View Single Post
Old 01-26-2022, 03:31 AM   #2
momaka
master hoarder
 
momaka's Avatar
 
Join Date: May 2008
City & State: VA (NoVA)
My Country: U.S.A.
Line Voltage: 120 VAC, 60 Hz
I'm a: Hobbyist Tech
Posts: 10,860
Thumbs up A Dead/DOA Gigabyte GP-P450b 450W ATX PSU – troubleshooting & repair

Since the PSU was totally dead (no 5VSB present at all), I started from the primary side first. The fuse, bridge rectifier, and NTC thermistor (placed between APFC boost diode and primary cap’s + leg) all checked out OK. APFC current sense resistor (0.05 Ohms I think, 3W MOX type) was also good. Thus, I applied 120V AC on the input of the PSU and checked the DC voltage on the primary Teapo cap, which was around 165-170V – so all good there. Then I checked primary-side auxiliary supply winding on the 5VSB transformer, and that was also OK (not open, at least), along with its rectifying diode. After this, I don’t remember what prompted me to check the output rectifier for the 5VSB, but when I did, I found it was shorted:
Gigabyte GP-P450b - faulty 5VSB rectifier.jpg

Well, that explains the lack of 5VSB and any signs of life.
Unfortunately, no matter how many PSU parts (and part PSUs) I have, I couldn’t quite find a rectifier/diode that I felt comfortable using here. The diode really needs to be Schottky type with low V_f. Otherwise, at 3 Amps, the power dissipation can get a bit too high. And since most large diodes (>2A If_avg) I had were just regular fast-recovery type with V_f of 1V or more, I decided to try something a little more… “deviant”. I’ll let the pictures below do the talking.
Gigabyte GP-P450b - experimental 5VSB rectifier (1).jpg
Gigabyte GP-P450b - experimental 5VSB rectifier (2).jpg

^ Ghetto-mod-thread nominee of the week?
It’s simply an ON-semi 4813NGH MOSFET wired as a diode. According to its datasheet, it can withstand only 30V DC, which probably isn’t optimal, considering the original diode was rated for 45V. In terms of current, though, the body diode on this MOSFET can do up to 29 Amps - provided the junction is kept cool! I didn’t do any junction to case/tab calculations as I should have, but I figured the 14 AWG wire soldered to the MOSFET tab and the surface area of the PSU’s PCB may be able to provide enough cooling for an Amp or two. So if nothing else, I installed this “diode” contraption in the PSU to see how it would work out.
Gigabyte GP-P450b - experimental 5VSB rectifier (3).jpg
Gigabyte GP-P450b - experimental 5VSB rectifier (4).jpg

Surely enough, the PSU sprung back to life, even with this hackjob repair.

But now to see if my theory would hold true with the power dissipation: run some load tests on the 5VSB and see how hot this “diode” would get. I put my K-type thermocouple on the tab and started applying a load (via NiCr wire element) on the 5VSB. With just 1.1 Amp of current, the tab of the MOSFET was already sitting at a toasty 75°C – and that’s with the PSU PCB completely out of the case and exposed to room temperature, which at the time of this experiment was a rather chilly 18°C (64°F.) In summer, I know my house can easily hit 30°C (86°F), which is a whole 12°C higher. Add to that 2-3°C with the PCB installed back into the case, and I guesstimated the resulting temperature could easily hit 90°C on the MOSFET’s case… which means I’ll probably be running close to the thermal limits of the device. Not good! Therefore, it was time to improve/revise (improvise? ) things. No point in trying out higher loads if 1.1 Amp already made the MOSFET/diode this hot.

