Damage microcontroller by driving FET directly?

[schnell]

Hello,

Will we damage our Freescale MC9S08AC60CFGE microcontroller by driving a FET directly from a microcontroller output pin, with no series resistor, (and no gate driver circuitry)?




We are powering solenoids via switch-mode solenoid drivers which comprise FETs being driven directly from a microcontroller.
-as stated , the FET gates are being driven directly from the microcontroller pins, with no series resistor used -and no specific gate driver circuit used. (-the microcontroller is supplied from 5V)

(there are actually four solenoids and so four of the same microcontroller's pins directly drive four different FETs)

Here is the basic switching solenoid driver schematic:

SCHEMATIC:

http://i53.tinypic.com/1zm1j08.jpg

-solenoid current is 220mA and switching frequency is 27KHz.


Page 304 of the MC9S08AC60CFGE microcontroller datasheet states that the maximum instantaneous current allowable in a microcontroller pin is 25mA

Page 304 of MC9S08AC60CFGE microcontroller datasheet:

http://i51.tinypic.com/4rtj00.jpg

I make it that the total power_fet gate capacitance (including miller effect) is around 2.2nF for a 48V bus.........C = Qg/V (values from datasheet, below)

The actual gate (Cgs) capacitance is Ciss - Crss = 316pF.

Our Bus is 36V , so we will have a total gate capacitance (Cgs) of around 1.8nF.


This is one of our sold products, but we had to stop producing it shortly after launch because a connector was not made of flame retardant plastic.
-we intend to re-launch in 6 months with the new connector.

I am told by colleagues that this product survived and passed "quite a lot" of qualification testing.

....but how can potential microcontroller damage be prooven?


MICROCONTROLLER DATASHEET (Freescale MC9S08AC60CFGE):

http://cache.freescale.com/files/mic...f?fsrch=1&sr=5

FET DATASHEET (STB16NF06LT4):

http://www.st.com/internet/com/TECHN...CD00002847.pdf

 

[MacGyverS2000]

It's often done, but I wouldn't consider it good practice. It certainly isn't done in a safety-critical system where you could "blow" an internal pin driver open/closed.

Besides, at a 25mA limit and 5V or less, is a 1/4W SMD resistor really that big of a deal?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

I run them directly with PICs all the time. That's one of the bennies of using a FET instead of a BJT. I do, howeve, always include a pull down on the gate as leakage current can turn on the FET a "little" which means "a lot" of heating in the FET - generally enough to toast it.

PICs are advertised to drive at 25mA though, and no "instantaneous" limits are put on them. 5V/25mA = 200ohms would provide the absolute limit. Of course you will pay for that with a reduced rise time on the load. 200 x 1.8e-9 = 390ns. Not much of a rise time penalty. Of course it will heat up the FET a little more than otherwise.

More conservative would be to run the resistor. But that has cost penalties and possibly space penalties.


Keith Cress
kcress -

http://www.flaminsystems.com

 

[2dye4]

I would never drive a FET that is feeding off the board into the wild without a series resistor between gate and pin, and perhaps clamp diodes to the logic rails rails.

It takes little energy to latch up the micro when a gate coupled pulse is driven back into the micro.

 

[Warpspeed]

I would not do it also from the view to possible fault escalation.

If the power FET dies gate/drain shorted, it could possibly kill the entire controller not just the one output driver, and perhaps lead to a highly destructive sequence of events.

 

[IRstuff]

Just because you've gotten away with it in the past is no guarantee of eluding dire consequences. The Space Shuttle program had two major disasters; in both cases, a major contributory factor was the complacency engendered by not being caught when not doing due diligence and not exercising due caution.

So you save on not putting in a couple of resistors; is that worth losing a few percent in field failures? It may be worth the trade, but you need to do the math before dismissing the alternative.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[schnell]

The thing is,
My boss does not want to add a gate driver, or series resistors, because he believes its needless overdesign.

-also, adding gate resistors alone will simply slow up the switching on of the FETs, and increase switching loss.

He points out that the FET's C(gs) is only around 316pF, (= Ciss-Crss) and that this is so little (its not that much more even when you factor in the 'Miller' effect)that even at 27KHz, its not worth adding gate resistors.

He says that the overall average gate drive current is very low.


However, i am saying that the peak currents are above 25mA, ...the stated "instantaneous" maximum current of this microconroller's pins

From pg 310 of the microcontroller datasheet , it appears that the output resistance of a pin is around 100R

Graph showing microcontroller pin output resistance (IOH)
...from pg 310 of MC9S08AC60CFGE microcontroller datasheet

http://i51.tinypic.com/4r9apv.jpg

Now here is the microcontroller pin current when driving these FETs directly, with a 100R series resistance added to the simulation (which uses the STB16NF06LT4 FET .model)

FET gate current with 100R series resistance (100R represents the microcontroller's suggested output resisitance):-

http://i53.tinypic.com/o60bkn.jpg

do you consider that this current is unnaceptable?

-bearing in mind that its 4 pins of the same microcontroller that are supplying this current.

