New Battery monitoring software update?

[whitex]

Fully agree with you here. OTA updates was either not on the list of requirements when the Taycan was in development or Porsche thought they'd be able to cobble something together later.

It's the latter. They stuck dedicated OTA hardware in the J1.1 Taycans thinking "it has connections to all different modules, it will be simple - like a virtual remote tech with a PIWIS". Sadly, it's really not that simple. See their own technical doc from 2019 for a glimpse into their original plans/ambitions.

Sponsored

 

[chun]

It's the latter. They stuck dedicated OTA hardware in the J1.1 Taycans thinking "it has connections to all different modules, it will be simple - like a virtual remote tech with a PIWIS". Sadly, it's really not that simple. See their own technical doc from 2019 for a glimpse into their original plans/ambitions.

I can add more details on this, directly from a gold porsche tehnician.

Most ECUs have the following problem: lack of memory for backup, so software would have to be overwritten - risk of failure is huge; and a lot of the ECU software is just files dragged and dropped in a folder through PIWIS, so not exactly a modern safe "install".

So it's not that they can't do OTA updates, they can. But provided anything goes wrong, that ECU is bricked most of the time.

 

[anonymouse]

Most ECUs have the following problem: lack of memory for backup, so software would have to be overwritten - risk of failure huge;

Wow. That just shows how incredibly naive the Taycan’s software writers must have been. Anyone with experience building programmable devices since about 1990 would know that these are critical things to get right.

 

[F1Ruaraidh]

Wow. That just shows how incredibly naive the Taycan’s software writers must have been. Anyone with experience building programmable devices since about 1990 would know that these are critical things to get right.

That's the difference with Automotive - until now with volume, memory was scraped down to the bare bones for everything. Not a penny more than needed.....

 

[F1Ruaraidh]

It's the latter. They stuck dedicated OTA hardware in the J1.1 Taycans thinking "it has connections to all different modules, it will be simple - like a virtual remote tech with a PIWIS". Sadly, it's really not that simple. See their own technical doc from 2019 for a glimpse into their original plans/ambitions.

Interesting doc, is there a link to it somewhere?

 

[W1NGE]

Absolutely, a car can cause a risk of life. Taycan is a 2 ton mass traveling up to 150mph - massive kinetic energy to say the least. Sadly, nobody gets it perfect every time. Hence recalls. Some are hardware, some are software, some are both. Car have become so incredibly complex it is impossible to get it right the first time. That said, you are still getting a significantly safer car, just with possible bugs. You could opt for a Ford Model T, very simple, all the bugs are worked out, but made of wood, no airbags or other safety features (all of those just add complexity, which adds to a chance of having to fix/recall them). Some of the cold war cars also were pretty robust, but again, not many features (safety of otherwise) compared to today's cars. Welcome to the 21st Century!

With all that, I will take a properly designed, implemented and tested OTA over dealer installed update, if for no other reason but uniformity/repeatability of the update process, which can be more thoroughly tested that way, vs. a myriad of ways techs may approach the update, or miss a step, etc.

The root cause of the OTA shambles is simply the myriad of suppliers Porsche used all requiring software input / outputs rather than the one thing Tesla did right and was make it 100% their own (the only positive thing I'd say about the brand).

Lack of enforced standards compounds the issues.

 

[F1Ruaraidh]

The root cause of the OTA shambles is simply the myriad of suppliers Porsche used all requiring software input / outputs rather than the one thing Tesla did right and was make it 100% their own (the only positive thing I'd say about the brand).

Lack of enforced standards compounds the issues.

So do Tesla. There are still lots of 3rd party ECUs on an S/X for example. Even the latest Cybertruck's Steer by Wire is ZF as another example.

Where Tesla win is iron enforcement of HW and SW standards rather than "give me what you've got for the least money possible" approach of traditional OEMs.

This is the route to OTA success

 

[whitex]

I can add more details on this, directly from a gold porsche tehnician.

Most ECUs have the following problem: lack of memory for backup, so software would have to be overwritten - risk of failure is huge; and a lot of the ECU software is just files dragged and dropped in a folder through PIWIS, so not exactly a modern safe "install".

So it's not that they can't do OTA updates, they can. But provided anything goes wrong, that ECU is bricked most of the time.

This is part of the issue. The biggest one is that Porsche didn't enforce standardized (to whatever spec, not necessarily some international body standard) OTA capability from individual vendors of parts. The lack of memory can be worked around, for example the OTA module could read out current firmware, then write the new one, if something goes wrong, write the old one back. There should be some recovery mode should power fail in the middle of an update, though with electric cars that can be mitigated against by ensuring there is sufficient charge in the main battery to run for hours before starting an update. There are tons of side issues, like security and anti-rollback, testing supported combinations of firmwares in the whole system, having actual tests to run after an update, not starting an update if expected conditions are not met - e.g. easy way to disable all Tesla OTAs is to disconnect one of the seat motors, OTA will come back with an ECU fault at start and will not apply, even if there is nothing to update for that ECU - this prevents an update potentially bricking a car. There is a myriad of other system and ECU level considerations, which I'm thinking Porsche engineers haven't quite imagined when planning Taycan OTA. I've worked with OTA, including automotive, and while it looks simple and easy to the end user, the details are are very complex, even more so if bricking it means a truck roll (i.e. tow a car to dealer) which is expensive, so multiplied by number of cars in the field could quickly become a huge liability (and cause a lot of unhappy customers, bad press, etc).

Full disclosure - I've never worked with Porsche. My information is purely from what I found on the internet (like that page from a document in post #151), and my own experience, including what I learned by playing around with a partial Taycan system I assembled on a bench at home from ebay parts while waiting 20 months for my Taycan during the pandemic (lots of time on hand, quarantine restrictions everywhere, impatiently waiting for car - makes for a good motivation to do that). So for example, I learned that the Porsche OTA FC unit is built by a 3rd party vendor (Harman/Becker - which also leads some ebay sellers to market it as an audio amplifier lol). Inside it contains a full fledged computing platform, complete with a dual code ARM SoC with external emmc (flash drive) for the OS and file storage, a decent amount of of DRAM, and a separate automotive microcontroller to offload some communications. Here is what it looks like inside (sorry for thermal paste, I realized I never took a picture of the whole board after I cleaned it off).

It is wired to various buses directly and via the gateway, so theoretically it can do anything the PIWIS software can do and then some. Unfortunately, Porsche software deployment even by human techs is still nowhere near flawless (i.e. techs need to debug, potentially work with engineering to resolve update issues), this means they cannot possibly codify this process for the OTA module to do do automatically.

Anyone know if the OTA FC module even exists anymore in J1.2 Taycans?

 

[whitex]

Interesting doc, is there a link to it somewhere?

AFAIK it was given to techs during initial Taycan training sessions. I found various parts on the internet over the years. There was even a guy on ebay selling copies for lots of money (I think he was asking $3,000 back then, way too much IMO) - I got a few pages from his ebay listing for example, including the complete TOC (see attached pdf). I found bits and pieces of it on the internet since.

 

[whitex]

Wow. That just shows how incredibly naive the Taycan’s software writers must have been. Anyone with experience building programmable devices since about 1990 would know that these are critical things to get right.

The guys who built programmable devices in 1990 have mostly retired by now.

 

[F1Ruaraidh]

The guys who built programmable devices in 1990 have mostly retired by now.

Still here!

 

[simcity]

AFAIK it was given to techs during initial Taycan training sessions. I found various parts on the internet over the years. There was even a guy on ebay selling copies for lots of money (I think he was asking $3,000 back then, way too much IMO) - I got a few pages from his ebay listing for example, including the complete TOC (see attached pdf). I found bits and pieces of it on the internet since.

Looks like the Porsche equivalent of the Audi Self Study Programme (for techs) for Taycan sister car, e-tron GT, which is a handy reference.

 

[f1eng]

The guys who built programmable devices in 1990 have mostly retired by now.

I retired from writing any software in 1988

 

[f1eng]

I retired from writing any software in 1988

Having written that I should clarify that I wrote the first software that earned me money in 1971, so at it for 17 years

 

[simcity]

Having written that I should clarify that I wrote the first software that earned me money in 1971, so at it for 17 years

You part timer! ?

Sponsored