Turbo Dynamics, simply a PID control system

and what is a PID system you may ask?

this PDF outlines their function, operation, and how each piece effects output versus a step function.

I know most people messing with turbo dynamics know this, but not everyone does and its good for people to learn

PID.pdf

Good!

We need this, thanks!

the notes are for a continuous time linear PID. A good starting point but we need to keep in mind the extra complexity added by sampling and the fact that we can make our control curves non-linear and clipped

I really appreciate someone else taking a scientific approach to tuning too

yeah, they are actually for using PID control to actuate a pendulum and speed control for a mobile robot, but its still a good intro

I can't take credit for writing it, thank the now-defunct tDEC program at drexel

the notes are for a continuous time linear PID. A good starting point but we need to keep in mind the extra complexity added by sampling and the fact that we can make our control curves non-linear and clipped

 

I really appreciate someone else taking a scientific approach to tuning too

I agree.

However, understanding the fundamentals is the first step toward understanding our real world control system.

Threads like these are disturbing. It is scary how little people really know about how their ecu works.

Just mess with the tables...make small changes...datalog and see what it does.

I don't think clipping is that big an issue. It is a safety net that I use but I try not to rely upon too much. I think the big issues with real world turbodynamics are:

1. Boost behavior is completely different below boost threshold. For example, on the LGT, it is ridiculously easy to stabilize boost if you start your WOT run at 4000 rpm but very different if you start your WOT run at 2000 rpm.

2. The actuator itself behaves as an accumulator and is like a secondary integral that responds based upon the previous WGDC and time.

3. The speed in milliseconds at which boost builds is dependent upon which gear you are in. The TD tables are all built around boost error as the independent varaible so d(TD)/dt will be different depending upon which gear you are in.

The 32 bit ecu definitely has more tables than you need to control boost. People are able to control boost with 16 bit ecus and they have fewer tables. The additional tables in the 32 bit ecus allow different strategies. I've experimented with several. Unfortunately, it still involves a lot of trial and error but in my latest one, I was able to control boost in all 5 gears. There was very little oscillation in 4th gear and about +-0.4 psi oscillation in 5th gear. That was my first iteration so I still need to work on it.

I am really surprised that no one has every asked how often does the ecu update TD. Without knowing that, we won't know how fast WGDC will build or drop. I estimate it at 70 ms, but I really need to set up a dummy rom with a max value of 1 for TD integral positive and check it. BTW, I think 70 ms is too slow for large turbos. I hope the 2008 STi has a faster ecu.

I agree.

 

However, understanding the fundamentals is the first step toward understanding our real world control system.

 

Threads like these are disturbing. It is scary how little people really know about how their ecu works.

 

 

 

I don't think clipping is that big an issue. It is a safety net that I use but I try not to rely upon too much. I think the big issues with real world turbodynamics are:

 

1. Boost behavior is completely different below boost threshold. For example, on the LGT, it is ridiculously easy to stabilize boost if you start your WOT run at 4000 rpm but very different if you start your WOT run at 2000 rpm.

2. The actuator itself behaves as an accumulator and is like a secondary integral that responds based upon the previous WGDC and time.

3. The speed in milliseconds at which boost builds is dependent upon which gear you are in. The TD tables are all built around boost error as the independent varaible so d(TD)/dt will be different depending upon which gear you are in.

 

The 32 bit ecu definitely has more tables than you need to control boost. People are able to control boost with 16 bit ecus and they have fewer tables. The additional tables in the 32 bit ecus allow different strategies. I've experimented with several. Unfortunately, it still involves a lot of trial and error but in my latest one, I was able to control boost in all 5 gears. There was very little oscillation in 4th gear and about +-0.4 psi oscillation in 5th gear. That was my first iteration so I still need to work on it.

 

I am really surprised that no one has every asked how often does the ecu update TD. Without knowing that, we won't know how fast WGDC will build or drop. I estimate it at 70 ms, but I really need to set up a dummy rom with a max value of 1 for TD integral positive and check it. BTW, I think 70 ms is too slow for large turbos. I hope the 2008 STi has a faster ecu.

I use clipping as a band-aid if I'm trying an experiment where I want very consistent runs. Other than that, I try not to use clipping.

I agree that the sampling interval is pretty important but I suspect the mechanical limitations of the pressure sensor are likely dominant. Most mems based pressure sensors for automotive use are BW limited to about 1kHz on their own. Once you add the plumbing the figure probably drops to low 10s of Hz. At that point one runs into a tricky situation with sampling. Not just aliasing (which really shouldn't happen in such a slow system) but the approximation to an additional 1st order pole due to the S/H (assuming the ECU designers just turned down the sampling rate knowing the external limitations).

Do you allow any deliberate overshoot with your setup? I did some experiments a while back where I deliberately set the boost target tables lower than my real targets and allowed the system to "overboost" to my real targets. The transient response was great on the street but caused significant problems on the dyno.

if you were really trick you could RnR the sh7058 and replace it with a faster-clocked variant

if Subaru has gone ROHS with their electronics, I don't think the board would survive. Even on a SRT machine, I roast half boards I need to pull large chips off of when the solder is Pb-Free. I bet they have some underfill under the chip to improve vibration tolerance...

