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

[marciolsf]

Well yeah, but it's not easy to explain the results in a way that would be uncontroversial. I was planning to make this a 200th page review post, but what the hell...

118 Users have attempted to fix 199 consoles on this thread. The actual number is probably higher, but they have only provided details on, claimed success on, or specifically mentioned 199.

• 26% (52/199) didn't say what model they had. One user said multiple of his consoles were backwards compatible but didn't specify the model. I had to put those in the "didn't say" category.
• 27% (54/199) were COK-001 (A/B models).
• 21% (41/199) were COK-002 or COK-002W (C/E Models).
• 7% (14/199) were SEM-001 (G models).
• 5% (10/199) were VER-001 (L, M, P, & Q Models).
• 4.5% (9/199) were DIA-001 (H Models).
• 4.5% (9/199) were DIA-002 (J & K Models).
• 2.5% (5/199) were SUR-001 (21xx Models).
• 2% (4/199) were DYN-001 (20xx Models).
• 0.5% (1/199) was KTE-001 (30xx Model).

166/199 consoles were Tantalum replacements (83%)

• 19% (31/166) were confirmed to work at least 2 weeks later.
• 55% (92/166) were last confirmed a YLOD.
• 25% (42/166) I marked down "inconclusive.

Inconclusive means the user claimed positive results, but never reported back to tell us if it kept working or not. Or after the replacement there were artifacts indicating BGA defects, false positives, or instability not worthy of calling it a positive result. Or the console was last reported working, but ~2 weeks have not yet past since the last update.

Before you argue marking consoles inconclusive is unfair, let me respond. Perhaps I should take everyone at their word when they say, "I did the tantalum thing and it worked!!!" Even if I did the inconclusive results + confirmed Positive results = 44% of the consoles tested. That represents all the consoles that claimed positive results, regardless of other factors. It assumes none of them will fail in the next few days when the board relaxes from the hot air they used to install the tantalums (false positives). It assumes that every positive case lasted, was thoroughly tested, and never got sold on ebay to someone who doesn't know about this thread. It assumes we got an update up on the progress of every console. Even if you accept all that absurdity…45% success rate still means the tantalum mod DOES NOT FIX the majority of consoles.

Now let's get real! Many of the inconclusive consoles probably did eventually fail, but we never got an update. Maybe it was sold. Maybe it got stored after "fixing." Another thing to note is the effect of "adulation." Users will comment on this thread simply to post positive results and thank the OP (hero worship). It's a type of response bias and will have a large influence on the statistics above, driving the percentage of success up.

On the other hand, many of the people attempting this mod are doing so with inadequate skills and equipment. Many use low quality capacitors, have cold solder joints, use too thin a gauge of bridging wires, don't troubleshoot the board before starting, and so on. These factors reduce the likelihood of a successful outcome. However, since anecdotal evidence from technitions like @squeept, @botakompong, @vyktormvmpay25 suggest that reballing is necessary in 90% of cases and that the tokins are very rare. These people have the equipment and skill to properly troubleshoot the console before attempting a fix. So if 90% are BGA defects or RSX issues, then any unskilled work in the area near the processors has a high chance of causing a BGA false positive. This means if you take any YLOD console and try the tantalum fix, it probably will work temporarily. But you won't know if it's a false positive or true fix until the console YLOD again. That could be in a few minutes, days, weeks. Generally if it lasts longer than a month with frequent use, it may have worked. But a false positive could revert in mont to a year too. It just depend on a lot of factors. This is why I always recommend you start with the SYSCON error codes first then proceed to troubleshooting. Then you know what was wrong to begin with and you're not chasing gremlins for months to come.

It's impossible to quantify the variables. However, ask yourself this, if 90% of YLOD consoles could be fixed with tantalum capacitors, is a 45% reported success rate reasonable? I think you could argue this either way. And that's why these results aren't very useful. I have been able to identify a number of myths...

