup:
I run a dry cell capable of consuming 350 amps, feed it with 14.2VDC from 220AMP Delco heavy duty alternator through #4 OFC wires that handle around 70 amp (this is the amps the cell runs at). Will upgrade my wires to #1 welding cables (will carry through up to 150 amps).

My KOH e-lyte is at 21%.
POWER MOSFET is ON Semiconductor NTB125N02R capable of driving 125A 24VDC @ 25C.

When I run it at 100% duty @ around 20KHz (it is that fast to handle up to 5 MHz) everything is just fine, being cooled by 12V 2" fan stays at 35C, but when I lower its duty cycle it immediately heats up to 102C and if I do not shut it off on time it ends up in a garbage bin.

Does anybody experienced the same sh1t?
Should I try lower switching frequency while lowering its duty cycle? I use microcontroller to drive the FET so I can preset any frequencies. Best freq. I found to be for my cell is around 3.6 & 23 Khz.

Also, I am playing with alternating two different freqs. every X seconds.

Any advice is appreciated!

Are you running the Mosfets at 24v as they are rated or at 12v? When running at 12 volts they will have half the current capacity so they are more like 62.5 amps. The lower the duty cycle the harder the mosfets are working.

[borescopeit]

Are you running the Mosfets at 24v as they are rated or at 12v? When running at 12 volts they will have half the current capacity so they are more like 62.5 amps. The lower the duty cycle the harder the mosfets are working.

No, I run these mosfets at 14V. Can you please prove your statement? I never never assumed that running lower voltage through the PFETs would derate amperage rating!

[bobsbbq]

Well I can't say with actual hands on experience, but as far a voltage and current goes there is nothing to prove.

Double the voltage half the current/ Half the voltage double the current. This is true is house wiring such as 110v vs 220v as well.

So without seeing the data information on the mosfets and their operating parameters I'm not sure. But I suspect if they are made for and rated at 24v input then the current is also rated at 24v input. If this is the case then you are cutting the voltage in half and effectively raising the current or in this case the current load of the mosfet is decreased.

I hope this makes sense.

[borescopeit]

Well I can't say with actual hands on experience, but as far a voltage and current goes there is nothing to prove.

Double the voltage half the current/ Half the voltage double the current. This is true is house wiring such as 110v vs 220v as well.

So without seeing the data information on the mosfets and their operating parameters I'm not sure. But I suspect if they are made for and rated at 24v input then the current is also rated at 24v input. If this is the case then you are cutting the voltage in half and effectively raising the current or in this case the current load of the mosfet is decreased.

I hope this makes sense.

All you are saying makes a lot of sense. This is exactly how Ohm's law states: I=V/R so R=V/I and V=I*R. P=V*I (I=Amps, V=volts, P= power in watts, R=resistance in ohms).

What I have discovered for myself from looking through a dozen of different power mosfet datasheets is that "Continuous Drain Current, VDS @ xxV" and "Power Dissipation, PD @ TC = xx°C" and "Static Drain-to-Source On-Resistance, RDS" are a few of most important parameters to take into account.

So, with my PFET (ON Semiconductor NTB125N02R: 125 AMPERES, 24 VOLTS, RDS(on) = 3.7M
(Typ)) I will take the following data into account:
VDS = 24V, PD @ 25°C = 113.6W, RDS = 3.7M, ID = 125A

Will run 14V @ 25°C, will not go over safely dissipated heat (Watt) to find out what load I can put on one mosfet:
ID = 113.6/14 = 8.11A (to make the FET stay cold)

If I want to run lets say 50 AMP through it, my heat amount will be at P = 50*14 = 700W (will cool with liquid hydrogen ). If I fail to cool down this much heat the FET will be kaput.

So, what I see from these calculations, probably the safest amps I can run through my FETs is around 65 AMP (100% duty cycle) with good 10000 rpm cooling fan. If I lover duty cycle the heat generation goes up x2 or even x3.
This is where my FETs start frying. In this case I will lower PWM frequency (on the fly) to ease up switching load on FETs...

[R&D4me]

Chiming in here, your title caught my eye.

The P-MOSs you're running have an 'avalance' (Drain time) that is between your ON/OF duty cycle. If you'res is 100% than you are emiting NO pulse at all.100% is 100% ON ALL the time no?

Also, the drain side of the FETs (Going to your electrolyzer) may be receiving some stored inductance (parasitic inductance possibly from the flyback diodes... ) that needs to be 'releived off' so to speak during each pulse cycle to lower their heat.

The Avalance is a property of the type of FET you're using; obviously without looking at the datasheet for your type, they have to be high-speed switching type. (No brainer excuse me if I don't check)

I was told by a smart guy...."an rc snubber to catch anything left over.
the resistance should be about twice the resistance of the load(YIKES!) you are switching, and the capacitor size determined by watching the switching waveform with an oscope."

I am currently working out a solution due to an increase in repairing them for friend with same problem. Here is my design, a modification of a practical kit I bought. I'd REALLY like to know what your solution was.....



This one should be good for a 60 continuous amp draw.(Fingers crossed) On the OScope 70~90% duty looks great. Will post the screenshot if anyone wants - it's avalance looks super clean but we'll see after hours/days of use the truth will always appear!

FYI- Just got this off Wiki: RE- {Flyback diodes}"The voltage spike across the switch (not the voltage spike across inductor) is usually the biggest problem in real circuits. So if you decide to draw only one voltage graph, this is the one I would prefer. Many transistor switches come with a built-in diode. (Yours does) Too often people incorrectly believe that diode can be used as the flyback diode in this kind of circuit. The "voltage across the switch" graph is the only one that clearly shows why they need *another* diode.
The graph of the voltage across the switch, on the other hand, is always positive."