Also kinda wonder if the software was written for a certain clock speed or if it timer event driven....

2005garnetGT.. are you from drexel?

no, I steal their old-ass documents.

yes, I go to drexel.

I agree.

 

However, understanding the fundamentals is the first step toward understanding our real world control system.

 

Threads like these are disturbing. It is scary how little people really know about how their ecu works.

 

 

 

I don't think clipping is that big an issue. It is a safety net that I use but I try not to rely upon too much. I think the big issues with real world turbodynamics are:

 

1. Boost behavior is completely different below boost threshold. For example, on the LGT, it is ridiculously easy to stabilize boost if you start your WOT run at 4000 rpm but very different if you start your WOT run at 2000 rpm.

2. The actuator itself behaves as an accumulator and is like a secondary integral that responds based upon the previous WGDC and time.

3. The speed in milliseconds at which boost builds is dependent upon which gear you are in. The TD tables are all built around boost error as the independent varaible so d(TD)/dt will be different depending upon which gear you are in.

 

The 32 bit ecu definitely has more tables than you need to control boost. People are able to control boost with 16 bit ecus and they have fewer tables. The additional tables in the 32 bit ecus allow different strategies. I've experimented with several. Unfortunately, it still involves a lot of trial and error but in my latest one, I was able to control boost in all 5 gears. There was very little oscillation in 4th gear and about +-0.4 psi oscillation in 5th gear. That was my first iteration so I still need to work on it.

 

I am really surprised that no one has every asked how often does the ecu update TD. Without knowing that, we won't know how fast WGDC will build or drop. I estimate it at 70 ms, but I really need to set up a dummy rom with a max value of 1 for TD integral positive and check it. BTW, I think 70 ms is too slow for large turbos. I hope the 2008 STi has a faster ecu.

I do not understand this mindset. Understanding does not equal a properly tuned car - for examples of that just look around here While understanding how things work is always a good idea, the notion of experimenting is a lot more useful in the real world than making fun of someone who has properly tuned the car.

You guys make it seem like tuning is a massively complex subject - the fact is that tweaking a factory turbo charged car is about as easy as it gets. If you want to pat yourselves on the back then get a standalone engine management system and tune it from scratch so that it idles and drives like a factory car in all conditions, start with a NA one and then work your way up to forced induction.

BTW, it is not Cobb's responsibility to teach everyone how to tune, although it would be in their interest to make it as understandable as possible from a marketing standpoint. Their program works, but to expect them to explain to you how tune is to expect Microsoft to explain how to write the great American Novel when yo by Word.

The simple fact is that the Subaru ECU uses several things to control boost. Why not talk about the details of how to change the turbo dynamics to sit your needs then engage in a discussion of how complex the system is, when indeed it isn't?

if Subaru has gone ROHS with their electronics, I don't think the board would survive. Even on a SRT machine, I roast half boards I need to pull large chips off of when the solder is Pb-Free. I bet they have some underfill under the chip to improve vibration tolerance...

 

Also kinda wonder if the software was written for a certain clock speed or if it timer event driven....

I dunno, call subaru and ask if their 2005 ecus are RoHS compliant?

they might be confused enough to answer!

I do not understand this mindset. Understanding does not equal a properly tuned car - for examples of that just look around here While understanding how things work is always a good idea, the notion of experimenting is a lot more useful in the real world than making fun of someone who has properly tuned the car.

 

You guys make it seem like tuning is a massively complex subject - the fact is that tweaking a factory turbo charged car is about as easy as it gets. If you want to pat yourselves on the back then get a standalone engine management system and tune it from scratch so that it idles and drives like a factory car in all conditions, start with a NA one and then work your way up to forced induction.

 

BTW, it is not Cobb's responsibility to teach everyone how to tune, although it would be in their interest to make it as understandable as possible from a marketing standpoint. Their program works, but to expect them to explain to you how tune is to expect Microsoft to explain how to write the great American Novel when yo by Word.

 

The simple fact is that the Subaru ECU uses several things to control boost. Why not talk about the details of how to change the turbo dynamics to sit your needs then engage in a discussion of how complex the system is, when indeed it isn't?

Are you actively encouraging ignorance?

I do not understand this mindset. Understanding does not equal a properly tuned car - for examples of that just look around here While understanding how things work is always a good idea, the notion of experimenting is a lot more useful in the real world than making fun of someone who has properly tuned the car.

 

You guys make it seem like tuning is a massively complex subject - the fact is that tweaking a factory turbo charged car is about as easy as it gets. If you want to pat yourselves on the back then get a standalone engine management system and tune it from scratch so that it idles and drives like a factory car in all conditions, start with a NA one and then work your way up to forced induction.

You are a real work. I don't understand your point.