Myths:

1. Tokins are prone to fail.
   • On Page 168 I wrote "A Retrospective Analysis of the NEC/TOKIN Proadlizer" in which I explored this topic in depth. I couldn't find any support for the idea that the NEC/TOKIN proadlizers are a defective technology. One would think that there would be ample information online about them, yet there is very little. No recalls, no technical analyses showing them to fail prematurely, nothing but the product information and the rumor. Either the internet was cleansed of the negative information, or it's a baseless claim.
   • SONY Replaced the NEC/TOKINs with Tantalum capacitors, so they must have been defective. That could be true. Or, they weren't needed. The slim models began moving to tantalum starting with the CELL side on the DYN-001 in 25xx models. It's important to note that model has HF bypass caps populated on the motherboard and 7x470uF TaPol caps. The HF MLCC are 30-40x 0.1uF on data lines connecting the RSX and IO, and 24x 10uF near the TaPol and on the CPU substrate itself (which previous models lack).That coincided with the move to a smaller processor manufacturing process. The Cell_BE CPU was reduced to 65nm starting with the SEM-001 (G models). G models still had the 90nm RSX GPU and both have NEC/TOKINs, but the CPU's MLCC bypass capacitors are unpopulated. This may mean that the lower power draw or smaller mnf. Process reduced the need for them. Or they were deemed unnecessary and removed as a cost saving measure. Also, the number of switching DC/DC VRM was reduced from 3 to 2 on the RSX. That changes it's RLC characteristics and suggests that the only reason for the extra one was for PS2 HWBC. Also the APS231 power supply was more efficient since it didn't need to operate as close to its maximum output power (285W). Therefore, G models should be more reliable. If they are not, then maybe the tokins are failing.
2. Heat Gun Trick (this does not tell you anything. If it does work, it more likely means you have a BGA defect requiring a reball/reflow to fix)
   
3. Tantalum capacitors are the "real" Fix for the YLOD. Reballing isn't necessary in the majority of cases. Total myth.
   
4. Electrolytic capacitors work (they don't. If they allow the console to boot, their high esr will burn them out soon afterwards. It's not a long term solution.)
   
5. Instant YLOD always = Short (a shor can cause an instant YLOD, but not all instant YLOD's are due to shorts)
6. A Random/Intermittent YLOD that occurs in games at certain points MUST be Bad NEC/TOKINs.
   • This one "seems" particularly plausible, but again all or nothing statments like this are never true. While this is the use case for the tantalum mod and it increase the likely hood that the tantalum mod would fix your YLOD, you should still get the SYSCON error codes to rule out a 3034 that would indicate a BGA defect. And if there are 1002 errors to indicate the tokins, then you can try this mod with fair certainty it is appropriate.

You see the problem is there's a response bias when the "culture" of the forum is pro tantalum fix vs reballing. Since the thread has shifted back and forth, this creates a situation where people only report success or failure to confirm the prevailing viewpoint. Our brains fail to equally weigh opinions that do not conform to what we expect or believe with those that do. This leads to a bias favoring expected results. "I reject your reality and substitute mu own!" Psychologists and neuroscientists call this "confidence bias." You might know it as seeing the world through rose tinted glasses, wishful thinking, living in our own reality. Basically, the fact people want so badly for this to fix their console, is why they tend to see the console's behavior in a way that tells them they should replace the NEC/TOKINs. Once they find evidence that can be explained by their expectation (pet theory), they see that as confirmation, Instead of evaluating how well the evidence fits the competing theory also.

The result of confidence and confirmation biases on this thread have artificially inflated the number of successful posts. People don't always respond unbiased. And I have to sift through the thread and infer the truith from what little details I'm given. The problem is that any work being done next to the RSX or CPU is likely to cause a false positive simply due to proximity. That and mounting pressures are different when you put the thermal paste and springs back on. There are too many variables to make it possible to decide. So I had to take them at their word most of the time. And because I can't separate the BGA false positives from actual success, the numbers aren't very useful.

...But yes. Technically speaking, 45% of the people who replaced their NEC/TOKINs with Tantalum caps "reported immediate sucess." That number doesn't include false positives that later failed. So at best, this mod is unlikely to fix you YLOD. More likely, it's very unlikely to fix your YLOD. The only way to know for sure is to get the SYSCON error codes.

