ECS 661GX/800-M7 Kills VRM MOSFETs

[c_hegge]

I aquired this board from work with a shorted VRM. After probingarund the VRM with the DMM, I found that MF3, a D412 BE5R2X, (circled in the VRM close-up) pic was shorted. After I replaced it with another D412, I tested the 12V CPU connector and the DMM no longer shows it as shorted. I attempted to fire it up with no CPU installed and the PSU fan just twitched quickly. After disconnecting it, I found that MF3 was shorted again. So I tried replacing it again with another D412, and same deal. The board killed that one too. So for the 3rd attempt, I also replaced the one next to it (a K3198) as well, although it didn't look shorted on the DMM. Once again the board killed the replacment MF3 instantly when powered up. The board seems to be killing ALL of the replacment FETs instantly. If you remove MF3 and power the board on it runs (although I never installed a CPU with the FET missing in case it overloaded the other ones). Any ideas as to what could be causing the baord to kill the FETs? The PSU is a re-capped Hipro HP - D3057F3H (like the one here) btw

[Uranium-235]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

top of my head, make sure none of those pins on the 775 are shorting

[PCBONEZ]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

Trace gate-pin back to controller IC and check that.
Also check the feedback path to the controller.

[severach]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

Did someone forget a capacitor at EC702?

See if it stays powered with the ATX-4 power cable disconnected.

I've seen 3 failures in dual MOSFET controllers. They might not work, work but overheat, or immediately burn out driven MOSFETs. Get data sheets until you find them. I hope your Alpha & Omega AOD412 are cheap.

The dual MOSFET controllers are often separate from the multi phase SMPS. If they were then I'd have expected to see 4 SOIC-8 chips, one per phase, or two larger chips, two per phase. I don't see likely chips for that so the dual MOSFET controllers are probably integrated into the multi phase SMPS chip.

Dual Buck MOSFET Driver SOIC-8:
OnSemi NCP3418 N3418, NCP3488 N3488
Richtek RT9603PS
Analog Devices ADP3418
Fairchild FAN5009, FAN5109
Intersil ISL6612, ISL6613

Dual Channel Dual Buck MOSFET Driver SOP-14 SOIC-14:
Richtek RT9602
Intersil ISL6614A
OnSemi NCP5358 (QFN-16)

[c_hegge]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

Did someone forget a capacitor at EC702?

I haven't removed any caps from the board, so the only people to forget caps were ECS.

See if it stays powered with the ATX-4 power cable disconnected.

Yes, it does

I've seen 3 failures in dual MOSFET controllers. They might not work, work but overheat, or immediately burn out driven MOSFETs. Get data sheets until you find them. I hope your Alpha & Omega AOD412 are cheap.

I have several other dead boards with D412s, so they're free (at least for now)

The dual MOSFET controllers are often separate from the multi phase SMPS. If they were then I'd have expected to see 4 SOIC-8 chips, one per phase, or two larger chips, two per phase. I don't see likely chips for that so the dual MOSFET controllers are probably integrated into the multi phase SMPS chip.

Dual Buck MOSFET Driver SOIC-8:
OnSemi NCP3418 N3418, NCP3488 N3488
Richtek RT9603PS
Analog Devices ADP3418
Fairchild FAN5009, FAN5109
Intersil ISL6612, ISL6613

Dual Channel Dual Buck MOSFET Driver SOP-14 SOIC-14:
Richtek RT9602
Intersil ISL6614A
OnSemi NCP5358 (QFN-16)

The controller is an Intersil 6614ACB. Is it possible that the controller IC is faulty? If so, how to test? it's the first time I've ever gone this far. Most work I've ever done on a CPU VRM was this repair on a gigabyte board (which still works great to this day).

[larrymoencurly]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

I had a controller chip damaged by a MOSFET shorting its gate. The replacement MOSFET ran really hot because the chip wouldn't turn it on completely

[PCBONEZ]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

The controller is an Intersil 6614ACB. Is it possible that the controller IC is faulty?

Yes
https://www.badcaps.net/forum/showpo...62&postcount=3

[PCBONEZ]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

They are all different.
To test it look in that PDF you linked to and figure it out.

One thing that helps is that more than one of the same circuit exist in the chip, so you can use the equivalent pins that go to the MOSFETs that aren't blowing to find what the expected values are.