I was just getting ready to write a response to bugblatterbeast about different strategies that I use to control boost, but apparently you don't think it's worthwhile.

PID feedback loop is well understood in industrial controls. However, how it applies to a turbocharged system isn't. I certainly didn't understand it and so I looked for info on it. It isn't available. I had to experiment with my car to find out what works and what doesn't work. Luckily, there are friendly and knowledgeable people like merchgod who can disassemble the rom and read the logic in it. That allowed me properly interpret the strange logic that I saw in the my datalogs..

Apparently, everyone knows how to prevent overboosting in 5th gear and still get full boost in 2nd gear.

kthanxbai.

You are right - you do not understand my point. I will take the blame for not being clear enough.

Sorry to have intruded here, please carry on and I will not interfere further.

You are right - you do not understand my point. I will take the blame for not being clear enough.

You are right. I overreacted.

the fact is that tweaking a factory turbo charged car is about as easy as it gets

I forgot that I'm not as smart as you. I found it quite complicated and did not understand it even though I was getting the results that I wanted. I had to ask someone who could read the logic to explain it to me.

BTW, can you explain to bugblatterbeast what happened here:

Do you allow any deliberate overshoot with your setup? I did some experiments a while back where I deliberately set the boost target tables lower than my real targets and allowed the system to "overboost" to my real targets. The transient response was great on the street but caused significant problems on the dyno.

I was just getting ready to write a response to bugblatterbeast about different strategies that I use to control boost, but apparently you don't think it's worthwhile.

I kinda like achieving target boost in 2nd and third gear, albeit at a hgiher rpm.

I thought that is what these forums are for, people that are enthusiasts and want to make things better. If you are happy with the level of knowledge or tune of car don't read it.

If you want to make your car better allow others to help.

You are right. I overreacted.

 

 

 

I forgot that I'm not as smart as you. I found it quite complicated and did not understand it even though I was getting the results that I wanted. I had to ask someone who could read the logic to explain it to me.

 

BTW, can you explain to bugblatterbeast what happened here:

one of the closed loop poles shifted to the RHP because of the lack of mass in the system when on the dyno (best guess right now). I now use different boost control strategies for road and dyno use......

With all due respect there are multiple elements to sucessfully tuning something. One method is to do careful analysis and dial things in through understanding of the system. Another method is to go by feel and intuition. Either method has its advantages and there is really no need to belittle one way or the other.

As a design engineer who started out electronics as a hobby I've seen and used both methods. While there are lots of "egg heads" who know all the theory but couldn't get a piece working to save their lives, there are also a lot of guys who muddle their way through a design without any understanding of what they are doing. The best approach IMHO is to take a bit of both. You need to understand what you are modifying to know the interactions and compromises you are making. At the same time, you need practical skills from field work to get the job done right. I'd rather encourage a bit of both as the results are usually better when the two methods are combined

I do not understand this mindset. Understanding does not equal a properly tuned car - for examples of that just look around here While understanding how things work is always a good idea, the notion of experimenting is a lot more useful in the real world than making fun of someone who has properly tuned the car.

 

You guys make it seem like tuning is a massively complex subject - the fact is that tweaking a factory turbo charged car is about as easy as it gets. If you want to pat yourselves on the back then get a standalone engine management system and tune it from scratch so that it idles and drives like a factory car in all conditions, start with a NA one and then work your way up to forced induction.

 

BTW, it is not Cobb's responsibility to teach everyone how to tune, although it would be in their interest to make it as understandable as possible from a marketing standpoint. Their program works, but to expect them to explain to you how tune is to expect Microsoft to explain how to write the great American Novel when yo by Word.

 

The simple fact is that the Subaru ECU uses several things to control boost. Why not talk about the details of how to change the turbo dynamics to sit your needs then engage in a discussion of how complex the system is, when indeed it isn't?

that was my guess. the car was able to spin up pretty quickly under certain loading conditions, then dwell at high loads (I was on a DD so we had control of the load). I hadn't compensated for the car's ability to switch from low revs to high revs so quickly.

another possibility was that the DD's attempt to emulate higher inertial through shifting the loading of the dyno was interacting with the car's boost control by shifting the load a lot.

I really didn't have enough time that day to hunt down the cause + I was really tired. I had been on the dyno close to 8 hours as I was tuning a couple Talons before I got to my own car.

one thing that did stick out in my mind was that the owner of the dyno mentioned having lots of trouble with the p-drive solenoids on his dyno. he had trouble getting the control slopes right on the dyno but they were ok on the street

I know that one, the opensource editors for subaru and honda are both 1000x better then DSMs

our DSM came is absolute last in the GRM $2007 challenge, because it wouldn't start for the autocross because it wouldn't start while hot; aaaand it wouldn't shift out of second, aaaannd it was still on the stock 13G turbo, aaaand my friends lost their wideband and my moates chip burner out of the back of their truck towing it down.

I see you've been down the same path.....

currently Talon number one has a walked crank and Talon number two has a chipped tooth on 1st.