Go Down

Topic: MOSFET, LEDs burning out (Read 1 time) previous topic - next topic


Hi again,

I made a simple circuit shown on this picture:

It is a logic level N MOSFET, driven from arduino PWM. Arduino and this circuit is powered from a 12V DC power supply, arduino and the circuit have common ground. There are about 99 LEDs wired. Fade in and fade out works nice...

Within two weeks, about 9 LEDs. I did not find this a normal situation so I did some measurements.
With arduino PWM pin set to 255, LEDs are on, and I measured about 12V DC between point A and B marked on picture, as expected.

With PWM value of 0, LEDs are off, and there is a sinusoid voltage with +7V max, -57V min, frequency of 50 Hz, between points A and B. Is this expected?

Are dying LEDs related to this or was it just a bad coincidence?



With PWM value of 0, LEDs are off, and there is a sinusoid voltage with +7V max, -57V min, frequency of 50 Hz, between points A and B. Is this expected?

This is a big clue to your problem, there should never be a sinusoidal signal in this circuit.  I think your power supply is failing and exposing your LEDs to a partially regulated voltage that's exceeded their maximum reverse bais voltage and in causing them to fail. 


Aug 09, 2012, 09:22 pm Last Edit: Aug 09, 2012, 09:28 pm by DVDdoug Reason: 1
Yeah...   And I don't see any negative voltage supply, so -57V is very weird.   You might have a measurement error.    You're using an oscilloscope?    The ground on your probe may be broken.

If there is really 57V (positive or negative), that's enough to blow-out most semiconductors.  :(   

If there's an inductor somewhere that you didn't show, that's one way to create a negative voltage in a switching circuit. 

Or, maybe your power supply is totally fouled-up! 

Have you calculated (or measured) the total current?   What value of the LED current-limiting resistors?  What's the forward-voltage spec on the LEDs?

If you've got 33 of these 3-LED strings, and 20mA through each string, that's 660mA total which the MOSFET should be able to handle... But, you might chack the spec just to make sure.   Make sure your power supply is rated for at least 750mA.

Within two weeks, about 9 LEDs.
LED failures are very-rare when they are operated in-spec.   Unless you bought a bag of cheap LEDs on eBay. :D   It's actually more common to get an LED that's dead-on-arrival, or that's wired in reverse internally.    Where I work, I see that in probably less than 1 out of 1000 LEDs.  I don't think I've ever seen anything returned for repair with an LED that had died, although I may have seen a failure or two during "burn-in"...   Or, perhaps we just didn't catch it until the post burn-in test.

If the MOSFET fails & sorts-out, you'll get full-brightness, but the LEDs won't be damaged.   


If the power frequency is 50Hz in your area, I think that radiated power-frequency noise might be what you're seeing on the voltmeter.  If you have the positive probe at point A, the LEDs in series will limit the positive-going excursion at A relative to B to three LED voltage drops - suspiciously close to 7V.  The only thing that will limit the negative-going excursion will be reverse breakdown of the LEDs.  With 99 LEDs, you have a lot of area covered, and your array may be fairly effective at picking up power-frequency noise.  If you're using a breadboard and jumpers, you could be picking up a whole lot of noise.

Some ways to check if whether it's power-frequency noise:

  • Take the MOSFET out of the circuit.  If it's radiated noise, the behaviour shouldn't change much.

  • Put a 10k resistor between A and B.  A strong reduction in the voltage reading supports the theory.

  • Put a 10k resistor between B and ground.  Again, a strong reduction supports the theory.

Do those tests without changing the orientation or placement of the circuit.  If it's radiated noise, it will be sensitive to location and orientation.  If you move the device around, you might see the voltage change when it wasn't really due to changes you made in the circuit.

After the in-place tests, test the voltage between A and B  with the gizmo in several different orientations.  If the voltage varies, that supports the radiated noise theory.


When the MOSFET is off point B is floating.  It cannot go below 0V or the MOSFET body diode will switch on, but it can go high enough to reverse bias the LEDs (normally LEDs are about 5V reverse voltage rated).  The sinusoid is at mains frequency of course as its pick-up from the mains.

The fact its going so high is odd - a lot of mains pick-up.  Long low-voltage cable runs parallel to mains wiring?  If so don't do that, its poor practice at best.

If the pick-up current is large enough it could be damaging the LEDs by reverse biasing them perhaps?  Try a 1k resistor from point B to ground and see what the voltage is then (yes all the LEDs will perhaps glow faintly then).

But before tinkering further with the circuit double check there isn't a serious safety issue with the mains (old, perished or damp cabling?  Earth fault?  live/neutral reversed?)
[ I won't respond to messages, use the forum please ]

Go Up