Author Topic: Mystery: who killed the MOSFET... ;)  (Read 9701 times)

0 Members and 1 Guest are viewing this topic.

Offline Sigmoid

  • Frequent Contributor
  • **
  • Posts: 488
  • Country: us
Mystery: who killed the MOSFET... ;)
« on: December 15, 2014, 03:23:51 am »
Hey,

I have this dead MOSFET on my Printrbot control board, most likely dead due to being connected to a too powerful fan. However, I'm not sure how I was supposed to see this coming, or what kind of calculation I can use to size the MOSFET (or the fan)...

So... here's the fan control circuit:


The MOSFET is an International Rectifier IRLML2060TRPbF, thermal dissipation 1.25W, continuous drain current 1.2A, pulsed drain current 4.8A at 20C.
http://www.mouser.com/ds/2/200/irlml2060pbf-90379.pdf

The fan is a Delta FFB0412SHN 12V high power fan, taking .5A current.
http://www.delta.com.tw/product/cp/dcfans/download/pdf/FFB/FFB40x40x28mm.pdf

Now, .5 is well below 1.2, let alone 4.8. After a 30 minute print with the fan running at about 50-75% (pulsed), the fan didn't stop when I tried to stop it - the MOSFET is letting current through in off mode. There IS some life left in it, if I set the fan to full power, it speeds up, and if I try to shut it off, it slows down significantly...

This was the first print with this MOSFET, as I recently installed it, replacing one that I accidentally shorted out with a multimeter probe. :) So I have considered a bad soldering job (I'm not used to working with SOT-23), but before the print, I did try turning it to full, and turning it back off, and the MOSFET was working at that time... So it died after (due to) being in use for a longer period.

So any ideas what I did wrong? I'm suspecting that the fet was actually soaking too much heat while switching... However I'm unsure how to calculate that sort of thing. What would be an acceptable heat dissipation value for driving this fan? Or what would be an acceptable fan for this MOSFET? :)
« Last Edit: December 15, 2014, 03:27:35 am by Sigmoid »
 

Offline IanB

  • Super Contributor
  • ***
  • Posts: 9767
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #1 on: December 15, 2014, 04:19:12 am »
Any concern about back EMF from the motor, given the pulsing?
I'm not an EE--what am I doing here?
 

Offline Sigmoid

  • Frequent Contributor
  • **
  • Posts: 488
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #2 on: December 15, 2014, 04:31:22 am »
Any concern about back EMF from the motor, given the pulsing?
Well if you check the circuit diagram, there's a Schottky there to protect the MOSFET from that sort of thing. :)

my crude calculations
assume rds on @ 5v = 0.6ohm
R-ja = 100c/w

assuming the mosfet actually pulsed at 0.5A
P = i x i x (rds) = 0.5 x 0.5 x (0.6ohm)
= 0.15w
Trise = 0.15 x 100 = 15oC

assuming the mosfet actually pulsed at 1A (due to inrush?)
P = i x i x (rds) = 1 x 1 x (0.6ohm)
= 0.6w
Trise = 0.6 x 100 = 60o
RDSon is around .5Ohm... So thermal losses when conducting are well below the threshold... I'm more worried about switching losses.
« Last Edit: December 15, 2014, 04:56:01 am by Sigmoid »
 

Online sleemanj

  • Super Contributor
  • ***
  • Posts: 2602
  • Country: nz
  • Professional tightwad.
    • The electronics hobby components I sell.
Re: Mystery: who killed the MOSFET... ;)
« Reply #3 on: December 15, 2014, 05:00:15 am »
Try putting the diode anti parallel across the motor.

Observe the crudely simulated difference between no diode, diode across fet, and diode across motor:

https://www.youtube.com/watch/7zGgETxvQc4
~~~
EEVBlog Members - get yourself 10% discount off all my electronic components for sale just use the Buy Direct links and use Coupon Code "eevblog" during checkout.  Shipping from New Zealand, international orders welcome :-)
 

Offline han

  • Frequent Contributor
  • **
  • Posts: 308
  • Country: 00
Re: Mystery: who killed the MOSFET... ;)
« Reply #4 on: December 15, 2014, 05:35:00 am »
Is the mosfet hot when running? (before it dies)
If the Gate driving voltage is below certain limit the mosfet became a steaming hot resistor and die
 

Offline Sigmoid

  • Frequent Contributor
  • **
  • Posts: 488
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #5 on: December 15, 2014, 05:36:19 am »
Try putting the diode anti parallel across the motor.

Observe the crudely simulated difference between no diode, diode across fet, and diode across motor:

https://www.youtube.com/watch/7zGgETxvQc4

Sounds good! :D And that sim looks very convincing... Thankfully the pin header is throughhole, so I can dead bug a diode across on the back of the board. :D

Is the mosfet hot when running? (before it dies)
If the Gate driving voltage is below certain limit the mosfet became a steaming hot resistor and die

Since it's in a not very accessible place, I haven't been able to check. After I powered down and checked, it wasn't very hot, but it could have cooled down in thate minute or so. I'm not sure about the gate voltage, it's driven from an Atmel mcu - my bets are on 5V. Since I didn't do anything to the gate side, only replaced the load, I don't think it's an issue.
http://reprap.org/wiki/Printrboard
« Last Edit: December 15, 2014, 05:43:19 am by Sigmoid »
 

Online tautech

  • Super Contributor
  • ***
  • Posts: 19670
  • Country: nz
  • Taupaki Technologies Ltd. NZ Siglent Distributor
    • Taupaki Technologies Ltd.
Re: Mystery: who killed the MOSFET... ;)
« Reply #6 on: December 15, 2014, 06:20:43 am »
 A gate waveform may show if the mosfet is switched fully on and not operating in the linear region where dissipation will increase.
VGS(th) is 1-2.5 V, are you sure you have sufficient drive voltage & current?
Avid Rabid Hobbyist
 

Offline Andy Watson

  • Super Contributor
  • ***
  • Posts: 1764
Re: Mystery: who killed the MOSFET... ;)
« Reply #7 on: December 15, 2014, 09:55:42 am »
Any concern about back EMF from the motor, given the pulsing?
Well if you check the circuit diagram, there's a Schottky there to protect the MOSFET from that sort of thing. :)
Schottky or Zener? If it is a Schottky it is not providing any protection against back-EMF, unless it has sacrificed itself! Have you checked D2?
 

Offline mikerj

  • Super Contributor
  • ***
  • Posts: 2576
  • Country: gb
Re: Mystery: who killed the MOSFET... ;)
« Reply #8 on: December 15, 2014, 10:09:30 am »
A tautech says, the gate waveform should be the first port of call.  If the MOSFET is being switched too slowly due to an inadequate gate drive, then it's power dissipation will greatly increase.

As an aside, unfiltered PWM isn't a very suitable method to control the speed of brushless motor fans.  I quick look through the Printrbot forums shows many people having problems with motor noise and fans failing to start at lower settings, exactly the kind of problems you might expect with this design.

If the internal brushless driver within the fan includes any significant decoupling on it's supply rails, then the peak current through the MOSFET could be many times higher than the steady state current.
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 16055
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Mystery: who killed the MOSFET... ;)
« Reply #9 on: December 15, 2014, 10:25:58 am »
What is drive voltage?

There is no current or power limiting here, you may well have had it fail due to inrush.  Maybe the fan has a bypass capacitor inside? Dumb enough to happen...

In a "dumb switch that just works" application, it's probably worthwhile to use a MOSFET that's overly large, large enough that you can ensure it will cause the fuse to open (like a self-resetting polyfuse).  You can also gain some leverage if the drain and fuse are thermally coupled (polyfuse is just a PTC with a sharp curve at 120C or so), though this may not be fast enough to prevent damage (if the fuse is heated to 120C, how much did the transistor's junction temperature heat up, above that?).

And to determine those ratings, sum up Rds(on), fuse resistance (minimum), fan resistance (if you can estimate it -- best guess at the minimum, just the wire resistance??) and power supply resistance (because if your supply is weak, you may well end up browning it out, too), and figure you need to withstand whatever short circuit current flows through that resistance.  Obviously the Rds(on) term depends on selection, but you can use the peak power theorem to see how to limit that.  Example: if everything else is 0.1 ohm, then a 12V supply will be capable of 120A peak and 1440W peak.  You might want to limit that to 14W peak in a beefy device (DPAK?), making some assumptions on how fast it will happen (a DPAK will take 14W or more for, probably tens of miliseconds, but only a watt or two continuous without extra heatsinking), which would require a 0.001 ohm transistor.  Well, that's quite unreasonable, so nevermind that, I guess.  But it goes to show how strenuous fault conditions are, if you try to get semiconductors involved.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline Sigmoid