I considered soldering a penny on the tab of the MOSFET for a lowly-low cost heatsink solution of $0.01 … but didn’t feel like removing the MOSFET to do that. So instead, I settled on a piece of copper sheet I already had cut out for something else that I didn’t end up using. Rather than it sitting and collecting dust in my scrap bins, I figured this was a better use… and it allowed me to solder it to the MOSFET without removing it from the PSU.
Gigabyte GP-P450b - experimental 5VSB rectifier v2.jpg

I then proceed to do the same 5VSB load tests again. With 1.1 Amps of load, this time the MOSFET’s tab only got up to 55-56°C at the same room temperature – almost a good 20°C lower than before. With 1.5 Amp load, the temperature rose to about 68°C. And with 2 Amps of load, the temperature was 80°C… which is still very high. But that’s to be expected. After all, the typical forward voltage of the body diode for the 1348NGH MOSFET, according to the datasheet, was given as about 0.9 to 0.95V. Thus, at 2 Amps of power draw, the MOSFET’s body diode was likely dissipating around 1.8 to 1.9 Watts… so it’s normal for it to run this hot. Although I could easily pull 2.5 and 3 Amps of current from the 5VSB, I didn’t try it for any extensive amount of time, because the P_d of the MOSFET’s diode would have been too great and likely caused it to overheat. Therefore, 2 Amps is all I could do with my hackjob – good enough for now!

Thermal dissipation aside, the 5VSB performed admirably well (IMO) in terms of efficiency. With the 1.1 Amp of load (@ 5.08V output), my Kill-A-Watt registered 7.5 Watts of draw from the wall. That puts the efficiency at a little under 75%. With 1.5 Amps, the K-A-W showed 10.3 W, dropping the efficiency to slightly under 74%. And with 2 Amps of load, the efficiency was back around 74.5% again… so overall, the 5VSB seemed to float very closely to 75% efficiency overall (as far as the K-A-W’s accuracy goes anyways), which is very good. In contrast, most 2-transistor self-oscillating circuits I’ve tested seem to do no better than 50%... and often times hover around 30-40% (i.e. about half of what the 5VSB circuit in this Gigabyte PSU got.) Imagine if I actually used a proper Schottky diode! Its lower V_f would probably improve the efficiency even further.

I guess that means I should probably also get some Schottky diodes/rectifiers on my next parts order. Since I don’t know when that will be (maybe soon maybe not), I decided to close up the PSU with this “temporary” repair. Of course, the label was also modified accordingly.
Gigabyte GP-P450b - experimental 5VSB rectifier label.jpg

Besides, the next time I open it again, I will probably upgrade/replace the output caps too. I don’t really trust those YC caps. And the “ChnCap” (China Cap? ) – who’s heard of that brand before? Definitely recommend a recapping to anyone who wants to use this PSU long term. In my case, since the PSU is still pretty new, I’ll just run it as-is with its original caps and kludged 5VSB. If the PSU proves to work reliably well over time, I’ll recap it and fix up that 5VSB “fix”. The (revised) warranty on these PSUs is 3 years, so I imagine the dubious output caps will probably make it at least that far – especially with lighter loads, which most of my PCs are.

That being all said, the only two other things that worry me about this PSU are 1) the big primary Teapo cap being rated for 400V and 2) the PSU not shutting down with a low line AC input. With the former, a new good quality primary cap would probably cost at least $4-5. As for point 2) above, I performed the same type of low AC line tests / abuse as I do on all my other PSUs now. The Gigabyte GP-P450b here didn’t want to shut down, even when the AC line dropped to 75V AC. Instead, the output on the 12V rail dropped to 11.3V, which is out of spec. I didn’t check if the PG signal dropped, but I suspect it likely did, which is how that should be at least. Nonetheless, I’d rather see the primary-side APFC and main PS circuitry shut down when the AC dips below 85-90V. I didn’t run the PSU long enough with this low AC line to see if something will blow up or overheat on the primary. It surely would be an interesting experiment! But on the other hand, I’m also trying to keep things out of the landfill, so I’d rather keep this PSU operational for the time being. And that’s where it ends with this one so far – another dead PSU saved (for now.)

As time goes on, I’ll try to update the thread if anything happens/changes to this PSU.
momaka is offline   Reply With Quote