PS3 nec caps/ylod

[Exclamation]

So it's come to my attention that around a year ago someone figured out PS3 ylod is generally because these nec tokein caps are prone to fail - at least where they are installed on fat ps3s - the caps are a bit weird as they are roughly like cpu sized caps and have 4 rails running parallel under them that they connect to. The fix they came up with was to mount 4 tantalum caps on the rails, and bridge the + rails, but the caps had to be mounted at an angle because they are a bit too long for the rail spacing.
My question: is there any reason these can't be ceramic or metal film or any other type of smd cap, and does anyone know the size code that would fit across these rails without having to be angled - I'm worried that I might accidentally get one pole of the caps connected to both rails if I tried to install at an angle - thanks if anyone has a solution for this!

Edit: here's a link for the original fix thread https://www.psx-place.com/threads/tutor ... fix.25260/

[RIP-Felix]

Firstly let me caution you against believing this is a miracle cure. It sound like it might be true, since reballing isn't a permanent fix, indicating it's not the whole story. However, there are many reasons that can cause a YLOD/RLOD and one answer doesn't fit all. I tried to fix a RLOD CECHA01 with Tantalums and didn't have success. ***EDIT: Turns out I had cold solder joints.*** You can read about it in this thread. I suggest you read my advice on removing the NEC/Tokins and soldering in the Tantalums, it's a real tedious job!

I've read some have had success and that NEC/Tokin caps have been known to go bad in other devices. It may be advisable to replace them as preventative maintenance, but this particular mod has not been fully investigated. It may not prevent a YLOD and a reball may still be necessary, or something else could cause it in the future. That's just the reality of the X-BOX 360 and PS3 era of consoles, they run hot and console life span suffers accordingly. IMO, this is the best reason for maintaining backwards compatibility indefinitely - So we don't have to rely on old hardware lasting forever.

EDIT: The more research I do into reballing the more I think reballing doesn't work. Electronics have perishable components. Not just on PS3/XB360, but on laptops as well. Hot running flip-chip BGA graphics processors just go bad over time, it's inevitable! I'm not convinced that the solder balls are the problem either, because reballs don't last. The heat from a reball may realign some molecules and restore function temporarily, but that's not going to fix the accumulated heat damage that naturally wears the chip out. I've heard anecdotally that reballs die after a few months (if heavily used), many die sooner. But I've never heard of a reball lasting as long as it did before, not in a console that saw the same amount of use before/after the reball. All computers today run hot. Because of this they will not last. Video game consoles of the past didn't run very hot. The 128-Bit era and before still function and have proven much more reliable, but eventually they will also die. 256-bit and up haven't been reliable. Heat accelerated their demise and all of today's consoles run hot AF. Video game preservation in it's original form is impossible for this reason. Emulation, either Hardware or software, official or unofficial, is the only way we'll be able to play retro games moving forward.

That's my definition of "Retro"- 6th generation & prior. From the 7th generation onward we entered an era that is not only difficult to emulate, but requires too much heat for the consoles to last into old age. They don't survive long enough to become retro anymore. So that's the cutoff. It also happened to coincide with the digital era, so the definition could also be considered the jump from analog to digital.

[Exclamation]

Firstly let me caution you against believing this is a miracle cure. It sound like it might be true, since reballing isn't a permanent fix, indicating it's not the whole story. However, there are many reasons that can cause a YLOD/RLOD and one answer doesn't fit all. I tried to fix a RLOD CECHA01 with the ceramics and didn't have success. You can read about it in this thread. I suggest you read my advice on removing the NEC/Tokins and soldering in the Tantalums, it's a real tedious job!

I've read some have had success and that NEC/Tokin caps have been known to go bad in other devices. It may be advisable to replace them as preventative maintenance, but this particular mod has not been fully investigated. It may not prevent a YLOD and a reball may still be necessary, or something else could cause it in the future. That's just the reality of the X-BOX 360 and PS3 era of consoles, they run hot and console life span suffers accordingly. IMO, this is the best reason for maintain backwards compatibility indefinitely - So we don't have to rely on old hardware lasting forever.

Oh for sure - it's just that I had it "reballed" years ago by some random repair shop - it only lasted a month at most. I'm sure he didn't have the proper station for a reball, so idk what he actually did, but my options are kind of limited as far as repairs go, so I figure the worst that can happen is I waste time and a bit of $$ on caps. Better than searching for an original release console that's still working imo, and maybe I'll have to do that anyway, but I just figure what the hell, why not try.

[RIP-Felix]

My question: is there any reason these can't be ceramic or metal film or any other type of smd cap, and does anyone know the size code that would fit across these rails without having to be angled - I'm worried that I might accidentally get one pole of the caps connected to both rails if I tried to install at an angle - thanks if anyone has a solution for this!

Tantalums have the advantage of being about 3 times smaller by volume for the same capacitance. ***EDIT: They have low ESR and are more reliable.*** In this case you need 470uF 6.3V caps that tend to be about 7.3mm in length, which is longer than the rails. The ones I bought are 4.3mm tall and that was too high for the RF shield to seat correctly. These might be what you're looking for, but 32 of them will set you back $50. They should be much easier to solder in. You can save about $20 with these, but you'll have to solder them in at an angle.

EDIT: This article explains it pretty good.

[Exclamation]

Tantalums have the advantage of being about 3 times smaller by volume for the same capacitance. In this case you need 470uF 6.3V caps that tend to be about 7.3mm in length, which is longer than the rails. The ones I bought are 4.3mm tall and that was too high for the RF shield to seat correctly. These might be what you're looking for, but 32 of them will set you back $50. They should be much easier to solder in. You can save about $20 with these, but you'll have to solder them in at an angle.

Ok - so I'm kind of a noob with electronics repair, so I need to ask this since I'm not sure: these are going in parallel, right? So if the nec caps are 1200uf 2.5v, can we not use 4x 330uf to replace them? And also, if they're available, doesn't voltage work the same way? So 4x .75v would be a closer rating?

[korpse413]

I attempted this but did not get positive results. If anyone out there has a proper reball station I am interested.