.

[c_hegge]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

OK. After examining the datasheet I came to the conclusion that the resistance should be the same between ground and pins 12 and 9 (UGATE1 and UGATE2). The DMM shows 0 between pin 9 and ground but shows about 10K between pin 12 and ground. Does this indicate a problem with the controller?

EDIT: After digging through my small collection of dead motherboards, I found one with a similar controller IC (ISL6614CBZ) The same resistance test on the other board shows that the two have a similar resistance to ground. Would it be OK to replace the 6614ACB with the 6614CBZ (assuming I can get them both off without damaing anything in the process)?

[PCBONEZ]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

It might but I can't see what you are doing.

The MOSFETs are connected to pins 9 and 12 through the board so MOSFETs in different condtions might cause different readings at pins 9 & 12.

Neither pin 9 or 12 should be at ground though.
At least pin 9 is a bad reading. I'm not sure if pin 12 has a bad reading.
-
I have a board here that uses HIP6602B which is the same general kind of chip and has the same pin-out.
Pins 12 and 9 (UGATE1 and UGATE2) both read 99k to ground.
- This board probably has different MOSFETs too so it isn't going to be the same reading as yours but obviously 0 ohms on pin 9 says something is wrong there.

You should be able to determine if the ground is inside or outside the IC by [carefully] lifting pin 9 so it is not contacting the PCB solder pad.
Then check to ground on both the solder pad and the lifted pin to see which way the ground is. [Towards IC or towards MOSFET.]
-
Pull off as much solder as you can with some wick.
Then work a micro-screwdriver [flat blade] under the 'leg' and *gently* turn the screwdriver to push the leg up as you heat with the iron.
It helps if you can turn screw drivers with your teeth.
[Not a bad time to have a 25-30 watt iron with a small tip around BTW.]

-
Also keep in mind that the drawing in the PDF is a 'reference drawing' and the circuit on your board is probably not exactly the same.
There might be some little SMD [or even full sized] parts on the board that aren't shown in the drawing.
.

Remember: "This is FUN!"
.

[c_hegge]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

Darn, back to square 1. I removed the shorted MF3 and re-did the test. The resistance is now 10K on both pins 9 and 12, so it looks like the FET was what was shorting it to ground. The controller IC also appears to be OK, since I get identical readings now on all silmilar pins (similar = similar function. ie, UGATE 1 and UGATE 2, PWM1 and PWM2, etc.)

[PCBONEZ]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

Check the resistors in the ISEN_ lines and the caps in the BOOT_ lines.
Also on those lines check for components that aren't in the reference drawing and if you find any check them too.

[severach]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

If so, how to test?

By deductive reasoning. The source and drain are hooked to 12v, Vcore, or ground. Plenty of power available but the MOSFETs are not being burned out from those pins or they would all burn out. The only way to burn out an individual MOSFET from a bunch in parallel is through the gate. From the ISL6614 sample schematic we see that the ISL6614 drives the gate directly with no intermediate components or branch circuits. This can be partially verified by a quick continuity check. Therefore if one MOSFET is being burned out then it's the attached gate drive from the ISL6614 doing it. Since most MOSFET gates can handle +-20v, it's not clear to me how the gate drive that only has 12v available to it can burn the MOSFET.

By more deductive reasoning. If the MOSFET short is only drain to source and does not include the gate then the control chip should not be damaged. If the low side MOSFET shorts then the control chip is not likely to be damage since Vcore is only available for a brief moment before the power is cut. The low side MOSFET only has ground available and grounding the gate drive probably will not damage the control chip. If the high side MOSFET shorts and the gate is included the control chip will probably be damaged as it tries to sink 12v power.

All the damaged gate drives I've seen have been from shorted high side MOSFETs where the gate is shorted too. It's a shame they don't put forth a little more effort to protect the gate drive of those dual buck MOSFET controllers from unexpected 12v power.

[PCBONEZ]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

The only way to burn out an individual MOSFET from a bunch in parallel is through the gate.

Not true.
They have individual PWM and Feedback and Gates.
.

[sydney0968513640]

Re: ECS 661GX/800-M7 Kills VRM MOSFETs

I have a board and after testing the vrm driver mosfets were shorted and overheating after injecting voltage of 12V.These driver mosfets are NCP3488 but i am asking if there can be an equivalent replacement.