The server that serves no more!

[TinkeringTim]

Hello everybody,

This is my first post here so hopefully I've posted the right thing in the correct place! I've been tinkering with electronics and computers for years but still very much an amateur. I love Dave's videos and they have inspired me get more in to troubleshooting as this is what I enjoy most. A friend has given me a 10 year old Dell PowerEdge server that won't power up. I thought it would be a fun project to try and repair because it's not worth much so doesn't matter if I kill it completely.

The symptoms are when the AC is connected an amber LED on the front of the unit  and on the motherboard illuminates. When you press the power button on the front, two relay clicks are heard coming from the power supply (like its tripping out) and then the power button starts to flash but nothing else happens.

After doing some research on the net and doing a visual inspections of the motherboard caps I came to the conclusion it was most likely the power supply at fault. It is not a standard PC power supply so simply swapping it is not an option. I've done a tear down of the PSU and found it to be full of dust and fluff etc. After cleaning it out I found 2 very charred looking ferrite coils. Everything else looks ok as far as I can tell. I've tested the coils for continuity and they seem to give a dead short. Each coil has 9 windings and they also seem to be wired in series. I'm not sure what they do or how to test them properly so any advice would be greatly appreciated. I have attached some photos for your perusal.

Cheers,

Tim.

[tautech]

Welcome to the forum Tim.
My approach is to first identify the SMPS controller IC, download a datasheet, look at "Typical Application" and work from there.
Often the typical application circuit is mirrored exactly in the circuit you are working on but not always.
Read the device description and usage notes plus some googling for better/full understanding.

Obviously the inductors have suffered one or several overcurrent events.

Build yourself a dim bulb tester for low start currents for power-up while repairing & testing

Be aware of the voltages involved.

[TinkeringTim]

Thank you for your quick reply and pointing me in the right direction.

I will report back how I get on.

Cheers,

Tim.

[kingofkya]

Unless its a custom shaped boards some of the 1u servers are you can generally plug in a normal atx supply and power it up. IT will likely also require and additional 8pin plug for additional cpu power if your supply dosen't have one Simple use that yellow and black 4pin connector(cometimes called a pentum 4 connector) it will work in most cases.

[Stonent]

Dell has made dozens of servers. Can you give the exact model?

[dr.diesel]

Kinda hard to tell from the pict, but looks like the transformer has gotten really hot as well, the yellow charred plastic?

[sunnyhighway]

Kinda hard to tell from the pict, but looks like the transformer has gotten really hot as well, the yellow charred plastic?

Both of them actually.
Not only the big one in the second picture looks overheated, but the smaller one at the bottom of that picture too.

[SL4P]

I reckon - apart from the fun/challenge of fixing the supply... you could buy another Dell PS or complete system on eBay for under $50 - probably within 100km of your location.

You may also find it is an OEM supply, and used in many different server applications.
(PSU only postage is much cheaper than the whole server!)

You solve the problem, have spares and maybe a reference to fix the dead PSU.

Just a thought. These units are a dime a dozen on the auction sites.

[T3sl4co1l]

Actually troubleshooting it could be a challenge, hard to say without a closer inspection.  The overall design does look roughly ATX-like.

Those coils look to me like saturable reactors, maybe -- these are sometimes used to ease the hard switching of rectifier diodes (more common back in the day, when diodes sucked more), or control the output voltage (rarely used for main supplies, but surprisingly common for supplementary supplies -- the average ATX unit uses one to derive the 3.3V rail from 5V windings).

Whatever they are, it's unlikely they're actually ferrite (black ceramic) -- that material isn't usually used in toroidal inductors, and doesn't usually have enough losses to fully nuke itself.  Metallic cores (ranging from powdered iron to metallic glasses!) are useful for inductors and saturable reactors, but they're more lossy, and in particular, are prone to runaway failure, where time*temp causes breakdown of the material, causing increased losses, causing... well, you know.

Or it could be good old fashioned cooking under load.  But this seems unlikely, given that the coils behind them are even larger, and likely carrying full load current, and don't seem to be suffering at all.  That's why I think they're doing something active, like diode switching, or voltage control.

The transformer looks okay: the tinge of darkness looks like ordinary transformer varnish.  Not that it couldn't be nasty inside, of course.

Tim

[CrabxCore]

If you get me the model number I will have a dig through my box. I have about 40 power supplies I took as payment for some soldering work from a computer shop. Also, I am in the US.

Crab