These are fascinating results, and the 45% "success rate" (if you can call it that) is better than I expected. Great work!!

I still don't think the mod is worth the trouble, however, for the majority of the users out there. Here's a few reasons, as everyone who has tried this has learned :

• The mod is difficult to get right
• Adequate parts are not cheap
• Proper tooling is expensive (good iron, good heat gun aka not harbor freight, fancy scope)

All of the above, for 45% chance it'll work, at best. These are not very good odds, specially for the amount of work required and money spent.

 

[Pacorretaco]

These are fascinating results, and the 45% "success rate" (if you can call it that) is better than I expected. Great work!!

I still don't think the mod is worth the trouble, however, for the majority of the users out there. Here's a few reasons, as everyone who has tried this has learned :

• The mod is difficult to get right
• Adequate parts are not cheap
• Proper tooling is expensive (good iron, good heat gun aka not harbor freight, fancy scope)

All of the above, for 45% chance it'll work, at best. These are not very good odds, specially for the amount of work required and money spent.

45 is not the "success rate". It's the rate at which people report "success" here.
Now this 45% Is actually worse than I expected.
Why? 45% is still a terribly inflated figure for the reasons mentioned before.
Namely, people who come here to post, are way more likely to post success, (or apparent success) and simply not talk about the failures.

Take me, for a funny example. I never believed this crap (ok maybe a little at first, but not after I joined this place).
But actually In that spreadsheet my success rate should be 100% !
Because even I only talked about the confirmed tokin faults I found and solved.
1 VER-001 board and a DYN-001 board.
Both are still working to this day. They were proper tokin faults and the success is genuine (and without removing old tokins mind you). 2/2= 100% right?

But I didn't come here to post about the rest of the boards I had where it clearly was not because of the tokins. And believe me everyone wants the myth to be true. Even I did a bunch of fake tokin fixes at first. But even those are not on the spreadsheet. They are in my closet.
And of course the ones where I didn't even touch the capacitors because I already knew they were not the problem.

I don't come here to talk about those.
In fact nobody who actually knows what they are doing normally would do that.
You don't see proper technicians here posting about every perfectly working tokin that they find.

Our data sample comes from the excited amateur, who posts success because they are happy. And posts failure if and because they think they did it wrong, need help or simply are disappointed!

Not from unbiased technicians...
They just stay out of this drama and keep working.

The % you are talking about is probably closer to 3% on old COK boards (the ones where most people want to try the magic on)

 

[RIP-Felix]

I thought it was interesting that 48% of the consoles people attempted to fix were backwards compatible A-E models. COK-001 and COK-002 motherboards are by far the most desirable consoles and are the consoles over represented in my DATA. They are the consoles with 90nm RSX GPUs and 90nm CELL BE CPUs too. That makes them the hottest and most prone to solder ball cracks. They draw the most power, which makes them more prone to voltage spikes that can cause fuses to blow and caps to short. In other words, the consoles people tried to fix the most, are the consoles most likely to have other problems.

On the other hand, J through 20xx models with 65nm RSX are less likely to have experienced a BGA defect and should therefore have a higher tokin to BGA failure Ratio. Some users have stated they tend to find more of these consoles with long YLODs and bad tokins. @ElGris has been outspoken about these models having a much higher chance of being fixed by replacing tokins. But these consoles are very under represented in this thread. Only 16% were J-20xx models.

Another interesting thing to note is that J-20xx models having a higher tokin failure rate bring up the total +% rate. 12/22 had a reported success rate. That's still only 55%.

If I only include backwards compatible consoles, models A-E and those consoles people said were backwards compatible, but didn't specify the model, then the number looks better. 55/95 = 58% of backward compatible consoles reported "initial success" or "Inconclusive results." That is, they turned on immediately after the mod, or went to a GLOD, or had freezing, etc. However, after 2 weeks of use that number fell to 39/95 = 41%. Remember, this 40% include consoles with GLOD, artifacts, freezing and other problems you wouldn't consider a "fix." The "confirmed" number of "fixed" backward compatible consoles is 17 (total), and they were fixed by 9 users. 186 pages and 2 years later, only 9 users have confirmed 18 consoles were working at least 2 weeks after installing tantalum capacitors. This number doesn't include the people who didn't come back, but who's console may have actually been fixed and I couldn't confirmed it. There are a lot of people who never came back to confirm the console was still working.

