Corsair SF450 PSU: can faulty -12V rail be removed?
I have a Corsair SF450 SFX power supply with slight water damage. 5VSB is present and it turns on via the ATX PS_ON pin, but immediately goes into protection mode (relais clicks on and off). During the brief moment before protection kicks in, all power rails are present except the -12V rail, which also went low resistance (24 ohm). I don't know if protection mode is triggered by over current on -12V due to the low resistance (the current limit is only 0.36A on that rail according to spec) or simply because the voltage is not present. I removed the IC generating the -12V (54231 buck converter) and the low resistance was gone (now kohms) but the protection behaves like before. -12V is not required on a modern PC, so I can live without it, but apparently the PSU controller won't stay on without it present. Is there some trick to force it to ignore the -12V or can I trick the controller by building my own cheap -12V converter and connecting it to the -12V line internally?
-
-
explain the water damage and how you dealt with it.
-
A few drops of watercooling liquid were spilled on it and apparently some made their way in. The PSU worked for a day, then it stopped and is now in the permanent protection state with missing (only) the -12V rail. I took it apart, cleaned all liquid stains I could find and checked for broken components, but nothing looks damaged on the surface. I assume the -12V buck converter blew due to over current, that's all. The rest of the PSU works, all voltages are present, it just won't stay on with missing -12V...Originally posted by stj View PostComment
-
Could you post some pictures of the insides of the PSU?
How is the 54231 buck converter producing the -12V rail? If it's from some higher negative voltage rail that's unregulated, then perhaps just dropping in a 7912 might be the simplest solution.
Or, if the PSU has an output inductor and is a forward converter type, you can add a winding with the same # of turns as the 12V rail (but just much thinner wire) and generate -12V off of ground.
Option #3... and this is probably the quickest way to test if the PSU is shutting down due to a lack of a -12V rail: take a 12V -UNGROUNDED- (2-pin AC plug) power adapter and wire it's positive (+) lead to the SF450 PSU's ground. Now take the negative (-) lead of the power adapter and wire it to the -12V rail of the SF450 PSU. Now plug in the power adapter first. Then plug in the SF450 power supply and try to power it up. With the -12V rail now generated by the power adapter, the PSU should be able to power up and not detect any issues, if the -12V rail really was the problem to begin with.Comment
-
I've already put the PSU back together for testing, so here's a picture of the insides from the internet and I marked the buck converter used for -12V:Originally posted by momaka View Post
It's very compact and a full inspection would require extremely tedious desoldering of the vertical daughterboards, which I wanted to avoid if there's any other way. The +12V are generated on the main board (underside), 5V and 3.3V are generated on the left daughterboard and the -12V is where I marked it on the main connector board (there are just caps and connectors otherwise). I strongly assume the 54231 buck converter is simply wired in "reverse" to the 5V rail to act as an buck-boost inverter (at least that's what I've read most PSUs are doing), so I don't think there's a higher negative voltage rail, but I can't really verify it without completely taking it apart. As I said, I simply removed the IC (gouged it off the board without taking it apart) and the low resistance was gone.
I had the same idea to inject external -12V and see if it will work. I wanted to do it with a battery and found an old 23A size 12V battery, but it can't even provide the <1mA quiescent current without dropping to 8V, so that didn't work. I don't have a proper floating power supply to test it with, but I tried it with some random power brick and measured only 2V difference between the SF450's ground and the brick's 12V - shouldn't be much of a problem, right? 😁 Anyway, long story short - it works! With the external -12V injected, the SF450 turns on and stays on and all the voltage rails are present.
So I've ordered a tiny buck converter module for 3 bucks, which I plan to bodge internally to the 5V line and run in inverter mode (should be MP2307 based):
I just hope it starts up fast enough (<200ms or so) before the protection triggers...Comment
-
Do you think that it would be a good idea to have a external voltage rail to power a rail that has an issue to leave there permanently I would do that to determine if this is the reason for it not functioning correctly but to leave it there permanently is hole another question I wondering if this is a good practice to use or follow ???Originally posted by momaka View Post
Comment
-
I have removed daughter boards before if you have a desoldering gun with the right size nozzle and use fresh solder before removing the solder from the soldering joints is not that difficult unless it is a multi layer board with large ground plains then it is a little bit more difficult to do but not impossible to doOriginally posted by Novgorod View Post
If you have the right tools to do this type of repair removing daughter boards
Last edited by sam_sam_sam; 08-23-2024, 11:35 PM.Comment
-
Of course the best way would be to replace the original blown IC on the original PCB, but that's a huge effort. The easy "permanent" fix would be to connect the $3 buck converter board internally to 5V and inject the -12V output into the respective rail (also internally). It's pretty much exactly how this rail is originally generated, just using a separate little buck converter board instead of the onboard one. For my use case it's perfectly fine and the PSU is even still ATX compliant with a functioning -12V rail. I think it's a very niche case where this particular rail gets blown, which is very easy to fix/bypass because it's low power and mostly irrelevant. If it happened to any other rail, I wouldn't be bypassing it and the effort to propery repair it would far exceed the value of the PSU ($80 or so)..Originally posted by sam_sam_sam View Post
I know it's possible with hot air or a good desoldering gun, but I don't have the right equipment at my current location, so it would be a huge delay and effort. I'll try the $3 solution first, for which I don't have to desolder anything..Originally posted by sam_sam_sam View PostComment
-
Ok, it's done and it works just fine! The PSU is back to doing its duty inside the PC with no issues. See gallery below - there's a short description for every photo.
The new buck converter board is wrapped with tape for insulation and stuck to the PSU wall with double-sided tape (even if the tape comes loose, the board is insulated so it shouldn't cause any harm).
Anyway, I'd say it was a full success with ~$3 of material cost. So if the stars align and you only blow the -12V rail, there's an easy fix, even on fancy modern SFX PSUs
..
Comment
-
Sorry, i totally don't follow how you can generate the -12v from a +5v source using the $3 buck module with a common gnd.Comment
-
-
It's a bit of an open secret that you can operate a buck converter as an inverter just by reversing its polarity. It's NOT a common ground, at least not with the rest of the system, that's the key. The only thing to look out for is the maximum voltage rating because it applies to the total potential difference from +5V to -12V (i.e. 17V), so a buck converter rated for 18V is just sufficient, but it won't work on +12V to generate -12V because that's 24V difference.Originally posted by Sam Scarbo View PostComment
-
it still doesnt make sense,
reversing the ouput polarity is one thing,
but raising 5v to 12v needs a boost, not a buck.
and dont trust those modules.
the pot is shit and can drift.
i use them, but i use a fixed resistor instead of the pot.
Comment
-
It's black magic, right? The buck converter "simply" creates a potential difference between the input and output pin with the output voltage being below the input voltage. If it can reduce a +24V input by 17V to make a +7V output, it can reduce a +5V input by the same 17V to get -12V out depending on what you consider a ground reference. It's all a matter of perspective, like the guy in the video explained.Originally posted by stj View Post
I know the built-in pot is not great but it's good enough for my application. The -12V aren't actually used anywhere, they just need to be present for the PSU controller to set the "pwr_good" condition. Some drift is no problem as long as it's not several volts...Comment
-
thanks for the info. one day, i will try to test this black magic out.Comment
Comment