[Bazarommcmullen]

Have you considered increasing the surface area of the heat sink to help disipate heat. The fan is moving aire but if the heat sink is not large enough to transfer the heat away from the mosfet then the fan is then inefective. You need the heat sink to disipate heat from the metalic surface and then the fan to move air over the heat sink to disipate excess heat. I have no clue what your design may be just trying to help point out something you may have overlooked.

[borescopeit]

Have you considered increasing the surface area of the heat sink to help disipate heat. The fan is moving aire but if the heat sink is not large enough to transfer the heat away from the mosfet then the fan is then inefective. You need the heat sink to disipate heat from the metalic surface and then the fan to move air over the heat sink to disipate excess heat. I have no clue what your design may be just trying to help point out something you may have overlooked.

I am gonna use 6" x 10" x 2" aluminum heat sink to nest 8 IRF1405 mosfets there.
Each mosfet is limited by its package's ability to dissipate so many watts. In my case the package is TO-220 (as all of us do) and I finally found right mosfet with excellent switching characteristics and max possible safe dissipation ability of 300 Watt.
So, will have to make up 8x array of 169A PFETs taking into account the 300W dissipation.

300W / 14V = 22A per PFET
22A x 8 = 176A of safe amperage draw w/o frying the nuts.

[DUG-KOH]

When you select a FET you should look at "Rds on". This is the resistance of the FET when it is fully on (at 25 deg C)
The first one you were using (NTB125N02R) is 0.0037 ohms. Respectable
The IRF1405 is 0.0049 ohms. higher but not unreasonably bad
What does this mean?

Here is a chart for power dissipated for the two different FET's at the listed current (100% on)

current AMPS NTB125N02R IRF1405
10 0.37 0.49
20 1.48 1.96
30 3.33 4.41
40 5.92 7.84
50 9.25 12.25
60 13.32 17.64
70 18.13 24.01

Keep in mind that FET's have a positive temperature coefficient. This means that when they get hot the "Rds on" goes up...(it's in the data sheet) That is why we try to keep them cool because when "Rds on" goes up, The power dissipation goes up and they get hotter and "Rds on" goes up...etc.

Paralleling FET's is a way to reduce the power dissipation per FET. It also reduces the current that each FET needs to pass. Usually a good idea.

(I've used 3 (three) SOIC-8 FET's to switch 28A using less than 3 square inches of circuit board area as a heat sink.)

One other factor to look for in a FET is gate drive requirements.

If you are using a PIC or Micro controller of some type to directly drive the FET, then the FET you select should be a "logic level" gate drive type.

The first one you were using (NTB125N02R) is a logic level FET. This means that it can be driven by a micro (5v). (Good choice)

The other one (IRF1405) is not a logic level FET. A 5V drive will not fully turn on that FET and will result in a higher Rds and therefore higher power dissipation. Not a good choice for direct drive...more on the solution later in this post

Another factor for gate drive is how much current is available to drive the FET. A FET will turn on as fast as the gate can be set to the required voltage.

One of the characteristics of FET's is capacitance. This is not too important when used as a switch to drive something like a relay. It is important when the FET is used as a PWM switch...less than 100% on time at higher frequencies.

The time that is used to charge the FET gate is reduced when more current is used to charge the gate capacitance.

Here is an approximate chart for one NTB125N02R

drive -amps switching time-mS
0.001 0.048
0.025 0.00192
0.1 0.00048
0.5 0.000096
1 0.000048

This chart does not take into considerations all of the factors for higher speed switching but shows the trend and is certainly valid for the lower drive levels.

What this means. Some logic chip can drive 0.025A. Assuming that the PIC or Micro-controller can drive the gate capacitance, using a single NTB125N02R at 20KHz would result at a single cycle of 0.05mS (50uS). Since there are two switching transitions for each cycle then 3.84uS is used for switching. This means that the switching time takes 7.2% of the total time. not too good...could be better.

The PIC or Micro-controller will have limits on its' drive capabilities. Look for high and low current output. The NTB125N02R Gate Charge graph (in the data sheet) shows that at 5V drive the charge is (aprox) 25,000nC. This translates to the equivalent of 5,000pF. Check the PIC or Micro-controller output and see if it likes to drive this much capacitance...some will limit the frequency rating of its' outputs. When Paralleling FET's this load is from each FET and would add up. (8 FET's:40,000pF)

If you cannot drive the gate with enough current it will not turn on fast enough. The time the FET takes to turn on is when it has a voltage drop across it (some where between 12v and 0v) AND is carrying current (between 0A and max A)

When this takes place many time per second then the power dissipated is very much larger than the fully on state. This is why we try to turn on the FET's as fast as possible.

The drive requirement issue is very likely why your first circuit with the NTB125N02R had problems when not at 100% on.

One good solution is to use a FET DRIVER. These take a logic level signal and "amplify" the signal with an amplifier that can drive a large amount of current into the FET gate. This circuit can be designed with discrete components or a single package IC may be used. Typically a 12V gate drive level and therefore logic level requirements for the FET are not required.

In many schematics you will also see a small resistor in the gate drive path (10-50 ohms). This is used for two reasons. The small wires that connect to the actual gate pad is very small and can only handle so much current. (in some of the data sheets). The other reason is to prevent high frequency oscillations during the switching period when the FET is in a linear mode. This can stress the FET and may produce EMI. (Electro-Magnetic Interference)

I design with at least 10R in the gate path.

I hope this helps everyone understand FET's a little better.