Okay, now we mire deep into the motivation of the reported sucesses, by calling out individual users. Beware, contoversy ahead! One of the 9 users confirming 4x BC consoles are still working was @Yugonibblit on page 129. He claimed fixing 10 consoles in response to a critical post by @Blubberz. This was during a time where people were still defending the OP and the Tutorial. So the culture of the thread was supportive of anti-critical statements. Users that had success were quick to come to the OPs defense. People were still investigating the mod, actively helping people perform the mod, and explaining their failures in the biased context of the"myth." In other words they saw evidence of false positives and assumed it meant the user must have soldered the caps on wrong, or didn't use enough bridging wires, or needed more tantalum, different tantalum, have a short somewhere, knocked a MLCC off the board that must have been important, anything to avoid the truth. Namely, that tutorial didn't work 100% of the time.

@Yugonibblit claimed success on 10 consoles, but only 5 were BC models he specifically mentions in his posts. On page 63 he mentions a G model and on Page 99 he mentions a B model. The other 8 are only mentioned in the retort post on page 129. I'm only picking out this one example to make a point. This is the type of post we tend to hear about these "other" console that are supposed to exist. They were not documented on the forum previously! So that begs strange credibility. They suddenly appear in defense of the OP and the Tutorial, and we're supposed to take himat his word? That doesn't mean that I ignored what @Yugonibblit claims. I took him at his word here and marked down 10 consoles. 1x B, 1x G, 1x H, 4x "BC" (backwards compatible, the models weren't mentioned, only said they were BC), and 3x "Didn't say". These were updated with added context from multiple posts over the many pages of the thread and I tried my best to figure out what models these 10x concoles were, from posts lacking details and unique identifiers that would allow me to figure out which console he's talking about. I had to infer that 5 of them were BC models from context in later posts. Then I had to assume the B, G, and H models were from those 10 and not one of the other 20 he was later talking about. Maybe I did it wrong, probaly I made mistakes because of this kind of thing. It was a PITA trying to follow the breadcrumbs people left me. For example, here's one of my notes I use to keep track of this kind of thing:

"I will claim 10 consoles that I did the NEC/TOKIN repair and all are still functional." [This is the first we're hearing of the other 8, he only describes 2 consoles prior to making this claim.] Later…"Yes I have five [BC models] that work flawlessly, keep in mind all have been completely delidid and new thermal past also, only one needed reflow, also all were done with 32 tantulum's at 470uf. And solid copper wire at least 14 to 16 gauge for jumper wire[pg129]...I have fixed about 20 that I can remember but some were not repaired by just tantalums, I had to reflow a few after all tantalums were replaced, I'm running a cechh right now for at least 3 months that only needed 4 440 caps.[pg158]"

So that's how I decided that 5/10 of those consoles were BC. You'll notice that he's talking about reflows, which clearly indicate BGA failures and represent false positives. This is just a sampling of the type of garbage I had to sift through. And now he's talking about 20 consoles??? WTF, how am I supposed to keep track of them also? I decided he didn't give enough context of those other 20 console to include them in my spreadsheet. So I stuck to the 10 he was talking about before. I assumed 1 was the B, 1 was the G, 1 was the H, 5 were BC, and the rest were "didn't say." The reflow I didn't include as a positive result. The other four I based the 2 weeks confirmation on this one post. And I took the same post as confirmation on the B model, which marked 54 days since the initial report that the reflow was a success. That was the reflow. I think it's generous that I took him at his word on the other 4 "BC models" that suddenly appeared on page 129. Of course if he wants to come back and give me details on those 10 consoles, or the 20 total he was boasting about, I'll gladly update my spreadsheet. I have solicited the thread multiple times for these types of details from users reporting success. Few times have I received a response.