 

[benta]

As itsmoked says, you need to include a pulldown.
When your microcontroller comes out of reset, the ports are usually set as inputs, which means the FET gate floats.
You have no control of what happens during the first milliseconds before your MCU is configured.

The series resistor is not 100% required, but the pulldown is.

Benta.

 

[schnell]

hi,

The actual overall product which contains these four solenoids costs $1100.

It is a "high quality" product at the luxury end of the market......generally customers are at least reasonably wealthy

-Given this fact, is it then still the general concensus that all that is needed is a series resistor and a pull-down?

I am trying to influence the management that we need a gate driver IC with four gate drivers, as well as a ~10K pull-down resistor conected to each FET gate..........
...as well as a ~10R series gate resistors for EMC reasons
.....as well as a Schottky from source to gate (cathode to gate) for ESD protection.......(this Schottky is also for clamping negative going transients that may get coupled through the drain-gate capacitance).

in my view , the gate drivers will pay for themselves because they will reduce switching losses, and thus allow us to use smaller, cheaper FETs.....
....at the moment, we are using four large D2PAKs when the current is only around 200mA in each solenoid (and therefore, also, in each FET)

 

[Skogsgurra]

If you take one step back and look at the situation, you will see two things:

First. It doesn't matter if the FET, the driver or the MCU or all is damaged. It is totally irrelevant to the customer. His problem is that the thing doesn't work any more.

Second. If the MCU fails during burn-in, you will win and your boss lose. If nothing happens during burn-in, your boss wins.

My conclusion is that you can keep the design minimalistic. Not to save a cent or two, but because there's no need to add components that are not needed. Even if you can see there's a risk for certain failure modes.

One thing that I would look deeper into is how to protect the transistor so that it will not be damaged from external abuse. That is something that will give your product a good reputation.



Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Warpspeed]

My boss does not want to add a gate driver, or series resistors, because he believes its needless overdesign.

The motto is "the boss is always right", and then HE wears any horrific problems that arise at a later time.

I have seen this same little game played out many times before in several places of employment. Go with the flow, but hang on to any e-mails or evidence, and cover your ass.

One day you will be the big boss, and you get to play this same game from the other side.

 

[2dye4]

Heed Bentas advice for the pulldown resistor. It is absolutely required.

And with the freq you are switching 27 khz a driver chip may well be called for.

This is one of those times when spice is a godsend. Put the exact model of your fet into LTSpice which i think you are using and simulate with the expected inductive load and various gate resistors to see what value resistor begins to degrade switching time. You can even capture your voltage and current waveforms to make a good estimate of the dissipation of the fet with the slowed switching times.

As for the politics, voice your concerns in an email then follow exactly what your boss commands to do.

 

[schnell]

I think the problem with just adding a gate resistor is that the internal resistance of the microcontroller pin has a wide tolerance band,

...from page 310 of the datasheet (graph above enclosed) you can see that the output resistance of the uC's output pin is anywhere between 83R and 125R.....on top of that there's tolerance in the FETs gate capacitance..........

..i think the most definite way to assure adequate FET drive, and to prolong the life of the microcontroller is to use a gate driver IC.

If nothing else, i am certain that a product with a gate driver IC would mean its microcontroller lasted longer than an otherwise equivalent product without the gate driver.

Do you agree?

 

[IRstuff]

You seem to be stuck on using a driver chip, but nowhere have you stated an operational requirement for the higher current drive.

Your stated requirement for 27kHz allows, for a square wave output, about 8 us for each transition, and even assuming an average of 1 mA driving the gate+Miller capacitance, the transition would be around 5 us. As a guess, the actual switching time would be less than 1 us. Do you actually have a need for a faster transition? Particularly given the disparity between the solenoid current of 220mA, and design current of the FET at 8A.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[schnell]

hi,

basically , i believe that we should stay under the datasheet (page 304) "instantaneous" maximum pin current value of 25mA........at all times...even for small instants.

-as is suggested, we could just put in a series resistor and get a slower transiton time....because the FETs we are using are huge and won't overheat.

....but to be honest, i believe we should use smaller, cheaper FETs, and use a gate driver.

...if we start trucking with slow transition times and series gate resistors, then we are going to be juggling around with component tolerances to ensure we don't overheat...and these tolerances aren't clearly stated, so i believe we should go for the definetely safe solution of using a gate driver.

But if 90% of the rest of the world are just using micros as direct power mosfet gate drivers, then i am going to feel a little left out....should i ?

 

[2dye4]

No, you are correct that you should not overload the pin even transiently.

Cheapest route is pick the lowest resistance that keeps the current under max and see if your FET is switching fast enough.

 

[schnell]

i am also noticing that 4 cheap, small FETS (IRLML2060) and two dual FET driver IC's (Clare IXDN602SIA) COSTS $1.36...............

..whereas....

..four of the FETs that we are currently using (STB16NF06LT4 by st.com) costs $1.54


So driving the FETs directly from ther microcontroller is actually more expensive than using a gate driver...(...since it forces us to use more expensive FETs)