How to find the faulty mosfet?

[madan1]

Hi everyone.
I friend asked me to try to fix his gigabyte ga-x99-ud4 motherboard which is not booting. After some probing, it turned out that the CPU power supply is shorted.. but not only vCore, but cpu +12V also ( the "atx +12v x4" connector ).
Basically almost everything around and under the cpu socket is shorted to ground.
+12V on the ATX 24pin connector seems to be fine. Indeed nothing on the 24pin connector seems to be shorted.

For the vCore rail there are 6x IR3550 mosfets and it looks like everything on them is connected in parallel ( except for the PWM pin which goes to the control IC and each mosfet has its own control line ). Unfortunately I couldn't find boardview or schematics for that or similar MB, so can not really trace the other mosfet pins, but the PWM on all 6 mosfets, shows the same resistance readings to ground.

So, I guess at least one of those mosfets has gone ( as far as I know, that's the only place where +12, vcore and ground meet ), but can not find which one(s). Visually all of them are fine.
I tried injecting power on the connector, but my bench power supply cannot go higher than 5A@0.5V (2.5W), which doesn't make any difference to the temperature of the mosfets .

Desoldering the mosfets right now is not an option, because my hobby hot air/soldering station just cannot apply enough heat .

So any other ideas about how to find the culprit?

[petehall347]

Re: How to find the faulty mosfet?

there is a new tool to find it . its on the forums . basically its a very accurate volt meter .

[Per Hansson]

Re: How to find the faulty mosfet?

Here is the tool petehall refers to: https://www.badcaps.net/forum/showthread.php?t=93031

[dicky96]

Re: How to find the faulty mosfet?

@madan1
Try using an ESR meter to find the shorted FET(s). Remove the CPU first.

Referring to the attached pic... You probably have an input inductor Lin from ATX_12V to the Drains of the high side FETs

Consider that Q3 is short and Q1 is good.
If you use a multimeter on ohms range and measure from the junction of Q3/Q4 to 12V you would see short as Lin is a very low DC resistance

If you use a multimeter on ohms range and measure from the junction of Q1/Q2 to 12V you would also see short as the path from 12V via Lin - Q3 (short) - L3 - L2 to Q1/Q2 junction is also a very low DC resistance.

And your mulitimter can't tell you which high side FETs are short.

Now make the same measurement using an ESR meter. Lin has some impedance and the ESR meter uses 100KHz test signal so you will now see some resistance from 12V to junction Q3/Q4.

If you now measure from 12V to junction Q1/Q2 you are again measuring the path 12V - Lin - Q3 (short) - L3 -L2 but you will see higher resistance as the ESR meter sees the impedance of all three coils in series (Lin+L3+L2). So you will see the lowest resistance on the shorted FET(s)

You can apply the same technique from Junction Q1/Q2. Q3/Q4 etc to find any shorted low side FET(s)

Give that a go and let me know the results

[piernov]

Re: How to find the faulty mosfet?

Remove them one by one, will be done much faster.

[dicky96]

Re: How to find the faulty mosfet?

In the OP he says desoldering the mosfets right now is not an option as is hot air station is not up to the job.

Quite what he intends to do if he finds a faulty mosfet by other methods, I am not quite sure.

[madan1]

Re: How to find the faulty mosfet?

Thanks for the replies.
No luck for now.

Well the first thing I got in mind was to remove them one by one until the short is gone, but my soldering station can not handle that amount of thermal mass.

I tried the ESR* test, but the tester just does not read any esr or capacitance between +12/ground ( two separate tests with probe on the +12 and then on gnd ) and the out, in and gnd terminals on the FETs.
If used in its default mode, only resistance is detected and again - the values are similar on all fets.
Then I used frequency generator - 100khz on +12/gnd and expected to see difference on the scope** when probed on the IN,OUT and GND fet terminals ( 200khz-1000khz switching frequency according to the fet datasheet, so I hoped that the filtering would be tuned to something like x00khz ), but again - identical readings on all.

