windows computer went kaputt - again

It was only in 2014 my Windows desktop computer went kaputt. I managed to diagnosed to the system power supply, or as I thought but the root cause was not entirely clear even though I managed to repaired it. I swapped out the power supply with another PC that has an identical unit. Strangely the donor PC works just fine with the "bad" power supply. That left me wonder if the power supply as after all not the cause of the failure. Fast forward to a few weeks ago, my same Windows computer died all a sudden while resuming from sleep.

The symptom was no POST code at all. Among the things I suspected were the motherboard, CPU, and the system power supply. To cut to the chase, I swap the power supply with the one I previously swapped out and placed into another Windows PC. Well this time there is still no boot with the motherboard, yet the same power supply works fine with the other similar but different motherboard. By now I was quite confident the likely cause are:

1. bad motherboard
2. this motherboard does not like something about the power supply because:

        a) the power supply has power sequencing problem
        b) the power supply has a voltage rail that is out of spec that the motherboard cannot tolerate

I tried to do some debug on the motherboard but it is just very difficult without access to a schematic and layout file. The pressure from not having a primary computer made me wanting to repair it as soon as possible. I decided to purchase an identical motherboard to avoid the challenge of having have to rebuild the Windows OS from scratch. At this point I was quite sure the problem is likely a bad motherboard. However, I cautiously not to ignore the possibility the power supply is bad, despite that fact my other motherboard works fine with it.

Finding the same motherboard that is now a few years old is not easy. I did eventually found a new one and it arrived in a few days. No luck! The new motherboard also has no POST. (I only has the CPU on the board to not risk damaging the memory and the graphics card). By now it is quite clear the problem is due to the system power supply, but what is it?

Checking the sequencing of the voltage rails with respect to the Power-Good signal is a real pain on the rear. I suspected that the 5V standby rail could be the problem as the chipset does not respond to the soft Power-On signal most of the time. It is rather easy to check the 5V standby with an oscilloscope. I broke out my trusty Tektronix 2465 300MHz analog oscilloscope.

Tektronix 2465 oscilloscope monitoring the 5Vsb power rail

What I saw on the scope was shocking. The 5Vsb rail has a significant AC ripple of 2.75V superposition onto the nominal 5V DC. This gives a peak of 7.75V! This is likely caused by a bad electrolytic capacitor. I cannot be sure but just a hunch. These two power supply has a 5 years warranty but it just ran out or about to run out. I decided to drive to Fry's to pick up a new power supply. Upon looking through the selections I was unimpressed with all and their inflated prices. The only one I like is the same Thermotake brand, and looks to be just a current equivalent of the same power supply. I decided to buy it. You must think I am mad.

Sure enough the motherboard booted with the new power supply. I quickly reassembled the computer and everything works. No data loss and I was happy.

POST code 27 is no system memory - exactly what I was expecting

I was not preparing to repair the power supply as they are quite cramp inside and are built not to be serviceable. Additionally there are live high voltage AC and DC voltage inside that can easily kill you due to the high power capability. Still I decided to give it a shot, not with a heighten expectation.

the big bundle of the conductor harness and the wire connections to the parts on the chassis make the debugging next to impossible

By cutting the cable tie of the main bundle I was able to tilt the main board sideways. I then trace the 5Vsb purple cable to the area of the PCB where the components for this supply rail are located.

I was able to tilt the main board sideways

I inspect the components in the vicinity trying to identify those likely are for this power rail. What I saw was quite likely the cause of the problem. An electrolytic capacity with a slightly but subtle bulge top.

the large electrolytic capacity by the big heatsink has a slight bulge on the top

I did a continuation check on the two terminals of the capacitor. It is connected across the 5Vsb and ground. It is certainly the bulk filter cap for this bad rail. It is secured with some latex like caulking for shock vibration reasons. I scrounge around the house for a similar capacity but to no avail. It is a 2200uF 6.3V. To me the voltage rating is too low to be safe but it is not surprising. I really don't feel like going on a trip to pick up a replacement. I scrounge around again this time looking for some device that might have a similar capacitor. This is one of many reason I have a hard time throwing away things I don't need. I knew the Microsoft model 80 USB speaker must have many inside. Too bad I threw out the whole thing two years ago when I could not repair it.

Eventually it dawned on me that a likely candidate is a very old ATX power supply - from circa 90s.

an old ATX power supply from circa 90s that I opened

Sure enough it has a bunch of electrolytic caps inside and the few smaller one would fit the bill. The challenge is removing them. They are all cemented together and to the PCB by the same latex caulking. Additionally they require a lot of care to remove them without causing damage.

the back side of the PCB; because they are throughhole and the leads were bent before soldering making them very difficult to unsoldered without causing damage

I managed to remove one with the soldering iron without causing damage. It is 16V so there is plenty of margin. The problem is the leads are very short and the diameter is larger than the supposedly bad 6.3V one. With a lot of juggle I managed to squeeze it in.

the 2200uF/16V cap that I scavenged off the old power supply

the supposedly bad one in my hand and the scavenged one that is larger

I reassembled the power supply and it is ready to be tested

Not wanting to risk damaging the new motherboard, I first hot-wired the power supply to switch it on. This is easily done by grounding the Switch-On signal. The voltage on the 5Vsb no longer has the AC component (ripple).

5Vsb after the replacement of the electrolytic capacitor

5Vsb before the replacement of the electrolytic capacitor

 I then proceeded to test it with the new motherboard and it booted

Enjoying the success of the repair I decided to also repair the other power supply which I verified also suffered from very large AC ripple on the same rail (even though the other motherboard tolerates it). Removing the other capacitors from the donor power supply calls for big fire power - a propane torch. I used a propane torch so I could melt the solder joints of a handful of capacitors simultaneously that are bonded together all at once.

more delicious duck confit; it has been my recent favorite dish as it is so convenient to prepare; in the photo too are my home made pickled baby cucumber

Success. I managed to remove all the large electrolytic caps off the donor power supply. There is no more 2200uF/16V however, but there are two 2200uF/6.3V ones. I used one to replace the one that is also bulged in the second Thermotake power supply.

 the same cap inside the second Thermotake power supply is also buldged identically

 here are the two buldged caps that are bad

the one on the right is taller but with identical capacitance and voltage

In retrospect I could have been more systematic in my fault diagnostic, and avoid the purchase the motherboard and the new power supply. However at the time, I really didn't feel strongly about the likely cause is the power supply as both works fine with a different motherboard.

Since I suffered from my PC went kaputt, I immediately better seeing this world's largest blue screen of death. Schadenfreude has never feel so good.