My reason for calling out @Yugonibblit isn't to make fun of him. It's to give an example of the kind of posts that made it into the 17 total consoles that were reported to have lasted more than 2 weeks. One of those was the OPs original C model. I'm not going to call anyone else out, but the other 7 users claiming success here are not super believable either. So we're really talking about very few credible success, and some are relying on "sucesses" people claim when they're coming to the OP's defense. Clearly bias is having an effect on the numbers.

I wonder how many people saw this Tutorial and bought a YLOD console off e-bay thinking they could flip it for some bank! And I wonder how many of them did thorough testing before selling it! Worse, I wonder how many poor saps on the internet bought a working console only to have a YLOD after 2 weeks! This is why you only buy sealed consoles people!

 

[Coro]

I am wondering if bad capatitors are causing my issue...

if i let my fat ps3 warm up, i rarely have a problem. but if i try to start Red Dead Dedemption too soon (running from iso on internal HDD), the ps3 will usually turn off suddenly with flashing red light afterwards. i also had a WWE game do this twice at the same menu section. i assumed an aging hard drive at first but have not checked it (it needs to be replaced with bigger drive anyway).

replacing caps is probably beyond my skills but i may be able to find someone locally to make the repair (using information from here if they are not familiar) if it is indeed a possible cause.

PLEASE HIT REPLY if you answer me, i fear i may forget that have posted here otherwise and miss a response. THANK YOU!

 

[RIP-Felix]

I am wondering if bad capatitors are causing my issue...

if i let my fat ps3 warm up, i rarely have a problem. but if i try to start Red Dead Dedemption too soon (running from iso on internal HDD), the ps3 will usually turn off suddenly with flashing red light afterwards. i also had a WWE game do this twice at the same menu section. i assumed an aging hard drive at first but have not checked it (it needs to be replaced with bigger drive anyway).

replacing caps is probably beyond my skills but i may be able to find someone locally to make the repair (using information from here if they are not familiar) if it is indeed a possible cause.

PLEASE HIT REPLY if you answer me, i fear i may forget that have posted here otherwise and miss a response. THANK YOU!

Start with the SYSCON error codes. They are easy and cheap to get and will tell you if the TOKINS are likely at fault. Based on what you've said, it could be the tokins, but you could also try another PSU. 2 users having similar issues were able to solve it replacing the PSU alone. 80 1002 are the error codes we believe indicate a tokins replacment. If there is a 3034, you need a reball/reflow.

 

[marciolsf]

186 pages and 2 years later, only 9 users have confirmed 18 consoles were working at least 2 weeks after installing tantalum capacitors.

I work with data for a living, and I wish some of my devs were as thorough in their data analysis as you did here!

As for the analysis itself, I think it's also telling that the majority of the "confirmed" were all by the same 9 people. If the fix were truly that universal, we should have seen at least a larger number of users reporting more success, and I don't think it's all due to incorrect ESR. We've seen every possible combination of tantalum replacement (I myself went through 2 different sets of caps), so there should've been a greater success rate than what we got.

 

[Coro]

Start with the SYSCON error codes. They are easy and cheap to get and will tell you if the TOKINS are likely at fault. Based on what you've said, it could be the tokins, but you could also try another PSU. 2 users having similar issues were able to solve it replacing the PSU alone. 80 1002 are the error codes we believe indicate a tokins replacment. If there is a 3034, you need a reball/reflow.

i had neglected to consider the psu but i will have to look for someone with a compatable one to try. i have another fat but appears to be incompatable on devwiki.

i can not investigate the syscon now due to lack of hardware. i need a flasher to fix a 25xx so i might as well get that first since i can then use it to experiment with new things.

thanks for the reply!

 

[DeadEnd]

Victor just replaced tokins on a dyn001 board that had a 1002 fault. Seems to have solved the problem indeed. The lowest value from a tokin he measured to be 981 uf, others had 998, 995, and 1003. Perhaps 981 uf is too low for a tokin in that case. Before replacing caps, he also reballed Cell and RSX, yet It had random YLODs. The console sometimes started, sometimes did not.

