PS3 Reball ps3 Cell/RSX

[haxxxen]

hi Vyktor, sorry for replying so late, but I have felt plain "dead" lately and had no motivation

I will definately send you my DECR-1400 these days

unfortunately, my CECHL is now broken as well, and it was again my stupid own fault, by messing with custom coolers. now my only PS3 left is my precious DECR-1000 and I have none other to make firmware tests

I have to get another one, so probably I will look for a Slim PS3 this time

there were 2 MLCC broken off which I have successfully replaced, but then I think I have messed up CELL or RSX again with custom coolers, so may I also send you my VER-001 mainboard?

for now I am completely done with any hardware mods, cause I only have destroyed lately

 

[vyktormvmpay25]

Ok you can send any board you want, as long as you did not try to reflow/reball is fine. That boards are a pain like this I've worked this weekend. It starts fine and running well in XMB hours and I can install as much as I want firmwares but in the moment I try to run any games will crash instantly with ylod. Been reflowed by someone and reballed twice by my thinking is rsx or imperfect connection under cpu. No same issue with a new smaller rsx and modchip. So it will be kept for experimental purposes and waiting my friend programmer to see if I can jtag this somehow.

 

[haxxxen]

Ok you can send any board you want, as long as you did not try to reflow/reball is fine. That boards are a pain like this I've worked this weekend. It starts fine and running well in XMB hours and I can install as much as I want firmwares but in the moment I try to run any games will crash instantly with ylod. Been reflowed by someone and reballed twice by my thinking is rsx or imperfect connection under cpu. No same issue with a new smaller rsx and modchip. So it will be kept for experimental purposes and waiting my friend programmer to see if I can jtag this somehow.

awesome, so I will send you both boards now, thanks. and no worries, cause I will never ever do sth like a reflow myself

the CECHL was definately broken by the coolers, cause it was running fine before, but after replacing coolers, it just blackscreened, but firmware is loaded cause I have some beep notification in sMAN added. though, you cannot shutdown with power button and it does some 2xbeep

it had some network error lately and after inspecting I just have noticed it was missing 2 mlcc's which I have replaced. then it worked again fine with no errors, but my stupid head wanted better cooling and added the custom coolers. what can I say, I hate myself for these stupid mistakes, which have broken the rule "never change a running system"

 

[kril89]

Vyktor, what equipment do you suggest for someone trying to get into reballing? I see that your setup isn't really available anymore. Is the IR6500 good enough or would I need something more powerful. Also I have access to a few dead motherboards that I could practice on before attempting to do this to a PS3.

 

[vyktormvmpay25]

Any bottom preheater around 2000 w and top heating of minimum 400 w would be able to desolder. Now it depends on experience of user.
Ir6500 get struggling to desolder or late profile will struggle to desolder ic for big boards. Some users here have done custom rework that work. You may calculate parts costs and create one better and powerful then what they sell.
I've been posting earlier a powerful dual controller with very accurate temperature sensor that can control dual heating. You could start reading about MSL profiles of reflow/reball of big companies and understand the time needed that board to reach temperatures in right time. Should not have more than 12 minutes for a total power profile. First study and collect informations. All here are as reference.
My advice ir6500 will be to slow for ps3 boards. Any laptop, pc will be fine.
Had ir6000, ir6500, Scottle IR v5 3 heating zones with 4800 w and sold for 1000 euros back in 2014. After Jovi but if this goes down I will probably exchange controller with one Omron with 4/5 zones which need a lot of testing and calibration (complex with many advance engineering advice from multiple friends have in another area that I'm working).
Been looking for Jovi RE8500 and seems oficialy not producing any.

 

[vyktormvmpay25]

Another thing is that I don't like when you wait 3 months a stencil and won't match with your ic.
This is shorted story, bought 3 stencils for ps4 cxd90025g from 3 different sellers and won't match with pitch of ic. Lucky that I have over 1000 stencils and found/impressed something.
First added 2 different places 2 balls on dried ic to keep stencil in place.

Then add flux over and add stencil

Added balls and heat with 220 dried first.

