Turbo Dynamics, simply a PID control system

[Deer Killer]

Seriously. I answered both your questions.

 

The thresholds and activations are clearly spelled out here:

 

http://www.romraider.com/forum/topic1814.html

 

The only thing that merchgod didn't give is the integrating time interval and I gave that to you.

 

I chose 0.77 psi because it is less than 0.97 psi. And 0.97 psi is in the rom. Frankly, it doesn't matter. You can pick 0, -5, -10. It doesn't matter. Once your target boost is below the disable, boost control is OPEN LOOP and the boost target is no longer used.

 

Forgot to add: The real target boost is in terms of absolute pressure. It can never be negative. The target boost as DISPLAYED in enginuity is in terms of MRP at sea level. It is only displayed like that for convenience. The ecu target boost is in terms of absolute.

 

"I estimated it at 70-80 ms for the 05 LGT ecu." This is not an answer. This might be interesting, it might be saying with your values it takes 70-80ms to fill the accumulator until you've reached the max correction threshold. This doesn't say much though. It's not clear if the integrator works on a window or gives equal weight to all values since the last time the accumulator was cleared.

 

An answer to the second question might be "The accumulator never clears unless you hit a cell which has less than the activation threshold", but I'm not hearing that either.

[mickeyd2005]

"I estimated it at 70-80 ms for the 05 LGT ecu." This is not an answer. This might be interesting, it might be saying with your values it takes 70-80ms to fill the accumulator until you've reached the max correction threshold. This doesn't say much though. It's not clear if the integrator works on a window or gives equal weight to all values since the last time the accumulator was cleared.

 

An answer to the second question might be "The accumulator never clears unless you hit a cell which has less than the activation threshold", but I'm not hearing that either.

 

I don't think you know how TD integral works.

 

Read merchgod's explanation. You'll figure out that you need to know the integration RATE. That's 70-80 ms.

 

As far as the second question is concerned... are you arguing for the sake of arguing? I answered your question and you gave it back to me. So, what's the problem?

 

Set your boost target to less than boost threshold to clear the TD integral. So, in the "landing" areas I set boost target = 0.77 psi. This is actually important if you do fast shifting. You don't want TD integral to already be wound up as you boost up again.

[mickeyd2005]

It's not clear if the integrator works on a window or gives equal weight to all values since the last time the accumulator was cleared.

 

I don't know why you don't read merchgod's explanation.

 

This last sentence you gave tells me that you really don't know how TD integral works.

 

TD integral does not learn past the reset. Once it is reset, it starts all over again from zero so there is no weighting involved.

[2005garnetGT]

you guys are funny.

 

I don't post my logs because my car runs fine and I am not looking for advice - I set it up 3 years ago and just drive it

I'm of the same opinion; I haven't changed my setup in 2 years and I havent logged it for over a year.

 

I made one change when the rev limit boost fuel cut resume param came out, and besides that I haven't changed a thing for 2 years

[SeeeeeYa]

No you don't!

 

What are we on this site for, pure entertainment? Friendship?

 

I am here to learn about my car, get advice about modify my car, etc... IMO it is why the site exists.

 

 

Some people like to talk about the king of cars and turbos that don't fit in their garage. Then they press repeat a thousand times. Why?

 

If you want to buy a car and drive it, then fine. If you want to be a main contributor in an enthusiasts site, be an enthusiasts, which to me generally means always tinkering, never satisfied, always trying to make it better.

 

I said "almost" .

 

Your reaction reveals an xray of my own inner, immutable, way of life. Sometimes it is important to clear the room of those who contribute nothing but noise.

[rao]

That's a great point - what could a person who set up there car exactly the way they wanted it and then enjoyed it as it performed flawlessly for years possibly add to any enthusiast site Only someone who is down with the struggle and needs a few years to tune their car has any knowledge about anything

 

Once again you guys are hilarious.

 

And one thing that I did learn on t his site is that the Legacy is the King of Cars!!!! It has been described as better than Corvettes, BMW's, Lamborghinis and pretty much every other car in the past or present

[SeeeeeYa]

That's a great point - what could a person who set up there car exactly the way they wanted it and then enjoyed it as it performed flawlessly for years possibly add to any enthusiast site Only someone who is down with the struggle and needs a few years to tune their car has any knowledge about anything

 

Once again you guys are hilarious.

 

And one thing that I did learn on t his site is that the Legacy is the King of Cars!!!! It has been described as better than Corvettes, BMW's, Lamborghinis and pretty much every other car in the past or present

 

Excellent! A perfect example of 'noise.'

[rao]

I aim to please

[SeeeeeYa]

raoThat's a great point - what could a person who set up there car exactly the way they wanted it and then enjoyed it as it performed flawlessly for years possibly add to any enthusiast site

Only someone who is down with the struggle and needs a few years to tune their car has any knowledge about anything

 

Once again you guys are hilarious.

 

And one thing that I did learn on t his site is that the Legacy is the King of Cars!!!! It has been described as better than Corvettes, BMW's, Lamborghinis and pretty much every other car in the past or present :)Today 08:32 AM

 

 

To paraphrase Socrates, "An unexplored ECU is an ECU not worth running.":lol:

 

Never read a book more than once?

 

Never questioned your own work?

 

Never found yourself smarter or more competent than you were three years ago?

 

Perfect at birth??

 

Do you ever read what you write?

 

"flawlessly" ???

 

It would help everyone understand your metronomic sarcasm if for once you posted your tune, your logs, some time slips, a time to speed.... or anything factual, really, about your self-described "flawless" tune. Provide your own methods for boost control, etc.

 

Join the conversation instead of spitting on it.

 

Some of us poor struggling incompetents might learn something.

 

Or not.

[rao]

You people really are funny. I have offered time and again to help people and share my tune. I have consistently offered ACCURATE, HELPFUL advice to solve all sorts of problems with these cars, common and otherwise. If you don't think that I add anything useful than ignore what I say. I have tried to save people a tremendous amount of money and aggravation. The ONLY other reason I participate in "technical" threads is to try (hopelessly I am sure, but I can't stop trying) to keep a lot of the complete nonsense from propagating so that newcomers won't be led in the wrong direction. I apologize if that offends you or anybody.

 

I never said or implied that I knew everything that there is to know or that there are not other approaches. I welcome an interesting discussion; not a bunch of people trying to prove to each other how smart they are .

 

To answer your specific questions:

 

I have read a book more than one time, I have read several more than once.

 

I do not believe that I was perfect at birth, or that I am perfect now, although that ultimately is more of a religious question that belongs in another forum.

 

I read what I write all of the time, mostly because I usually get paid for what I write

 

My car runs flawlessly; there have been several other people that have driven my car (I always let other people drive it if they want to and one of the measures that I use for success is that I don't need to give them ANY instructions before hand), some of them have reported their experiences here, maybe a search would help you I didn't realize that I need an evaluation by internet experts to determine if my car is running well Maybe if I blew the motor or spent several years working on the tune I would be more qualified

 

I have never claimed that my car was fast, quick or anything else, only that I like it.

 

If you don't see the humor of someone saying, in the midst of a highly technical discussion of a complex system, while making the argument that in depth analysis of a wealth of data is necessary to then say that you can just ignore a large pool of relevant data, then I really can't help you

[mickeyd2005]

rao, what are you doing in this thread?

 

1. offering advice

2. seeking advice

3. thread crapping

 

The logs are not for you. They are for other people to learn by observing others.

[rao]

I will stay out if this thread, until something else that is funny happens

 

Have fun

[Deer Killer]

I don't think you know how TD integral works.

 

Read merchgod's explanation. You'll figure out that you need to know the integration RATE. That's 70-80 ms.

 

As far as the second question is concerned... are you arguing for the sake of arguing? I answered your question and you gave it back to me. So, what's the problem?

I know how integration works, I know how PID is used to achieve good control with fast response. Searching 3-4 forums mostly just end up in finding a post by you saying how "someone doesn't know how TD integral works", which may be right, but still...

 

http://upload.wikimedia.org/math/d/5/9/d593c27abdc1aecffb56d06d2a9ba8e3.png

Are you saying t=70ms always? That's not what merchgod appears to be saying. What it sounds like you're saying is with the current sample rate it takes that integral 70ms to hit max correction.

[ScoopMan]

Wow, I read this whole thread and I still don't know what PID stands for...

 

I am a professional engineer with a masters degree in structures and this discussion is way over my head. I guess because I design targets rather than weapons, as the joke goes... You guys are DEFINITELY weapons designers!

 

Maybe being intimidated by this level of knowledge is a good thing, as it will tend to keep me from even TRYING anything out of fear of blowing up my engine. On the other hand, I would very much like to get on the ramp to this knowledge in a safe way. At the very minimum, I want to know what I DON'T know, if you know what I mean.

 

But even if I play it 'completely safe' and have everything done by a professional tuner, I cannot help wondering how sophisticated THEIR knowledge of these issues are. I suspect many of them are 'twiddlers'...

[mickeyd2005]

I know how integration works, I know how PID is used to achieve good control with fast response. Searching 3-4 forums mostly just end up in finding a post by you saying how "someone doesn't know how TD integral works", which may be right, but still...

 

http://upload.wikimedia.org/math/d/5/9/d593c27abdc1aecffb56d06d2a9ba8e3.png

Are you saying t=70ms always? That's not what merchgod appears to be saying. What it sounds like you're saying is with the current sample rate it takes that integral 70ms to hit max correction.

 

I am saying that it is not a TRUE indefinite integral. Even if you know the theory, you need to know how Subaru implemented it on their ecu. It is actually a SUMMATION and the interval between summations is 70-80 ms. However, Freon said it is 15 hz (67 ms).

 

I am positive that my understanding of turbodynamics matches merchgod's because he is the source. Read post 19. That was an answer to my question regarding time interval.

 

http://www.romraider.com/forum/topic1814-19.html

[SeeeeeYa]

Wow, I read this whole thread and I still don't know what PID stands for...

 

I am a professional engineer with a masters degree in structures and this discussion is way over my head. I guess because I design targets rather than weapons, as the joke goes... You guys are DEFINITELY weapons designers!

 

Maybe being intimidated by this level of knowledge is a good thing, as it will tend to keep me from even TRYING anything out of fear of blowing up my engine. On the other hand, I would very much like to get on the ramp to this knowledge in a safe way. At the very minimum, I want to know what I DON'T know, if you know what I mean.

 

But even if I play it 'completely safe' and have everything done by a professional tuner, I cannot help wondering how sophisticated THEIR knowledge of these issues are. I suspect many of them are 'twiddlers'...

 

Of this I, for one, am certain.

 

But I am a student here myself, like most. And it has been invaluable in developing my tuning, leverageing me away from the, at best, crudely effective half truths of Cobb's information.

 

By listening, by experimentation using the ideas in this thread, e.g., and by pursuing the goal of a tune with polish instead of mere pedestrian functionality today I have a tune that delivers most of what I want. As my understanding grows I'm sure that will improve even more. As it is, were I to stop development right this moment I'd have no problems. But for me, like most of the active members here, close enough is never enough.

[NSFW]

Wow, I read this whole thread and I still don't know what PID stands for...

 

[...]

 

But even if I play it 'completely safe' and have everything done by a professional tuner, I cannot help wondering how sophisticated THEIR knowledge of these issues are. I suspect many of them are 'twiddlers'...

 

The person who started the thread apparently did, and assumed everyone else did too, hence the "simply a PID control system." You have a very good point about that being a bogus assumption, and I think a lot more people might benefit from having it explained. So here's PID in a nutshell:

 

PID stands for "proportional, integral, derivative," and it refers to a control system that uses those three concepts to come up with the amount of feedback to apply to a system to reach the desired state (in this case, the "desired state" is the boost target).

 

All of the feedback terms are based on the "boost error," or the difference between the target boost and the actual boost... A PID feedback loop does three things with the error, adds those three things up, and applies the result as feedback (in our case, they control the wastegate). The goal is to drive the error to zero as quickly as possible without overshooting (overshoot = boost spike). Generally, overshooting is just the first sign of an oscillation that can last for a while if the gains are too high. (Gains will be explained in a moment.)

 

Total feedback = P feedback + I feedback + D feedback

 

Proportional feedback means: look at the error, multiply that by a "proportional gain" constant. The further apart the actual and target boost are, the more feedback gets applied to bring the back together. But with proportional gain alone, actual might never reach target, since the amount of feedback gets smaller as they get closer.

 

Intergral feedback solves that problem: look at the actual boost, look at the target boost, subtract them, multiply that by an "integral gain" constant, then add that to an "accumulator" which contains the sum of all of the previous errors and use the accumulated error as feedback. So if target boost stays lower than actual boost for a while, the integral term "winds up" and eventually supplies enough feedback to make up the gap left by proportional feedback alone.

 

The accumulator is conceptually similar to, but not quite identical to, a mathematical integration of the error over time. The accumulator adds the errors up over short periods of time, so it's just an approximation of a true integral.

 

Derivative feedback: look at the error in this iteration, subtract the error in the previous iteration, multiply that by a "derivative gain" constant. The cool thing about the derivative is that it changes sign as the error gets smaller, so it effectively puts the brakes on the total feedback to prevent overshooting the target.

 

I don't know if our ECUs actually use a derivative term, I haven't seen it mentioned in the maps.

 

And here's a much more elaborate description:

http://en.wikipedia.org/wiki/PID_controller

[NSFW]

Oh, and as for tuners being twiddlers, I suspect you're right in some cases. It would be hard to be a good tuner without understanding what's going on, what's open for tinkering, what should be left alone, etc. I'm sure twiddlers can get it right, but I wouldn't trust one to tune my car, and I don't know how to tell a twiddler from a scientist, and that's why I want to understand this myself.

 

This one thread, more than anything else, convinced me that I had to learn enough about tuning to be able to figure out whether my tuner understands this stuff:

http://www.mp3car.com/vbulletin/engine-management-obd-ii-engine-diagnostics-etc/110137-starting-bussines.html

 

And it took 10 posts before someone pointed out that "being a tuner" requires "knowing about tuning."

 

Later there was a thread (on RomRaider's forum, I think) where someone posted a map in which negative gain coefficients had been made positive. Probably wasn't the same person, but I suspect it was the same mindset.

[Deer Killer]

An approximated integral is still an integral, the parts of integration still have meaning. The details matter. http://www.math.umn.edu/~ivkin/GE/RSum/OLD/index.html

 

The 2D maps provide the derivative component when set properly. I wonder if it's easier to tune both tables linearly, then pull down the inner vectors.

 

Which also reminds me - how many "tunas" understand calculus? I'm not exactly a math whiz myself, but most people can barely do algebra.

[ScoopMan]

Thanks for taking the time for the clear explanation NSFW, it was extremely helpful. It has been pointed out in many threads that I have read that boost control is the toughest part of tuning to nail down.

 

It would seem to me that like many other things in engineering, if one is conservative enough, with a big enough safety factor, one ends up with a very reliable system, like the oem setup, as a perfect example. Maybe some of it is a cya 'ignorance factor', but really, so what, as long as you have confidence in the safety of the solution. Obviously, the more one understands the problem, the more tightly it can be dealt with. In my own work, I know the limits of my knowledge, and when I realize that it is below a certain comfort threshhold on a given problem, I go outside and find someone with the knowledge that I lack.

 

At my present knowledge level (low), I would be quite happy with a conservative (and simple) approach to boost control. I am stock right now, but am trying to educate myself as I am contemplating a new turbo.

[LittleBlueGT]

are you arguing for the sake of arguing?

 

 

Hmmmmmmmmmmmmm.

 

I think you are making things harder then they are. Log integral and if it is spiking and coming back down it obviously ramps up to aggressively. If it takes too long to accumulate to the needed number, then it needs to be adjusted. (this is all assuming that WGDC initial is in the ballpark)

[LittleBlueGT]

Wow, I read this whole thread and I still don't know what PID stands for...

 

I am only an engineer at heart.

 

I couldn't get it either. Once MickeyD2005 helped me, vis e-mail, it all made perfect sense.

 

It is easier for many (like myself) to see, rather then talk about.

 

If you want I will gladly help you understand it, it is not that hard once you get it.

 

All you need is a few logs to look at and a phone.

 

pm me if you are interested.

[Deer Killer]

Hmmmmmmmmmmmmm.

 

I think you are making things harder then they are. Log integral and if it is spiking and coming back down it obviously ramps up to aggressively. If it takes too long to accumulate to the needed number, then it needs to be adjusted. (this is all assuming that WGDC initial is in the ballpark)

But that isn't right. TD Integral _will_ spike if it's done right. The proportional and differential take care of that. First the proportional should be adjusted until it damps the oscillation, but you will still overshoot, and then differential should be tweaked by changing the slope of inner TD integral vectors. Changes in slope in the proportional are yet another differential factor which probably isn't necessary and complicates tuning. Still thinking about it.

[LittleBlueGT]

More then one way to do it I guess.

 

1) I start with WGDC needed to hit and keep target boost in top gear, make that the initial. Do this with 0 proportional, and up hill if you want (and in the cold)

 

2) Then figure out how much WGDC to hit target boost in 2nd gear, again with 0 proportional. Spool will definitely be down. This can be done with 0 integral as well, makes it easier to find. Again, just looking for what is needed to hold steady state boost, not spool the thing quickly.

 

3) whatever the difference between 1 & 2 is will be your needed integral range (WGDC needed for 2 - WGDC initial = max needed integral, but may want to add a bit more to it for various driving circumstances)

 

4) tune integral so that it keeps 5th nice and steady (should really be showing little to no integral once spooled), but allows the integral to accumulate to hold boost in 2nd

 

5) now start with proportional on a non-linear curve (this takes the longest time IME) to get the best spool possible (might have numbers as high as 50 in the last column for proportional) but have it start tapering off rapidly as target boost approaches to avoid overboosting.

 

This is a summary of what I have found that worked.

 

Sometimes, many times you talk over my head. When you have your theory working (if that happens) post up some graphs and logs, I find things much easier to decipher that way.

[mickeyd2005]

But that isn't right. TD Integral _will_ spike if it's done right. The proportional and differential take care of that. First the proportional should be adjusted until it damps the oscillation, but you will still overshoot, and then differential should be tweaked by changing the slope of inner TD integral vectors. Changes in slope in the proportional are yet another differential factor which probably isn't necessary and complicates tuning. Still thinking about it.

 

The TD integral will never spike.

 

There is no differential in our ecu.