-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
MCU fails for the second time
- Thread starter andyrebele
- Start date
-
- Tags
- mcu Model S touch screen
Sudre
Active Member
I don't want to high-jack this thread but I figured putting the MCU failures in one place is a good thing.
My MCU appears to have failed. Feb 2013 Model S60, Original unit. I think I have around 65K miles. Out of warranty.
It was first thing in the morning. I got in the car, stepped on the break and noticed my music wasn't playing. I hit the skip button (I was listening to music on my USB stick) to see if it'd find another song. It informed me there was an error. I didn't think much of it. I figured maybe my USB drive failed. When I got to the end of the alley I turned on my blinker and there was no sound. Before I went any further I reset the MCU. It never came back on. I was still in gear so I was able to drive to work. Tesla is looking at it now so it might not actually be a failure. I have an Ipad as old as the car and it seems to be doing just fine. If the chat about it failing from flashing or logging is true then this is going to be about an every 5 year thing. A very expensive maintenance issue.
If the unit has failed I hate to say it but I will probably exit the Tesla car market. I have already spent 3K on the A/C compressor DCDC converter this year. Add to that the yearly service costs and I am spending the same cost as buying a new car (that is not a Tesla). The car is too expensive to repair and the failures are too frequent. They have replaced almost everything on my car except the charger and the battery and I bet one of those is next.
I am extremely leery about buying a Model 3 now. I am an avid Tesla bull and hate to be thinking along these lines.
Am I just really unlucky? because the repairs are not slowing down. An MCU failure on a Model 3 is the end of the car and a stranded situation. Plus that cost is about 10% the cost of the base model car.
2017 compressor replacement
2017 DCDC converter replacement
2017 passenger front Door handle backing failure
2017 replaced inner frunk switch
2016 Drive Train replacement
2016 front louver assembly replaced
2016 fuse to 12v jump post blown, replaced (knew it existed but never looked for it much less used)
2016 replaced missing (never installed) condensate drain hose. repaired battery corrosion.
2015 rear axles replaced
2015 repaired rear light condensate issue
2015 replaced driver side rear door handle
2014 replaced driver side front door handle (can't wait for the others to go)
2014 replaced UMC (it melted the inside connectors. I caught it well before serious failure) old adapter.
2014 charger door had a loose control wire. It woudln't open.
2013 Replaced HV junction box.
2013 Install C & D battery & replaced connectors on DC/DC converter
2013 replace CONNECTOR, HV, COOLANT HEATER
2013 replaced 12V TPMS ECU
There are a few others I remember but I must not have saved the report.
and the usual assortment of shimmers, washers and clips and 12 volt batteries.
My MCU appears to have failed. Feb 2013 Model S60, Original unit. I think I have around 65K miles. Out of warranty.
It was first thing in the morning. I got in the car, stepped on the break and noticed my music wasn't playing. I hit the skip button (I was listening to music on my USB stick) to see if it'd find another song. It informed me there was an error. I didn't think much of it. I figured maybe my USB drive failed. When I got to the end of the alley I turned on my blinker and there was no sound. Before I went any further I reset the MCU. It never came back on. I was still in gear so I was able to drive to work. Tesla is looking at it now so it might not actually be a failure. I have an Ipad as old as the car and it seems to be doing just fine. If the chat about it failing from flashing or logging is true then this is going to be about an every 5 year thing. A very expensive maintenance issue.
If the unit has failed I hate to say it but I will probably exit the Tesla car market. I have already spent 3K on the A/C compressor DCDC converter this year. Add to that the yearly service costs and I am spending the same cost as buying a new car (that is not a Tesla). The car is too expensive to repair and the failures are too frequent. They have replaced almost everything on my car except the charger and the battery and I bet one of those is next.
I am extremely leery about buying a Model 3 now. I am an avid Tesla bull and hate to be thinking along these lines.
Am I just really unlucky? because the repairs are not slowing down. An MCU failure on a Model 3 is the end of the car and a stranded situation. Plus that cost is about 10% the cost of the base model car.
2017 compressor replacement
2017 DCDC converter replacement
2017 passenger front Door handle backing failure
2017 replaced inner frunk switch
2016 Drive Train replacement
2016 front louver assembly replaced
2016 fuse to 12v jump post blown, replaced (knew it existed but never looked for it much less used)
2016 replaced missing (never installed) condensate drain hose. repaired battery corrosion.
2015 rear axles replaced
2015 repaired rear light condensate issue
2015 replaced driver side rear door handle
2014 replaced driver side front door handle (can't wait for the others to go)
2014 replaced UMC (it melted the inside connectors. I caught it well before serious failure) old adapter.
2014 charger door had a loose control wire. It woudln't open.
2013 Replaced HV junction box.
2013 Install C & D battery & replaced connectors on DC/DC converter
2013 replace CONNECTOR, HV, COOLANT HEATER
2013 replaced 12V TPMS ECU
There are a few others I remember but I must not have saved the report.
and the usual assortment of shimmers, washers and clips and 12 volt batteries.
Dr. Smoke
My former name was Edmond. Before that, Quantum`.
Yes black screen can mean a corrupted filesystem, or worn-out flash. ext3 isn't fully journaled like ext4 or zfs, so it can get damaged by a reset (two-finger-salute) or by disconnecting the battery at the wrong instant. Hell, even ext4 is touchy.
Lots of bad MCUs are floating around and the fix isn't difficult for someone willing and able to do a little surgery. I haven't seen the spec sheet of this part but some can be reflashed by just clipping on, if there is a disable pin to disconnect it from the bus. If not the chip can be unsoldered by heating one side with an infrared microtorch, lifting with a feeler gauge, then heating the other side. Pop a new flash in the jig designed for that package, and reflash the image appropriate for that car. Use a rosin pen to grease up the pads and pins, and re-heat each side respectively with the microtorch. (Mind where pin 1 is!) It tends to self-center due to surface tension. Use a pair of Coil glasses to look for solder bridges and you're done.
And right, Ingineer has made alot of improvements to the firmware, including ameliorating logging. Tesla now does less logging too; it was necessary in Early Days.
Lots of bad MCUs are floating around and the fix isn't difficult for someone willing and able to do a little surgery. I haven't seen the spec sheet of this part but some can be reflashed by just clipping on, if there is a disable pin to disconnect it from the bus. If not the chip can be unsoldered by heating one side with an infrared microtorch, lifting with a feeler gauge, then heating the other side. Pop a new flash in the jig designed for that package, and reflash the image appropriate for that car. Use a rosin pen to grease up the pads and pins, and re-heat each side respectively with the microtorch. (Mind where pin 1 is!) It tends to self-center due to surface tension. Use a pair of Coil glasses to look for solder bridges and you're done.
And right, Ingineer has made alot of improvements to the firmware, including ameliorating logging. Tesla now does less logging too; it was necessary in Early Days.
Last edited:
Both ext3 and ext4 are fully journaled, whatever you mean by that. In other words they both guarantee metadata consistency in case ther was no hardware failure.
ext4 has a bunch of other features that are mostly unrelated to journaling (like delayed allocation, extents, ...)
If now is less logging than before, I am scared to think how bad it was early on.
That said I compared 7.0 to 17.x and I don't see appreciable reduction in logging. In fact I think it even increased some.
Dr. Smoke
My former name was Edmond. Before that, Quantum`.
I am not having a 'who has the longest peter' contest, so common around here. {pff} When an ext3 system crashes or is reset, the data content of files or directories being modified will be affected. And depending on what that was, it can be fatal to the system. ext3 journaling can unroll attempted changes, but it can't create something that isn't there.
Also ext3 does not do checksumming when writing to the journal. How can it be called full journaling when you have no integrity protocol. ('HOW can ye have any pudding when ye don't eat yer hmeat?!') So what if a reset while the journal is being written? ext3 is nice and I'd used it for many years, but it is not ZFS.
Be frightened I guess. I haven't investigated the inner workings of Tesla myself but Ingineer has, and I have plenty of reason to have confidence in what he says. Best to check with him.
Last edited:
ext3 cannot do unroll, neither can ext4. Similarly ext4 cannot create what is not there on disk (neither can zfs).
Similarly when you write into a file, no matter the fs, since posix does not provide you with any particular transaction API, if you have a power loss the failure mode is undefined!
This could lead to files that are truncated all of a sudden, data not making it or making it in partially and such depending on many factors. On both ext3 and ext4!
ext4 does not do checksumming of journal records either (by default, anyway, I am sure I saw a patch to add it as a special mode - but it was for performance, not consistency, jsut to avoid a small journal header sb update write).
There's no argument that ZFS is more advanced in many ways compared to ext[34] (still not perfect, mind you, and a lot of advancements come at quite a price!), but I specifically disagree with your claims about ext4 journal superiority over ext3.
Hm, I guess I don't want to derail this thread too much so we might start another one if you really want to explore inner workings of journaled filesystems
Last edited:
lunitiks
Cool James & Black Teacher
So, questions from a guy in the market...
A lot of these issues seem related to 1st gen hardware.
Is the hardware different in the AP2.5 hardware?
Will it still have "soldered on, overused flash"?
Are you guys able to "simply" get the control and screen from a "salvaged" tesla.
Are there tesla pick-and-pull type junk-yard services?
A lot of these issues seem related to 1st gen hardware.
Is the hardware different in the AP2.5 hardware?
Will it still have "soldered on, overused flash"?
Are you guys able to "simply" get the control and screen from a "salvaged" tesla.
Are there tesla pick-and-pull type junk-yard services?
hw2.5 is about autopilot only, the actual bigscreen computer is the same 2011 vintage Tegra3 with tiny overused 8G flash.
There are salvages that dismantle teslas and you can get parts from them, the trick is in being able to copy data from your unit to the new one and concealing this fact from Tesla (because they take a very dim view of such things)
The AP2.5 hardware is completely separate from the MCU. While we don't know for 100%, I am 99.999% sure that the emmc is still soldered on.
Sure you can "simply" get a new MCU board from a salvaged Tesla, but you would have to get Tesla, or a hacker, to re-program before it would work in your car. (Someone on here put a new board in his car and took it to Tesla and they refused to program it.)
I'm not aware of a Tesla "pick-n-pull" junk yard, but sometimes they end up in a junk yard, but a lot of them are bought at auction by someone that needs parts and they sell the extra parts online.
Dr. Smoke
My former name was Edmond. Before that, Quantum`.
I didn't even read it because this is a tiresome attitude, verygreen.
It's the reason I don't hang out here.
Barklikeadog
Active Member
I went ahead and called out the important part of your post.
The eMMC is not coming off the board, and a new one ain't going on. Unless @Ingineer knows something several of us do not. Please let me know before I go swap another MCU onto my screen, since as of a couple weeks ago, I'm not allowing my MCU to reboot since it gets stuck in an error loop and may fail completely.
My hypothesis for my memory going back on *THIS* MCU is : Sitting in the sun at the body shop for 4 months
Hi @demundus, If it is failing eMMC I certainly can help you. I'm repairing automotive electronics for many years. Replaced eMMC on my CID and know internals of Tesla very well. Could do either whole CID swap for you if you obtain it elsewhere (will "marry" it to your car) or fix the existing one. @verygreen could confirm it since we were collaborating on uncorking project for some time.
Similar threads
