When you click on links to various merchants on this site and make a purchase, this can result in this site earning a commission. Affiliate programs and affiliations include, but are not limited to, the eBay Partner Network.
I actually pulled off the freeway on the way home to log my last settings. I will post it up in a few days as my wife got home from a long trip yesterday and I need to get my ancient laptop on the internet to post logs and such(lazy). Really my closed loop is almost there, I just need to log a few more passes or good on ramps.
I am still missing gears once in a while, so I will probably try a different fluid and the brass bushing soon.
This is my most recent decent log, my current PID's are p16 i20 d122 but this log is from earlier at p16 i20 d121.
This was my lasy drag log
Thanks to DKMakinson for suggesting how to make my AFR and boost targets easier to read.
Going over some good 12 second logs I noticed that my boost targets are a bit low below 6000 rpm. So I made some adjustments to my target table before shutting my computer off. There was traffic so I couldn't try the new PID'S yet. Almost there though.
If you wanna keep your current settings and tweak them, I'd increase the "I" term and lower the "D" term. I assumed the 18 and 35 limits are derived from open loop testing, but that's not optimal for spool since you never actually hold the wastegate completely shut. Which kind of defeats some of the benefits of having PID control vs open loop.
If you want to experiment, I would change the solenoid freq to 20, the control interval to 50ms, move the "sensitivity slider" all the way to the left (off), P = 80, I = 15, D = 7 and change the min/max limits so that it can pin the wastegate shut to maximize spool. Drive it, watch it overshoot and hopefully settle, look at the logs and go from there. The logs you posted you never hit target so you got a lot of room to improve.
I just noticed that TS has incorrectly labeled "Derivative Gain" as "Differential Gain". If you click on the question mark, it correctly discusses derivative. EGO and CLI are labeled correctly.
This is a photo not a screenshot, so it's a bit fuzzy. But this is my boost control, targeted to hold boost flat. MAP sensor is the Red line, RPM is White, GREEN is TPS. The MAP sensor has absolutely zero filtering, so that's the raw signal you see. As you can see, no overshoot, no undershoot, very flat.
I will try your settings, but if they aren't close I am going back to my method. I have played with this a ton. Seriously, Wednesday night was the first time since I modified my wastegate can that I have not overshot my boost targets with wild oscillations in closed loop.
I also disagree with you about pinning my wastegate shut. Partly because I was thinking about something you posted about keeping the wastegate shut with the spring. So my wastegate pressure is ~11psi, it doesn't need any help from the EBC until 175 kpa or so. I can't run more than 20% boost duty at lower rpms or I overboost. I used to need to run 100% boost duty at 7000 rpm to reach 19psi. Now I can hit 21psi at 7000 rpm with only 34% boost duty. But regardless, I always need more boost duty at higher rpm to overboost than at lower rpms.
I think only a small amount of boost duty is required to assist the spring at 200 kpa, I know this actually. My actual open and closed duty at this frequency is 10 and 60%. I never need more than 35% up top now, which had always been when I had trouble reaching my boost target before, up top. That is when you need maximum force holding the wastegate shut not at low rpm while spooling, in my current opinion.
Just so there's no confusion, the top image is my most current log the bottom image is with slightly different PID'S from the dragstrip. I am hitting my boost targets now and actually do have a little overshoot. I have been trying to get my modded 2 port wastegate can tuned properly for the last few weeks. It is a bit trickier than a crappy 2 port can or a single port can or a 3 port solenoid. The only similar can I have seen is the turbosmart dual port, otherwise no comparison.
Also, your log looks like a spool log. I am much more concerned with fast stabs at the throttle in the 80-120 kpa range around 4500-5000 rpm. Spool and dyno conditions are much easier to control with closed loop EBC than what I am talking about.
My setting will not be close since most are guesses, but they would be a start at setting it up to ultimately get to target boost faster. Just like where you are at now, it will likely take a few trials/review logs before it's dialed in.
Your current setup in my opinion is technically working (it does control boost), but the way it is tuned is not giving you the best possible response you could achieve. In your datalog you posted, your boost control was really holding you back through all of 1st, all of 2nd, and all of 3rd. 4th it overshot a bit and then trailed the target a bit, but overall not too bad. The first 3 gears, that boost control is really holding you back, you're a few PSI away from target at times!
I've spent more time than I will admit with electronic boost control, so I know it can be tricky getting it tuned just right.
In the log I posted above, the wastegate was pinned shut until I was 20kPa from target, then the control system took over and brought boost to target. No under shoot or overshoot, and it held target extremely well for EBC.
