Mainboard Suicide after Sleep

[TonyBe]

Hi guys I may as well join the party here

I'm the 'dude from badcaps' that Rasz mentioned who has been looking at the same problem.  My board has been sitting on one side but in light of all the information here i will have another look at it

Mine is the Z87 Extreme 4

Boardview is here

Hi and welcome.

So your Board died in wakeup from sleep as well?

Regards

[dicky96]

Actually no. Or at least I don't know. 

I buy motherboards and computers from the local flea market just to see if they are working or if I can repair them.  I generally look for old motherboards and computers, Pentium 4 socket 478 and earlier (I've picked up some Pentium 3, Slot 1 and even 486 motherboards) and the occasional rare things like a Core 2 Quad/Extreme board with both AGP and PCI-e.  I also buy old GPUs. 

Other than that I buy any gaming motherboards or recent boards.  What I don't bother with is stuff like Dell, HP, Compaq et al and Core 2 Duo/|Pentium D which are everywhere.

I fix this stuff in my spare time when I have no paying jobs in the workshop.  What I fix, or was already working anyway, I either sell on ebay to buy more retro PC stuff for myself, or keep for upgrading my own retro rigs of which I have several now.

This motherboard came from the flea market so I don't know it's history.  It does have the same symptoms though, powering on/off every few seconds.  it does the same if the ATX is forced on but at a higher cycle rate.  To save repeating all the info you can find it here

https://www.badcaps.net/forum/showthread.php?t=79449

[dicky96]

I had another look at my board

I have the same signal SLP_S3# going high low pin 64 on SIO.  On my board this signal only connects to the PCH/North Bridge and a the gate of a FET (AQ3) that looks like it connects into the audio circuit and from the signal name on one on the centre pin DEPOP_CON I think it is meant to stop the speakers popping at power on.

Here is the datasheet for my Z87 PCH.  I tried removing the heatsink and applying pressure to the BGA but no difference.
https://www.intel.com/content/www/us/en/products/docs/chipsets/8-series-chipset-pch-datasheet.html

Has anyone tried lifting pin 64 of the SIO (which is conveniently situated in a corner ) and using a pull up resistor to drive it to 3.3V to see what happens?   That would be my next thought when i am a tthe workshop again tomorrow

Rich

[dicky96]

I'm also thinking this could be a SIO related problem or something entirely different.  As it affects different chipsets (Z87, Z97 at least) it seems a bit odd both chipsets would have the same problem and also wouldn't we be finding the same fault on motherboards from other manufacturers that use those intel chipsets?  So probably we should be looking for some common factor.

Firstly, do all the affected boards have the same SIO?   Maybe the PCH is detecting something that it thinks represents a fault but don't you think it is odd that it would signal that fault to the SIO using SLP_S3# ? 

Also some of the things I tried on the badcaps thread.  Forcing the ATX PSU on does not stop the on-off-on cycling of Vcore but it does increase the frequency of the cycle.  Removing the BIOS makes no difference so it's not like the CPU is signalling the reset. 

On the north bridge the 3.3V and 1.1V supplies are cycling on and off. It seems to me that shortly after the 1.1V Supply to the PCH comes up, it sets SLP_S3# low which powers down the PSU, thus removing 1.1V supply from the PCH so SLP_S3# goes high again and the PSU powers back up.

Forcing SLP_S3# high by lifting pin 64 of SIO has got to be the next thing to try.  If that doesn't cause the power cycling to stop (of if it stops the cycling but the board still won't boot so possibly the timing of  SLP_S3# going high is important) I'm inclined to lash together a simple latch circuit on breadboard, trigger it from the pad of SIO 64 going high and use it to latch pin 64 high.   The aim of this would be to determine if the motherboard can boot, i.e. the hardware is actually good.  If it can, then it seems the PCH is detecting a phantom fault.  This doesn't' fix it but it could be a useful insight

Rich

[TonyBe]

Hi,

well, that's great dicky. Really appreciate your work. Good to know, that they are guys out there like you (especially when the 96 is your year of birth ^.^)