[RIP-Felix]

...these are going in parallel, right? So if the nec caps are 1200uf 2.5v, can we not use 4x 330uf to replace them? And also, if they're available, doesn't voltage work the same way? So 4x .75v would be a closer rating?

Tantalum is expensive. So fewer caps is less expensive. The space requirements are tight and the number of capacitors that need to be soldered in is 24 or 32. Fewer parts means fewer chances of screwing up, less to fail, easier to solder, and cheaper.

Voltage in capacitors doesn't calculate like capacitance, nor like resistance does in resistors. Don't get hung up on it. Basically, a higher voltage rating is fine, going lower is not. It's usually fine to use higher capacitance, a lower one may lead to problems. Capacitance scales up in parallel. If you stack caps in series, they loose capacitance. So yeah, 4x 330uF in parallel = 1320uF or 3x 470uF = 1410uF. However, you also have to factor in the tolerance. So as long as the -% is greater than 1200uF, we should be fine. Let's work out the math(s):
4x330uF -10% = 1320uF - (0.1 x 1320uF) = 1188uF.
3x470uF -20% = 1410uF - (0.2 x 1410uF) = 1128uF.

Both are slightly under the capacitance rating, but the likelihood that they'll operate that far towards the -% extreme is minimal, and you can add another to compensate (assuming it'll fit). That's not the problem...Getting those Nec/Tokin caps off is. It's a PITA and risks tearing the pads. Soldering in those Tantalums is very difficult without shorting the rails too. Earlier you asked about the rail spacing. I just went and pulled out my failed attempt. The rails are about 4mm from outter edeg to outer edge with maybe 1mm in between. I also noticed my soldering was very $h!tty, so I got curious...I have more experience under my belt now, and better tools. I just double checked my work and found I managed to short the rails. ***EDIT: I later found there is supposed to be about 2.5mOhms of resistance between rails and multimeters will read that as a short (buzz) when set to continuity. That's misleading. I did have cold solder joints, from not getting enough heat to properly adhere to solder. That's a potential reason it failed. Also the caps I used had too high an ESR.*** I remember checking continuity with my multimeter, but the probe can sometimes be insulated by soldified flux and I must have been too quick about checking my work. That would explain why it didn't boot! So obviously I need to revisit this myself. I'm not saying it will work, but my impression before was under the assumption I hadn't screwed up, which I see now I clearly did...lol.

Sigh...

So I'm rethinking my cap choice entirely. I need to remove my caps and start over. All the 330uF 2.5v+ @10% on Digikey are 6mm or longer and 470uF 6.3V+ are all 7.1mm or longer. I do have an idea that'll probably work, but for how long IDK. The idea is to solder a resistor leg to them on both sides. Then bend it around both sides toward the center and solder the legs to the rails. This way The problem with this idea is that the caps are meant to be soldered to a pad. Its heat dissipation characteristics and longevity are based on that assumption. So I'm not sure how doing this might effect them. It'll probably be fine, but IDK for sure.

However, I'd like to use caps that doesn't require me to fit them on at an angle or have to resort to soldering on a janky resistor leg solution like above. Enter a new contender. These are 3.5mm in length, which should give just enough room for them to be soldered to the rails in strait. ***EDIT: I updated the link to a similar package that has a much lower ESR. The original NEC/TOKINS had and ESR of 1.5-2.0 mOhms @100kHz and it's advisable to get as close of a replacement as possible. Also, 4 of these can be soldered next to one another per rail. S up to 8 could be soldered on each NEC tokin. The more of these in parallel the lower the ESR too, because the ESR of 2 identical caps in parallel halve. So 8 should have an ESR of about 5.6mOhms @100kHz. It's still High, but it's much, much closer!*** That should be perfect! 50 are only $27, so that's quite reasonable. I need to order some other parts, so I think I'll pick these up as well. I already have the 470uF caps I installed previously, so maybe I'll try the resistor leg hack while I wait for them to arrive.

Now is the time I should ask about your soldering equipment. If you intend to go ahead with this project, I recommend at least the following:
1) A temperature controlled Soldering station with a variety of tips.
2) Good Rosin Paste Flux, Desoldering Braid, Flush Cuts, Solder. etc.
3) A pair made solid of addamantium!

If you don't already have all the soldering equipment, or intended on getting into modding/electronics projects, then this may be more expensive than it's worth. If all you have a is a radioshack special (15/30W soldering iron) FORGET ABOUT IT!!! That thing is trash, throw it away now! Rethink your modding path, because this is way too difficult to attempt without adequate tools.

[Exclamation]

3) A pair made solid of addamantium!

Hehe - yeah I've got all the stuff besides the caps - I know these nec caps are a major pita to remove, but anyway - been doing a bit more digging and had a look at the service manual for the original boards, as well as the data sheet for the nec caps, and found a couple things.

1 - the 4 nec caps ea for the CPU/gpu are all connected in parallel. This could be why some people are having problems when only replacing 1 of the caps with a roughly equal rating set of tantalums: they may have had more than 1 of the nec not working right and basically just still not getting enough power until they over spec the 1 that they are replacing - I'm thinking it's possible to have almost the equivalent of all 4 nec caps located on only the pads for 1 or 2 of the original caps - might not be good for the pads themselves though to carry that much current and maybe why there's so many caps to begin with?

2 - the esr could be a factor as well - the info we can see on the nec package, as far as I can tell, does not tell us *exactly* which cap is used, but it does narrow it down to 2 - their esr is either 1.5 or 2 mOhm (I probably wrote that abbreviation wrong since I am a noob, I swear, but I do like to and am trying to learn) (edit: I did have that abbreviation wrong because I'm a noob - fixed)- this is important since CPUs love power like our brains love oxygen - they simply do not work right without enough of it. I'm wondering if that could be the cause of a lot of the failed attempts on nakedsnake's thread? - I noticed a lot of people on there talking about super cheap caps - like 10 for 5.00 or maybe even cheaper - they may not have the esr to deliver power quickly enough so they are just as bad as having the original caps not working right.

