iPad Air2 image corruption after normal boot, but looks fine in recovery mode?

[JMC]

Hi,
I come from a background of Desktop/Laptop PC repair and only been dabbling in Phone/Tablet repair for a short time and practising as much as i can. The videos that iPadRehab make have been a huge help to me and i'm pleased to say I've had some success with what I've learnt and bought most of the equipment. Usually when there is a clue to help me follow up a fault i'll happily spend hours working out whats wrong, but this time i have a fault that seems to contradict itself and i am in need of some expert guidance.

I have an iPad Air2 that was completely dead but got very hot on the back side. I prodded around the logic board with my finger to find that one corner of the PMIC chip was extremely hot. After watching a Rossman video a while back i put a few drops of IPA alcohol around the PMIC and let it warm up. Two small caps at the corner of the PMIC were bubbling! I tested with a multimeter that these two caps were showing a dead short, so i removed them. Short tested gone! The iPad turned on after doing this and seemed to be fine. I didn't glue the iPad back together incase i had further problems and it was a good job i didn't. I turned it off and then tried it again the next day, but after about 5 minutes there was a pop/spark and crackling so i quickly disconnected the battery.

The PMIC was red hot after trying the iPad once more and the caps at the corner of the PMIC were showing short (including the pads of the cap i removed).

I've done a little bit of BGA practicing but i will admit that this PMIC was scaring me. I bought two PMIC's just in case and began to fit the new chip. Everything seemed to go ok. I only pulled one NC pad while braiding (the lift was perfect to my amazement). Everything else was intact.

I must also say that i removed a number of caps below the PMIC to make it's removal easier. I guess this is a n00b decision to make but i didn't feel confident enough to lift the PMIC without being able to slide a thin bga flat blade underneath it. I'm pretty sure if i tried using a pick at the edge i would have screwed the board up from my past experience!

After putting all of the small caps back that i removed to aid the PMIC replacement, the iPad turned on! Fantastic! ...but after the apple logo the graphics went all weird. Most of the screen was white, but i could see black blocks in one corner (imagine minecraft), graphics artifacting and a portion of the background image in another corner. I could put the iPad to sleep and wake it up just fine, and it would appear that touch was working because i could turn off the iPad by guessing where the "Slide to power off" button was. Occasionally if the iPad is left on a few minutes and you keep waking it up, the screen "tries to start working" and will refresh small parts of the screen (at like 1 frame every 10 seconds or so). When that happens i can catch a glimpse of the battery symbol reading 76% and other things.

Now to the part what really confuses me. Coming from a laptop repair background i would naturally come to the conclusion that either the LCD is at fault, or perhaps during my rework i have damaged important data or power lines for the LCD function. However, when i decided to try booting into Recovery Mode... the picture is perfect! As this iPads data is backed up to iCloud i also attempted to Restore the iPad. During the restore, the apple logo appeared flawlessly, the progress bar appeared normal and moved smoothly. I got no weird graphical glitches or artifacting. After the restore though, the iPad is no different. During the boot process the Apple logo appears normally, but as soon as it switches into the OS, the graphical mess begins.

The fact that the LCD function appears normal in recovery mode has completely baffled me. Why would this be? Is there some LCD/graphics mode switching going on that is producing a fault in normal mode and not in recovery mode?

From my limited knowledge, my understanding is that the recovery mode (along with the bootloader(s)) are burned into the CPU's firmware, whereas iOS and user data reside on the NAND memory. So it is also going through my head, is this corruption of the data on the NAND? If this were true, then why didn't a restore fix it though?

I won't rattle on anymore as my thoughts are going off on tangents, but this is a weird one for me and any help on where to begin would be greatly appreciated.

P.S. i should also mention two more things. Once is that the iPad appears to turn itself on about 2minutes after connecting the battery, which doesn't seem normal to me. Also, after several minutes of the iPad being switched on, it can sometimes reboot itself.

James.

[crea2k]

Dfu mode itself is built into the cpu, as if you remove the nand it'll still go into dfu mode, but it never starts restore as it needs the nand for that so couldn't say if the progress bar is built in too or not. I'd hazard a guess though and say it probably is as the progress bar appears when flashing the nand even before any data has been written to the nand so it can't display something from software if it's not written the software yet.

As for your problem I'd check all of the voltages from the pmic. Also I'd replace the caps on the shorted line as if they were on an image line it may be interference.

[JMC]

Thanks i've got my ZXW yesterday so i will go and check all of the voltages i can and report back.

[JMC]

I've taken some volt measurements from the PMIC, but i'm just seeing if i can find out what the CPU and GPU readings should be. I don't have a known good board to compare unfortunately.