Back to Topic: i've allready used a scalpel to cut a trace very close to the PCH since on my board, several transistors for voltage regulators are switched by the Signal, so i really need to make sure, that only the PCH is deconnected.
I then soldered a thin wire to pin 64 and feed in a 3V3 Signal so simulate a "healthy" SLP_S3#. When pressing the power switch, the on-off symptom does no longer apear. The PSU stays on, the CPU Fan starts spinning slowly and keeps a low speed. But nothing else happens. No CLK Signal @ BIOS Flash. Therefore, i think, that the HBC or the SIO Chip still prevents the MB to properly start booting. Reason being (possibly), that SLP_S5# is also faulty..? never checked that sadly.

I like the idea of finding an common fault... but at the moment, i'm having this issue with the 1.05V Rail at the HBC... and if i cannot find a faulty Cap or something like that, the fault must be inside the HBC. And if this is the case, i'm no longer able to fix the board since the HBC-Chip isn't available or even if so, would be so expensive, that it would not make any sense to spend the money in suspect only.

I'm still stuck on a missing 1150 CPU for further repair work.. but i'm trying to find a common point of all the observations, which clears the possible fault. If there is a bad MLCC on the 1.05V Rail, why should it only clip the DC/DC regulator and not the external Lab Supply,... when feeding externaly, the voltage is stable, so even if there is a faulty cap, it should at least boot up with a higher current on that rail,.. so hm... i'm not quiet sure. :/

Regards
Tony

[dicky96]

OK that's also interesting to hear you tried that.

I may go ahead with my idea to lash up a latch circuit on that pin tomorrow and see what, if anything, occurs

96 is not my birth year by the way, it comes from an email address i set up back in the day many moons ago and somebody else already had dicky69......   

Unfortunately that isn't my birth year either lol

Rich

[Rasz]

but at the moment, i'm having this issue with the 1.05V Rail at the HBC... and if i cannot find a faulty Cap or something like that, the fault must be inside the HBC
If there is a bad MLCC on the 1.05V Rail, why should it only clip the DC/DC regulator and not the external Lab Supply,... when feeding externaly, the voltage is stable, so even if there is a faulty cap, it should at least boot up with a higher current on that rail,.. so hm... i'm not quiet sure. :/

are you still talking about PU9? this is standby supply, powers lan controller and pch in deep sleep
in S0 1.05v is generated from ram power supply (PU3) in linear fashion using power mosfet Q33 (AP9T16) and LM324 as a regulator
for 1.05 to work:
ram supply must be ok - DRAM_PWROK, GR749 between bios chip and sata connectors
LM324 regulation must be ok - 1.05V on Q33

Im in same boat, no spare cpu to test my broken board while in quarantine

[dicky96]

OK so  I spent some more time on this

I lifted pin 64 of the SIO and soldered a 470R resistor to the pin but unfortunately while I was trying to attach the other end of the resistor to 3.3V I broke the pin off the SIO.  Shpould have done the soldering the other way round lol.

Not to be deterred I tried it anyway and find the PSU powers up and stays on.  I now have 1.1V and 3V to the PCH and Vcore to the CPU.  PSON# is low so  I can only assume the broken pin on the SIO is floating high.

However I have no POST.   My POST test card says FFFF or some other random reading which generally suggests to me that the CPU is not running at all

I referred to the datasheet for the Z87 PCH and traced out some signals.   PWROK is good so the PCH is obviously happy that all voltage rails are up.  It's 3V on the PCH and it drops to 1V on the other side of resistor R8614 which signals Power OK to the CPU

But PLTRST# and PLTRST_PROC# are low.  On my board PLTRST# goes to a vacant resistor space which should connect the signal to the SIO and doesn't seem to go anywhere else and as the resistor is not present and I can't see any signs of damage or soldering I assume this signal is not used.  But PLTRST_PROC# is asserted so the CPU never gets out of reset.

Right now I need to order a replacement SIO NCT6776D and I think a PCH as well.  Both are available on aliexpress with good reviews from other buyers and total costs about 20 euros for the two.  I will order today but given the lockdown, Correos (Spanish post) is rather slow here right now.  I did get a delivery form the UK yesterday so obviously they are working to some extent.

The only other thing I can think to try is to lift one of the resistors between PCH and CPU and connect PLTRST_PROC# to PCH 1V supply via a resistor  - to see what happens.... 

Rich

[Rasz]

PWROK is good so the PCH is obviously happy that all voltage rails are up.  It's 3V on the PCH and it drops to 1V on the other side of resistor R8614 which signals Power OK to the CPU

good for standby onlystarting the boot process

But PLTRST# and PLTRST_PROC# are low.  On my board PLTRST# goes to a vacant resistor space which should connect the signal to the SIO and doesn't seem to go anywhere else and as the resistor is not present and I can't see any signs of damage or soldering I assume this signal is not used.  But PLTRST_PROC# is asserted so the CPU never gets out of reset.

how about RP63 PWM_PG on boardview, SYS_PWROK W31 in intel datasheets?
reset wont go up without it

[TonyBe]

Hi guys,

well that's interesting. I really wonder if the RST issue is the same on my board, since the PSU stays on as well, when feeding the SLP_S3# signal externaly. Have you probed the SLP_S5# signal as well? Would be interesting to see if the PHC does send out this signal in a proper fashion or if it signals "power off" as well.

With the change of SIO and PHC, things getting really exciting  right now. 
That way we'll know if the issue is caused by a faulty IC or a peripheral,... at least dickys issue. I'M definetly stay tuned for that update

Thanks guys, has been such a valuable discussion.. even till here!

Regards

[dicky96]

@RASz
On my board RP63 PWM_PG on boardview (SYS_PWROK W31 in intel datasheets) does not connect to anywhere else apart from the pin on the PCH according to the boardview, so I can't probe it and would have to assume it is unused

@TonyBe
Same applies with SLP_S5# signal. it does not go anywhere according to the board view so again seems to be unused.

I put the board back on the bench today and Vcore is going on-off-on where as I swear the other day it was permanently on at about 1.7V  The only thing I did was to put the heatsink back on the PCH before I put the board on one side waiting parts, but taking it back off today made no difference.  I could be wrong about vcore being on yesterday.  3.3V and 1.1V to the PCH are both on (measured on the capacitors mounted on the BGA substrate.

TonyBe - what do you have on Vcore when you force SLP_S3# high, and do you have 3.3V and 1.1V to the PCH?

It will be some time for the PCH and SIO to arrive from Aliexpress :-(  We are 3 weeks into our lockdown tomorrow, with at least 9 more days to go after that and the postal service here seems to be not functioning.

[TonyBe]

Hello again,

this Lock-Down really start's getting harder and harder,... hopefully everthing will be over soon.

Regarding the voltages:
i'm not sure if I can measure those voltages when no CPU is loaded in. I can turn on the PSU manually but since all of those regulators seems to have EN Logic, i'm not sure if the results are valid.
But what i've measured before the CPU was plug on the new MB: the 3.3V were there and also quiet clean. The 1.1V (or in my case 1.05V) were very messy, as described above... But: I'm not sure how the switching between the voltage Rails is done when the MB starts to boot. Maybee those weird ripple effects were due to bad loading condition when the StandyBy Rails are "floating".

I'm still looking for a good 1150 Socket CPU which I can buy for further use (if the MB Repaiur should be succesful).

Regards

[Rasz]

@RASz
On my board RP63 PWM_PG on boardview (SYS_PWROK W31 in intel datasheets) does not connect to anywhere else apart from the pin on the PCH according to the boardview, so I can't probe it and would have to assume it is unused

you can probe it on the resistor marked RP63, its used to enable CPU Vcore sense Vcore PGOOD signal directly from ISL6379

@TonyBe
Same applies with SLP_S5# signal. it does not go anywhere according to the board view so again seems to be unused.

PCH uses SLP_S4# instead

I put the board back on the bench today and Vcore is going on-off-on where as I swear the other day it was permanently on at about 1.7V  The only thing I did was to put the heatsink back on the PCH before I put the board on one side waiting parts, but taking it back off today made no difference.  I could be wrong about vcore being on yesterday.  3.3V and 1.1V to the PCH are both on (measured on the capacitors mounted on the BGA substrate.

PCH detecting problem with Vcore and cutting power?

[dicky96]

I had forgotten Tony doesn't have a CPU for his motherboard.  Mine will not turn on at all without a CPU.

I'll have a look tomorrow to see what is going on with the Vcore,  I should be able to determine what signal is causing it to switch on and off

Rich

[dicky96]

OK, I've been doing another few jobs but I have come back to look at this again

The VRM controller for Vcore is an ISL6379.   I can see various signals flicking on and off (VCORE EN, VC RDY) as I would expect as Vcore is also doing that too - but nothing seems easy in this world as I couldn't find a datasheet for ISL6379. 

Without that it is hard to determine which signal is dependant on which other ones being present first.  In other words is the Vcore being switched on and off by another chip (PCH?)or is the 6379 detecting an error when it does power up Vcore and shutting back down again

I don't suppose anyone could find a datasheet for this chip?

Rich

[Rasz]

solder wires to both and put scope/LA probes on them, the second one should never go out on its own

Once the output voltage is within the
proper window of operation, VR_RDY asserts logic high.
When OCP or OVP occurs in VR0, VR_RDY will be
pulled to low.

close https://xdevs.com/doc/_Datasheets/_DC_DC/isl6366.pdf

would be sad if it turns out to be CPU power section problem

[dicky96]

Hi Rasz

I had another look.  Unfortunately the datasheet you gave me is not similar to my chip I think - the pin outs are all different for a start, looking at the board view

I have VCORE_EN pin 3 of ISL6379 going on and off

Vcore is going on and off and looking at all the ISEN+/- they are all going on and off as well and all seem to be behaving identically (same voltages during the brief on phase) 

I can't get my head around how that happens, or what is going on with dual mosfet QT19  - 2N7002BKS.  VCORE_EN connects to pin 6 of QT19, but pin 5 is high 3.3V constant and all the other pins are 0V.  If VCORE_EN is supposed to be an input why isn't it shown connecting to anywhere else because it can't be controlled via QT19 unless I am losing my head lol? Could you look at the boardview for the Z87 Extreme4 and the 2N7002BKS datasheet and see if you can figure out how this is supposed to work?

However that is just for my own interest, as there is something wrong with Phase 1 of Vcore on my board!.  The waveform on the MOSFET gate looks different from all the other phases and so does the output to the inductor.  If I can find someway to balance my phone in front of my scope I will record some video

Rich


However, at the moment

[TonyBe]

I can't get my head around how that happens, or what is going on with dual mosfet QT19  - 2N7002BKS.  VCORE_EN connects to pin 6 of QT19, but pin 5 is high 3.3V constant and all the other pins are 0V.  If VCORE_EN is supposed to be an input why isn't it shown connecting to anywhere else because it can't be controlled via QT19 unless I am losing my head lol? Could you look at the boardview for the Z87 Extreme4 and the 2N7002BKS datasheet and see if you can figure out how this is supposed to work?

Hi, i'm back for some more guessing

i've checked the Board View from Z87EX4.. When highlighting the VCORE_EN Signal, I can see that this signal connects to QT19, the ISL and several more Caps and also (!) resistors. Since i expect the input impedances from QT19 and the ISL to be very high, i really think that the signal comes from the resistors.
Further: Resistor RP839 is a 12k one pulling the Signal to GND. RP837 Pulls the Signal to ATX+12V. Since this is an 100k resistor, the 12V will be devided to around 1.18V which fits the logic levels quiet well.

So my gues here is, that VCORE_EN is naturaly pulled HIGH and gets switched LOW by one of the transistors. Checking the datasheet of QT19, this seems to be confirmed. Pin 6 of QT19 is the drain of a n-channel FET. Since pin 2 is the gate of the transistor, this pin is the one you really interested in. Checking this pin, you'll see, that it connects to pin 3 (the drain of the second package FET) and to a +12V pull-up resistor. This seems to form a simple FET inverter circuit.
And well,... now i understand why you just checked pin 5 *fail*

The point is, i can only think of QT19 to switch the EN-Signal. Since Pin5 is High, the Transistor should not ground the EN voltage, and therefore, the EN should be high. Have you tried to lift Pin 6 of the SOT-363? If the Transistor is dead, the enable signal could be tight to ground, resulting in no enable for the ISL...?