Anyway - as far as tolerance goes, the original caps are 20% so I don't think it would hurt to put the same back in, especially if over specced, and would be probably the best place to try to keep cost down - esr though I don't think is able to be compromised and is probably a good idea to overspec that if possible - will be looking more into that over the next couple of days and hopefully can find something that's not like 5.00/cap - I had issues with mouser.com not loading/freezing last night so had to give up for the night and it's already late tonight, but I'll let you know what I find, and please let me know what you find as well - I kind of just want to do this repair for exp at this point, and to have a spare as well I guess since I still have the slim PS3 that was like my 2nd or 3rd replacement years ago, and just got a release model ps2 about a week ago - but these things will die eventually so it will be nice to have exp working on things as much as possible

[Exclamation]

All the 330uF 2.5v+ @10% on Digikey are 6mm or longer and 470uF 6.3V+ are all 7.1mm or longer. I do have an idea that'll probably work, but for how long IDK. The idea is to solder a resistor leg to them on both sides. Then bend it around both sides toward the center and solder the legs to the rails. This way The problem with this idea is that the caps are meant to be soldered to a pad. Its heat dissipation characteristics and longevity are based on that assumption.

It's possible to make new pad locations as well I found out, since the copper for the + extends quite a bit on either side of the pads. Just have to use a fiberglass pen and remove some of the resin to expose it and you've got a new location to solder to the same plane. Also I didn't see or forgot by the time I got there on my last replay that we were both considering esr properties - so yes, please let me know what you find there!

[RIP-Felix]

Yeah, I saw a post on PSX Palace that slapped ESR strait back into my consciousness. I did not take that into account last time. So I just ordered 100 of these.
They have an ESR of 45mOhms each but you can fit 8 of them per nec/tokin. If you replace 1 nec/tokin with the maximum of 8 of these that's a combined ESR of 5.6mOhms. 16x of these caps in Parallel is 2.8mOhms ESR, 1.9 for 24x, and if you replace all 4 with 32 of them it drops to 1.4 mOhms. You'll have way more capacitance than necessary, but it's the only way to get the ESR that low.

Now my board already has all of the nec/tokins removed, so I'll start with 24x (3x per rail) on each of the RSX and CPU sides. I bought 100 because I was anticipating 48 being the number I would need per PS3, and I want enought to fix my working unit when it starts to act up. The busted unit was for learning and parts to keep my working unit alive. But hey, if I can get it running again that'd be pretty cool!

[Exclamation]

They have an ESR of 45mOhms each but you can fit 8 of them per nec/tokin. If you replace 1 nec/tokin with the maximum of 8 of these that's a combined ESR of 5.6mOhms. 16x of these caps in Parallel is 2.8mOhms ESR, 1.9 for 24x, and if you replace all 4 with 32 of them it drops to 1.4 mOhms. You'll have way more capacitance than necessary, but it's the only way to get the ESR that low.

Ok, but does that not mean that since they had 4 in parallel, if they were all working, then the esr would be around .37 mOhm? Don't know how much to be concerned about that, but given that caps can't be tested while in circuit, and these caps can't really be removed without being destroyed - idk the best course of action. Leaving at least 1 seems to mean you don't have to bridge any + rails, and we may not be able to get esr low enough per pad space with small enough caps... Thinking...

[Exclamation]

If you replace 1 nec/tokin with the maximum of 8 of these that's a combined ESR of 5.6mOhms. 16x of these caps in Parallel is 2.8mOhms ESR, 1.9 for 24x, and if you replace all 4 with 32 of them it drops to 1.4 mOhms.

I found these: https://www.mouser.com/ProductDetail/Pa ... cHog%3D%3D
They may be a little more appropriate and only need 6/nec -330uf 2.5v and esr is 9mOhm - if I'm correct, then that makes pretty much an exact replacement with 6/ea nec location and gives 1980uf 2.5v and 1.5 esr - also is in the 3528 size so won't have to be angled either!

Also found these with the same specs but they're about 1.00 more/cap https://www.mouser.com/ProductDetail/KE ... sB1g%3D%3D

[RIP-Felix]

Ok, but does that not mean that since they had 4 in parallel, if they were all working, then the esr would be around .37 mOhm? Don't know how much to be concerned about that, but given that caps can't be tested while in circuit, and these caps can't really be removed without being destroyed - idk the best course of action. Leaving at least 1 seems to mean you don't have to bridge any + rails, and we may not be able to get esr low enough per pad space with small enough caps... Thinking...

Yeah, you are correct, thanks for catching my error. Yeah so 4 nec/tokins in parallel should have had an esr of 1/(4/1.5) = 0.375mOhms. Yeah, I think Sony was trying to eliminate the ESR as best they could. From reading about it, one reason is that the lower the resistance the lower the heat produced (important in this application). Higher ESR will limit the current draw, which the CPU and RSX chug like beer. It's possable that ESR of that circuit is one of the boot checks, and if it's too high the system will YLOD. Plausable enough, but then why are people claiming sucess with higher ESR caps doing this mod?

Using the caps you linked:

Parallel Resistance Equation = 1 / [(1/R)+(1/R)...]
= 1 / [(1/9mOhms)+(1/9mOhms)...] = 1/(24/9mOhms) = 0.375mOhm

So yeah, the math(s) check out. Very nice find indeed! The higher capacitance shouldn't matter, but theoretically 6x 220uF caps of the same rating would be closer in capacitance, if that matters (it shouldn't). The 220uF are slightly cheaper, but the leakage current is 55uA vs. the 330uF's 165uA. I'm not an Electrical Engineer, just a hobbyist that likes to mod consoles, so I'm not sure how that might affect the circuit in parallel. The NEC/TOKIN datasheet says they had a leak current of 300uA. There are other specs., like the dissipation factor, I'm at a loss with too. From what I gather thees factor into the impedence for a given frequency. Changing the frequency, changes the ESR. So the frequency characteristics curve can look different even if the ESR at 100kHz matches up (which is what we've been doing). I'm thinking of making a spreadsheet to graph these and compare caps to the NEC/Tokin originals. This is foreign territory to me, so I could be speaking out my ass. Point is, there's more to these equations that we're not taking into account. And unless we know what's really going on, we're just guessing.

