PS3 (Research/Experimental) - NEC/TOKIN Capacitors Replacement - YLOD

hoyohoo said:
Hey guys, I am running into issues, as changing the 4 NECs on the back side did not help me. Could it be that the wires I am attaching are too thin? Is there a high current flowing here? I will try 1,5 mm² Wires now, however, the pins of the Caps I am soldering are very thin too, so, will this even make a difference?
I would buy the tantalums, if I only knew that the PS3 will definetly work, but I got nothing till now so I am very unsure...

PS: Or, do I have to fully assemble the ps3 again? I am always testing with only the PSU and the Touch bar in place. Not even the FAN is plugged in...
Click to expand...

Show us a clear/close up picture of the caps on the board.
 
squeept said:
When you're all done with creative writing class, maybe you could start the scope data collection? No? Cool.
Click to expand...
Look i am trying hard to understand what you mean but your message is not clear.
Is the point you want to state that the real fix is a reball? Or you are pointing out the obvious, that the real issue of ylod is the tokins?
Can you please make it clear for everybody to understand your point of view!

Στάλθηκε από το POCOPHONE F1 μου χρησιμοποιώντας Tapatalk
 
Cultists win this week! Opened up my deliveries from the weekend and found a CECHK01 with bad TOKINs on the GPU only. You're up to 2 in 42!

Just noticed I forgot to add the details to the previous image: it was a CECHA01, 3 second YLOD, pressure test did nothing, heat test did nothing, and it would never boot for any reason. I didn't get "after" images because it popcorned during reball - I don't sell CECHA01 without reballing the GPU even if they work perfectly, plus I wanted to add to my mental notes another notch in the "heat from reballing doesn't bring them back to life" column. Thus is life.

Now, for this one: CECHK01, random length YLOD and would even boot for a few seconds sometimes. Pressure test did nothing, heat test had no discernible effect. Not reballing this one since CECHK01 aren't worth enough to justify the time. Worked fine after replacing half, but as you can see below, there was still quite a bit of that pulsing "sawtooth" wave left for me to feel comfortable. Replaced the rest, and it almost looks normal.

Bad GPU TOKIN:
LNit0tL.png


Bad GPU TOKIN half replaced with tantalum:
5Xcc84J.png


Bad GPU completely replaced with tantalum:
XhY3sfB.png


The set I burned looked somewhere in between the dead CECHA01 and the dead CECHK01. I can't seem to find the burned set image now since I purged some poorly written notes, so *gasp* you'll just have to trust me. I'm guessing the CECHK01 looks so much smoother and more pronounced since the TDP is lower, so there's less switching noise.
 
Last edited:
squeept said:
Cultists win this week! Opened up my deliveries from the weekend and found a CECHK01 with bad TOKINs on the GPU only. You're up to 2 in 42!

Just noticed I forgot to add the details to the previous image: it was a CECHA01, 3 second YLOD, pressure test did nothing, heat test did nothing, and it would never boot for any reason. I didn't get "after" images because it popcorned during reball - I don't sell CECHA01 without reballing the GPU even if they work perfectly, plus I wanted to add to my mental notes another notch in the "heat from reballing doesn't bring them back to life" column. Thus is life.

Now, for this one: CECHK01, random length YLOD and would even boot for a few seconds sometimes. Pressure test did nothing, heat test had no discernible effect. Not reballing this one since CECHK01 aren't worth enough to justify the time. Worked fine after replacing half, but as you can see below, there was still quite a bit of that pulsing "sawtooth" wave left for me to feel comfortable. Replaced the rest, and it almost looks normal.

Bad GPU TOKIN:
LNit0tL.png


Bad GPU TOKIN half replaced with tantalum:
5Xcc84J.png


Bad GPU completely replaced with tantalum:
XhY3sfB.png


The set I burned looked somewhere in between the dead CECHA01 and the dead CECHK01. I can't seem to find the burned set image now since I purged some poorly written notes, so *gasp* you'll just have to trust me. I'm guessing the CECHK01 looks so much smoother and more pronounced since the TDP is lower, so there's less switching noise.
Click to expand...
This is actually very interesting... I've been waiting to buy a scope for a while now, but I'm admittedly very inexperienced with one. I have a few questions, based what I can tell from those images:
  • The bad tokins are throwing the line all over the place, but what is that measuring? Capacitance? Frequency? Load?
  • It makes sense that the scope read stabilized as you added more tantalums. How different is that final measurement from a "known good" set, but with the tokins?
  • And to top off the list of basic questions -- where in the board did you hook up the scope?
 
It's just voltage versus time. I measured at the vias at the side of the TOKIN facing the chipset. My guess (and what others have already guessed) is that they're shorting out under load. Or opening? Whatever lets you see sawtooth after the filter cap. Where's that guy that knew what he was talking about earlier?

The known good images are about 20 or 30 pages ago. I haven't bothered saving any more images of those since they're pretty much all identical, some have lost a little bit of capacitance so they have different sized spikes, but that's about it. It's an entirely different waveform (or lack thereof).

It actually stabilizes as I remove the TOKIN's, not as I add more tantalum. Adds more weight to the shorting out theory. I'll get images in between those steps next time, but you might be waiting awhile.
 
Last edited:
squeept said:
It's just voltage versus time. I measured at the vias at the side of the TOKIN facing the chipset. My guess (and what others have already guessed) is that they're shorting out under load. Or opening? Whatever lets you see sawtooth after the filter cap. Where's that guy that knew what he was talking about earlier?

The known good images are about 20 or 30 pages ago. I haven't bothered saving any more images of those since they're pretty much all identical, some have lost a little bit of capacitance so they have different sized spikes, but that's about it. It's an entirely different waveform (or lack thereof).

