Adaptive Learning

[EVI]

Could someone please explain what adaptive learning learns and what it adjust.
Thanks

[joecar]

Generally speaking:

It learns the shift "feel" and time, and trims it by controlling pressure and rate of pressure (ramp up/down).

There's more to it than that, but that is as much as I know.

[EVI]

Thanks for the reply but what is the target? Does it try to adjust to a pre-programmed target, like AFR and if so can I change the target? Should I disable adaptive learn for up-shifts if I want full time firmer up-shifts?

[joecar]

The targets are typically these:
- shift time tables,
- shift pressure tables;

No, you don't necessarily disable adaptive learn for firmer shift (the shift will be hard rather than firm)... instead you modify the targets...

WARNING:

The 6L80/6L90/6T75 transmissions operate on a clutch-to-clutch basis...
i.e. the TCM controls the timing between the releasing component (i.e. for the gear it's coming from) and the applying component (i.e. for gear it's going into)...
if this release/apply timing gets messed up the transmission will suffer internal damage.

You will need to get more guidance on how to do this safely from someone who has gotten good results (I haven't had the chance to play with these).

[EVI]

OK, I don't mean to beat a dead horse, I just want to thoroughly understand so I can be a better tuner. That being said, I still don't really understand why we need adaptive learn. If the target for shift time is say .500 sec then why not just command .500 sec like on the 4l60e. There are no variables like baro or air temp or humidity like with AFR. It seems to me that if you don't disable adaptive learn, then whether your tune is for a hard shifts or soft shifts you would end up with the same feel as you started with after adaptive learn gets the values back to whatever it's target is. Don't give up on me now Joecar.

[CV8IIM6]

I’m certainly no expert but here’s my understanding:

If you command a shift time of say 0.325 seconds and a base line pressure of say 1250Nm, when the TCM performs that shift it may only achieve a shift time of .600
This could be because of individual clutch characteristics, line pressure may not be high enough or perhaps torque reduction is not right etc. etc........

If adaptive shift is OFF the TCM won’t know and/or care that it’s not achieving the desired shift time and will continue getting it wrong.

With adaptive shift ON the TCM monitors “actual shift time” and compares it to “desired shift time”
Then if they don’t match the TCM will take steps (like adjusting line pressures) to correct the discrepancy.

I have found it generally takes about five shifts in each cell to learn the tune, this can take several drive cycles or up to 2 weeks.

I think the 4L60E has to learn to some extent too but someone who thoroughly understands the internal workings of both transmissions may be able to shed some light on why it seems the 6L80E has to learn so much more.

[oztracktuning]

Turn off adaptive learning and face high risk of the numbers not being anywhere near right and that will be costly.
These TCMs are able to do incredible adaptation when the calibrations are right. They can be improved in their shift feel incredibly and last. But turning off adapts is a dead end road.

[Patrick G]

Any idea why the 2-3 shift max adapt volume (D9264) is 40 while the rest of the shifts are 200? Would this narrowed learning volume explain why some of us have light load 2-3 shift flares, but solid heavy load 2-3 shifts? Would opening up the adaptive volume to 200 help the 2-3 shifts?

[joecar]

Any idea why the 2-3 shift max adapt volume (D9264) is 40 while the rest of the shifts are 200? Would this narrowed learning volume explain why some of us have light load 2-3 shift flares, but solid heavy load 2-3 shifts? Would opening up the adaptive volume to 200 help the 2-3 shifts?That is very odd...

I wonder if there are any T43 calibrations having D9264 set to other than 40 cm³...?

[joecar]

OK, I don't mean to beat a dead horse, I just want to thoroughly understand so I can be a better tuner. That being said, I still don't really understand why we need adaptive learn. If the target for shift time is say .500 sec then why not just command .500 sec like on the 4l60e. There are no variables like baro or air temp or humidity like with AFR. It seems to me that if you don't disable adaptive learn, then whether your tune is for a hard shifts or soft shifts you would end up with the same feel as you started with after adaptive learn gets the values back to whatever it's target is. Don't give up on me now Joecar.If shift learning is turned off, the trans will just command each shift using line pressure from the shift pressure tables (shift pressure vs engine torque) without regards to timing or ramping...