Hey could you do me a favor? The NEC/TOKIN rails furthest from the CPU and RSX are showing continuity on on my board. The rails closest to the chips are isolated from each other, as I would expect. I can't see any reason why these rails would be shoring and I don't have another board to test if this is normal. Maybe POS & GND for that end of the NEC/TOKIN is supposed to be continuous? Anyway, would you mind testing continuity once you get the NEC/TOKINS off and clean the pads. I need a sanity check here. I did knock off 3 tiny caps by the CPU when removing the the nec/tokins, but I don't see why that would cause a short (especially since I cleaned their pads to be sure there wasn't any broken pieces of the cap left on there).

Also, let me know if you find a schematic anywhere (COK-001). I'd like to replace those caps I knocked off the board, but don't know the values. I could pull an adjacent one, but rather not since they're so small nor assume it'd be the same. I'm guessing 100pF, but that's a shot in the dark.

[RIP-Felix]

Just found a note in the NEC/TOKIN datasheet that mentions how they should be connected for 2 different applications. "An effective decoupling performance can be achieved by not separating Vcc layer. This method also makes surface mounting of loads(CPU etc.) efficient." That lead me to this video. That is probably exactly how SONY used them, as decoupling (noise filtering) caps for the CPU/RSX. It explains their close proximity to those ICs. Still learning...

EDIT: Found this. Useful for calculations, but still looking into it..

[Exclamation]

The NEC/TOKIN rails furthest from the CPU and RSX are showing continuity on on my board. The rails closest to the chips are isolated from each other, as I would expect. I can't see any reason why these rails would be shoring and I don't have another board to test if this is normal. Maybe POS & GND for that end of the NEC/TOKIN is supposed to be continuous? Anyway, would you mind testing continuity once you get the NEC/TOKINS off and clean the pads.

You have a + and a - rail with continuity? And this is after removing everything from all 4 sets of rails for CPU or gpu? I would think there shouldn't be any continuity there either - will check that for sure once I get mine stripped down.

Also, let me know if you find a schematic anywhere (COK-001). .

For sure - found a free copy here: https://elektrotanya.com/sony_playstati ... nload.html

Just found a note in the NEC/TOKIN datasheet that mentions how they should be connected for 2 different applications. "An effective decoupling performance can be achieved by not separating Vcc layer.

I saw that too, but I really don't know what it means - I did find a thread with a guy trying to repair a laptop with the same method, and at one point someone asked him about the + rails being connected somehow on the board, and suggested he bridge them - I'm thinking I'll order an nec cap off eBay for testing since that step of the fix never made sense to me - wondering about the resistance through the nec caps from + to + and if it wouldn't be better to have a resistor or 10 as part of the bridge?

[RIP-Felix]

I saw that too, but I really don't know what it means - I did find a thread with a guy trying to repair a laptop with the same method, and at one point someone asked him about the + rails being connected somehow on the board, and suggested he bridge them - I'm thinking I'll order an nec cap off eBay for testing since that step of the fix never made sense to me - wondering about the resistance through the nec caps from + to + and if it wouldn't be better to have a resistor or 10 as part of the bridge?

Yeah, so the reason I thought my soldering jobs was bad was because I tested continuity with a multimeter and it was showing a short between +/GND, which I assumed. However, now I'm thinking it was because I bridged the + rails like the guide says to do. Since the lower rails are continuous with GND on my board, jumping the +rails caused them all to short.***EDIT: Lol! That's not it and they are not shorted. Just a small resistance, which is normal.***

Now the the datasheet for the NEC/TOKINS shows the + rails on the capacitor are internally connected. Same with the GND. So that's probably why the guide says if you remove all the NEC/TOKINS that you need to bridge the + rails. Otherwise to leave 1 in place and add enough capacitance with the remaining 3 pads. However, I don't think that's the whole story. The 2 examples in the data sheet for a filter circuit and decoupling circuit are why. I thought that note in ex.2 (decoupling circuit) was probably how SONY used them, because it mentions a "CPU" use case. But from testing with a multimeter I know the Vcc layer on the motherboard's CPU +rail is separated (no continuity between rails). So that make this a filter circuit. By using a jumper wire between + rails makes it a decoupling circuit. I need someone to chime in as to what the difference is.

As for internal inductance between the +rails, I would expect that to be very low since these are LOW ESR caps and any significant resistance would be counterproductive. Isn't the Idea for decoupling caps to get ESR down as much as possible, because it would increase resistance to current draw (bad for current hungry ICs), generate more heat (which that part of the board already has enough of), and decrease power efficiency (not good for the power supply or the environment).

I used that simulation on Panasonic's website to graph the ESR frequency characteristics of the cap you linked. Just put "ETPE330MA9GB" into the part number and you can generate the graphs. You can compare other caps on the same graph too. This shows the ESR across all the useful frequencies.

Sanity check:
I'm assuming you want to choose a cap with the lowest ESR for the frequency range you want to pass. If high ESR, in high heat uses where there is expected to be induced ripple currents, such as a CPU demanding current from a DC-DC switching voltage regulator, isn't the capacitor much more likely to fail if it keeps resisting current flow of normal frequencies. Correct me if I'm wrong, but shouldn't the ESR valley line up with the DC-DC switching Voltage regulator's frequency range, so it wont impede normal current delivery? In this case that range is 300kHz- 1MHz. I'm gather that low ESR in this range is very important. This could be why SONY choose low ESR caps and placed so many in parallel. Placing them so close to the CPU improves filtering of decoupling caps.

Now the caps you linked are most efficient from 100kHz - 1MHz, with an ESR of 8-10mOhms, but that rises to about 15mOhms by 20MHz and plateaus there about. That's what makes it the low ESR version. The other one (no "A" in the part number) keeps rising with higher frequencies. The NEC/TOKIN data sheet shows the S21 resonant frequency at about 100MHz. But that graph and the ESR vs. frequency are not the same. I found the calculation to convert from S21db to ESR and made an excell table to get the ESR from 100KHz - 1MHz. The data sheet shows S21 of -82db to - 85db, which corresponds to an ESR of 2.0mOhms(@100KHz) to 1.4(@1MHz). 2.0mOhms matches the data sheet value for 100kHz, so I trust my math(s). What this means is:

• - 4x NEC/TOKINs in parallel have an ESR range of 0.35 mOhms(@1MHz) - 0.50 mOhms (@100KHz).
  - 24x 330uF caps you linked have an ESR range of 0.33 mOhms(@1MHz) - 0.42 mOhms (@100KHz).

