Go Down

Topic: Basic MOSFET Problem (Read 3570 times)previous topic - next topic

CraigKC

Jun 04, 2012, 06:20 pmLast Edit: Jun 08, 2012, 05:07 pm by CraigKC Reason: 1
I'm trying to use my first MOSFET to turn on/off a strip of LED lights with an Arduino Nano (5V).  When the gate is high the strip lights perfectly and voltages look correct, but when the gate is low the LED strip is dimmer but still partially lit and the measurement of drain -> vcc is ~3.5v.  I'm expecting 5V.  I'm only powering a small strip that uses ~80mA now but will be used for strips that consume ~200mA to ~2A in the future.

Vishay Si4778DY MOSFET
SOIC-8
Logic Level Gate
25V / 8 Amp
Datasheet: http://www.vishay.com/docs/69817/si4778dy.pdf

Circuit Details
Schematic: http://i.imgur.com/IC3ST.jpg

* Note the breadboard has the 10k pulldown on the gate side of the 300ohm series resistor.  I tried both locations to verify gate was getting a strong pull to gnd.

Gate High
GND -> VCC = 5.031V
GND -> Gate = 4.846V
Source -> VCC = 5.016V
Drain -> VCC = 5.014V

Gate Low (GND)
GND -> VCC = 5.031V
Gate -> VCC = 5.024V
Source -> VCC = 5.03V
Drain -> VCC = 3.519V (This is the problem / LED strip connected here)

My problem is that when the gate is low the drain is only partially closing.  I'm sure there's something in the datasheet that explains this but I don't know which spec to start looking at.  I believed that using a logic level MOSFET would be simple so I'm not sure which direction to look in.

To see a related post of mine from a few days ago with more usage questions/advice see here:
http://arduino.cc/forum/index.php?topic=107598.0

Grumpy_Mike

#1
Jun 04, 2012, 06:50 pm
You have no current limiting resistor in that LED. It could be that the leakage current of the FET is keeping it on.
Put R1 directly onto the gate, that is the other side of R2.

Docedison

#2
Jun 05, 2012, 07:03 am
Just at looking at your measurements I see an error... The Source Must be Grounded that you measure 4+ volts there indicates that the circuit is improperly wired. with the gate and source at the same potential the Mosfet cannot conduct. The position of the pull down isn't important... the gate resistor is 300 ohms and the pull down is 33 times as large. The pull down is for drivers that go "Tri-State" (open circuit), it keeps the gate from 'wandering' as the local electrical fields change. Without it the connecting wire will act like an antenna. That is something you really don't want.  IMO

Doc
--> WA7EMS <--
"The solution of every problem is another problem." -Johann Wolfgang von Goethe
I do answer technical questions PM'd to me with whatever is in my clipboard

Grumpy_Mike

#3
Jun 05, 2012, 10:07 am
Quote
My problem is that when the gate is low the drain is only partially closing.

No your problem is when the gate is high the voltage in the drain should be zero. When the gate is low the voltage on the drain should be at the supply voltage.
You need to connect all the drain pins together, and all the source pins together. You are not doing this.
What voltage is your LED being powered from? You need to make sure that the -ve of this supply is connected to the arduino ground.

dc42

#4
Jun 05, 2012, 05:51 pm
On the face if it, that arrangement should work (I'm assuming that the LED strip includes a series resistor). It's best to put the pulldown resistor on the microcontroller side of the gate series resistor (as you currently have it) to avoid reducing the gate voltage when the pin is driven high; but with the values you are using (300 ohms and 10K), the gate voltage will only drop by 60mV if the resistor is in the other position, so it makes little difference. I don't think that connecting all the source and drain pins will make any difference as they will be connected internally, although when you come to switch larger currents, you should definitely connect all the pins so as to share the current between them. It could be that the mosfet is faulty - maybe you damaged it during soldering (remember, mosfets are static-sensitive).

What happens if you disconnect the gate wire from the Nano and connect it directly to ground instead? What sort of meter are you using to measure the drain voltage? If it's an old analogue meter that has a significant current draw, then you can't expect it to read 5V on the drain when the mosfet is fully off.
Formal verification of safety-critical software, software development, and electronic design and prototyping. See http://www.eschertech.com. Please do not ask for unpaid help via PM, use the forum.