It actually stabilizes as I remove the TOKIN's, not as I add more tantalum. Adds more weight to the shorting out theory. I'll get images in between those steps next time, but you might be waiting awhile.
Click to expand...

The shorting out theory is really interesting... It might explain why, even when a sufficient number of tantalums have been installed, that the unit would still fail to boot (assuming that the rsx and gpu are healthy, of course).

I guess one way to know for sure (the hard way, without a scope) would be to remove all 8 tokins, all tantalums, then test for continuity as we install the tantalums? That wouldn't do any good, though, if the rsx needs work, but I can't think of any way to test the rsx, other than pulling the whole thing out and inspecting the pads.
 
@hoyohoo
Those capacitors won't work as their esr is sky high and usually CPU's & GPU's require very low esr to work. They look like fakes nippons tbh.

Στάλθηκε από το POCOPHONE F1 μου χρησιμοποιώντας Tapatalk
 
demetris said:
@hoyohoo
Those capacitors won't work as their esr is sky high and usually CPU's & GPU's require very low esr to work. They look like fakes nippons tbh.

Στάλθηκε από το POCOPHONE F1 μου χρησιμοποιώντας Tapatalk
Click to expand...
Ok, thats exactly why I first asked here, if they will work or not ... Well, this sucks.
I used these, the datasheet can be downloaded on the same side too (it is called "Datenblatt" and it is english):
https://www.conrad.de/de/p/yageo-se...0-f-16-v-20-x-h-8-mm-x-12-mm-1-st-445373.html
 
marciolsf said:
I guess one way to know for sure (the hard way, without a scope) would be to remove all 8 tokins, all tantalums, then test for continuity as we install the tantalums? That wouldn't do any good, though, if the rsx needs work, but I can't think of any way to test the rsx, other than pulling the whole thing out and inspecting the pads.
Click to expand...

There's no way to make sure that the RSX is working. There are only ways to show it is dead. Unfortunately, without Sony giving us the JTAG data, there's a lot of guesswork involved.

No test not under load of any of the bad caps has showed any measurement different from the good caps with my equipment. Even the self-healing burn marks don't tell you anything - from my extremely limited sample size of 2 confirmed bad sets, one set was burned pretty nasty, and one set looked pristine.

I really can't think of any way to test them on the board without removing a ton of other components first. I'm sure I could come up with a simple circuit to test the caps off the board now that we know they have to be under load to get ugly, but (seriously not trying to be an asshole here) nobody else here has the tools to remove them intact so that seems pointless.

I just checked Amazon, there are great USB scopes coming in just under $60 US. Anyone reading this thinking about doing the repair is already about to tackle a project pretty far above your random hobbyist. You're also about to spend $10 on capacitors blindly, and an hour or two dicking around - you can justify the cost of a cheap scope, and if you're doing this kind of shit you're going to use it again. I'm even seeing cheaper DIY kits, just make sure it can do 50mv/div or lower and 1us/div or lower so you can see the sawtooth wave clearly.
 
Those are some interesting scope images. 100 mV is quite a bit a ripple. A short or excessive leakage current probably describes this best. A short would cause a voltage drop in excess of the design ripple voltage since the capacitor has effectively become a resistor. Also, I don't think a capacitor failing in an open state would describe what is happening here. I'm assuming that this is being tested at idle so a capacitor failing open should not affect ripple too much since the circuits are designed for highest current draw.

@squeept You may be able to test them in circuit for leakage or shorting if you disconnect all of the current sense resistors or inductors. This wouldn't be perfect, but it might confirm shorting of those capacitors (or possibly ceramic ones). Obviously, confirming that the RSX doesn't conduct much when it is off would be necessary. As far as I know, most of the shorting when the circuit is off comes from the mosfets.
 
sautzer said:
If I'll buy a PS3 with the YLOD but there was an attempt of reflow, does that involve the success of this guide?
Click to expand...
Yes, don't buy it. Or you might need a reball or even new RSX or/and Cell chips

Στάλθηκε από το POCOPHONE F1 μου χρησιμοποιώντας Tapatalk
 
Moltiplier said:
Those are some interesting scope images. 100 mV is quite a bit a ripple. A short or excessive leakage current probably describes this best. A short would cause a voltage drop in excess of the design ripple voltage since the capacitor has effectively become a resistor. Also, I don't think a capacitor failing in an open state would describe what is happening here. I'm assuming that this is being tested at idle so a capacitor failing open should not affect ripple too much since the circuits are designed for highest current draw.

@squeept You may be able to test them in circuit for leakage or shorting if you disconnect all of the current sense resistors or inductors. This wouldn't be perfect, but it might confirm shorting of those capacitors (or possibly ceramic ones). Obviously, confirming that the RSX doesn't conduct much when it is off would be necessary. As far as I know, most of the shorting when the circuit is off comes from the mosfets.
Click to expand...

The waveform didn't change whether it was on the console that was only ever 3 second YLOD (I have no idea how much stuff the system tries to power up in that 3 seconds) or on the one that would boot completely up for a few seconds here and there. No data from attempting to run a game, it didn't stay on long enough.

The instantaneous and some of the short spikes on a console completely missing half of the TOKINs (and when completely changed over to tantalum) would get over 200mv. The system never flinched playing a game like that. It only seems to matter when you start seeing a clear sine wave on the CPU or that weird sawtooth on the GPU.

RSX chip itself is like 3 ohms from Vin to GND (off the board), so no dice there. Either the intact caps or the chip have to come off, and that's not helpful for anyone here. I took a quick look on aliexpress, and here's a janky looking standalone pocket scope with probe for under $40 US that goes to 2mv and 1us - https://www.aliexpress.com/item/32836776371.html Who wants to be a guinea pig?
 
You must log in or register to reply here.

Similar threads

Back
Top