 

[vyktormvmpay25]

Fix and few other things are inside this link
http://s.go.ro/9zaqzt7z
This is one rare case, I did not had any of 1002 before in UART log, all like 3034, 4041/42 before reball cpu and gpu , then all ffff after assembled and tested in a day. Second day ylod, removed hdd forced start with fan cleaning test(eject keep preparing in the moment of ac start , third time will start, then running fine all day. Third day it didn't work at all. Did not ever did this before exchange in a slim, could be board that was like in a wet environment.
It was 2100mf what I seen on board, removed only plastic and seen that oxid taught is better to exchange. After exchange I get like 5000uf reading on board, 3000uf on cpu, not very accurate on board but it was started straight.

 

[Pacorretaco]

Did not ever did this before exchange in a slim, could be board that was like in a wet environment.

So you are reporting this is the first tokin fault you found ever?
In hundreds of slim boards...
Even I thought this was not THAT rare.

I found and solved only one too, but I only had 1 slim from family member instead of hundreds because I'm not not doing this professionally like you.
But now sounds like I was just "lucky"

 

[vyktormvmpay25]

Yes first time, even seen long time this kind of reports, always checked out on board after desolder, gpu and cpu, got similar values, like 2500uf on gpu line and 2800uf cpu line, second day 2100 on gpu on board.
So yes it can happen.
I always hate phats, now have to give extra care to those nec's and I will always open while on board.

 

[Pacorretaco]

Yes first time, even seen long time this kind of reports, always checked out on board after desolder, gpu and cpu, got similar values, like 2500uf on gpu line and 2800uf cpu line, second day 2100 on gpu on board.
So yes it can happen.
I always hate phats, now have to give extra care to those nec's and I will always open while on board.

Interesting. I wonder how the capacitors looked in my slim with this problem.
(Here is my post talking about it btw:
https://www.psx-place.com/goto/post?id=284538#post-284538)

Because I didn't remove them, in theory I could still measure the aggregated capacitance on-board and also remove the plastic to see if they look oxidized at the edges like yours. Because we don't even know if it's significant or just coincidence. After all, very few real, credible cases have been reported, and fewer still with meaningful experiments/tests.

Then again, because the machine is still working perfectly, I don't think I will run and do that yet. Unless I change my mind for some reason of course.

 

[vyktormvmpay25]

As long as you don't get random ylods like I got, then nothing to worry. Like I've said it was king of wet environment and radiator was starting as aluminium white oxid. Changed that.

 

[MYS]

Just to update on my result:
2x CECHE (1 US and 1 Jap) still working since I change to Tantalum capacitors.(over a year at least to date)
1x CECHA still working since I change to Tantalum capacitors (over a year at least to date)
2x CECHH is a failure. Both work for about 6 month after changing to Tantalum capacitors then YLOD. Replaced with "new" Tokin from AliExpress, worked another 3 month (didn't record the actual duration) but YLOD again. Now nothing works. I don't have the tools to do re-ball so leaving them as it is till I find the mood to "play" with them (check syscon)

 

[RIP-Felix]

Victor just replaced tokins on a dyn001 board that had a 1002 fault. Seems to have solved the problem indeed. The lowest value from a tokin he measured to be 981 uf, others had 998, 995, and 1003. Perhaps 981 uf is too low for a tokin in that case. Before replacing caps, he also reballed Cell and RSX, yet It had random YLODs. The console sometimes started, sometimes did not.

Good to know. Yes as @Pacorretaco said, we've rarely had a confirmed bad tokin. In the past, I have said that reballing first is really the only way to rule out a BGA failure, but now that we have ssen the 1002 error in at least 2-3 consoles with confirmed bad tokins, I am ready to say that the SYSCON can confirm bad tokins without a reball first. That's exciting!

@vyktormvmpay25 A DYN-001 (20xx model) is the only model of PS3 that has both TaPol (CPU side) and Tokins (RSX side) onboard. Can you confirm the CPU side has TaPol caps? They should be Black/grey caps with a white/silver bar at the positive end (with a + inside). The silk screening should be oriented in portrait. AlPol have a solid grey + bar and the markings are oriented landscape (like the caps you used to replace the tokins).

