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

[marciolsf]

I may not be as big of an expert as @squeept or @RIP-Felix, but I've sided with the "tantalums are not a permanent fix" team. Why? Well, I've spent close to a year working on my cecha and was an early adopter of the syscon method. At squeept's urging, I bought a cheap pocket scope and while it wasn't as good as a full bench scope, it was good enough to test most components to some degree or another. I then went ahead and tested and documented every single IC on the board, broke and replaced a few caps and fuses, tried multiple versions of the tantalumns, bridge wires, and it always, ALWAYS came back to the same error -- the dreaded 3034.

At this point, it just became a matter of simple elimination. There's absolutely nothing else wrong with my board, so it has to be bga. I debated getting it reballed just to rule that out as well, but instead I bought a near-mint 2501A, stashed away my cecha01 and haven't looked back. I might get my fatty reballed at some point, and I'm fairly sure it'll work ok after that, but for right now I'm good.

 

[ElGris]

I may not be as big of an expert as @squeept or @RIP-Felix, but I've sided with the "tantalums are not a permanent fix" team. Why? Well, I've spent close to a year working on my cecha and was an early adopter of the syscon method. At squeept's urging, I bought a cheap pocket scope and while it wasn't as good as a full bench scope, it was good enough to test most components to some degree or another. I then went ahead and tested and documented every single IC on the board, broke and replaced a few caps and fuses, tried multiple versions of the tantalumns, bridge wires, and it always, ALWAYS came back to the same error -- the dreaded 3034.

At this point, it just became a matter of simple elimination. There's absolutely nothing else wrong with my board, so it has to be bga. I debated getting it reballed just to rule that out as well, but instead I bought a near-mint 2501A, stashed away my cecha01 and haven't looked back. I might get my fatty reballed at some point, and I'm fairly sure it'll work ok after that, but for right now I'm good.

The only real question here is how much a BC PS3 could last with a full BGA change AND a full tantalum replacement. Even if they're chinese or not, even if you use only a couple of jumpers or no.

I got many tantalums destroyed for no reason while I was using some fats, but there's could be a reason for that: Too much heat from my heatgun, and bgas trolling people since 2000s. We still don't know if a fat fails after a fix exclusively due bgas or no. Nobody tested this yet, so we are not sure if tantalums killed those mobos or bgas got damaged (or healed) after applying heat to the board.

Maybe this fix works, and it does really fine, but we are not gonna find a definitive answer until we don't see a full reballed BC CECHmodafkingA with all NECs reaplaced and many months of testing.

 

[Pacorretaco]

I may not be as big of an expert as @squeept or @RIP-Felix, but I've sided with the "tantalums are not a permanent fix" team. Why? Well, I've spent close to a year working on my cecha and was an early adopter of the syscon method. At squeept's urging, I bought a cheap pocket scope and while it wasn't as good as a full bench scope, it was good enough to test most components to some degree or another. I then went ahead and tested and documented every single IC on the board, broke and replaced a few caps and fuses, tried multiple versions of the tantalumns, bridge wires, and it always, ALWAYS came back to the same error -- the dreaded 3034.

At this point, it just became a matter of simple elimination. There's absolutely nothing else wrong with my board, so it has to be bga. I debated getting it reballed just to rule that out as well, but instead I bought a near-mint 2501A, stashed away my cecha01 and haven't looked back. I might get my fatty reballed at some point, and I'm fairly sure it'll work ok after that, but for right now I'm good.

The sad thing is... Even syscon can't really tell you it's a BGA defect. The issue doesn't necessarily have to be "under" the chip. Nothing is guaranteed, and people get too focused on the lesser of the bad guys, while the meaner 2 are still laughing.

But thanks to @squeept we have reasons to believe it can very well be the case.

@ElGris ... You say it yourself. No need to use heatgun if you only want to mess with the tokins.
When you do that, maybe the job appears easier, but you make troubleshooting harder, because you will never know what happened.

People don't mention this enough, but even a measly 50 degrees C is enough to make or break a precarious RSX connection.