In the 6L80 this will not be good (release/apply of components is performed by multiple hydraulic circuits, so release/apply timing is a function of the TCM calibration/algorithm)... if the release/apply components are "stepped" off/on (respectively) at their natural physical speed (which is fast), then component binding would occur (the releasing component takes longer to release than the apply component takes to apply... in effect this would give you a trans brake effect for a brief moment), this will quickly and efficiently damage internal components.

In the 4L60E/4L80E this will be fine (release/apply of components is performed by a single hydraulic circuit, so release/apply timing is a function of the hydraulic/mechanical physics (i.e. the volume of the accumulators, the size of holes in the VB separator plate, the servo piston sizes))... there will be no release/apply component binding... if the shift pressure tables have sensible pressure vs torque values, each upshift can be very quick/hard... the shift pressure tables by themselves command line pressure with no timing trimming, this does not constitute learning (you can see this when you set the shift time tables to zero or when you disable the MAF for example).

Adaptive learn makes an attempt to converge to the values in the shift time and pressure tables... you adjust these to get firmer shifts, but you have to know how much adjustment you can get away with...

And then there are other tables to consider such as max adapt volume as PatrickG pointed out.

Why not just command 0.500 s on a shift with no learned timing feedback/trimming...? Commanding 0.500 s is quite slow and may actually be ok... but commanding 0.100 s is fairly quick and may not be ok (miss-timing of release/apply components)... on the 6L80 shift learning is everything.

I'm not giving up on you (I don't have one of these to play with)... I am seriously saying you have to be careful with this.

[oztracktuning]

The 40 may be an error or it may be for a reason, but it is possible to get rid of that 2-3 flare by changing the tune calibration. Changing that 40 is one of the things that i do. But many other things are done as well and i think it can get absorbed and fixed by the learning. My TT 6L doesnt flare anymore 2-3

[Patrick G]

The 40 may be an error or it may be for a reason, but it is possible to get rid of that 2-3 flare by changing the tune calibration. Changing that 40 is one of the things that i do. But many other things are done as well and i think it can get absorbed and fixed by the learning. My TT 6L doesnt flare anymore 2-3Not an error. Every single 6L80E I have ever tuned has the same 2-3 adaptive at 40. Every G8 GT, G8 GXP, 2010 Camaro, Yukon Denali, Cadillac Escalade, Cadillac XLR-V, Corvette, and Silverado I've tuned has it stock at 40. Very strange. I'll open it up and see if it helps with the pressure and shift timing changes I've made for the 2-3 shift.

[oztracktuning]

Thats even more the reason it could be an error. They probably just transfer previous settings into any newer calibrations and things like that never change. But on the other hand it could be there for a reason.

[joecar]

It can't be an error, some enterprising engineer at GM Powertrain would have picked it up by now...

[oztracktuning]

You are probably right , however there are errors made and left. eg have a look at the high octane table in any LS3 tune, there are errors in it.

[CV8IIM6]

It can't be an error, some enterprising engineer at GM Powertrain would have picked it up by now...

I assume your being sarcastic?

I flash updated calibrations into various on board modules every single day at work because of enterprising engineers f!@K ups. Ha Ha (no offence intended to anyone)

ECM, TCM, PCM, BCM, CIM, RFA, REC, RSE, SRS, ABS, EBCM, HVAC, IRC, IPC....you name it, they'll f!@k it! :doh2:

Having said that, engineers at holden have told service managers back in April 2009 that the intermittent 2-3 shift flare and subsequent thump on following gear change is due to faulty hardware inside the box. Hardware changes were made and vehicles built after March 2009 should not be affected.
Unfortunately dealerships are NOT allowed to simply replace transmissions on earlier vehicles for this complaint as holden is still trying to find the most cost effective solution(cheapest way out of the shit).:secret:

I hope they are wrong again (early on there were TCM calibration updates and various other excuses as to what causes this problem).

I'll try changing D9264 and see what happens.:unsure:

[joecar]

Yes, I was being sarcastic... :hihi:

You left out SDM... :angel_innocent:

[EVI]

Thanks guys, you've given me a better understanding and that's exactly what I was looking for.

[A6sportsracer]

From what I have read the adaptive learning also keeps the shift firmness the same from new to 100,000 miles, adjusting volume and pressure to maintain the programmed shift times. I don't think there is any advantage to turning it off.

I have also adjusted the 2>3 shift volume to 200 getting a better 2>3 shift.

If you adjust the shift times to tight you will get a DTC indicating pour shift performance or desired performance could not be achievced...because the solinoids or the fluid flow could not achieve the desired time

[Tre-Cool]

digging up a slightly old thread, but can anyone explain to me the difference in the dvt controlls for the t43 in regards to:

Adapts Reset vs Adapts Preset
Fast Reset vs Fast Learn

i did the top 2 before i left work today and my shifts went to shit, it only got better as i got home. roughly 50km's of driving.

good point so far is i have noticed the flare between 2nd-3rd is almost gone.

[Tre-Cool]

oh. thanks to who ever added in the self cleaning parameters to the t43 cal too as i have a large trans cooler it's take a long time for my car to reach the default 70c temp setting. i can now drop it down and perform the test.

[gmh308]

Hopefully this makes sense in general terms... :)

Adaptive learn is intended to normalize the trans physical state so that the electronics can control it consistently. i.e. either through factory tolerance stack up or wear and tear over time the dimensions of the clutches, both the fibre line and the metal plates change, as do ports/orifices in hydraulic components.

Learn factors ensure that these changes or differences always look the same to the electronic calibration/operation and that over 100,000 miles or more, provided the wear has not exceeded the compensation limits that learn provides the TCM will compensate for wear by correcting the calibration parameters with the learn factors and then piston fill rates/volumes for clutch operation are adjusted by the stored learn factors.

[gmh308]

digging up a slightly old thread, but can anyone explain to me the difference in the dvt controlls for the t43 in regards to:

Adapts Reset vs Adapts Preset
Fast Reset vs Fast Learn

i did the top 2 before i left work today and my shifts went to shit, it only got better as i got home. roughly 50km's of driving.

good point so far is i have noticed the flare between 2nd-3rd is almost gone.

Adapts reset clears adapts to "zero"
Adapts preset sets adapts to a pre-calibrated factor that the factory worked out for a new trans.
Fast learn runs the built in learn sequence while standing still in a workshop so that a 50km drive is not needed. :)
Fast reset - faster version of adapts reset? Or maybe a TCM reboot. Not mentioned in any service docs :).

[gmh308]

I assume your being sarcastic?

Having said that, engineers at holden have told service managers back in April 2009 that the intermittent 2-3 shift flare and subsequent thump on following gear change is due to faulty hardware inside the box. Hardware changes were made and vehicles built after March 2009 should not be affected.


LOL yes. They finally figured out after 3 - 4 years of A6 trans's that a seal was faulty or badly installed on some (a lot?) trans's.

On the 08 6L90E I run there arent any flare issues. Very lucky I woul say. Have not changed the 2 > 3 Max Adapt Shift Volume. Still stock at 40. Though this is the same on all A6's normally from the factory at 40 as Patrick G pointed out, GM did change it to 50 for the CTS-V 6L90E. Some people change it to 200 and claim a difference. Maybe it does help with the flare issue due to the bad seal. My understanding of these numbers is that they are simply limits for the adapt compensation. The 2 > 3 shift uses 2 oncoming clutch sets and one offgoing so the number is not about a single clutch, and all the pistons have different fill amounts which are nowhere near 200cc even if they are dead empty which they are not. Lucky if they exceed 100cc on the larger diameter clutches like the 1-2-3-4 clutch.

[TBMSport]

If shift learning is turned off, the trans will just command each shift using line pressure from the shift pressure tables (shift pressure vs engine torque) without regards to timing or ramping...

In the 6L80 this will not be good (release/apply of components is performed by multiple hydraulic circuits, so release/apply timing is a function of the TCM calibration/algorithm)... if the release/apply components are "stepped" off/on (respectively) at their natural physical speed (which is fast), then component binding would occur (the releasing component takes longer to release than the apply component takes to apply... in effect this would give you a trans brake effect for a brief moment), this will quickly and efficiently damage internal components.

In the 4L60E/4L80E this will be fine (release/apply of components is performed by a single hydraulic circuit, so release/apply timing is a function of the hydraulic/mechanical physics (i.e. the volume of the accumulators, the size of holes in the VB separator plate, the servo piston sizes))... there will be no release/apply component binding... if the shift pressure tables have sensible pressure vs torque values, each upshift can be very quick/hard... the shift pressure tables by themselves command line pressure with no timing trimming, this does not constitute learning (you can see this when you set the shift time tables to zero or when you disable the MAF for example).

Adaptive learn makes an attempt to converge to the values in the shift time and pressure tables... you adjust these to get firmer shifts, but you have to know how much adjustment you can get away with...



The explanations put forth for Adaptive Learning (while informative) still didn't provide me with a full grasp as to what it does. So I went searching for more info. As taken from a Corvette site.

"Below is the actual description from Service information. Just remember, under normal circumstances, the transmission will default to second gear in certain conditions but LEARNING IS CONSTANT>

The 6L80 transmission utilizes a line pressure control system during upshifts to compensate for the normal wear of transmission components. As the apply components within the transmission wear or change over time, shift time (the time required to apply a clutch) increase or decreases. In order to compensate for these changes, the transmission control module (TCM) adjusts the pressure commands to the various PC solenoids, to maintain the originally calibrated shift timing. The automatic adjusting process is referred to as "adaptive learning" and it is used to ensure consistent shift feel plus increase transmission durability. The TCM monitors the A/T input speed sensor (ISS) and the A/T output speed sensor (OSS) during commanded shifts to determine if a shift is occurring too fast (harsh) or too slow (soft) and adjusts the corresponding pressure control (PC) solenoid signal to maintain the set shift feel.

The purpose of the adapt function is to automatically compensate the shift quality for the various vehicle shift control systems. The adapt function is a continuous process that will help to maintain optimal shift quality throughout the life of the vehicle."

Based on this definition and the highly modified tunes that I am playing with, I am finding some trends and patterns and struggling to develop a theory of its application and operation. I believe that the higher the Adaptive Learning numbers (D9500 and on) are, the more aggressively the transmission will attempt to soften shifts...felt by the end user as a delay in upshift. Obviously, lower Adaptive Learning number (D5000 and on) are, the less aggressive the transmission's attempt to soften shift..i.e.quick shifts.

How all of that convergence translates into a process that can explain the affects of line pressure...well I'm at a loss. Maybe by setting a higher line pressure target to reference, knowing that it takes extreme driving conditions to achieve, the transmission sees minimum clutch wear and keeps a rather minimum adjustment of clutches engagement. Conversely, by setting lower line pressure targets means more rapid clutch wear and and therefore constant adjustments to the clutches engagements. Perhaps at this point it refers to other T43 settings, which being vehicle specific, may be set up for lazy/clutch protecting shifts based on engine load, vehicle speed and TPS. Or maybe it adjusts all/some other T43 shift settings to those Adaptive Learning line pressure targets. So the smaller the standard deviation for ALL T43 settings the more consistent the shifts will be under all driving conditions, rather than hit or miss.

At the end of the day, I think the key for any 'tuner' to understand about these transmissions is that they are EXCEPTIONALLY intelligent. And with all 'intelligence' it requires a great deal of information AND accurate application of that information. That said, all the various parameters that we tweak and adjust throughout a calibration have to be considered within the context of the entire calibration. In other words, making an XYZ adjustment to a given parameter for a 2010 Camaro may not effectively give the same results for a 2008 Escalade as the T43 looks at that particular parameter as a whole of its hundreds of other data parameters. Changes therefore need to be applied to broader methodology, rather than change this cell to this or that.