The reason I ask, is because the pictures I used on the DevWiki are all I had available to me when I was making up my model comparison spreadsheet, which lists these differences by model. This allowed me to see incremental changes in the filter design. I was hoping to infer information about the engineering thought process...

The NEC/TOKINs have great noise attenuation (rejection/reduction). Tantalum is the next best, but not as good. And AlPol is the next best to tantalum. Because I only based the spreadsheet off the pics on the DevWiki examples of each motherboard revision, it's possible that SONY used different caps (TaPol or AlPol) for different batches of 20xx models. If there are DNY-001 boards with AlPol, then that information disproves my Hypothesis.

My "Hypothesis" is they wanted the extra ESR performance of TaPol to ensure the new processors would work. First they changed the CPU to 45nm using TaPol, but left the RSX alone. When that worked, they removed the IOR sw VRM and changed the CPU filter caps to AlPol on the 21xx models. Then, the RSX to TaPol on the 25xx model. When That also worked, they changed to all AlPol on the 30xx models, now thet all the IOR switching VRM were gone (which are noisy). AlPol is cheaper and not a conflict mineral like Tantalum is. So if the VRM are less noisy and they can get good enough filtering with AlPol, that's a good move.

However, we need to be careful to attenuate the filter with our choose in caps based on which model we're working on.

If your DYN-001 has AlPol caps on the CPU side, it tells me that they didn't need the extra ESR performance of the TaPol caps in the first PS3 model to have a 45nm CPU, which is one of the only models that had Tantalum caps. The other PS3 models to have Tantalum caps was the JTP-001 (25xx models), which was the first to have a 40nm RSX. If these models can have either TaPol or AlPol, that's important to know.

This is why I recommend people always use TaPol instead of AlPol when replacing the Tokins. They are the best performing alternative. And I put 6x 0805 MLCC pads on my Tantalizer PCB so you can add 10uF and 22uF caps for added noise rejection in models that might need it. Or you can just add a bunch of extra caps to to squash the frequency response curve. It's wasteful, but works so long as you don't go too far over the total capacitance specification (4000uF for 0D108 Tokins and 4800uF for 0E128E Tokins).

 

[vyktormvmpay25]

On cpu they were white. Box full assembled, let on room cold in workshop, still 10° in night, in the morning still staring. Need to test to see if was bad joints. It seems not. Random ylods were relate to this 1002 error.
A simple test I did not did to heat board at about 80 and start till it gets around 50. I assume heating a little bit those nec's will start board if this error appears in UART screen very often.

 

[RIP-Felix]

Just to update on my result:
2x CECHE (1 US and 1 Jap) still working since I change to Tantalum capacitors.(over a year at least to date)
1x CECHA still working since I change to Tantalum capacitors (over a year at least to date)
2x CECHH is a failure. Both work for about 6 month after changing to Tantalum capacitors then YLOD. Replaced with "new" Tokin from AliExpress, worked another 3 month (didn't record the actual duration) but YLOD again. Now nothing works. I don't have the tools to do re-ball so leaving them as it is till I find the mood to "play" with them (check syscon)

I updated your positive results for the 2x E and 1x A models (first mentioned on Page 89)...

Just to share: I had fully replace NEC Tokin for 3 Ps3 (CECHA01 Jap, CECHE01 Jap and CECHE06 USA) with 32 piece of 470 uf Capacitor. All 3 set are working very well since then[...}

Those 3 consoles did make it into my Long term success column.

As for the H models, I previously only had you down for the one H model??? Are you sure you meant to say 2x H models? I don't remember seeing a second H model mentioned?

Spoiler: Sounds like you're only talking about one H model

Just to update, I somehow manage to confirm, capacitor replacement (tantalum) might not work with newer PS3 with 65nm process (eg. CHCEH model). I bought some NEC Token (0E128) from Aliexpress and replace all 8 pcs today and PS3 came back to life (have to remove 32 capacitor and used some solder paste to fix in the NEC token).
Newer process might need "cleaner" current to work thus direct capacitor connection might not work. I had tried 4 difference capacitor combination before deciding to test out replacing NEC Token.

