Killed a Dimension 4550

[bgavin]

I took in a working board, recapped it, and killed it.

It had four bad Nichicon HM (no surprise) in 3300/6.3v. I recapped with MCZ, same spec. DOA after recapping.

The board will power up from the push button. The PSU lights and stays up. DVM measurement at the ATX-24 shows all good. A known good PSU is the same result. Definitely a board problem.

I noticed these four were very difficult to desolder. Kinda like the iron wasn't hot enough. I run at 450C for desoldering, so this should not be a problem. I cleaned holes, etc.

I disassembled my Hakko iron tonight, and made sure everything was snug. I'm not sure if something was loose or not... long day, and all that. The iron seems to be taking more time getting up to LED Off (at temp). Might just be wishful thinking on my part.

I will touch all the points again, just to see if there is a cold flow. After that, I suppose I can try out the VRM diagnostic PDF I picked up here.

[kc8adu]

Re: Killed a Dimension 4550

iirc that board is set up for 8 or 10 mm parts and if you are not careful you might get them in wrong.
i have had 2 sent to me after a failed attempt by the customer.
both had all in backwards.

[bgavin]

Re: Killed a Dimension 4550

I'm good there. When I remove a cap, I clean out the holes and verify the stripe location.

The job was only 4 caps. High power magnification shows one with less than stellar soldering technique.

I pulled the board off the Dell metal tray in preparation for resolder tomorrow. The push-pin HSF mount just plain sucks. Problem is, the board mounts to the tray with the push pins. Doubly lame. Otherwise, I have a number of high performance coolers in socket 478 that I could use. Too much fiddly for a low-perf board. It sure ate up my day off, though.

[bgavin]

Re: Killed a Dimension 4550

Resoldering the four new caps was no help.

I removed everything from the board, and it will not power up without a processor in the socket. I installed an old P4 for this, and it never got warm.

I was able to detect a very slight high frequency buzzing at the VRM caps. I cannot tell if input (3300) or output (560uf). My stethoscope filters out high frequencies, so it was unusable with the ear pieces. I unscrewed the tip and jammed that in my ear... the MOSFETS are the noise souce.

I'm no electronics tech, so I have no idea if this is Good or Bad. I've been trying to follow AKOR's vrm guide, but I don't know where the VRM inputs, outputs, etc, are located.

If anybody has a PDF with a picture, etc, so I can identify the VRM a bit more, please post or email.

[gg1978]

Re: Killed a Dimension 4550

I don't exactly have a picture of the 4550 motherboard in front of me, but since you've mentioned it's Socket 478, i would venture a guess that the VRM is on the motherboard between the CPU socket and ATX back panel.. You can check for Vcore at the output side of any of the VRM phase inductors.. I would guess it to have a two or three phase VRM. One of the inductors is part of the +12V input filter, and is typically smaller in size than the VRM phase inductors..

Try finding the VRM controller chip, it'll be in the same general area. It'll probably not have more than 24 or 28 pins, SOIC package.. Get the VRM controller datasheet, and start probing the pins.. Make sure the VID inputs are valid, and the VRM controller is not signalling a short or other fault..



Also, is the northbridge and or southbridge getting warm?? Also might want to check the VRM itself for heat, if something's getting really hot, there's probably something wrong.. If the CPU heatsink isn't getting hot, the VRM shouldn't be getting hot either..

[gg1978]

Re: Killed a Dimension 4550

Alright, i took a quick look at ebay to see if anybody had a picture of the board..

http://cgi.ebay.com/Dell-Dimension-4...ayphotohosting

The VRM phase inductors are the two large red ones.. Two phase VRM, i typically probe Vcore at the inductor output side, which is closer to the CPU socket and caps..

[bgavin]

Re: Killed a Dimension 4550

I have a tutorial thread running in the Power Supply section. There are high-res photos of the 4550 VRM there.

I don't know which components are which in the VRM. There are a number of IC, and I know I'm looking for a "buck controller", but I can't identify it.

Tinkering with a working KG7 board, I found the MOSFETS divided into two functions: Vcore and Vdimm. There are six total, three for each function. Is this 3 phase?

The ATX12V is apparently the +12v bulk supply to the VRM. There is a toroid inductor close to this connector, and the 2200/16v filter caps are also close. I assume the 12v goes to the input side of the D2PAK MOSFET. This board isn't working right, so any measurements are suspect without knowing the correct values.

