M-DCT Lag issue. Owners please post your car info here

[dar2008]

Originally Posted by swamp2

But "magically" your good fortune lasted ~10 days and RJs only overnight or so... Perhaps BMW even wrote the version of the software you have just for you and they wrote it with a virus that gets rid of the lag and then sneaks it back in later on purpose just to anger Audi fans.

Are you mkoesel in disguise or twins separated at birth?

Originally Posted by Radiation Joe

I've been staying out of this little love fest until I get some more miles on the car...

Has anyone heard anything more concrete from BMW themselves on this issue in the last few days? or is it still the "that's how it's supposed to behave" message

[swamp2]

Originally Posted by dar2008

Are you mkoesel in disguise or twins separated at birth?

Sharing of the same logical, reasonable points of view is deserving of sarcasm with a negative tone? Oh well, not in my opinion.

[McCartney]

I just want to put in my 2 cents here,

It is very possible for a gearbox to "learn" about a driver over an extended period of time. Put it this way, 5 megabytes of data to hold data that can be processed in a very specific manner and put through a few Bayesian probability schemes can "learn". The data can be taken at specific points and if the right algorithm is in check it can be efficiently implemented by doing a simple dynamic programming algorithm with a memoization table which allows us to store certain values at a rev/gear/speed and thus over a long period of time can be implemented.

Everything is a computer, EVERYTHING. Including the transmission. Please look up assembly language and compilers to realize that any sort of electronic device is taking instructions from somewhere and processing them, the DCT is a much higher level of this. I wouldn't be surprised if it had a C compiler or the-like.

[jml]

Originally Posted by McCartney

I just want to put in my 2 cents here,

It is very possible for a gearbox to "learn" about a driver over an extended period of time. Put it this way, 5 megabytes of data to hold data that can be processed in a very specific manner and put through a few Bayesian probability schemes can "learn". The data can be taken at specific points and if the right algorithm is in check it can be efficiently implemented by doing a simple dynamic programming algorithm with a memoization table which allows us to store certain values at a rev/gear/speed and thus over a long period of time can be implemented.

Everything is a computer, EVERYTHING. Including the transmission. Please look up assembly language and compilers to realize that any sort of electronic device is taking instructions from somewhere and processing them, the DCT is a much higher level of this. I wouldn't be surprised if it had a C compiler or the-like.

??

[waremark]

Originally Posted by footie

I can't recall whether it was Waremark or not but one of the other UK member also said that his DCT took close on 300~400 miles before the lag reappeared so maybe there is a difference between the setups.

Yes, I did not experience the lag at all for the first 500 or so miles after my October software upgrade to 31.2 - I thought my car did not have the problem, although as a reader of this forum I was looking for it. I would be amazed if I had managed to drive in a style which avoided it for so long, so I think something changed over time - no idea what caused the change.

Currently, I generally drive round the lag, essentially waiting for the clutch to finish re-engaging after a downchange to 3rd or 2nd gear before going back on the gas. I think the problem is always present if you don't drive round it. However, the lag is the only problem with what is otherwise a brilliant transmission and it hardly bothers me.

My personal speculation (if you don't like speculation ignore this) is that the software upgrade increased the time the clutch takes to re-engage when downchanging to 3rd and 2nd in order to make downchanges smoother (downchanges as you brake to a rest are smoother than before the software upgrade) and also probably to eliminate the engine stalling issue. The extra time taken to re-engage the clutch created the lag issue. I suspect that whoever is working on the problem is finding it difficult to come up with a better compromise. Incidentally, I don't like the sound of the noises reported by the couple of forum members who have had a recent upgrade, so I will not be hurrying to get that upgrade.

As I mentioned before, the senior tech at my dealer told me that I was literally the only customer who had raised any concerns of any kind about DCT. It would appear either that our cars are all unlucky (unlikely, I think) or that most DCT buyers accept the transmission as it comes.

I consider comparisons with other brands to be pretty irrelevant. For example, Porsche: it is so much easier to make smooth manual shifts in Porsche's than in M cars that the auto clutch mechanism must have a much easier job to do. And so far as the VW group is concerned, short of the Bugatti, I did not think they had a twin-clutch gearbox which copes with the power level of our cars. Don't they use an SMG style system in the R8?

[swamp2]

Originally Posted by McCartney

I just want to put in my 2 cents here,

It is very possible for a gearbox to "learn" about a driver over an extended period of time. Put it this way, 5 megabytes of data to hold data that can be processed in a very specific manner and put through a few Bayesian probability schemes can "learn". The data can be taken at specific points and if the right algorithm is in check it can be efficiently implemented by doing a simple dynamic programming algorithm with a memoization table which allows us to store certain values at a rev/gear/speed and thus over a long period of time can be implemented.