I too tired 2R5TPE470M9 before but they didn't work on CHCEH models[...]I ordered more NEC Token from Aliexpress and plan to change all 3 set back to NEC token when I have the time.

I had succeed in changing 3 PS3 (backward compatible) NEC tokin (8x) with 470uf (32x) and 6 wires as bridge. But I couldn't repeat the magic with a CHCEH model. The motherboard is smaller and the NEC tokin point is much shorter so I had to remove some solder mask to place in the capacitor. I still couldn't get solve the ylod issue. Any idea apart from NEC tokin, what else could I try out? I already test out (by taking parts from another working unit) power supply, Blu-ray, fan, HDD and still ylod for this unit.

I think (based on wat i had done so far on 4 PS3), it as higher chance of working if you replace back using NEC Token. Direct tantalum cap might not work on these newer version of PS3 as the current might not achieve what is required by the processor.

Interesting to know how the SYSCON would show with all the Tokin changed to Tantalum cap (470uf). I had 3 BC PS3 with all Tokin changed to Tantalum cap (470uf x 32 pcs) and all are working quite well (at the moment), but not sure if such changes would slowly kill off the processor and gpu due to "noise". The other PS3 which is non-BC died to YLOD after 8 months of changing all the Tokin to Tantalum cap, which I do not know the reason as I don't have the equipment to check SYSCON. I tried replacing the Tantalum to Tokin I bought off Aliexpress and the system worked for a month plus and it died again (YLOD). Hope your result would share some light to the failure I encountered so i can bring that unit back into working condition (maybe).

 

[RIP-Felix]

On cpu they were white. Box full assembled, let on room cold in workshop, still 10° in night, in the morning still staring. Need to test to see if was bad joints. It seems not. Random ylods were relate to this 1002 error.
A simple test I did not did to heat board at about 80 and start till it gets around 50. I assume heating a little bit those nec's will start board if this error appears in UART screen very often.

Yeah, that matches the wiki photo and what I thought. Thanks for confirming that. You might want to consider adding 12x 22uF and 12x10uF MLCCs in future repairs of 0d108 tokins on RSX side of 20xx boards. They lower the impedance from 1-10Mhz to acceptable levels. This is just to better emulate what Sony did from 21xx boards onwards. Based on my oscilloscope measurements 100KHz - 10MHz is the range of switch mode noise generated by the VRM.

Here's the bad tokin waveform on my CECHA01 (PS3#7). Frequency of the switch mode noise is 485kHz:

And here it is after TaPol. CPU (Yellow) = 18x270uF TaPol and RSX (Blue) = 12x 470uF TaPol. The remaining noise is in the 1-10MHz range, especially on the CPU (Yellow). The 270uF B-case caps actually allow more of it through. So I recommend sticking with 470uF:

The noise above 10MHz has very low amplitude.It's adequately decoupled by 0.1uF MLCCs on the motherboard or reduced by common mode chokes.

So, 470uF caps took care of 100KHz -1MHz, but the remaining 1-10MHz noise can be decoupled using 10uF and 22uF MLCCs. Theoretically they'll flatten it below 1mOhm. I recommend keeping the range of 100kHz - 10MHz below 1mOhm impeedance and ESR in order to guarantee efficient PSU and filter performance. You can use KEMETs K-sim to figure out any combination of caps you want.

 

[vyktormvmpay25]

Yes is lot easier adding 12x 22uF. This was something at fast move still took half of day apart of test in menu,start stop .Thank you.

 

[MYS]

I updated your positive results for the 2x E and 1x A models (first mentioned on Page 89)...

Those 3 consoles did make it into my Long term success column.

As for the H models, I previously only had you down for the one H model??? Are you sure you meant to say 2x H models? I don't remember seeing a second H model mentioned?

I did change for 2 set, but one set belong to my friend (so didn't bothered to mention it then). But both set was switch to "new" Tokin about 2+ month apart but seems like both set had the same result within months. I bought lots of "new" tokin upon first success and guess those Tokin will be sitting in the cupboard for sometime...