 

[marciolsf]

The sad thing is... Even syscon can't really tell you it's a BGA defect

Yep, that's true... What I really hope is that one day we'll manage to reverse-engineer the syscon and add debugging back. From what I understand, debugging is present in the development syscon, but it's removed once it goes to production. With that, we should get a much better understanding of what is going on, and maybe also understand some other "mysteries", like bit-training, etc.

Short of that, the best we can do is correlate specific errors with specific hardware failures, which is exactly what @squeept is doing. I was able to contribute to that a bit, with my very expansive set of 1 unit to work with, but what we really need is lots and lots of machines to go through and document.

 

[squeept]

Unfortunately, I bought a lot of previously butchered consoles recently so I'm not getting good information. I'm going warranty seal hunting later.

 

[MadMaxGR]

Guys a noob question and little off subject. I got a ps3 fat A01 on hand. It has already talantum capacitors on the top side of the board next to GPU. I think it was wrong TOKIN replaced, so i will try to replace the ones beneath the GPU.

In this case do I have to put wire bridge on any side if the TOKIN of CELL CPU are intact?

Sent from my MI MAX 2 using Tapatalk

 

[Revak3115]

People don't mention this enough, but even a measly 50 degrees C is enough to make or break a precarious RSX connection.

50 degrees.. i really doubt that, do you have proof? the RSX runs at over 50c by design so that doesn't make sense to me.

 

[RIP-Felix]

In this case do I have to put wire bridge on any side if the TOKIN of CELL CPU are intact?

I'm going to get alot more technical than your question. So in short, up until now I have reccomend that at least 2 Tokens should stay for continuity. You should start adding bridge wires immediately, but defiantly after 2 IMO.

However...and BTW this is not a noob question at all...I just got thinking about this myself after watching a totally unrelated video on why coax is 50 ohms. The guy's voice put me to sleep about halfway through, so I can't remember why coax is 50 ohms, but it made me realize there could be an inductance mismatch that leads to reflections. The tantalums and bridge wires we add to the circuit have some inductance of their own. It probably doesn't match the tokins we replaced. It's probably higher. So now you could have a inductance mismatch between the filter and CPU/RSX, leading to reflection. I wonder if this could shorten the life of the tantalums and GPU/CPU? I'm just leaning about it now, so this it probably complete bS, but there's defintely alot more to it than just capacitance.

If anyone recalls when I first joined this forum back on page 137 that PS3#1 shorted when I only placed one conductor after removing all the tokins with TaPol caps.

Spoiler: This was the result

I assumed that was simply a matter of the amount of AWG22 solid core conductor not being enough to handle the current the chips draw. It's minuscule resistance (impedance) was enough to heat it to the point the solder vaporized and no longer conducted electricity, causing a YLOD (yes, the tantalums allowed that A01 to boot after each and every attempt, making me think it was going to work). Now, the NEC/TOKINs have a lower impeedance to a larger frequency range than any other solution, which is why they are desirable. They probably don't heat up as much as a TaPol array would. I'm wondering if there's any way to be sure our tantalum/bridge wire impedance matches the tokens they replace, along the entire frequency range they would expect to be exposed to. Anyway, it's just more reasoning for why I think the best replacement for a proadlizer is a proadlizer and why this Tantalum replacement idea is a bad one.

 

[Pacorretaco]

50 degrees.. i really doubt that, do you have proof? the RSX runs at over 50c by design so that doesn't make sense to me.

Yes, maybe I'm going with extreme example but the key word was "precarious".

I had a console that would artifact incredibly predictably as the GPU reached about 60ish degrees. Then if let cool down, it would work again until this temperature is reached again. To the point where I decided to put a K type thermocouple and I could watch the temperature rise and clench my butt at the the precise moment the image would freeze.
Maybe not what you'd call science but @squeept has also a video of a console that would respond to him tapping it with a finger on the top.

The point is, these types of failures are incredibly fiddly. Be it BGA or uglier things, they can behave unpredictably with the slightest changes.

So if you use hot air you aren't always just doing what you think you are doing. That's what I was trying to say

 

[MadMaxGR]

I'm going to get alot more technical than your question. So in short, up until now I have reccomend that at least 2 Tokens should stay for continuity. You should start adding bridge wires immediately, but defiantly after 2 IMO.