Really fun to check all this stuff.

See my reconstructed schematic attached. I start thinking, that this logic is working and all that you see is the on/off of the 12V ATX Voltage which results in the Pull-Up resistors have nothing to pull against.

Regards

[dicky96]

Looking at  that diagram it seems to me that OTE_GATE1 going high would ground the gate of QT19A causing VCORE_EN to go to whatever is the voltage on the junction of the 100K/12K resistors - which quick mental arithmatic says is gonna be near to 1.2V

OTP_GATE1 going low would connect VCORE_EN to whatever voltage is on OTP_EN, basically.  As QT19A is effectively in parallel with the 1.2K resistor that voltage is going to be something less that 1.2V

If i remember correctly from this morning VCORE_EN was indeed switching between about  1.2V and 0V.  But all the other pins of QT19 were static as I posted earlier

I definintely have something wrong with Vcore phase 1 so I am going to concentrate on that tomorrow.  At first it looked from the boardview that PWM 1 - 6 from the ISL6379 go nowhere, which does not make sense, until I clicked on the Net names and found they go to six ISL6625ACRZ mosfet drivers on the other side of the PCB.  I will have a look at that area. I suspect that the output mosfets of Vcore phase 1 are getting no high side (and maybe low side) gate drive.

I can imagine how that may cause the ISL6379 PWM controller to have a hissy fit and repeatedly shut down Vcore

[TonyBe]

Hi

Looking at  that diagram it seems to me that OTE_GATE1 going high would ground the gate of QT19A causing VCORE_EN to go to whatever is the voltage on the junction of the 100K/12K resistors - which quick mental arithmatic says is gonna be near to 1.2V

OTP_GATE1 going low would connect VCORE_EN to whatever voltage is on OTP_EN, basically.  As QT19A is effectively in parallel with the 1.2K resistor that voltage is going to be something less that 1.2V

If i remember correctly from this morning VCORE_EN was indeed switching between about  1.2V and 0V.  But all the other pins of QT19 were static as I posted earlier

Yes, that fit's my opinion about the circuit quiet well. The only thing i would add is that OTP_EN is more likely another "Flag" to controll VCORE. It's like a weird "AND" circuit for turnung off VCORE_EN since both, OTE and OTP needs to have a specific state to make VCORE_EN=LOW.

Anyway: I have good news about my MB..
Instead of buying a CPU, i've just figured out, how the CPU_OCCUPIED signal is put together by the MB and add a single Jumper to let the SIO believe in the presence of a CPU.  ^.^
I then connected the PSU, pushed the Powerbutton and yes, indeed: the MB cycles through PSU on/off sequence as before.
Reason for that quick fix was a new assumption that i was burning for to check. With the boardview files in hand, i found the datasheet for the DC/DC regulator (RT8065ZQW) which was connected to the inductor i've desoldered (mentioned earlier, see pictures). As described above, feeding the 1.05V externally has not made a change for me back then. But since i'ce never had a datasheet for the regulator, i couldn't check for any powergood signals or similar.
Well indeed: with my Lab PowerSupply turned to 1.05V, the PowerGood of the RT8065ZQW gives out 0V. I litterally turned the Lab Supply a tiny tiny bit higher to 1.08V to get this Signal coming up to 3.3V.

Reconnecting everthing together, pushing the Powerbutton and there we go... a constantly spinning PSU Fan and healthy SLP# Signals. 

Now i'm really heading for a new CPU to try this puppy...
Still don't know what really causes this 1.05V Rail to getting loaded to that heavy ripple. Could be the DC/DC Regulator, or a faulty Cap... don't know. The current consumption really jumps up and down quiet heavily.

Regards

EDIT: just found out, that i can sample this part.. so i'll do this and then replace it. Let's see if this works.

[dicky96]

Well done Tony

I think my bord may have a different problem to yours as on mine there is no voltage regulator getting hot in that area.

It is possible that I have another pin with an active signal on QT18, and maybe I just didn't get a good connection with my scope probe and thought I had 0V.  I will check again out of interest.  It doesn't make much sense otherwise.

I'm fairly convinced now, that I have a VCore VRM Problem, Phase 1.

I would have done more on this today but the battery on my Van has given up the ghost. It's been acting up for the last week, not charging properly. I had it on a trickle charger on Monday night, and it was OK for 2 days but yesterday I didn't have enough power to start the van again, I had to bump start it.  So I trickle charged the battery again for 18 hours but this morning it still doesn't have enough power to start. Of course it is Good Friday, plus all the shops here are closed anyway apart from food and chemists, due to the lock down.  So the only way to get a battery is to email or message one of the car parts shops, who then get back to you and give you a time to go and collect it.  So no chance to get a battery or go to the workshop today.  Probably they will be closed until Monday (we have Thursday Friday Easter holidays here)

[TonyBe]

Thank you, would never found that problem withour your help guys, so shout out to you!

Well that's a classic murphy isn't it
Hope you'll find a replacement very soon. I'll stay tuned to this thread so no worries.

Have a great time (as far this is possible at the moment)

Regards

[dicky96]

OK so I got my van back on the road.  The shop was closed to the public but they offer a delivery service instead.

I had a look at the motherboard, the MosFETS are marked K03J4 3G2 62D1 lhigh side and K03J5 3C5 60G1 low side. 
The MosFET drivers are ISL6625ACRZ

Looking at Phase 1, it has a longer burst of PWM pulses from the controller ISL6379CRZ  than the other phases, then the Vcore shuts off.  The other phases all look identical

Checking the ISL6625ACRZ on phase 1, I have high side gate drive but no low side gate drive (actually I do have some low side gate drive but it is very much attenuated compared with the other phases)

Also the waveform on pin 8 ISL6625 (Phase) looks horrible compared to the others

I measured the DC resistance from the Ugate and LGate phase 1 and they read the same as form the gates of the other phases.

So it looks like I have a defective ISL6625ACRZ on phase 1, with no low side gate drive.  But it could be defective FETs I guess though I can't see any difference on the diode range of my multimeter when compared to the other phases

I can get the ISL6625ACRZ cheap enough, and also the ISL6379CRZ is not expensive if I want to order that as well.  However i can't find any MosFETs with those numbers... does anyone know what they are?


I suppose I could swap the MosFETS from phase 1 to another phase to eliminate them

Rich

[PKTKS]

MOSFETs (compared to other ICs) are much more easy to replace

You need to find a "compatible" or even "superior" one with
the original proper specifications.

ICs are not that kind to replace.

Paul

[TonyBe]

I can get the ISL6625ACRZ cheap enough, and also the ISL6379CRZ is not expensive if I want to order that as well.  However i can't find any MosFETs with those numbers... does anyone know what they are?

Hi Dicky,

i'm not sure if i'm getting your question right, but the FETs at the Phase-Inductors are completely identified in the Boardview file... You've mentioned
"K03J4 3G2 62D1" and "K03J5 3C5 60G1". That fits the Boardview Information quiet well:

-  N-MOSFET RJK03J4DPA    WPAK(3) RENESAS/30V/4.6MOHM
-  N-MOSFET RJK03J5DPA    WPAK(3) RENESAS/30V/7.2MOHM

But even with that ID, i wasn't able to find a datasheet for that. So i've checked the naming conventions from renesas on their webpage and find out, that the "J" in "03J4" is kind of weird. They used to place a "0" there. So i've replaced the search term with RJK0304DP and there you'll find some information.
Here you go with at least one Datasheet https://www.renesas.com/us/en/doc/products/transistor/003/rej03g1352_rjk0304dpbds.pdf. This Part matches the mini discription of boardview quiet well.

Maybee this helps you a little further my friend

I've received my DCDC replacement IC few minutes ago, but of course: murphy hits me hard. The heating element of my hot air rework station has blown up yesterday... so one more week for getting new rework station. Hm.
I'll post an update if i'm done replacing the regulator.

Regards!