Docedison

#5
Jun 06, 2012, 07:58 am
In All the books I read throughout my days... There was always the notion that an  N channel enhancement mode Mosfet will not conduct if the gate is at the same potential as the gate, are we looking at some new physics here??? I might add that I personally never found one that did unless it was shorted, something that doesn't seem to be the case here I quote from the initial observations...
"Gate High
GND -> VCC = 5.031V
GND -> Gate = 4.846V   <<<<<<<< Good gate voltage device should be fully on
Source -> VCC = 5.016V <<<<<<<< The same voltage, the device is cut off and this reading is I believe wrong as it duplicates "Gnd -> Vcc"
Drain -> VCC = 5.014V   <<<<<<<< This one doesn't look right either looks like the neg probe was grounded instead
If the gate were destroyed by static (I've done for a few myself) the device 'usually' is partially turned on, it depends on how much energy is coupled into the gate. When ALL the possibilities are != then there is something missing. In my case all too frequently it was a bad measurement or "I Thought I did"... and usually was wrong. The guy chose a good fet, Real Low Rdson .023 ohms from the data sheet, a little hard to mount and when the PCB is fabbed there should be a LOT of copper under the drain leads as that is the heat sink... Not that a real lot would be required because of the low on resistance and low (relatively) gate drive required. From the Mfr's data sheet a Vgs of 3 V is good for 12 A drain current... That's a lot of lights, a whole lot.  IMNSHO...

Doc
--> WA7EMS <--
"The solution of every problem is another problem." -Johann Wolfgang von Goethe
I do answer technical questions PM'd to me with whatever is in my clipboard

MarkT

#6
Jun 06, 2012, 12:01 pm
You've confused people by measuring voltages w.r.t. Vcc - please measure w.r.t. ground.  And measure when shorting the gate to source directly (then we aren't being confused by any bugs in your code).

You _should_ be seeing the Vds to be less than 5V when off, this is quite normal as the leakage current through the MOSFET will cause some forward voltage across the LEDs.  But you shouldn't be seeing any appreciable lighting up of the LEDs.  Add a 100k resistor between drain and Vcc to bypass the leakage current - if the LEDs still light then the MOSFET leakage is well out-of-spec (you have fried the device).

Ways you may have fried the device in include static electricity and poor soldering technique (soldering should be fast with a hot iron).  MOSFET gate oxide is a few nm thick - easy to puncture by stray static electricity if you don't take precautions (for instance don't wear nylon).
[ I will NOT respond to personal messages, I WILL delete them, use the forum please ]

Docedison

#7
Jun 06, 2012, 09:59 pm
If you DO see any "Leakage" the part or the measurement is defective, From the Data sheet for the Mosfet
"VDS = 25 V, VGS = 0 V, TJ = 55 °C = (MAX) 10uA" and there is no typical value for this parameter because at the extreme (131 Deg f) the Worst case "pass" criteria is 10 uA, that is they are sold as another type of part whose criteria isn't as demanding, if that criteria isn't met in testing. This is not to say that all parts aren't made alike. The errors creep in because parts are "sampled" from each production run and tested.  However once again I got off the point... Typical Mosfet Leakage for "normal" parts is effectively 0 at room temp 20C or 68F deg. That you measure a differential across a small value part does not necessarily indicate leakage but rather the small current drawn by the meter, the proof is to disconnect the lead in question and measure the difference in voltage as the difference is now the combined leakage of the other parts connected and PCB leakage etc. I used Mosfets in production for many years and found very few with leakages of more than 10 - 20 uA, above that the parts were usually damaged and I had them replaced as they were an accident looking for a place to happen. For casual or better 'Hobby" use even that leakage in a switch isn't anything to be concerned about... It will either fail due to the damage done to cause the leakage (Not usual) or it will work for years, in my experience and opinion based as I say in my tag line, "... on Many Years of Mistakes". I learn much from my mistakes as they are generally uncomfortable and discomfort (pain) is the Best Teacher. It is said that If you don't learn you are doomed to repeat them... IMNSHO

Doc
--> WA7EMS <--
"The solution of every problem is another problem." -Johann Wolfgang von Goethe
I do answer technical questions PM'd to me with whatever is in my clipboard

Go Up