Everything is a computer, EVERYTHING. Including the transmission. Please look up assembly language and compilers to realize that any sort of electronic device is taking instructions from somewhere and processing them, the DCT is a much higher level of this. I wouldn't be surprised if it had a C compiler or the-like.

Well we are certainly getting sidetracked OT on this point of adaptivity. It is an interesting discussion.

I have always posted about the importance of software for DCTs - it is just about as important as the mechanisms and it can serve to highly customize the DCT experience or to adapt it for a sports car or to an economy car.

I never claimed adaptive programming is impossible or non existent. It is a mature field and can range from look up tables to actual code altering code. I only claim that there is no REAL EMPIRICAL evidence besides this "the lag was not there and then it came back" thing to point to the DCT software being long term adaptive. Similarly, what kind of sinister, performance reducing adaptivity would this be? It just makes no sense - it almost requires a programmer wanting to make sinister code. When we do not have access to the engineers themselves, the use of the scientific method to ground hypothesis about the transmission simply demands that our insights be based on OBSERVATION. Observation is the essence of science and of understanding. So I repeat again - there is NO evidence of any long term adaptation by the M3 ECU nor TCU. Another point is that it doesn't need the great extra complexity on top of 11 existing Drivelogic modes.

I would be willing to bet there is no C compiler in any of the ECU or TCU software. You just do not need it, so it won't be there.

[footie]

waremark and myself have roughly the same opinion of the experiences with our upgrades and what is causing the lag we are experiencing now, we may differ slightly on why it's such a problem for them to cure. I believe that it's maybe a interaction flaw between the car's ECU software and the gearbox which originally caused the stalling and every upgrade since has brought about different problems which can't seem to be overcome without the return of the stalling. It's probably the lesser of the two evils, entering a roundabout with some lag or stalling.

OK, my opinions maybe a little biased due to my experiences with other brands, very one who has had years of experience else where will feel the same way, especially if those experiences were problem free. I just laugh at my continuing battle to fight off this labeling when everything I am saying is relevant, i.e. other brands not/never suffering this problem, in fact, if Ferrari doesn't have the same problems then it's even more damning for BMW. It/the problem may even stem from there efficience drive system, who knows.

I like everyone else have learned to work around the problem but with a £57k car I wouldn't expected to have to that all, should you either?

[McCartney]

Originally Posted by mkoesel

Why would you want a compiler in a device has no reason to store source code? I suppose you could do runtime evaluation (like, Eval(code)) but that is a memory and processor intensive thing to be doing in an embedded application. I would say that any code being executed by the ECU is already compiled, and is simply highly parameterized and process conditionally. No need to be rewriting instructions here.

I'm sorry, what i should have stated more eloquently is I believe that the information that is being held from day-to-day is compiled. An on-the-fly compiler would be inefficient, like you said it would be processor and memory intensive in something that's embedded. What I should have said is that I'm sure there is a higher level language that is used to run the algorithm on the information that's collecting during the operation of the vehicle, which is then compiled periodically and put in a register of the microprocessor that's responsible for handling the DCT, which we know uses assembly. (Don't need to tell you mkoesel you seem well versed), but for other people, think of a compiler as a translator of a programming language (that's not assembly, or VHDL) to a machine-language (which is microprocessor dependent, the machine language used that is).

[alito]

Guys. Really. Stop arguing with each other. We all know it has a lag. We all know it was after the software update. We all know BMW has not fixed it and won't until it has to. We all know BMW does not give a shit and it thinks we are all ignorant people with money to buy their cars. This forum should be to keep each other up to date not argue about wether the trasmission, ECU or whatever learns shift patterns. I just sent a letter to BMW-NA and filed the complaint 10265062 at the site.
I will drop off my car next week at the dealer to have a formal complaint. If no response soon, we can all write to a large newspaper and coordinate it so that it gets printed and BMW learns. If BMW would just comunicate with its customers other that give the impression we are idiots, it would not be so bad.

[jml]

If someone can make it happen dramatically on command, then video it and upload to youtube. Then we should all post it to our facebook accounts and any other more public forums. You'd be surprised the kind of reach you can get that way.

[ygtbsm94]

Sorry, I forgot to change my personal info.

Changed!

[FunRevn]

Looking for an update... I got flashed in November after a SIB notice. I think Im 31.1 A question for those that have been flashed more than once. Is there any improvement over 31.1 ?? from what I have read the lag is still there but are down shifts in general improved??

Thanks.

[alito]

why can't I go to page 18!!