L8103 = PPVDD_CPU = 0.84v
L8106 = PPVDD_GPU = 0.91v

L8150 = PP095_SOC = 0.94v
L8108 = PP1V8_S2R = 1.79v
L8151 = PP1V2_S2R = 1.19v
L8152 = PPVDD_SRAM = 0.95v

L8111 = PP3V3_S2R = 3.3v (C81A2 cap missing on line, it blew away during PMIC rework woops. I will replace this later)

C8177 = PPVCC_MAIN_CPU = 3.87v
C8182 = PPVCC_MAIN_GPU = 3.87v
C8187 = PPVCC_MAIN_SOC = 3.87v
C8190 = PPVCC_MAIN = 3.87v

C81FF = PPBATT_VCC = 3.91v

[crea2k]

The voltages should be on the schemetic

[JMC]

Yep you're right, apparently the GPU, CPU and PPVDD_SRAM lines should be 1.1v. So it looks like i might have a problem here, although why everything seems ok in recovery i don't know. I might probe those voltages while it's in recovery just out of curiosity to see if there is a difference. I'm assuming that on a good board those voltages should stay reasonably static at 1.1v as per the schematic then?

When I've probed laptop CPU voltages in the past they can rise and fall depending on throttling/clockspeed etc, but this isn't a laptop it's all new to me.


EDIT:

Now this is interesting. In recovery mode, CPU voltage is 0.90v, but GPU is 0v!

What the hell? Does the CPU do software rendering or something instead when in recovery, and turn off the GPU? I know i'm dreaming up new ideas here but there is certainly a reason why there is 0v on GPU in recovery... A borked GPU causing artifacts in normal mode? i hope not! It would explain why the first portion of a normal boot with the apple logo looks ok before it takes full advantage of the onboard GPU perhaps.

[Jessa_the_Professa]

Hey there, welcome to the forum.
Yes—as you’ve discovered GPU is not active in recovery and my guess from the description would be GPU processing or GPU line out problems. The voltages you listed all sound normal although you could compare to a known good board. You can have normal apple logo and battery icon that are independent of GPU processing to generate the image at passcode lock screen.

the broader question is what happened to this iPad at the second failure?

PMICs almost never drop dead unless they have physical damage like a crack. Getting a vccmain short on caps adjacent to pmic is common after drop, or short on caps by WiFi, but not having an internal short to the PMIC that develops during use after clearing a vccmain short on a local cap. That's really weird.

So how did this turn into a GPU problem? One idea is that because the PMIC is full of vccmain inputs that all are being used to generate various PMIC outputs, that any internal short of main to ground within the PMIC may have also caused an internal bridge of main to any other line---say PP_GPU. If that was the case, then if the order of events was main bridges to PP_GPU, then to ground, there would be a split second where main voltage went into GPU (which, btw is just a part of CPU chip) and then stopped the instant main shorted to ground. This would mean that for a second, the power hose of vccmain blasted into the delicate GPU and caused some damage. This is just a theory.

Another theory is that the reason the iPad developed the original vccmain cap short was drop--which may have also made a cold joint at GPU. You could try pressing on CPU/GPU to see if that resolves your image problem.

You could also check at the lcd connector to make sure this isn't something simple like an image clock or data line choke knocked off the board at connector or an image reset line fault---diode mode at lcd connector.

I'd pull that PMIC back off this board and reball it. iPad AIR PMIC are expensive and almost nobody bothers with that, if you have invested in some you'd be better off using them to solve a PMIC charging problem later on.

[Jessa_the_Professa]

One more thing---in voltage properties, the voltage listed is the MAX voltage that line will ever see. For example, batt_vcc is 4.6v in voltage properties, but we know that batteries are 3.8-4.2v in day to day operations. So don't think that 1.1v in voltage properties means "you should have this" it means 'you could have this' 0.8-0.9v as a day to day voltage on those lines is perfectly reasonable.

[crea2k]

It's an air 2, the pmic for those are really cheap, you can get 5x air 2 pmic for the same price as 1 air 1.

[JMC]

Hi Jessa,
many thanks for your help. It's so disheartening when you seem to have fixed something as simple as some shorted caps and then it decides to disintegrate a few hours later I've been kicking myself all day wondering if ive created the GPU problem now since replacing the PMIC but i have been so particular about it. I made sure CPU was well heat shielded etc.

It's really odd that the PMIC went short like that afterwards. It was definitely the fault of the PMIC the second time around though as the short disappeared immediately after i removed it.

I've checked the LCD connector this evening. It looks like this iPad uses an embedded displayport that uses 4 sets of wires. I've made sure there is continuity between the connector and all of the edp filters. From what i see the signal then travels through individual capacitors (well i know technically not through them but i suppose this must be an AC signal?). Then they arrive at the GPU (inside the CPU). LCD voltages all look to be present.

The interesting thing i saw was that as well as having 4 sets of EDP data lines, there is an AUX EDP pair of data lines too. I wonder why that is?

My gut feeling has been saying all day the GPU has took it's last breath. I'm going to take your advice and salvage those expensive PMIC's and perhaps just buy a used board to make the iPad usable again.

Many thanks for your advice once again. I'm not sure how many UK Brit's you get coming to your forum, but you are doing a great service,

I have an old iPad 4 that has an iTunes error that i might revisit shortly, so i will post some details about it soon.

James.

[crea2k]

I'm based in the uk too, if you get the pmic off aliexpress you can get them for £12 each which is pretty cheap compared to the £33 for the air 1.