However...and BTW this is not a noob question at all...I just got thinking about this myself after watching a totally unrelated video on why coax is 50 ohms. The guy's voice put me to sleep about halfway through, so I can't remember why coax is 50 ohms, but it made me realize there could be an inductance mismatch that leads to reflections. The tantalums and bridge wires we add to the circuit have some inductance of their own. It probably doesn't match the tokins we replaced. It's probably higher. So now you could have a inductance mismatch between the filter and CPU/RSX, leading to reflection. I wonder if this could shorten the life of the tantalums and GPU/CPU? I'm just leaning about it now, so this it probably complete bS, but there's defintely alot more to it than just capacitance.

If anyone recalls when I first joined this forum back on page 137 that PS3#1 shorted when I only placed one conductor after removing all the tokins with TaPol caps.

Spoiler: This was the result

I assumed that was simply a matter of the amount of AWG22 solid core conductor not being enough to handle the current the chips draw. It's minuscule resistance (impedance) was enough to heat it to the point the solder vaporized and no longer conducted electricity, causing a YLOD (yes, the tantalums allowed that A01 to boot after each and every attempt, making me think it was going to work). Now, the NEC/TOKINs have a lower impeedance to a larger frequency range than any other solution, which is why they are desirable. They probably don't heat up as much as a TaPol array would. I'm wondering if there's any way to be sure our tantalum/bridge wire impedance matches the tokens they replace, along the entire frequency range they would expect to be exposed to. Anyway, it's just more reasoning for why I think the best replacement for a proadlizer is a proadlizer and why this Tantalum replacement idea is a bad one.

Basically I do not need the bridge wires in my case, since the CPU TOKIN are intact.

Right?

Sent from my MI MAX 2 using Tapatalk

 

[Pacorretaco]

Basically I do not need the bridge wires in my case, since the CPU TOKIN are intact.

Right?

Sent from my MI MAX 2 using Tapatalk

Not right.

There's one tokin array for CPU, and another one for RSX (GPU).

Not sure if I'm interpreting your situation correctly but,
If you removed all 4 tokins from RSX, then yes you need to restore the continuity, hence the bridge wires.

 

[MadMaxGR]

Not right.

There's one tokin array for CPU, and another one for RSX (GPU).

Not sure if I'm interpreting your situation correctly but,
If you removed all 4 tokins from RSX, then yes you need to restore the continuity, hence the bridge wires.

Yes removed all four. So I need to run bridges on both sides.

Sent from my MI MAX 2 using Tapatalk

 

[RIP-Felix]

Sorry, I assumed you knew the RSX has its 4 tokins and the CPU has its. They are electrically isolated.

Yes, the more bridge wires the better. Tokins are internally connected along the + rails, but they conduct alot of current. All four share the load. Forcing all that current through fewer tokins will shorten the remaining tokins lives. So bridge wires help spread the load.

 

[DeadEnd]

In addition to solder issues and capacitors, what about micro fractures in the layers inside the board due to thermal stress? With uneven heat changes and all, the board warps/expands/contracts. So how can anybody know 100% that there are no issues with the internal layers, where a certain trace may be slightly breaking and eventually losing contact, which in turn causes strain to RSX... Just a theory.

https://www.comsol.fi/paper/fractur...-due-to-thermal-stress-on-soldered-pins-15533

 

[marciolsf]

The only real question here is how much a BC PS3 could last with a full BGA change AND a full tantalum replacement. Even if they're chinese or not, even if you use only a couple of jumpers or no.

I meant to follow up on this sooner -- that's the other question, isn't it? I think Squeept said he's using regular leaded solder, which from his experience is working well so far -- how many years would say, @squeept ?

My guess is that it's more likely that the tantalums will perish long before the reball does, based on the research @RIP-Felix has done indicating that the tantalums are not necessarily a close enough match to the tokins.