[alito]

what the?!*%& Anyway, I learned today from a BMW "person" that the trasmission keeps track of you habits but is can be preogramed to shift more appropirately based to YOUR habbits but not necessarily on its own. It may help. I'm going to try it. It is about a 3 hour thing.

[swamp2]

Originally Posted by alito

what the?!*%& Anyway, I learned today from a BMW "person" that the trasmission keeps track of you habits but is can be preogramed to shift more appropirately based to YOUR habbits but not necessarily on its own. It may help. I'm going to try it. It is about a 3 hour thing.

Sound interesting but highly unlikely. Without this coming from a very good source (factory or supplier direct) I seriously doubt this.

[mkoesel]

Ok, I waited a few days to post to make sure I am not crazy.

My transmission has defininitely changed its behavior. It now does the following:

- Rejects downshifts to 2nd or 1st when the car is just started and warming up. I have verified 100% that it is not due to potential overrev. I.e this is low RPM shifting.
- Downshifts less smoothly, accompanies by a loud clunk, when doing multi-downshifts off throttle. It's not all the time though, and I can't duplicate it on demand without a few tries.
- Downshifts more smoothly (less violently) under heavy throttle in mid range gears.

This is all S5 behavior - I rarely using anything but S5/S6. The thing is, this all comes just as my drving habits have changed. I had been working from home for the past year so I had not been doing any rush-hour commuting. I am now commuting in heavy traffic, with lots of stop-and-go, more "panic braking", and less predicable gear changes in general.

I am not saying that the car has learned anything, necessarily. However, something has changed. It could be that he transmission has just loosened up some, but that explanation does not really cover the rejected shifts. However, it might be that there is some logic built into the transmission to try to protect itself. In other words, it might change its shifting patterns if it thinks that it is being used in an unpredictable manner that could cause premature wear or poor performance (poor shift times). Even if so though, I think that the algorithms are flawed, and are misreading or misinterpreting driver behavior.

Anyway, if something like this exists, this could potentially be the source of the low-speed lag as well.

Just my 2 cents.

Oh, and I should add that I remember reading a couple posts from others about their transmission beginning to reject downshifts and "clunk" as time went on. I never paid much attention to them though. Now I am curious. I'll try and find them and post links.

[gr8000]

Mine has started to reject upshift commands occasionally a month or so ago. I am not sure if this has to do with any sort of adaptation of the DCT-algorithm or whther the upshift paddle has gone bad. I will continue my observations...

[swamp2]

Originally Posted by mkoesel

Ok, I waited a few days to post to make sure I am not crazy.

My transmission has defininitely changed its behavior. It now does the following:

And you did not get any recent software updates correct?

Perhaps some will think this is a semantic argument, but it really is not. Perhaps the explanation is that the transmission behaves differently when driven differently. Well that is obvious, it does. But the question is why do you ascribe some long term adaptation to the trans when it is you that has changed? Here is the big question. If your driving style reverted back to your old style would the transmission also "revert" back. My point of view on this is again that the unit has no capability for long term/driver specific adaptation.

[CSanto]

2009 M3 DCT:
never lagged, not once, not ever

[footie]

Originally Posted by swamp2

And you did not get any recent software updates correct?

Perhaps some will think this is a semantic argument, but it really is not. Perhaps the explanation is that the transmission behaves differently when driven differently. Well that is obvious, it does. But the question is why do you ascribe some long term adaptation to the trans when it is you that has changed? Here is the big question. If your driving style reverted back to your old style would the transmission also "revert" back. My point of view on this is again that the unit has no capability for long term/driver specific adaptation.

We are all going very close to the situation that I suggested a few weeks/months ago when I said the gearbox software was a learning software. And in this I don't mean holding on to a gear just because you have floored the throttle, I am convinced that the software adapts to the driving style and adjusts to it's changes.

Remember I said that after I got the upgrade done I exclusively drove the car in D-mode for those first two weeks, trying to reproduce the lag but couldn't, yet when I started to introduce the S-mode again and make to shift decisions instead of the gearbox software the lag slowly but surely started to return and within about 4~5 days it was back to normal. Guys I continue to believe that it's a lot more advanced than you think and maybe not for the better because of it.

P.S.

I am really looking forward to the summer and getting to enjoy the M3 to the full. It really is an incredible machine and much better than I ever believed it would be.

[swamp2]

Originally Posted by footie

We are all

"All",

[alito]

Originally Posted by swamp2

Sound interesting but highly unlikely. Without this coming from a very good source (factory or supplier direct) I seriously doubt this.

Coming from the service manager of a large BMW dealer. Take it as you wish. I am going to try it.