Because tiny ic are worst to handle watch videos and like or disagree. Finally should have it, if any ball will fail like mine add back stencil, add a bit of flux and eventually push it back down to get right soldered shape.
Quite hard to work on and making video with phone in good quality.
http://s.go.ro/x4bivpiy

There are 2 flux types, both can do same job but one won't smell bad. You probably know which is the best when you will use it.
It does not matter the expiry date on flux or alloy balls. All it matters to start that broken device.

 

[kril89]

Any bottom preheater around 2000 w and top heating of minimum 400 w would be able to desolder. Now it depends on experience of user.
Ir6500 get struggling to desolder or late profile will struggle to desolder ic for big boards. Some users here have done custom rework that work. You may calculate parts costs and create one better and powerful then what they sell.
I've been posting earlier a powerful dual controller with very accurate temperature sensor that can control dual heating. You could start reading about MSL profiles of reflow/reball of big companies and understand the time needed that board to reach temperatures in right time. Should not have more than 12 minutes for a total power profile. First study and collect informations. All here are as reference.
My advice ir6500 will be to slow for ps3 boards. Any laptop, pc will be fine.
Had ir6000, ir6500, Scottle IR v5 3 heating zones with 4800 w and sold for 1000 euros back in 2014. After Jovi but if this goes down I will probably exchange controller with one Omron with 4/5 zones which need a lot of testing and calibration (complex with many advance engineering advice from multiple friends have in another area that I'm working).
Been looking for Jovi RE8500 and seems oficialy not producing any.

This is a lot to take in but I think with enough research I will be able to do this. I've got access to a bunch of old electronics to practice on before I start trying to repair my BC PS3s.

 

[vyktormvmpay25]

As this unit is starting a game but will fail in game after few minutes like 2~3 minute, today I can say all ps3 will have same problem until 28nm (did not had time to test yet)
It is all about imperfect connection with both related with deformation in time.
I've just desoldered cpu and rsx, reball both with lead free.
Work well in a game for an hour.
Everyone who is interested to do reball must do on both if delid don't help or high temperature on one of them.
Mostly 3034 and some 404x errors will tell that this is an issue.
Why both because I see empty mother becoming straight by itself after desolder both.
Begin with leaded balls and if you succeed with that you can play with lead free alloy.

 

[DeadEnd]

Maybe my method also helps someone, here you can see it from beginning to the end using a blue jig.

 

[vyktormvmpay25]

Got it working that 4004 board that was failing.
I've been testing another 28nm motherboard but it was reballed before, compared with one dead rsx I've been keeping for years as example, on mine there is that square pad that is not soldered, looking on rsx on this board it has alloy on that square pad so I figured it wos reballed quite professionally to be observed. Took a photo just after I desoldered.
Looking on pads on motherboard on both nothing unsoldered.
It was with error on 0611802 related to RSX int.
Then after reball same Rsx that board come with, same error no start.
OK took out to investigate.
I took one ihs from old models and added to this by keeping fingers to his rams and just to have thermal contact to its die.
In a quick test without radiators any ps3 with ihs on ic's should start for 10 seconds test.
It starts for 10 seconds and both getting worn, errors didn't come.
Concluded that rsx rams are unsoldered for sure, can say which one but sometimes even measuring only power lines won't tell exactly is good. Tested with and without pressing rams so confident about this.
This is the reason why reballed/reflowed mobo should not work on. I didn't realise that was repaired before until I've seen rsx out of board. This was very professional repair.
Tomorrow will be fixed, I have another working motherboard that was reballed and I will move rsx to this.
Anyway all stencil for this rsx seems to have issues with pitch matching on market . It is like one side won't match exactly, so I've use it and had to push down few balls while melted, still can do the job for 3 dollars.

 

[kril89]

Got it working that 4004 board that was failing.
I've been testing another 28nm motherboard but it was reballed before, compared with one dead rsx I've been keeping for years as example, on mine there is that square pad that is not soldered, looking on rsx on this board it has alloy on that square pad so I figured it wos reballed quite professionally to be observed. Took a photo just after I desoldered.
Looking on pads on motherboard on both nothing unsoldered.
It was with error on 0611802 related to RSX int.
Then after reball same Rsx that board come with, same error no start.
OK took out to investigate.
I took one ihs from old models and added to this by keeping fingers to his rams and just to have thermal contact to its die.
In a quick test without radiators any ps3 with ihs on ic's should start for 10 seconds test.
It starts for 10 seconds and both getting worn, errors didn't come.
Concluded that rsx rams are unsoldered for sure, can say which one but sometimes even measuring only power lines won't tell exactly is good. Tested with and without pressing rams so confident about this.
This is the reason why reballed/reflowed mobo should not work on. I didn't realise that was repaired before until I've seen rsx out of board. This was very professional repair.
Tomorrow will be fixed, I have another working motherboard that was reballed and I will move rsx to this.
Anyway all stencil for this rsx seems to have issues with pitch matching on market . It is like one side won't match exactly, so I've use it and had to push down few balls while melted, still can do the job for 3 dollars.

What stencils do you currently suggest?

 

[vyktormvmpay25]

Any universal 0.6mm stencil that will cover ic. Use 055mm balls.

 

[vyktormvmpay25]

Few days working on different super slims models. Must recap all measurements and testing points for each model. The worst case is where 28nm parts are introduced.
In different cases cpu is going to die like most of ps4.
Anyway foud what can be fixed and not for hole ps3 units now.
In 8 super slim only 3 repaired with different parts ported to boards.

 

[vyktormvmpay25]

An entire collection
On board with X or ? where I see less then 2.7 ohms I don't even bother to desolder from board. I've had already syscon UART test with different errors like 5fff or 1802, even it says is rsx, no is not about rsx in all cases when 2 ohms are on cpu side I'm afraid is dead so no need to work without any reason. Just keep it for scraps.

 

[sandungas]

@vyktormvmpay25 how many stencils do you have for CELL ?, can you confirm if this is the complete collection ?
From left to right...

CXD2964 series (90nm)
41x41 pad array (hole 19x19)

CXD2981 series, CXD2989 series, CXD2990 series (65nm)
41x41 pad array (hole 15x17)

CXD2992 series, CXD2996 series (45nm)
41x41 pad array (hole 17x17)

CXD2999 series (28 nm ?)
42x42 pad array (hole 18x22)

Btw, just ignore the rotation of the names in your stencils... as you can see every manufacturer writes the name in the stencil with an arbitrary rotation or at the other side
See how i had to rotate the third stencil (45nm) to match with the others at left (90nm and 65nm)
And the last CELL stencil... im not sure if the way how im rotating it in this image is correct (im trying to align it with the others), and im not sure if the CELL is 28nm, i never seen one delidded

 

[vyktormvmpay25]

I will add later photos with all usually stencils and last 28nm (good matching holes). Been trying 3 different sellers on eBay and found it on Aliexpress.
About 3000 series and further Cell I have not attempted to delid anymore as for 3 tests gently handled all went broken, probably NSC can delid with heat, I'm afraid to test any.
Edit
@sandungas I got photos for ps3 daily used stencils. They are right as you mentioned
http://s.go.ro/5e118ky0
Best 28nm rsx stencil bought from here
https://www.aliexpress.com/item/32882410138.html?spm=a2g0s.9042311.0.0.5bb94c4dq1231w

Edit2
As a real advice I have to say 28nm models have something else then reball issues. As probably stated before they have either a minor fix, wifi , AV/HDMI ic etc
Or cases where 5FFF can't be fixed ,1802(still can not be sure ) and much more that are related to attempts of delid cpu (i dont test delid for models kte and further as i see good temps) ,different cases where cpu/rsx act as a dead part .There are situations where brick will act as glod and easy to understand where SB debugging won't load.
Why not a reball issue ,those are small ic area wont have those issues ,unless dropped unit .

 

[sandungas]

I will add later photos with all usually stencils and last 28nm (good matching holes). Been trying 3 different sellers on eBay and found it on Aliexpress.
About 3000 series and further Cell I have not attempted to delid anymore as for 3 tests gently handled all went broken, probably NSC can delid with heat, I'm afraid to test any.
Edit
@sandungas I got photos for ps3 daily used stencils. They are right as you mentioned
http://s.go.ro/5e118ky0
Best 28nm rsx stencil bought from here
https://www.aliexpress.com/item/32882410138.html?spm=a2g0s.9042311.0.0.5bb94c4dq1231w

Edit2
As a real advice I have to say 28nm models have something else then reball issues. As probably stated before they have either a minor fix, wifi , AV/HDMI ic etc
Or cases where 5FFF can't be fixed ,1802(still can not be sure ) and much more that are related to attempts of delid cpu (i dont test delid for models kte and further as i see good temps) ,different cases where cpu/rsx act as a dead part .There are situations where brick will act as glod and easy to understand where SB debugging won't load.
Why not a reball issue ,those are small ic area wont have those issues ,unless dropped unit .

Thanks for the confirmation about the CELL stencils, i was doing some images and i needed a 100% confirmation
My only doubt is the rotation of the last stencil for CELL, do you have some motherboard using it where the CELL was removed ? (to see the pads of the board with a mark for the "A1" pad in a corner)

Another btw, you said you broke some of that newest CELL models when trying to delid them, do you still have the broken pieces ?, my only interest is to see the size of the silicon DIE so it doesnt matters if is broken
I can see the substrate is smaller than previous versions, this can be seen very well in your photo by comparing them, they also reduced the "pitch" in between BGA balls (previous models needed 0.6mm balls, and the last one 0.5mm)... but they also increased the size of the ball array (from 41x41 in previous models up to 42x42)
Are a lot of changes, so it would not surprise me if they also reduced the DIE technology... from the previous 45nm to... dunno 35nm (or 28nm like the latest RSX models)

Anyway, if you or someone reading us finds a photo of one of the last CELL models delidded please post it here and/or advise me to wikify it

 

[vyktormvmpay25]

I will add photos of an reballed already 3010 error for cpu. Can try delid on this didn't get it working after reball both still same error after errlog clear 00000000.
It's from a scrap pqx001 Probably is a delid issue and accidentally I fix it.
Now not sure who came with that idea of aluminium from a can of Coke but that is for sure 0.1mm the only solution to delid it on board, may try first delid on board, if I use heat to it can do imperfect connection but will be another error on log and probably fixed, I will try painting knife on this as well

Spoiler

Pretty sure they left that empty space for expension and heat release out of chamber.
To bad my camera usb port is faulty on main ic and can't fix it. Locked on hdmi, when will die completely will buy another.

Spoiler

It's a mix of flux and ipa there beside silicone hard to understand atm.
Can use painting knife as well. Now going to heat to understand the stage of temperature and its soldering method

Spoiler

It didn't work for me, won't work for nobody, unless ic gets to much heat and gets desoldered by itself from ihs which is quite impossible in my opinion, cpu gets fried internally and won't be desoldered from ihs.
Removed silicone was easy, heat at bottom 320 and top up to 370 (done in stages 15 minutes) and didn't got it off

Spoiler

Something that it came in my mind is something like this fridge but haven't tested it yet, I may test it with a normal 3000 rsx and cpu in normal fridge not sure if it will work. Diodes/transistors/ic will survive better on cold then heat probably.

 

[sandungas]

Ohh well, i was not asking you to do it, but mostly asking if you had the broken pieces of the previous attempts to make photos of it, or if someone found a photo of it in google, my help requests for wiki doesnt uses to be so invasive/destructive, lol
But now that you are at it... can you upload a good quality photo of this area ?, it needs to be good enought to allow to see where are located the "missing" pads

About the DIE size, hmmm, i made an sketchup with your images to compare them, are scaled to match with the size of the stencil
The image made in a image editor (not a photo) is from wikipedia and represents the CELL 45nm
And the photo of the broken CELL... well... im not 100% sure but it seems to be 45nm too

 

[vyktormvmpay25]

Good later going to remove sata port and 12v pin in order to scan on printer scanner.
Edit
@sandungas
Tried to print it but not getting it to right distance so still with phone

Spoiler

Only few caps got good resolution.