  • Frequent Contributor
  • **
  • Posts: 488
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #10 on: December 15, 2014, 12:19:03 pm »
Schottky or Zener? If it is a Schottky it is not providing any protection against back-EMF, unless it has sacrificed itself! Have you checked D2?
D2 - a Schottky - seems alive. Yea it won't do much in terms of forward voltage spikes, you're right. The simulation @sleemanj posted has also shown that. :) I think I'll bodge a diode across the motor, put in a new transistor and see what happens.

As an aside, unfiltered PWM isn't a very suitable method to control the speed of brushless motor fans.  I quick look through the Printrbot forums shows many people having problems with motor noise and fans failing to start at lower settings, exactly the kind of problems you might expect with this design.
Yea, I've noticed that. :) I'm trying to figure out how much it can be improved without replacing the board. And mind you, this is the IMPROVED circuit. The previous board revisions didn't have either the polyfuse, or the Schottky.
 

Offline janoc

  • Super Contributor
  • ***
  • Posts: 3254
  • Country: fr
Re: Mystery: who killed the MOSFET... ;)
« Reply #11 on: December 15, 2014, 02:04:04 pm »
A better solution - get a variable speed fan that is designed for PWM speed control.

Advantages - no back EMF issues, no FET to blow up, no whining and unreliable function at low speeds (as you aren't PWM-img the entire BLDC driver!). 

Disadvantages - the fan costs a little more and you will likely need to adapt the PWM frequency/duty cycle range in the firmware, as the fans tend to be quite specific in what they need.

E.g. this one: http://www.digikey.com/product-detail/en/OD4028-12HB5/OD4028-12HB5-ND/2706610
 

Offline Sigmoid

  • Frequent Contributor
  • **
  • Posts: 488
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #12 on: December 15, 2014, 03:37:06 pm »
A better solution - get a variable speed fan that is designed for PWM speed control.

Advantages - no back EMF issues, no FET to blow up, no whining and unreliable function at low speeds (as you aren't PWM-img the entire BLDC driver!).

I wonder how I could modify the board to drive that... The fan header on the board is constantly fed with +12V, and it's the ground wire that is switched through that MOSFET...
 

Offline jlmoon

  • Supporter
  • ****
  • Posts: 608
  • Country: us
  • If you fail the first time, keep trying!
Re: Mystery: who killed the MOSFET... ;)
« Reply #13 on: December 15, 2014, 04:14:14 pm »
Try putting the diode anti parallel across the motor.

Observe the crudely simulated difference between no diode, diode across fet, and diode across motor:

https://www.youtube.com/watch/7zGgETxvQc4

I was looking for the Schottky across the load as well.  If I recall a must when using inductive devices as a load.
Recharged Volt-Nut
 

Offline mrflibble

  • Super Contributor
  • ***
  • Posts: 2024
  • Country: nl
Re: Mystery: who killed the MOSFET... ;)
« Reply #14 on: December 15, 2014, 04:43:04 pm »
So any ideas what I did wrong? I'm suspecting that the fet was actually soaking too much heat while switching..
Colonel Mustard did it. In the library. Bastard just keeps blowing up mosfets. :-BROKE
 

Offline leppie

  • Frequent Contributor
  • **
  • Posts: 269
  • Country: za
Re: Mystery: who killed the MOSFET... ;)
« Reply #15 on: December 15, 2014, 04:58:11 pm »
So any ideas what I did wrong? I'm suspecting that the fet was actually soaking too much heat while switching..
Colonel Mustard did it. In the library. Bastard just keeps blowing up mosfets. :-BROKE

Colonel Mustard did it. In the library. With a shorted cap.

FTFY ;p
 

Offline MacAttak

  • Supporter
  • ****
  • Posts: 682
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #16 on: December 16, 2014, 04:41:28 am »
If this is the transistor that I think it is, then I'm impressed in your ability to replace it. Not only is it a really small part, but it's damn near buried underneath the plastic shroud of the fan connector itself.

I've seen a number of people who have managed to pop that tiny fet, and they often end up just tossing the board and buying a new one because it's so hard to repair.

In general I'm not a big fan of the printrboard design - lots of little crappy things like the reverse pin numbering on the endstops, the ease in which the two large mosfets could short their heatsink tabs together (bridging the extruder and bed heaters), the total lack of extra GPIO exposed (and the extra pins that are exposed can't be used with PWM because they share a timer with the fan PWM). And when dealing with a cheap clone it's even worse. The clones will often use cheaper 2-layer boards with a different trace layout, and will usually swap the mosfets for ones that aren't properly spec'd for the application (which often causes them to overheat), etc.

edit: actually from looking at your schematic snip you seem to have a newer board (revision F or later), which moved the transistor and connector around and added some protection. So maybe some of the other things I mentioned have also been fixed - I've only dealt with the older boards.