Now I'm totally out of ideas... probably will pass it to someone with an adequate hot air station to get those little **** out of the board.



*indeed my ESR tester is one of those https://www.banggood.com/Original-Hi...-p-986954.html
**diy DSO138

[dicky96]

Re: How to find the faulty mosfet?

@madan1
Hmm I am a little surprised about the results of the ESR test. I'll deliberately put a short on a motherboard vcore high side mosfet tomorrow (solder a bit of wire D to S) and then try this again myself to see if i can detect which phase has the short

I'll put the video on my youtube chanel

My ESR Meter is a different model but I also have one of these on order which looks similar to the one you have but much cheaper
https://www.aliexpress.com/item/4000...247a4c4dRyTRFv

I intend to do a video on this $6 tester when it arrives

[stj]

Re: How to find the faulty mosfet?

check resistance between ground and the gate pin on each fet

[madan1]

Re: How to find the faulty mosfet?

check resistance between ground and the gate pin on each fet

Thanks for the idea.
I tested pin5 against gnd and 12v and this time I found an odd one.
Pic1 is what I got on 5 of the fets, and pic2 is the reading from FET3.

When I got the motherboard I tested all FET pins with multimeter, but I guess it is not good enough for that particular job.



By the way, I also tried injecting a more powerful pulse signal on the rail ( and load connected in series with the MB ) and hoped to be able to detect its path with an improvised small antenna connected to a scope... the damn thing rang around the whole board .. I guess the whole ground plate became antenna
Then I wanted to try to detect the magnetic field around the conducting path, but don't have small enough ferrous particles on hand.. and am too lazy at the moment to grind some...



p.s. Do you think a mainboard would run without one of those vcore FETs? From what I've read, they are there to dephase the ripple and not that much for the power need. Is there feedback from them or some kind of ripple monitoring by the controller? Just curious...

[dicky96]

Re: How to find the faulty mosfet?

@Madan1
OK I checked out the method of using an ESR meter to find the high side MosFET short again and I don't understand why it didn't work for you as I have had plenty success with this method but I would be very interested to hear why

I made a short(ish) video describing this fault finding technique - if you get time to watch it I would be interested in your feedback

https://youtu.be/i8TudEfD2Ao

Rich

[CapLeaker]

Re: How to find the faulty mosfet?

not sure if EDS is still making the LeakSeeker 89.

[dicky96]

Re: How to find the faulty mosfet?

p.s. Do you think a mainboard would run without one of those vcore FETs? From what I've read, they are there to dephase the ripple and not that much for the power need. Is there feedback from them or some kind of ripple monitoring by the controller? Just curious...

That's not a good idea, though I have seen it done on GPUs. Anyway, depending on the VRM controller, it may just shut Vcore down after a second or two if one phase isn't working, or is not working correctly.

[dicky96]

Re: How to find the faulty mosfet?

So I decided to see if the ESR meter technique would also work on a GPU where there is a very low resistance (less than one ohm) from Vcore to Ground

Here is the result
https://www.youtube.com/watch?v=5cG2aJwNoAw

[madan1]

Re: How to find the faulty mosfet?

@Madan1
OK I checked out the method of using an ESR meter to find the high side MosFET short again and I don't understand why it didn't work for you as I have had plenty success with this method but I would be very interested to hear why

I made a short(ish) video describing this fault finding technique - if you get time to watch it I would be interested in your feedback

https://youtu.be/i8TudEfD2Ao

Rich

Just finished the video.. could it be that in my case all 3 - Vcore, ground and +15 are shorted ( hi and low phase I guess )?
Also, my ESR meter seems to be not working when there is no capacitance ( I'm attaching an image from testing a cap on a board ).

not sure if EDS is still making the LeakSeeker 89.

100+ us bucks on the fleabay... I guess if someone runs a pc/laptop repair shop is a good investment, but for the hobbyist, who rarely play with such hardware is lil'bit on the expensive side.