Regarding spring pressure, if your wastegate regulates boost to 11 PSI, it likely starts "cracking" open around 8 PSI and by 11, it has opened enough to stop the boost from rising. This assumes it's just hooked to the boost source of course. It's purely mechanical, but to describe it as if it were controlled by EBC It's setting the frequency to 1000Hz, P = 20, no I, no D term. It just reacts instantly and cautiously to error. In my opinion if you want to stick with your current settings, like I said more "I" term and less "D" term. Dropping the D term is going to make it overshoot, so you'll need to raise the P term to compensate.
Also, your log looks like a spool log. I am much more concerned with fast stabs at the throttle in the 80-120 kpa range around 4500-5000 rpm. Spool and dyno conditions are much easier to control with closed loop EBC than what I am talking about.
Here's a log of it to redline, then a gear change to the next gear. For a transient. Admittedly this was an in-the-works log and it wasn't as fine tuned then as it is now. Still relatively flat, even during the gear change. You'll note after the gear change there is no spike, or dip.
Fair enough, I will try dropping the d term but I have had to inch it up for a while now. I am not messing with the frequency and interval though. I am not done playing with this and I have had my single port can behaving perfectly in the past, so it's not like I have never done this before.
Here's a log of it to redline, then a gear change to the next gear. For a transient. Admittedly this was an in-the-works log and it wasn't as fine tuned then as it is now. Still relatively flat, even during the gear change. You'll note after the gear change there is no spike, or dip.
How do you have no dip in MAP?
My boost target is lower after a gear change, so my MAP would always drop some. But you are right that I am not hitting my target in lower gears.
Since you said you're not changing the freq or interval, I'm curious as to why?
Let me elaborate, maybe this will help.
You have the frequency at 39Hz (1/39 = 25.6 ms) but you have the interval set to 15ms. So if you calculate say, 60% duty, then that would be on = 60/100 * 25.6 = 15.36ms, then off = 40/100 * 25.6 = 10.24ms. Thus for 60% duty, with a 39Hz freq and 15ms control interval, you need to turn the solenoid on for 15.36ms out of 15ms, and then.... OH WAIT, time to do control interval again!!!!
Hmm, wonder why the system only works from 10% to 60%???
EDITED for fixing the 20 to 40, math was right but I typo'd wrong number.
My boost target is lower after a gear change, so my MAP would always drop some. But you are right that I am not hitting my target in lower gears.
I have amazing EBC. And flat foot shifting was enabled. All my testing was on street tires, so I had to do 3rd gear pulls (6 speed at the time) to prevent wheel spin. I tested it in 2nd and it worked fine there too for fast climbing boost, but because of wheelspin the datalogs were messy and I would end up peddling it or something. On the top graph, green is TPS. On the second graph, green is fuel I think. Sorry for crap pics, I took them a long time ago and don't have the logs to get better info for ya.
Ok, I admittedly suck at math and you are losing me with some of this.
I did a bunch of tests with my solenoid to see the actual open and closed duties for different frequencies. For 39hz it is fully open at 60% and fully closed at 10% under all conditions. I know this, I have posted the results in my build thread.
I can try a different interval, but it's not like I haven't tried different frequencies and intervals before.
Ok, I admittedly suck at math and you are losing me with some of this.
I did a bunch of tests with my solenoid to see the actual open and closed duties for different frequencies. For 39hz it is fully open at 60% and fully closed at 10% under all conditions. I know this, I have posted the results in my build thread.
I can try a different interval, but it's not like I haven't tried different frequencies and intervals before.
I believe that 60% is fully open with your settings, the math proves it. Look again at my math.
With the 39Hz, 15m control interval, assuming 60% duty cycle, you're telling the computer to do this:
"You have 15ms to do the following. Turn on the solenoid for 15.36ms, then turn it off for 10.24ms. After a total time of 15ms has went by, do it again."
You're asking the solenoid to do 25.6ms worth of work in a 15ms interval. Which is impossible.
So you are suggesting that with an interval closer to 30ms I will have more luck? I basically wanted the fastest interval possible for the algorithm to decide what to do. I chose 39hz because it was the fastest frequency with a reasonable range, approximately 10-60%.
So you are suggesting that with an interval closer to 30ms I will have more luck? I basically wanted the fastest interval possible for the algorithm to decide what to do. I chose 39hz because it was the fastest frequency with a reasonable range, approximately 10-60%.
Yeah, something like that. It's at least possible to do 25.6ms of work in 30ms time.
05-25-2017, 01:10 PM
0