« Last Edit: December 16, 2014, 04:45:54 am by MacAttak »
 

Offline janoc

  • Super Contributor
  • ***
  • Posts: 3254
  • Country: fr
Re: Mystery: who killed the MOSFET... ;)
« Reply #17 on: December 16, 2014, 04:43:35 pm »

I wonder how I could modify the board to drive that... The fan header on the board is constantly fed with +12V, and it's the ground wire that is switched through that MOSFET...

Cut the ground trace, connect it permanently to ground, bypassing the FET. And then use the connection that drives the gate of the FET to drive the PWM pin of the fan instead. If the connector doesn't have 3 pins, you may have to bodge something up there, but it shouldn't be too hard.

 

Offline kingofkya

  • Regular Contributor
  • *
  • Posts: 143
  • Country: us
Re: Mystery: who killed the MOSFET... ;)
« Reply #18 on: December 16, 2014, 07:14:05 pm »
A better solution - get a variable speed fan that is designed for PWM speed control.

Advantages - no back EMF issues, no FET to blow up, no whining and unreliable function at low speeds (as you aren't PWM-img the entire BLDC driver!).

I wonder how I could modify the board to drive that... The fan header on the board is constantly fed with +12V, and it's the ground wire that is switched through that MOSFET...

Easy attach the input or output if you fixed it of that fet ti the pwm pin normally blue on 4 wire pc fans. and provide the 12v on red and gnd on black. yellow is rpm feed back that you generally dot'n need need.
http://savedonthe.net/image/986/pinout.png

http://www.ebay.com/itm/like/121507041076?lpid=82
 

Offline Leadfootin

  • Contributor
  • Posts: 29
Re: Mystery: who killed the MOSFET... ;)
« Reply #19 on: December 22, 2014, 01:41:01 am »
A back diode on the motor is an absolute necessity. The DV/DT across the mosfet is great enough to damage the gate junction during turn off. I failed a 20A 600V IGBT with a 9V battery and 15mh inductor making a 500 volt temporary regulated boost converter power supply on my bench, shorting the gate thru without the device even getting warm. The IGBT had an internal diode configuration same as your FET.
 

Offline Rupunzell

  • Frequent Contributor
  • **
  • Posts: 349
Re: Mystery: who killed the MOSFET... ;)
« Reply #20 on: December 27, 2014, 06:52:20 pm »
Schottky diodes have a break down voltage of 50 to maybe 70 volts.
Inductive spikes can be a LOT more than that.

The bigger the FET, the more difficult it becomes to drive for a given output.

Most DC fans have an internal controller which does not require inductive spike suppression. If there are inductive spikes, more than just a diode should be used to control these high voltage spikes.


[/quote]
Well if you check the circuit diagram, there's a Schottky there to protect the MOSFET from that sort of thing. :)

{quote}
 

Offline Leadfootin

  • Contributor
  • Posts: 29
Re: Mystery: who killed the MOSFET... ;)
« Reply #21 on: December 27, 2014, 11:36:47 pm »
The Schottky is reverse biased to the supply voltage only, conducting the inductive spikes and as a bonus greatly increasing the low speed torque if DC motors. The faster the back diode the cooler it and the misfit will operate.
 

Offline Rupunzell

  • Frequent Contributor
  • **
  • Posts: 349
Re: Mystery: who killed the MOSFET... ;)
« Reply #22 on: December 28, 2014, 01:59:35 am »
Schematic does not show a motor, the load appears as a FAN which may or may not be inductive. As previously mentioned, DC fans today have internal controllers which tends to not produce inductive spikes at the power input due to the internal circuitry used to drive the motor. Diode across the FAN (load) would be OK except that may not be the reason why the mosfet failed. If one has a scope capable if extremely fast writing speeds and probed across the source-drain of the mosfet, that would tell more about what spikes-transients could be happening.

More suspicious would be the lack of a gate resistor or impedance controlling network as driving a mosfet directly from logic or similar can result in HF oscillations which are load-layout-power supply decoupling-ground dependent. What does the rest of this circuit look like and what about the layout and powering and load connections?

These all matter and could be the difference between stability and HF mis behavior.
 

Offline SaintGimp

  • Contributor
  • Posts: 17
Re: Mystery: who killed the MOSFET... ;)
« Reply #23 on: January 13, 2015, 05:22:13 pm »
Oh hey, I have exactly the same problem with a Printrbot Simple Metal that I got in December.  I have the Rev F control board but I gather from various forums that the Q3 fan MOSFET still gives people a lot of trouble even on this rev.

In my case the PWM control wasn't working well at all.  At 100% the fan would turn fine, but drop that down to 99% at it would slow to less than 10% of its former speed and emit a horrible whining tone.  In the process of trying to analyze the problem I unplugged and replugged the fan cable on the board while the board was powered up (doh!) and discovered afterward that the MOSFET was blown and the fan is now on all the time.

I've got a replacement component coming from Mouser right now but I'm also trying to sort out what kind of bodge I can put on this thing to make the fan work better and the MOSFET more survivable.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf