Pontiac Solstice Forum banner

21 - 33 of 33 Posts

·
Super Moderator
Joined
·
15,961 Posts
The tune is data driven. The ECM uses a set of rules to determine where in the various data tables it pulls its settings then applies them. It doesn't "learn" because it can't chang any of its own software or data. Once the "tune" is installed and the ECM boots up, if it's an initial start it goes to the "baseline " initial point and "looks" at the data coming from the MAF on start up during the calibration part of the start sequence and adjusts the data points accordingly. If it's a restart- which is almost always the case, the EVM uses its last stop point as its start point then adjusts for the MAF input

All engines that are in mobile devices have some kind of systems that allow them to self adjust for changes in air density. They must adapt to altitude changes and temperature variations that can affect the fuel air mixture

The ECM can only adapt to changes in perveived environment. No learning going on

As the temp goes up, it will change fuel timing and quantity to match changes of air density to fuel delivered The ECM does this for all the parameters it can affect. No learning going on

This ability to change its operating parameters is limited by start cycle. There are two opposing needs in balance. It must be able to adjust parameters but if it is seeing faulty data it may adjust too much and cause running issues. So the amount of change is limited by "start cycle". 5 or 6 cycles caused the stock tune to eliminate the gains from adding a high flow cat

What all tunes do is mace changes to the data used by the ECM to manage the engine, for example adding more fuel at some settings or adding more timing. The tune "enhances" performance by modifying data in tables

The people who designed the engine and the tune initially hit a very conservative power target that was adequate and had. Very large safety factor for reliability. When the GMPP tune followed later it changed the data to achieve a significant improvement in power, at a still high acetylene factor though it was reduced. See failing intercoolers and the odd broken block.

Other tunes that go to higher power levels do so by changing the data tables and come with some level of increased risk

In addition to changes to the data tables, some tunes modify the rules built into the code. GMPP added the no lift shift for example. The changes in RPM limit are I believe data driven

Unless you replace the operating system on the car, all tunes use the same code. They make data changes based on the experience and skill of the tuner

All tunes bring added risk to the longevity of the powertrain

The more power, the greater the risk

So, one could make a case that all tunes are identical in a very fundamental way, the differences are does a specific tuner really know the car well enough to improve performance with an acceptable level of risk

And

Can I maintain this tune

I chose the RPM tune first and had a bad experience. This was in 2009

I then went to the GMPP tune because it is still maintained by them and while it makes comparable performance improvements it's the least risk solution

I was willing to pay more up front with the knowledge it would be more likely to avoid breaking something later that cost more
 

·
Premium Member
Joined
·
365 Posts
...if it's an initial start it goes to the "baseline " initial point and "looks" at the data coming from the MAF on start up during the calibration part of the start sequence and adjusts the data points accordingly. If it's a restart- which is almost always the case, the ECM uses its last stop point as its start point then adjusts for the MAF input.
Call this nit-picking or semantics, but I thought that was the definition of learning.

Anyhow, it does not explain what I think I experienced - a noticeable increase in performance after about 500 miles of driving. My friend said he experienced the same. Or did he just plant that idea in my head and then I imagined it happened?
 

·
Super Moderator
Joined
·
5,944 Posts
Call this nit-picking or semantics, but I thought that was the definition of learning.

Anyhow, it does not explain what I think I experienced - a noticeable increase in performance after about 500 miles of driving. My friend said he experienced the same. Or did he just plant that idea in my head and then I imagined it happened?
I don't think that you would see a step-change in performance, but I certainly believe that over the first 500 miles the long-term trims are going to optimize and that you will see an improvement during that time.

And yes, that is the definition of learning.
 

·
Super Moderator
Joined
·
15,961 Posts
The trims vairy on a “continuous “ basis but are always performing within the constraints of the code. It’s not artificial intelligence which can “learn”. It’s not my definition of learning when it’s tsking a pre selected input value and giving a pre defined output value.

The code is fixed. The values are fixed. The “tune” matches the values of the inputs to the allowable values of the outputs. Nothing happens that is not defined in the code and the data. It’s a basic machine not a learning device in my definition of learning.

If you go back and read the threads about “learn down” the GM engineers did not want to use the term learn but let people use that short hand to make their understanding easier.

NOTHING happens in the code, the data or the algorithms in the code that is not written down and stored. It’s not “making decisions”. Or figuring things out. It’s a Moore machine. If this, then that.

The normal tune. achieves performance gains by changing the data that defines “then that” to achieve gains. An advanced tune may make changes to the actual code, but to do so means you would need the ability to read the compiled code or have access to a version you could operate on to make changes then be able to compile the chanced code and install it.
 

·
Super Moderator
Joined
·
5,944 Posts
All learning isn't the same, and the definition can change with the application.

The ECM starts out with default values, and uses short-term and long-term trims to adjust its operation to produce the desired result based on measurements that it makes. Those trims are retained for use the next time it runs because it has learned its environment and the ways that its system varies from the assumed nominal system. It has therefore learned.

It is no different than your behavior when you perform a particular task, but aren't happy with the result. You may do exactly the same thing the second time you do it, but vary the force or speed that you do it. Your "program" hasn't changed, but you learned that your initial parameters weren't optimum, so you adjusted them.

The fact that unpowering the ECM for a period of time "resets" it to the defaults demonstrates that it has "learned" something, otherwise it couldn't "forget" anything.
 

·
Registered
Joined
·
3,770 Posts
I'm not saying that anyone should be able to copy and distribute somebody else's work, and I certainly don't have an issue with profiting from one's labors. I don't think everything should "free." If I feel a product isn't worth the asking price, I simply won't buy it. I've never asked anyone to just give me a tune for free.

However: I shouldn't have to cough up more cash and start over with another vendor if Trifecta blows me off - a pattern that they've demonstrated multiple times as documented elsewhere. If *I* want to modify *my* tune that I paid for, on *my* car, I shouldn't need someone else's permission.

And, of course: because their tune is locked, you can't just overwrite it. We recently dealt with a friend's car that had a Trifecta tune and were unable to overwrite it. I'm not even talking about reading the thing - we couldn't *write* to the ECM. The friend in question took the ECM to GM, who attempted to reprogram it, and wound up with a basically useless device.

That is, in technical parlance, Not Cool.

We went back and forth with Trifecta over a couple of weeks - because they don't publish a freaking phone number and take their sweet time responding - and were basically told to eff off.



I work in software these days - if piracy were as rampant as everyone thinks that it is, I wouldn't have a job. That said, software patents are BS. The people that steal the tunes weren't going to pay for them anyway. Point of evidence: see all of the other tuners that do NOT lock their tunes and are (gasp!) still in business.



People also blame Starbucks for spilling coffee on themselves. People put blocks of dry ice on their engine to cool it off between drag runs. People chop off their hands with table saws. Ask Gilbert or Martin when the last time was that they got sued over a blown engine.

The idea that locking a tune reduces or eliminates liability is not supported by evidence.

But, hey - to each his/her own.
This is very interesting... :lurk: Did you know that there is a push from GM to not allow anyone to tune their cars? If this gets passed in courts, and you take your car for service at a dealer, and they find out that you have touched the tune, you get a HEFTY fine!!

Don't believe me? Look for John Deere. They have been taking farmers to court over the "tunes" they're putting in their farm equipment, and it's setting a precedence that GM can follow through with as well as other OEs. If it passes, it would treat your vehicle software like it does any other software...that you decompile and modify it. It will also treat your vehicle, the way John Deere has won in court, that you merely "lease" the car, whether purchased or not, for it's usable life and aren't allowed to modify it in any way.... John Deere technically now still owns all farm equipment "sold" to people now.... Do I agree with it? Yes...it's software. If we want to have laws governing computer software that was painstakingly developed, why shouldn't it apply to vehicle software?
 

·
Premium Member
Joined
·
1,660 Posts
Yeah, I’ve read that here and elsewhere.

That’s going to be very difficult to enforce. They could do it during annual safety inspections, I suppose.

I’ll just buy a second ECM.

As if I needed another reason to not let anyone else work on my cars...

At the end of the day, there will always be options. Standalone harness and ECM are easily available and not crazy expensive. A bit much for a $7k car, but for any serious application, a couple of grand isn’t unreasonable.


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
3,770 Posts
All learning isn't the same, and the definition can change with the application.

The ECM starts out with default values, and uses short-term and long-term trims to adjust its operation to produce the desired result based on measurements that it makes. Those trims are retained for use the next time it runs because it has learned its environment and the ways that its system varies from the assumed nominal system. It has therefore learned.

It is no different than your behavior when you perform a particular task, but aren't happy with the result. You may do exactly the same thing the second time you do it, but vary the force or speed that you do it. Your "program" hasn't changed, but you learned that your initial parameters weren't optimum, so you adjusted them.

The fact that unpowering the ECM for a period of time "resets" it to the defaults demonstrates that it has "learned" something, otherwise it couldn't "forget" anything.
I completely agree John!!

Newer vehicles DO learn. My Charger's TCM learns. Otherwise it wouldn't leave itself in 6th gear when I'm trying to get fuel mileage. It has learned my driving habits (yeah...I'm a bit heavy footed in it!). It knows that I like power and will downshift manually if I have to, so now, it seems to keep itself in 6th until I hit 50mph. Then 7th at 50, and 8th at 65+. And that's in economy mode. Switch to performance and it's 1 gear lower and never hits 8th gear. My understanding from the Charger forums is that it "develops" a sort of graph and tries to follow it...much like programming the AFR in our cars.

Also with AEVs coming, AI is being implemented more and more, even though we are still years off from full AI.
 

·
Registered
Joined
·
3,770 Posts
As if I needed another reason to not let anyone else work on my cars...

At the end of the day, there will always be options.

Sent from my iPhone using Tapatalk
There will not be options if GM and John Deere get their way. Nobody will be able to access any of the data that's stored in your ECM except a dealership. That information will be sent ONLY to them...and nobody else will have the tools to fix your car. If you guys don't want this, I would highly suggest getting involved in the fight!! Follow MEMA news and start watching articles about it.
 

·
Super Moderator
Joined
·
15,961 Posts
All learning isn't the same, and the definition can change with the application.

The ECM starts out with default values, and uses short-term and long-term trims to adjust its operation to produce the desired result based on measurements that it makes. Those trims are retained for use the next time it runs because it has learned its environment and the ways that its system varies from the assumed nominal system. It has therefore learned.

It is no different than your behavior when you perform a particular task, but aren't happy with the result. You may do exactly the same thing the second time you do it, but vary the force or speed that you do it. Your "program" hasn't changed, but you learned that your initial parameters weren't optimum, so you adjusted them.

The fact that unpowering the ECM for a period of time "resets" it to the defaults demonstrates that it has "learned" something, otherwise it couldn't "forget" anything.
I agree. You stated the behavior well. We just differ on what we call "learning" In my world, I do software design for complex space systems, code that "learns" modifies itself to achieve outcomes that were not foreseen by the coders. The code creates the rules that allow opportunities for self writing code to exist, but you cant always predict where the software will need to go under all circumstances.

In an air breathing earth bound human controlled vehicle, as opposed to an orbiting autonomous vehicle, the allowable "learning" is highly constrained. To me its the difference between creating boundary conditions and allowing the code to travel where it needs to go to achieve its objectives, and creating a set of rules with no boundaries and the code can modify itself as it sees fit.
 

·
Super Moderator
Joined
·
5,944 Posts
I agree. You stated the behavior well. We just differ on what we call "learning" .......
Seems like a contradiction. Is it only learning if you gain a new skill or ability? Or does learning also include gaining the knowledge to do the same thing better?
 

·
Super Moderator
Joined
·
5,944 Posts
This is very interesting... :lurk: Did you know that there is a push from GM to not allow anyone to tune their cars? If this gets passed in courts, and you take your car for service at a dealer, and they find out that you have touched the tune, you get a HEFTY fine!!

Don't believe me? Look for John Deere. They have been taking farmers to court over the "tunes" they're putting in their farm equipment, and it's setting a precedence that GM can follow through with as well as other OEs. If it passes, it would treat your vehicle software like it does any other software...that you decompile and modify it. It will also treat your vehicle, the way John Deere has won in court, that you merely "lease" the car, whether purchased or not, for it's usable life and aren't allowed to modify it in any way.... John Deere technically now still owns all farm equipment "sold" to people now.... Do I agree with it? Yes...it's software. If we want to have laws governing computer software that was painstakingly developed, why shouldn't it apply to vehicle software?
I have been following the John Deere situation and unless I missed this getting overturned I think you are behind the times.

In October 2018 the Library of Congress adopted an exemption to the DMCA specifically to make it legal to circumvent access controls for "non-infringing" purposes such as repair.

Further, Minnesota seems to be about to pass a Right-to-Repair law of its own.

I don't think that either of these compels manufacturers to provide the tools or information needed to do it, but they do make it legal.
 

Attachments

21 - 33 of 33 Posts
Top