Assuming 100KHz-1MHz is the operational range we're concerned with, then those caps should be ideal. However, the NEC/TOKINs ESR @100MHz really drops (0.1mOhms per cap)! The 330uF you linked jumps up to 15mOhms each. So if the 100MHz range is important then it shakes up this way:

• - 4x NEC/TOKINs in parallel have an ESR range of 0.025 mOhms (@100MHz).
  - 24x 330uF caps you linked have an ESR range of 0.625 mOhms (@100MHz).

So if the new caps operate in the 100MHz range, then the ESR would be 25x higher than it would have been otherwise. However, I think the low value ceramic caps in the High pass decoupling filter are meant to prevent that kind of high frequency noise. Speaking of which, I think those 3 tiny caps I knocked off of my board while removing the NEC/TOKINs are part of the high frequency filter. Checking the schematic you linked revealed they're C6164, C6151, & C6165 (0.1uF 10V X7R 1005). There are 36 of them in parallel (3.6uF total), so missing a few is probably not the end of the world. They are fed +1.0v by the Vsw lines on IC6104, IC6105, IC6106. which are IP2003ATRPBF DC-DC switching voltage regulators.

I could really be talking out my ass! Someone please jump in if I am. I'm trying to piece this together as I go, no better way to learn EE than wanting to understand this stuff to fix a busted CECHA01! Does this story make sense so far? I'm way out of my league here.

[Exclamation]

I could really be talking out my ass! Someone please jump in if I am. I'm trying to piece this together as I go, no better way to learn EE than wanting to understand this stuff to fix a busted CECHA01! Does this story make sense so far? I'm way out of my league here.

I definitely am in both cases lol - nice findings though!

I was just thinking a couple tweaks might be to add multiple bridge wires on each pad rather than a single one on each or a single one per 4 nec. And also it it would help with esr, maybe just using larger size replacement caps even if they don't quite clear the emf shield, exposing more of the copper planes with a fiberglass pen to make soldering easier, and drilling some holes in the shield for clearance.

[RIP-Felix]

I was just thinking a couple tweaks might be to add multiple bridge wires on each pad rather than a single one on each or a single one per 4 nec. And also it it would help with esr, maybe just using larger size replacement caps even if they don't quite clear the emf shield, exposing more of the copper planes with a fiberglass pen to make soldering easier, and drilling some holes in the shield for clearance.

Yeah, you can try that. I had only one jumper wire bridging each the CPU & RSX, but it was a thick gauge solid core. The pads themselves are super thick, so I don't think there's problem with forcing the current through one spot. However, if we want to rule it out and keep everything the same then a jumpier on each pad would be identical to the NEC/TOKINs. Unless the thermoplastic insulation on the conductor melts under the heat, it shouldn't hurt. I can't know it won't be a problem, but highly doubt it would be. If that happens it could short. I looked this up and it should be fine up to 90C. The CPU & RSX dies themselves can get up to around 90C or so, before the fans kick into full rev. If it keeps heating, presumably up to the 105C max rating on alot of electrical components, the PS3 should overheat and shutdown. However, the area adjacent to the heat spreader shouldn't be as hot as the die itself. So I think the PS3's built in heat protection redundancy would prevent this.

[RIP-Felix]

Hey could you do me a favor? The NEC/TOKIN rails furthest from the CPU and RSX are showing continuity on on my board. The rails closest to the chips are isolated from each other, as I would expect. I can't see any reason why these rails would be shoring and I don't have another board to test if this is normal. Maybe POS & GND for that end of the NEC/TOKIN is supposed to be continuous? Anyway, would you mind testing continuity once you get the NEC/TOKINS off and clean the pads. I need a sanity check here. I did knock off 3 tiny caps by the CPU when removing the the nec/tokins, but I don't see why that would cause a short (especially since I cleaned their pads to be sure there wasn't any broken pieces of the cap left on there).

Actually, the rails closest to the CPU and RSX chips are the ones with continuity and the rails furthest don't. I said it backwards. It's the same form both chips. I wonder if that's on purpose? Is it possible that the +rail on the NEC/TOKIN that we thought were internally connected are not? The data sheets shows them being connected and SONY probably decided there is no need for extra copper traces or VIAs to bridge them, when the caps would do the job for you anyway. That's the most likely explanation for there being no continuity between the outer + rails.

But the circuit diagram for a "decoupling" circuit would have them connected, whereas a "filtering" circuit doesn't (ex.2 vs. ex.1). Is it possible that the design of the NEC/TOKIN makes it possible to have both a decoupling and filter circuit in one cap? A simple continuity test between +rails on an intact NEC/TOKIN would answer the question. Continuity=NO! If yes, then bridging the rails would remove the filtering circuit and make all the caps act as a decoupling circuit (which I think is correct).

I still don't understand why the +/GND rails closest to the CPU/RSX are shorted!!! YTF would that happen? Could a resistor have blown and be shorting, a cap, or IC? What when blown, shorts a circuit? Or is that normal in a decoupling circuit? I'm struggling with this part.

***EDIT: Before you remove your NEC/TOKINs, PLEASE TEST CONTINUITY! IF all 4 rails are showing continuity then I can be sure this is normal. Also, the caps you linked came in the mail last night and I soldered them in this morning. I used a 22AWG solid core conductor to bridge the +rails on 1 of the pads for each the CPU & RSX. Now all 4 rails are showing continuity with one another, as expected with my board. I suppose I could open my working CECHA01, but I'd rather not. It's running fine and I go by the addage, don't fix what ain't broke! So if you could check your board before removing those nec caps, that would help me out. I am now reading 2.2ohms across +/GND and 0.3 GND/GND. My continuity buzzer goes off, but there is added resistance, which might be too small of a difference for my meter to recognize?***

[RIP-Felix]

I made a Flow-chart of the components in the power conditioning circuit upto the CPU/RSX, based on the PS3 scematics and part data sheets. I mostly googled around for what each is generally used for and guessed why they are being used where they are to "attempt" to make sense of the circuit. This is the fruit of that labor. I think it makes sense, except the purpose of the NEC/TOKINS at the end. The mod turns them into Decoupling caps, but SONY's schematic has the acting as filter caps. It would be nice if someone more experienced in EE could chime in. The image is high res, so I put it in spoiler quotes to keep the thread tidy.

