mosfet driver failures

[geoff07]

I have an led lighting system at home, powered by 12v batteries floated off the mains, and controlled ultimately by an 18M2 (we get a lot of power cuts). The picaxe outputs drive two 4427 2-channel mosfet drivers, which then drive the four mosfets. So far so good, nothing difficult about the circuit or the software.

But in the last four years I have now had two 4427 failures, suggesting that mosfet drivers are vulnerable to something. The datasheets don't suggest any particular design requirements. Does anyone with experience of such beasties have any words of wisdom regarding circuit design, to protect the drivers? Or have I just been unlucky?

I'm about to redevelop it for an X2 as I want to add some features that need the timing capabilities of an X2, so now is a good time to get this right.

 

[Dippy]

I've used these and similar for motor driving, but this would be no more than 20 minutes continuous.
Never had a problem.

Please post your exact schematic with component values and types. If possible post your PCB artwork so we can see placement.

You said "12v batteries" i.e. plural. Are you running 2 x12V in series?

 

[Circuit]

Your demonstrable expertise within the forum suggests that operator design is less likely than a failure attributable to the chip. Therefore I would ask which manufacturer are you using; Microchip, International Rectifier, Micrel, Maxim - and dare I ask, from where did you obtain the ICs? Are you sure they are the genuine article and not of questionable parentage as so many seem to be reported these days? As an aside, why on earth does Chester, of all places, suffer so many power cuts?

 

[premelec]

Not to be too OT but since you are redesigning - do you really need these drivers - what you describe may not need fast gate driving and a simpler drive circuit may work for you - and be more resistant to transients... also look carefully at your circuits to see if possibly opto isolation could help - in what mode do the drivers fail [e.g. on or off permanently]- perhaps a current limiting resistor, perhaps ferrite beads on gates [MOSFETs can get nasty].

 

[geoff07]

Thanks guys, it seems there is not yet any particular advice.

- The current board was a prototype on stripboard, its replacement is not yet drawn fully. You couldn't get a much simpler chip than a mosfet driver so I don't think there is any point in making a drawing of the old setup. I think it originated in the days before I used Eagle for everything so is probably in an old notebook somewhere. It was running well for several years so I don't think it is the design, unless there are specific things not in the datasheet that cause stress.

- the batteries are in parallel! I have 180Ah of 12v lead-acid, which powers not only the lights but also all the picaxe-based gubbins I have that run the house (garage door, heating weather compensator, solar water pump, etc.)

- The failed chips were both Maxim, bought from a reputable company (I forget which but probably RS or Farnell). The replacement batch is by IR, also from one of the big companies, which worked when swapped in, hopefully they will last. The recent initial failure was a cause of strange resets, but then it went hard with one channel permanently on.

- We are in a village just outside Chester, and unlike most of the village, our power comes on overhead cables. Trees seem to play a big part in the outages, which number several each year and may last a few hours. I decided to go for led lights on 12v after a particularly stressful outage on Christmas Eve caused by an exploding pole-mounted transformer.

 

[BeanieBots]

Most FET drivers require some resistance between OP and FET gate to protect them.
Do you have any resistance.
I have a similar setup to yourself but as already suggested, I don't use FET drivers for LED control. Just bare logic level FETs even for PWM control.
Do you have very long leads to the LEDs? Could be flyback voltage coming back through the FET. Fit a clamp diode anyway.

 

[geoff07]

Thanks, very interesting. The Maxim datasheet does not mention resistance, and as the purpose is to charge the mosfet gate capacitance as fast as possible I didn't add any. The IR datasheet does show a resistor but it isn't specified or explained. What value would you use? I don't want to extend the switching time too much as I don't want to have to put in big heat sinks. Only one channel has PWM control and that does have a decent heatsink.

I'm using the drivers because I'm switching maybe 2A on a channel and can't afford to lose much voltage so want to push them hard on (also same point re heatsinks).

There are long leads -probably 8m of 2.5mm twin+earth is the longest. But flyback generally implies inductance, and I didn't expect that. But I will add some diodes to the new board.

 

[Hemi345]

When I was looking at the drivers for a 24V motor project, I normally saw 20 to 30 ohms on the FET gate.

This article has some prettt good info on drivers:
http://www.electronic-products-design.com/geek-area/electronics/mosfets/using-mosfets-as-general-switches

 

[premelec]

That's a good article Hemi... Geoff, as an initial approximation consider 1/2 the rise time to be R x C [the 'time constant']. So if you want 2 microsecond rise time and gate capacitance is 2nf R= .5T/C = (10^-6)/(2 x 10^-9) = 500 ohms minimum [if I got that right!].