The dead board has a noticeable high frequency whine. The working KG7 does not. The VRM is different, but both have MOSFETS. I was able to trace the whine: it is coming from multiple MOSFETS on the dead board. CPU, NB, SB, VRM are all cold.

[gg1978]

Re: Killed a Dimension 4550

I went ahead and posted in your other thread.. If there's any other questions, feel free to ask.. I guessed correctly that Intersil made the VRM controller IC, but then again they do make a lot of them....

As for the Abit KG7 it has a three phase VRM, with one top and one bottom FET per phase.. The six FETs are Vcore FETs only, Vdimm is towards the front of the board on the KG7..

[gg1978]

Re: Killed a Dimension 4550

I would probably check the VRM controller first.. If everythings cold that doesn't bode well though.. I'd go ahead and check for Vcore, Vdimm, and the northbridge and southbridge voltages.. See if any of the VR's on the board are making voltages to make stuff work..

I suspect the VRM controller for the CPU is entering a shutdown state, and that can be signalled to the rest of the system via the shutdown or fault pin, which indicates a fatal VRM condition. Might also check to see if maybe you accidentally unsoldered a pin on the VRM controller, as it is right there, and you did have to use a good bit of heat to unsolder the Vcore caps, which just happen to be right next to it..

[bgavin]

Re: Killed a Dimension 4550

The audible switching noise in the FETs is unusual to my ear. They are all doing the same thing, the Top group around 2.4v and the Bottom group at 0 output voltage.

The board died after replacing the four caps marked "R" in the photo. I though about collateral damage from heat, and inspect the feet of the IC close to those caps. I see no lifting under high magnification. I suppose it could be damaged by associated heat.

I use a Hakko station at 450C for desoldering. This board took an unusually long time to desolder the caps. I attributed that to my tip being loose. After replacing the caps, I disassembled the tip for a look-see. It seemed loose, but could have been my imagination. I recapped the KG7 later that day, and it works fine.

VRM, NB, SB, Memory are all cold. The board will not power up without a processor, so I used a known-good processor to let the board come up as far as it will for testing. The Intel VRM design guide says a processor is not required to activate the VRM, but I needed one. I dunno... this is my first foray into VRM diagnosis.

[kc8adu]

Re: Killed a Dimension 4550

sounds like an output short.
the singing is the tipoff.
wont be the first time i have seen a bum cap from new.
the one i had was a panasonic fm that popped after running a short time.
the jacket was backwards on another in the bag.
since it worked before you recapped its 1 of 2 things.
an error on your part or bad cap.
its the stupid shit that will get you.look it over real good.

[Wizard]

Re: Killed a Dimension 4550

Get your ohm-meter and ohm across replaced capacitors. (Rem ember CPU removed).
Usually appox 20-30 ohms or more for good ones). If shorted, pull the replaced caps
and retest.

Ideally, I rather lift the coil for the VRM first after marking it to put back in right way around and determine which one is shorted. That one is between CPU and the VRM.

Cheers, Wizard

[bgavin]

Re: Killed a Dimension 4550

I remember the reversed jacket story... I should be checking both for the stripe and the leg length. Live and learn.

The 3300/6.3 MCZ I'm using is all from the same batch (AFAIK). I bought a 200pc bag from Wai Fong, and only have two left.

I did remove the MCZ farthest from the ATX12V connector, and put the Fluke across it. It seems to charge, same as any other cap. I really don't know how to verify a cap with a DVM. Suggestions?

My next free time slot is Thursday, so I will desolder all four and measure. If they are all identical in performance... I must have killed something else. I will also DVM compare them with the two remaining new caps I have.

[Wizard]

Re: Killed a Dimension 4550

If it charges up, then there are no shorts on that, WHICH caps that got replaced?

Could you take a pic and mark them?

Um, The buzzing, squeak etc is overload noise of the VRM. I had FET ICs/transistor blow due to broken VRM controller IC. This happened in my Dell D600 was given to me dead, I'm now typing this on it). Fix was find compatible FET IC (totem pole two FET in one IC) and VRM buck voltage regulator controller stolen from junked notebook (Compaq 900 series AMD type good RIDDENCE!)

Make sure memory slot are good, slots as well.

Cheers, Wizard

[bgavin]

Re: Killed a Dimension 4550

The high-res photos are in the tutorial thread. The replaced caps have "R" on the top. Four 3300/6.3v MCZ were installed to replace defective Nichicon-HM.