Spoiler

I can confirm by testing with a multimeter that SONY did not connect the +V rails.

Spoiler

So according to the reference on the data sheet that's a filter circuit design. However, by adding a jumper between Vin/Vout rails we are transforming it into a decoupling circuit.

Spoiler

However, SONY's schematic shows 2 capacitors separated by an internal inductance between +rails for each NEC/TOKIN.

Spoiler

Now I'm wondering if it's not a good Idea to use a jumper between +rails? Perhaps the Inductance marked on SONY's schematic is just indicating the inherent ESL of the cap? Or was the NEC/TOKIN serving 2 purposes? Could one rail be a filter cap design and the other a decoupling circuit? Could that explain why +/GND on one side was showing continuity & the other side was isolated? Perhaps we should be using an inductor between positive rails?

[Exclamation]

But the circuit diagram for a "decoupling" circuit would have them connected, whereas a "filtering" circuit doesn't (ex.2 vs. ex.1). Is it possible that the design of the NEC/TOKIN makes it possible to have both a decoupling and filter circuit in one cap? A simple continuity test between +rails on an intact NEC/TOKIN would answer the question. Continuity=NO! If yes, then bridging the rails would remove the filtering circuit and make all the caps act as a decoupling circuit (which I think is correct).

I still don't understand why the +/GND rails closest to the CPU/RSX are shorted!!! YTF would that happen? Could a resistor have blown and be shorting, a cap, or IC? What when blown, shorts a circuit? Or is that normal in a decoupling circuit?

Will check it out for sure - I can't imagine that the +/- pads should be connected in any way other than through a cap - are you getting close to 0 resistance there? If so I would think that is from a direct short and makes any caps useless because of electricity following the path of least resistance. I did end up ordering a tokin cap off eBay to try to understand it better - the listing had a pic of the underside of the cap though too so even just seeing that and some of the removal videos has helped seeing them basically disected. The underside has 2 outer rails like you would expect if you only ever saw a bare Sony board - the middle part going to the - pads is actually 1 big contact that takes up all the space between the inner pads. The inside looks to be just layers of foil/conductive material separated by layers of semi conductive material. So basically it looks like we always thought - there is reason to connect the + pads.

I still don't understand why the +/GND rails closest to the CPU/RSX are shorted!!! YTF would that happen? Could a resistor have blown and be shorting, a cap, or IC? What when blown, shorts a circuit? Or is that normal in a decoupling circuit?***

Afaik almost any type of component CAN blow shorted - I imagine this would be a pita to track down aside from just spotting something on the board that's exploded/burned :/

Perhaps we should be using an inductor between positive rails?

I had seen that symbol on the schematic but wasn't sure what it was, and honestly wasn't even sure if that was on the + or - side because I suck at reading schematics and am way out of my league here - but thats what I was getting at earlier talking about adding resistors in the bridge wire - once I get that nec cap in is that something I could test with a basic multi meter? I'm guessing not, but I may have someone I could send it to for further testing - and would be glad to send it out to anyone who has equipment to test it with - my league only covers the very basics + soldering + common sense + curiosity lol.

[RIP-Felix]

Will check it out for sure - I can't imagine that the +/- pads should be connected in any way other than through a cap - are you getting close to 0 resistance there? If so I would think that is from a direct short and makes any caps useless because of electricity following the path of least resistance. I did end up ordering a tokin cap off eBay to try to understand it better - the listing had a pic of the underside of the cap though too so even just seeing that and some of the removal videos has helped seeing them basically disected. The underside has 2 outer rails like you would expect if you only ever saw a bare Sony board - the middle part going to the - pads is actually 1 big contact that takes up all the space between the inner pads. The inside looks to be just layers of foil/conductive material separated by layers of semi conductive material. So basically it looks like we always thought - there is reason to connect the + pads.

Yes I'm reading 2.5 ohms across +/GND on the rails closest to the CELL/RSX chips. It's 0.4ohms from GND/GND. Both are sufficiently low enough for my multimeter's continuity to buzz. A forum post led me to this, which details the NEC/TOKINs construction and performance benefits compared to other capacitors. I was reading it when you respond. Cheaper, fewer components to install, better performance, extremely low ESR, and good temperature performance. Sounded like a breakthrough technology, as all industry papers like this make their products sound like. Hindsight being 20/20, it turned out to be a bad decision.

When you get that NEC/TOKIN cap in, test the resistance over the +rails and across your test probes to see if there is any significant difference. If not, then the jumper wire is necessary.

Afaik almost any type of component CAN blow shorted - I imagine this would be a pita to track down aside from just spotting something on the board that's exploded/burned :/

A resistor will short if it blows, but it has to literally burn to a crisp first (like incandescent hot)! That would leave a scorch mark I'd see or smell. I inspected the board thoroughly and found no burn marks or smells. I did notice continuity across the 0.001Ohm resistors that separate the coils from the capacitors (see previous flow-chart). However, they look fine and all of them read continuity, which is what you would expect if the multimeter reads 2.3 Ohms as a short! So that doesn't really mean anything I don't think. You might have a test of them too when you get a chance, just to verify my results are the same on your board.

I had seen that symbol on the schematic but wasn't sure what it was, and honestly wasn't even sure if that was on the + or - side because I suck at reading schematics and am way out of my league here - but thats what I was getting at earlier talking about adding resistors in the bridge wire - once I get that nec cap in is that something I could test with a basic multi meter? I'm guessing not, but I may have someone I could send it to for further testing - and would be glad to send it out to anyone who has equipment to test it with - my league only covers the very basics + soldering + common sense + curiosity lol.

It's the symbol for impedance. It's shown on the +1.0v line. The dotted line signifies the capacitor itself and appears to indicate some internal impedance between cathodes (+), but it's not specified there. The datasheet for the NEC/TOKIN does specify the dissipation factor at 10%, which is related to it's impedance, but that's not specifically between the cathodes. Maybe that's just a standard symbolic practice or something, IDK. Wht I can tell from that schematic is that the +/GND are completely separated by the capacitor (Bypass circuit). Current must flow through the cap to ground:

There is a difference between a bypass cap and a decoupling cap, and I'm not sure that decoupling is what's going on here. That link say's it's usually to isolate AC from DC and that's not what's going on here. I guess we're just trying to give noise a bypass?, so it doesn't enter the CELL/RSX. Maybe "Bypass" is the better term here.

Now, if there is continuity between the + rails on that NEC/TOIKIN the following is is a mute point. Jumper time! But if there is any significant resistance, then maybe we should NOT be using a jumper? This is based on the assumption that the rails closest to the CELL/RSX are have a low resistance for a reason, not that board is fried like an egg (which is a definite possablity). I need someone with a COK-001 to confirm. Now there was a post by Delita over a year ago with the same behavior:

Spoiler

I get continuity between these following pads color coded, and it doesn't matter which pad, top or bottom, they all share.

That post kinda got swept under the rug and forgot about. I haven't read the whole forum (It's 134 pages long!), but I didn't see any follow up over the next ten or so pages. I'm still reading through the forum. In later posts, there is talk about tying things like I mentioned above, so I'm going to keep reading over their results to see if anything stands out...but with 134 pages it might take awhile to digest!

[RIP-Felix]

...I didn't see any follow up over the next ten or so pages. I'm still reading through the forum. In later posts, there is talk about tying things like I mentioned above, so I'm going to keep reading over their results to see if anything stands out...but with 134 pages it might take awhile to digest!

Found another post that talks about the continuity. Basically that's normal for very small resistance. A 0.001Ohm resistor will buzz if you use continuity and it's hard to even measure with a multimeter, because that's not the normal range general use multimeters continuity are designed for. Resistance is what should be read, and it should be higher than it read across GND rails. So the fact I measured ~2.5Ohms across the +/GND rail vs. 0.3 Ohms GND/GND is normal.

Mechanical stress fatigue on BGA through thermal cycling is still a thing, especially on high temp electronics. However, it doesn't necessarily explain why the PS3 fails when other devices using similar BGA solder chemistry, thermals, and board construction seem to be fine. I will say the more I read the more amateur this hack feels. I'd like to see a signal analysis of the Vout to the CELL/RSX with & without load, before & after the cap replacement. There doesn't seem to be any direct evidence of abnormal current noise leading to the YLOD through direct measurement. It's all circumstantial and anecdote. 'I did the mod and it worked!' 'I did the mod and it didn't work'. But that doesn't prove the NEC/TOKIN proadilizer was the source of the problem in the first place. Only direct measurement would do that. I'm not sure how/if this could be measured. I'm definitely still skeptical.

I also worry about the strain placed on the power supply by people using caps of unknown ESR. Replacing the original part with one that "best" matches the original, spec. for spec., helps to remove the guesswork, but not all of it. How does that affect the temperature of the caps? Should the Tantalums be cooled? I mean, the NEC/TOKINs had a larger surface area in contact with the board to dissipate heat. Bumping up the voltage rating and lowering the ESR can help with that, but now you're departing from spec. And just because something works, doesn't mean it's safe or will last. Plenty of unknowns for Murphy to leverage.

[RIP-Felix]

IT'S ALIVE!

I decided to man up and just give it a try. A few days ago I VERY carefully installed these 330uF 2.5v 9mOhm Tantalums (6x per NEC/TOKIN they replaced). I replaced all 8 with a total of 48 Tantalums. I used a hot air rework station (320C) propped a few inches above the area I was working on, to preheat the area and make soldering to that ridiculous ground plane easier. This time I wanted to be sure there were no cold solder joints, even at the risk of overheating nearby components. With the hot air a few inches away and the motherboard dissipating the heat, I really don't think that was a problem. It was just enough, with rosin flux, to get nice silver fillets and no semisolid/oxidized/hard solder blobs (you know what I mean). This time I put my best work into it.

I was still unsure about whether or not the + rails should have a conductor bridging them, based off the SONY schematic which shows some impedance between cathodes. I'm still interested in a resistance measurement from @Exclamation between + rails on a new old stock NEC/TOKIN he ordered off e-bay, but that may take awhile to arrive. I decided to "assume" it's internally connected like everyone say's and use the jumper (just a single core 22AWG conductor). I waited a couple of days while I read more and let it dry (I used 70% IPA to clean up, I ran out of 99%).

This morning I decided, "ah screw it!" So I assembled it and powered it on. I was a little more than surprised, when there was NO YLOD!!!! Now, I haven't stress tested it yet. I still need to delid the RSX, replace thermal paste, Pads, reassemble, and try by best to torture it. I have a fan controller modded inside with 3 thermistors I can place around the motherboard for direct temperature measurements. I plan to place one under the RSX heat spreader close to the die, another on the CELL processor, and one on the Tantalum caps, to be sure they are not overheating. If the Tantalum caps need cooling I could add a thermal pad so they can dissipate heat through the RF shield like the inductors and DC-DC voltage regulators do.

There's still work to do, but so far so good. There's hope for this hack yet!

[Gunstar]

IT'S ALIVE!

Really happy for you and I hope it stays fixed! Keep us updated on the long term stability of this.

[Exclamation]

Mechanical stress fatigue on BGA through thermal cycling is still a thing... ...But that doesn't prove the NEC/TOKIN proadilizer was the source of the problem in the first place... ...Plenty of unknowns for Murphy to leverage.

Oh for sure there's plenty of other things that can cause ylod, but without a way to read error codes, or a logic probe, or much knowledge to do testing, it seems like a good try - especially if a system is otherwise dead to you.

IT'S ALIVE!




This morning I decided, "ah screw it!" So I assembled it and powered it on. I was a little more than surprised, when there was NO YLOD!!!!

There's still work to do, but so far so good. There's hope for this hack yet!

Hell yeah! That's badass to hear!

. I'm still interested in a resistance measurement from @Exclamation between + rails on a new old stock NEC/TOKIN he ordered off e-bay, but that may take awhile to arrive.

It will be a little while cause it's coming from overseas - but uk mail does get here a bit faster than China mail so hopefully not much more than another week or 2.

[RIP-Felix]

Really happy for you and I hope it stays fixed! Keep us updated on the long term stability of this.

Will do!

I've heard that GT6 is a good game to stress test with, as under normal circumstances it pushes the RSX to the limit. So I need to order a copy, I like GT anyway. I need more thermal paste too (I like MX4).

My working PS3, main workhorse (also a CECHA01) had the thermistors & fan controller on it. Since I needed to disassemble it to get them off, I decided to probe the NEC/TOKINS while I was in there. So I can confirm a few things for testing purposes:

• - If you set your multimeter to continuity, all 4 of the NEC/TOKIN rails will buzz. In fact everything after the DC-DC switching voltage regulators will buzz. It makes sense for a 1v circuit, because all the resistors and inductors are of such low value they don't provide enough resistance to for the multimeter to recognize the difference. You need to be measuring resistance to see the difference, but it's small by design.
  - Between GND rails there's a resistance of about 0.4 Ohms (could be within the error/precision resolution of my multimeter).
  - Between +/GND there is about 2.5 Ohms.

So this is what you can expect from a normal/working COK-001 motherboard.

Because it buzzed continuity on my initial attempt, I thought I had bunked the mod, but now I know that's normal. I think the reason it didn't work was because I either had cold solder joints or caps of too high an ESR. Probably the former, because I've seen other have success with the same caps I used that time. The ESR was ridiculously high, so they probably would have made the Power supply work harder, heated up more themselves, and then not have lasted as long. I think these are going to be a much better choice, because they match or are slightly better than the NEC/TOKINs in ESR. The higher total capacitance should also reduce wear. They have the same voltage rating (2.5v), but that's fine because the CELL/BE processor has +1.0v and the RSX eats +1.2v. So that voltage rating is doubly sufficient for the voltage they see. The only thing that could reduce their lifespan compared to the NEC/TOKINs, that I can think of, is heat. They're rated for 105C, so anything under that should be within tolerance.

[RIP-Felix]

Yesterday I delided the RSX. Hot air gums/softens the thermal adhesive. A stiff, but thin, plastic card beneath the side closest to the CPU and a little leverage pops the heat spreader off. The side adjacent to the CPU doesn't have any SMD component under the spreader. If you try the other sides, you could knock components off the substrate. Don't force it all at once, apply steady pressure and slowly add more until air gets under the edge and work its way along the surface of the chips. If you go too fast, suction will add unnecessary force and possibly tear the memory chips. Once the heat spreader is off, getting the thermal adhesive off the chips was a PITA! "Goof Off" works to dissolve the adhesive, but it's slow. You need to get as much off with a razor blade first, then dissolve the rest with cotton swabs dipped in the goof off. ***EDIT: The four memory chips are the ones with the adhesive, NOT THE DIE in the center. Don't use a razor blade on the blue DIE!!! The thermal compound on the die will come off with IPA.*** It help to use heat to soften the adhesive, but keeping heat on there while you brace with 2 hands, so as not to slip is hard. I tried IPA, it doesn't cut it. WARNING! It's very easy to slip with the razor blade and knock off some nearby caps, but if you plan your cuts and brace your hands right, it can be done:

Spoiler

You can see how much trouble I had originally getting the NEC/TOKINs off in the picture above. Look at all those scratches..lol! Luckily none of them managed to bridge circuits, or cut traces. They just scoured away some solder mask where it doesn't matter. I also applied new thermal pads with better performance. I just orderd more MX4 thermal interface compound, but it'll be 2-5 days before it arrives.

I also ordered Gran Turisomo 6, which is known to kill fat PS3s! This game is so taxing on the RSX that there are many complaints of the console powering off mid-game, in otherwise functioning fat PS3s. Rumor has it that's a sign the console is on it's way to a YLOD, but not quite there yet. In any case, if this game plays fine for a few hours it's a good sign the YLOD is fixed and the Tantalum caps are sufficient to stabilize power to the RSX.

Something else of note is that the heatsink for the CPU and RSX are not connected. When you take it off you'll notice that the RSX side is clipped in, but not actually connected (thermally) to the LARGER CPU side. The RSX gets 2 heat pipes and a smaller heatsink than does the CPU, which has 3 heat pipes and more surface area. No wonder the RSX runs hotter! I wonder why SONY didn't want them to share the same heatsink, and why the RSX side is soo much smaller?

Speaking of the CPU, I didn't delid it. There's too many SMD component nearby and I didn't have a tool with the right angle to get under there without damaging the area. I decided it wasn't worth the risk. I'll keep an eye on the temperatures of each with this:

Spoiler

It's just a simple PC fan controller, with 3 thermistor probes I can place anywhere I want. It sits under the smoked black top cover, so when it's off you can't tell it's there. When the console is on, it shines through with the temps. I don't use the fan controller, Instead I let the PS3 handle that fan ramp:

Spoiler

Before you ask, yes that's hot glue ("douche goo" as Lord Voultar calls it). In this case it's easy to remove, which is why I used it. In my defense, I performed this mod many years ago - before I developed an ethic against physically modding the case. In this case it's completely hidden, so I don't mind too much. That and it was a trash bound anyway, so I reserve the right to hack the SHIZ out of it!

[RIP-Felix]

18x 270uF per chip, 2.5v, 6mOhm ESR ETPSF270M6E is equivalent to 4860uF, 0.33mOhm ESR. The 4x NEC/TOKINs were equivalent to 4800uF, 0.375mOhm ESR.

They are cheaper (from arrow) and would better match the capacitance of the original caps, without sacrificing ESR. The reason I bring this up is, I've been looking over the schematic more and it's possible the NEC/TOKINS are part of a "RLC band pass filter" on the +1.0v(BE)/+1.2V(RSX) output of the DC-DC switching voltage regulator. If that's the case then changing the capacitance of the filtering caps WILL change the resonant frequency and Bandwidth of the filter. I'm still unsure how this calculates given the specific way they are arranged, or if that's in fact what we have there. This is kinda advanced (for me), but interesting stuff! Still learning...

[RIP-Felix]

I attempted this but did not get positive results. If anyone out there has a proper reball station I am interested.

Would you mind posting the caps you used (for posterity)?