They’re here. Let’s begin. Look at that 6800uF, voltage upgrade and only slightly taller. 2200uF, I wish I could have got a taller one but there just wasn’t one available in that dimension from Rubycon. The 1000uF went superwide with the extra 5mm to 7.5mm lead (leg) width. It also gained 4mm of width, which was OK as this was as dual 12mm/16mm width silkscreen with both leg sizes available.
Nothing too interesting happening with the 330uF and 220uF capacitors.
47uF, 1uF bipolar/non-polar were all pretty standard swaps. The 0.47uF bipolar/non-polar got switched to a multi layer ceramic capacitor (MLCC). Going by the silk screen it looks like they intended to use ceramic but ended up putting on an electrolytic bipolar/non-polar at the time.
Speaking of those two, let’s put them on the board immediately. They go on one of the sub-boards, just need to make sure I add some hot glue to the top of the 1uF non-polar capacitor otherwise it will short the back of the board in front of it - that would be bad. The MLCC is visible at the front.
Now for some shots of the newly installed Rubycons:
Absolutely stunning. But we had a little issue with the hot glue holding the really big (non-replaced) capacitors on the high voltage side of things.
The isopropyl alcohol and CRC 2-26 have caused it to detach.
I am terrible with the hot glue gun.
At least they won’t take off into orbit now when we flip the switch.
With all the capacitors in, the sub-boards/daughter boards can be soldered back in.
See that short I was worried about? The hot glue wasn’t holding the boards together, it was stopping the capacitors shorting out on the other board.
Looking good enough, the glue dried before I could jam the board up against it. It’ll do the job it needs to do.
That’s it, we’re done.
Well, I guess we need to reconnect the ground, +5V and multi-colour cable to the sockets.
That’s right, it was so long ago I forgot about this mess. Time to get the AC area hooked back in and probably earth too.
That looks about how I remember it being.
I don’t want this to be exposed when I turn it on. Although I triple checked the components, I just don’t trust old power supplies.
I feel safer already. But how should we test? I don’t want to risk the main board just yet.
Maybe, we’ll just try something like this. I am a little concerned that there is no load. I should probably add some load to the rails. Then again, I could also just flip the switch.
+5.00V with no load. Let me just zoom in on that.
+5.02V, even better. But wait, the fan doesn’t start up? It tries to tick over but doesn’t engage. The fan runs off the +12V rail. That’s not a good thing, right? Let’s check the voltages, remembering that +12V was perfect before work began…
+6.60V is as high as the +12V rail would read in the 20 seconds it was powered up.
And the lowest we saw was open loop (O.L) - there was a whole range of voltages in between.
Troubleshooting time. What can we test without taking this thing apart again, knowing it is a pain to disassemble. Resistance. Let’s check the ohms. 242 ohm between +12V and GND. All other rails (including the negative rails) were showing ≥0.5k ohm resistance; many resistances more… is this an issue? Have we shorted ground to 12V somehow? Is a newly installed capacitor faulty?
I stared at old photos of the underside of the board, tracing where +12V and ground may have collided for about an hour and then decided to call it a night. The long walk home gave me time to run through some ideas. What if 240/250 ohm is normal for this board? What if we need load on the rails for them to engage? There is a +12V headed to the main board. We don’t have the main board attached. I did some digging, this is what I found.
Source 1 - The symptoms are close enough, 250 ohm resistance, fan powers on briefly then powers off.
I don’t think that you need to worry as long as the resistance you measure doesn’t show a real short circuit. I quickly measured the impedance shown between 0 and 12V on two good ATX power supply that I have lying around here - I also got 250 Ohms on one and 120 Ohms on the other.
Source 2 - Clicking noise, trying to switch on. Sounds familiar.
The PSU will only switch on with a certain load attached to it. Usually it is sufficient to hook up a hard drive or an old floppy drive in order to get a computer PSU to start properly. Without load it will either not start, or try to switch on but make a clicking noise. Without load it is also not possible to measure any output voltages.
Source 3 - This is the light bulb moment.
PC power supplies use a technique called switching to generate their DC voltages. Due to the manner in which these sorts of power supplies function, they need to have a load, meaning something that draws power from the supply, in order to function properly. A power supply that is turned on with no load attached will either fail to function or will function improperly. Better-quality supplies will detect a no-load situation and shut down, but cheap ones can be damaged. This is why you should not “test” a power supply by just plugging it in with nothing attached to it.
Whoops. Already did that, unless you count the multimeter as load…
In the early days of the PC, power supplies often had considerable load requirements, both for the +5 and +12 voltages. The +5 voltage requirement was easily satisfied by connecting the power supply to the motherboard, but the only devices that draw +12 V consistently are hard disk drives.
We are certainly not connected to the main board. What if it is as simple as that?
Thus, people who tried to troubleshoot power supplies without connecting a hard disk drive could be tricked into thinking the supply was bad. Also, systems with no hard disk drives would have a problem starting up unless equipped with a “dummy load”.
Well then, consider me tricked! This machine would have shipped with a hard drive. That said, it might be new enough to not require the +12V dummy load. I need to plug that power supply into the main board tomorrow morning… for now I will sleep and dream of Microsoft Solitaire. 