 

[BeanieBots]

Good article Hemi, covers what many of us keep saying.
Although we keep banging on about driving the gate hard (because most don't) as the article explains, it can be driven too hard.
I always go for a resistance between 10R and 100R to avoid VERY rapid on/off times mainly to reduce RFI.
Premelec's calcs look fine but if you end up with 500R, it might be worth asking if a driver is really required. (perhaps to get the voltage rather than the switch time).
2A is not a very large current and I would not use a driver for 2A of LEDs but I would use a logic level FET.
I have some 3A applications using PWM @ 18kHz with no driver and no heatsinks.

I'm quite confident that if you fit 10 - 100 R between driver and FET, your problems will go away (and your wireless / bluetooth bandwidth will probably also increase!).
I would not expect you to able to see any increase in FET temperature with the R in place even with PWM @ 50% @ 2A.

8m of cable will present a fair amount of inductance (relative to very rapid on/off times) but there will also be some capacitance.
The addition of some R will slow down the on/off times to level that will render their influence to a minimal amount.

 

[geoff07]

Lots of good info, as always here, thanks people.

 

[Dippy]

Ooops this post went wrong,

 

[Dippy]

That's why I was hoping for a circuit schematic.

Then others could look at the specifications and determine peak currents and average power etc.
Sometimes you really have to dig out the calculator before applying the soldering iron
Motor circuits can be even trickier with those nasty reactive loads kicking around.

Anyway, I hope the others have given you food for thought.

 

[Goeytex]

Specific recommendations require knowing what MOSFETs are being driven, pwm frequency, ,etc. The more info available the more specific and relevant the recommendations can be. In a VERY GENERAL sense, from what I can gather from the limited information provided, I agree with others that if a driver is used (4427) there should be a series resistor between the the Driver output and the MOSFET gate. Not knowing the MOSFETs being used and assuming that the driver supply is about 12V, I would recommend a series resistor of 15 to 30 Ohms. This should limit the peak current to about 600 ma, which should be more than enough to efficiently drive a well selected modern Power MOSFET with a reasonable amount of gate capacitance.

If not already implemented, ceramic bypass capacitors should be placed close to the driver supply pin. Maxim suggests 4.7uf in parallel with a 100nf s does MIcrochip. These should be Low ESR types. An 25V X7R MLCC for the 4.7uf cap should work fine.

Maxim also recommends a ground plane, but in absence of that, separate ground returns for 4427 inputs and outputs are recommended. If I were designing a similar circuit, I would place the driver as close to the MOSFETs as possible to keep the traces short to minimize stray capacitance. Long leads from the driver to the MOSFET can be problematic.

I would also consider some kind of transient suppression circuitry to suppress transients at the battery if it is connected to a mains based charger. Having designed and built a few system of this type, I can verify that batteries can produce some nasty spikes when voltage is rapidly removed, as in the case of a power failure. A battery can act as both a capacitor and an inductor, and along with the leads can be a complex device. A bidirectional Transil/Transorb diode can be very effective protecting the driver from these transisents.

While I am a fan of MOSFET Drivers (when properly implemented), another alternative as previously suggested is to eliminate the driver and use Logic Level MOSFETS. Modern Logical Level FETs can be highly efficient if correctly selected. Look for a gate threshold voltage of 1.5V and as low RDSON and gate charge as possible. This may mean the use of an SMD MOSFET, but that should not be a deterrent for since many of these come in packages that are hand solderable. If designed properly and with a current not exceeding 5 amp the FET should barely get warm.

Good Luck

 

[Circuit]

This has been a most useful thread to observe. A question arises in consequence; if it is prudent to limit the current to the MOSFET gate from the OP of the driver, is there a circumstance where it would also be sensible to limit the output of a microcontroller when driving the gate of a logic-level MOSFET directly? I don't know what sort of peak currents are involved in charging the gate of a logic-level MOSFET but is there ever an issue?

 

[geoff07]

In view of all the remarks I shall have another go with logic level mosfets. I have been put off in the past by SMD and by datasheets that don't seem to bear out the logic level claim. However, I have found some STP36NF06L in TO-220 format at RS so they are now on order and I will ditch the drivers. I had not seen driving the gate of a mosfet as a particularly challenging task, expecting any driver precautions to be mentioned on the data sheet of the simple chip designed for the job.

Clamp diodes seem like a very good idea on the power, there can be a lot going on on that supply, I will investigate the possibilities.

Thanks again for the inputs.

 

[BeanieBots]

... is there a circumstance where it would also be sensible to limit the output of a microcontroller when driving the gate of a logic-level MOSFET directly?

YES, on absolutely EVERY occaision!
NEVER drive a FET gate directly from a micro output.

Although a FET gate is regarded as a very high RESISTANCE, it has capacitance (and a few dynamic impedance changes such as Miller effect) which means when it is switched it can require several AMPs of drive current. A micro cannot provide anything like enough current to drive a FET gate fast.
Saying that, many applications do not require fast, so 'direct' (as in no driver) can be OK.
However, nasties can come back from the gate and destroy your micro.
So always fit a resistor.

 

[Circuit]

YES, on absolutely EVERY occaision!
NEVER drive a FET gate directly from a micro output.

Although a FET gate is regarded as a very high RESISTANCE, it has capacitance (and a few dynamic impedance changes such as Miller effect) which means when it is switched it can require several AMPs of drive current. A micro cannot provide anything like enough current to drive a FET gate fast.
Saying that, many applications do not require fast, so 'direct' (as in no driver) can be OK.
However, nasties can come back from the gate and destroy your micro.
So always fit a resistor.

Wow! Definitive statement, thanks. I am just realising that must the be case - but it also means that the schematic on page 8 of PICAXE MANUAL 3 is erroneous; it shows the gate being driven directly from the PICAXE output. Some time ago I had, I now realise foolishly, tried to fire up a logic level mosfet directly from an 74HCT595 shift register; now I know why I fried the chip! Now my next question; what resistor value should be used? I guess that for a 5V PICAXE this should be the "standard" 330R giving a maximum current of 15mA. (180R for 3.3v). Is this a reasonable "standard" value to use for driving the gates of all logic-level MOSFETs, or am I again missing a trick here? Going through my notes, I have found a reference circuit where a MOSFET is driven through 100R, but this would, surely, overload the chip output at 50mA?

Edit; looking at the 74HCT595 datasheet, it claims "high current 3-state outputs" etc. and that is why I mistakenly equated it to a microcontroller output. In fact the datasheet states +/- 6mA output driver at 5 volts. Therefore I should be using 820R to connect it up to a MOSFET Gate; am I right at last?

 

[Goeytex]

The peak current drawn by the FET gate only lasts a very short time. This is the time it takes to "charge" the gate capacitance. It can be from a few nanoseconds to a few microseconds depending upon the current available from the driving device and the gate capacitance of the FET being driven . With a Picaxe operating at 5V and the current limited to 25ma the resistor would theoretically be 200 ohms. I have had no problems using a 180R gate resistor.

Since a PIC cannot really source or sink 25ma (in most cases) many folks use lower value resistors than the "standard" with no ill effects. Note that this is peak current and not continuous current.

 

[BeanieBots]

There is no "correct" resistor value because it depends on many factors.
One factor is to protect whatever is driving the gate. The value will depend on gate capacitance which is a function of supply voltage and switching current.
(and also load impedance angle. That is it's dynamic relation to inductance and capacitance)
Another important factor which many dont' care about (until it causes THEM a problem) is radiated noise. A higher value will reduce HF noise.

All that said, you can't go far wrong with values in the range you suggest.

As for the Rev-Ed boards, not only do they not have resistors but they also use non logic level FETs. (though I think this might have changed recently).
The use of IRF rather than IRL FETs is actually a useful thing on those boards. It limits the abilitly of the FET to conduct beyond about 3A which prevents the tracks from frying under fault conditions. (target audience = schools).
Also, I doubt many applications of that board use PWM drive which is when all these issues manifest themselves to the greater extent.

Consider it like wearing a seatbelt.
It has very little effect on the journey or driving experience.
It is not required when manouvering in a car park.
It would none-the-less be good advice to ALWAYS wear one when in a car.

There will always be somebody that will boast about how they drove all the way around the car park without a seatbelt and had no problems, therefore seatbelts are not needed and should not be fitted to cars.

Some things such as decoupling caps and FET gate resistors should just be second nature.

 

[Circuit]

Beanie, Goey, many thanks. Your clarifications have been immensely helpful. I am about to start some new circuit construction using MOSFETS and I now know, better anyway, what I am doing.