 

[RIP-Felix]

To be clear, my research was to find the best case scenario for tantalums. My array tested as good in my YLOD console as the Tokins on a stock working console. However the console needed a reball, so I was not able to thoroughly test it. In theory, TaPol caps should be fine, but it's not how the filter was designed by SONY from the begining. One would do things differently if you were designing around tantalum from the begining. Given the unknowns and untested nature of using Tantalum long term, I still say Tokis are the best replacement

As for lead solder. It has better elasticity. It's more durable than lead-free. It should last longer than the console would otherwise have. Having said that, if it lasts 15 years, that would be a miracle. The bumps under the die are still lead-free and susceptible to failure, other components could go bad, the heat from the reball is hard on the chipset, so long term reliability is hard to estimate. It's definitely not going to last as long as your N64, SNES, Genesis, Wii, and so on.

 

[squeept]

Yeah, the correct answer is.... no goddamn idea. I've had a few reballed consoles come back after a few years, but for the most part I never hear from the customers again. I'm guessing that these systems don't get a ton of hours these days like when they were new, so there's that, too.

There's mountains of academic junk about how long dies, bumps, and balls will last, but when it comes down to it, every device and chip is wildly different. There are just too many variables to make any real guess other than "it's going to crap out eventually." I'd be inclined to agree that 15 more years without needing a new RSX would be pretty shocking for any 90nm system that still gets used frequently, regardless of that console's history, maintenance, cooling modifications, or repairs.

 

[Pacorretaco]

Yeah, the correct answer is.... no goddamn idea. I've had a few reballed consoles come back after a few years, but for the most part I never hear from the customers again. I'm guessing that these systems don't get a ton of hours these days like when they were new, so there's that, too.

There's mountains of academic junk about how long dies, bumps, and balls will last, but when it comes down to it, every device and chip is wildly different. There are just too many variables to make any real guess other than "it's going to crap out eventually." I'd be inclined to agree that 15 more years without needing a new RSX would be pretty shocking for any 90nm system that still gets used frequently, regardless of that console's history, maintenance, cooling modifications, or repairs.

This reminds me btw, do you normally care to check the "usage time" which is displayed on WebMan settings?
I'm not sure how relevant or even accurate this may be, but the longest I've seen on a 90nm RSX is 250 days on a H model, still working with the warranty seal intact (and Cell needing delid, which may be significant).

Coincidence or not, about all 5 C model consoles that I could check, had begun giving problems at around the 150d mark.

Maybe good thing to check on those consoles that come back, checking before and after. It should show how much they were actually used before coming back.

(Btw This is just some recent underdeveloped thought of mine but, could it be possible that the 90nm RSX present on H models is somehow revised or something? After all, they were the last ones to incorporate it.
But of course If I'm not mistaken it's not like H models aren't prone to YLOD. They very much still are, so faint hopes there)

 

[Pacorretaco]

I meant to follow up on this sooner -- that's the other question, isn't it? I think Squeept said he's using regular leaded solder, which from his experience is working well so far -- how many years would say, @squeept ?

My guess is that it's more likely that the tantalums will perish long before the reball does, based on the research @RIP-Felix has done indicating that the tantalums are not necessarily a close enough match to the tokins.

Well yes, we don't really have that much information about this sadly. And I'm in no way qualified to make claims about this.

But, maybe It would be a good time for the man @Naked_Snake1995 himself to chime in.
He had a C model with tantalums, and we are not here to deny his experience.
He reported success, and not exactly short term. I personally wouldn't be that surprised if his system is still running fine.

Probably from that, after all, he must have taken the strong conviction needed to make the original post... and no reasons to deceive anyone.

There's still much value in the original post, besides being overly exaggerated and containing a couple of very questionable affirmations. Which may make it difficult for newcomers to see the bigger picture... at least before they do irreversible alterations to their systems.

Cheers

 

[squeept]

This reminds me btw, do you normally care to check the "usage time" which is displayed on WebMan settings?

I actually don't do anything with custom firmware on these systems, I sell them bone stock. Since it's my job and not just a side hustle, I can't risk getting kicked off of any sales channels or getting sued. They target the "for profit" folks. I know as long as I don't include any pirated software I should be fine, but it still scares me. I think the newest thing I've ever sold with any kind of jailbreak / modchip is a PS1. Figured that was old enough that nobody would bother...

Maybe I can time myself doing an install then revert to see if it makes sense to add to the workflow. It would be nice to check that number and look over the temperatures during stress testing.