Hello,

First, thanks for this long discussion thread talking about tuning EBC. I've made great progress tuning the built MSM engine in my '04 MSM because of that discussion. I limited the Integral factor winding up by reducing delta so the loop starts later, but mostly by setting boost/RPM targets based on real world data. I recorded a 5th gear, WOT run from 2k RPM until 18 psi. Just when I thought I had EBC fairly dialed in I'm starting mostly over. I had the slider sensitivity set at 450 with PID numbers that worked well based on tuning with my boost gauge. Then I recently learned while reading data logs that boost was quickly oscillating around target by ~2 psi, maybe 3. I could feel something was off but my SpeedHut digital boost gauge seems to have more non-adjustable damping than I realized. It was showing the initial over shoot but not the oscillations, if the needle was moving it was such a tiny oscillation that I was unable to be see it bouncing down the road under boost.

The last two days I have been off work, driving back and forth on a few, safer test roads with lower traffic and no side roads. It has a few block long road into a developing industrial road that closes in another two weeks, probably to put up more buildings. It also has onramps nearby where I can circle back and forth to avoid being a stationary target if anybody was to take notice and call police. Twice in the last month I've tested and tweaked the "Boost Control Bias Duty Cycle" table, those RPM columns are also set based on real world data so I can hit and hold target in open loop accurately with minimal drift off target. What is confusing is watching the Boost DC in data logs, I can see when it hits delta but it's not dropping immediately back to the table numbers. It still overshoots a bit (.5-1 psi), then drops about 1 psi below target (am still working on tuning these issues slowly adding more Integral and a little more Derivative has been working) but the loop settles on DC numbers that are much higher than open loop requires to hit target.

I'm tempted to change the lookup table to match the numbers where the PID loop stabilizes but I just lowered DC by 2-3% a few weeks ago to minimize the initial spike. The new, higher flowing FM manifold and outlet pipe needed less DC than before the upgrade. Today I switched back and forth from CL setup mode and advanced modes after dropping sensitivity back to 375 due to the quick oscillation around target as the popup guidance says to do. Sensitivity is now around 340-350 and that oscillation is gone. Once the loop stabilizes (one smaller up/down wave), DC is +10 more than the lookup table at all RPMs up to redline.

Any thoughts as to why there's such a difference between OL table numbers (same as CL setup mode) and where the PID loop stabilizes under sustained boost? I set target down to 16 psi, made progress tuning out the oscillation and overshoot, then dropped to 14 and reworked the DC table with more PID adjustments. Then I set target to 12 psi and played with delta to reduce winding up the I term in higher gears. I might also raise the 12 psi target RPM in the DC table, raising that column slightly will reduce the wind up but also soften spooling in lower gears, right? Lower delta helped for now but it's an idea I hadn't considered while working with the PID numbers.

I'm taking a break to work around the house and after lunch I'm heading back out for more tuning runs. I might also download the SDcard logs to check how each P, I, D term is reacting but that's getting into the deep end of the pool. BTW, still using a serial port on MS3Pro = . I knew tuning to hit and hold maximum target was difficult even with an accurate DC lookup table but tuning CL EBC for all boost targets is another level. My targets for good street control are 7 psi under 30% throttle (using the GT2560R 7.5 psi can), then 8, 9, 10, 12, 14, 16, and 18 psi). I don't want to be cruising along in higher gears and ask for 10/12 psi only for the PID loop to go crazy and BAM, hit boost cut. I'm getting close but it's taking a LONG time to dial it in!

Thank you,
Jess

 

No Bueno, with target set at 14 psi, I tried and tried adjusting PID and sensitivity but made no progress and had to stop. It's fully sunny and 85 degrees, the shade trees I was using this morning are no longer shaded areas and my frustration was mounting. I have decided to try the settings that used the 450 sensitivity slider setting, I am wondering if they might just a higher I term/setting, maybe a lot more. It also probably needed a higher D setting. I'm reverting back to settings that were more stable other than the oscillation after slightly overshooting. The dip that occurs after the overshoot turned into a constant oscillation around target but it was focused and consistent. I tried slider as low as 300 altering PID as I thought might help to no avail, the results were heading the wrong way or at best, making little difference.

With lower sensitivity the loop seems to adjust DC more slowly. I tried raising the DC lookup to be closer to what I saw in the data logs while targeting 14 psi, I still have no idea why the PID loop settles at higher DC than what works in OL or CL setup mode. I was unable to tune out the overshoot, nor the dip. It was taking about one second to stabilize, +1.5 psi, then -1.5 psi before sticking more closely to target. In higher gears, starting at lower RPM, CL EBC was working pretty nicely. In 2nd gear, or 3rd mashing throttle around 4k RPM was showing that spike/dip. Shifting hard into 3rd under boost would easily spike ~3 psi over target. Raising the delta setting decreased that spike (and/or lowering DC back to what works in OL), and of course, between shifts it goes back to max DC (set to 62%) so with boost climbing faster, by the time the loop started working it was too late to avoid the spike. Too high delta means the I term will still ramp up too quickly in higher gears resulting in spikes or hitting boost cut.

I'm leaning towards the higher sensitivity setting helping solve all these issues, it just needed more time tuning I and D settings. I deleted most of the logs, I need to label them even better than I was doing...I really wish there was a place to write and save notes when starting a log for later reference. As for now, I am recording 2-3 WOT runs to target between my stopping places, then analyzing the most recent log and making minor tweaks. Wash, rinse, repeat, if anybody has a tip or better way I'm all ears...this is getting old.

 

I guess I'll just talk about what I'm learning and seeing as I work, hoping for a little input from anybody that has been in this situation. Maybe something will ring a bell and I'll get a little feedback that might help. Yesterday evening I turned the tide, started seeing progress again. I'm going from memory working on my newer laptop so I can't see my notes or check things that are on the tuning laptop.

I left the PID settings at 350 sensitivity slider (gain?), P=193, I=80 or 85, and D=50 or maybe 60 with a delta of about 24 kPa. I set target at 183 kPa (14 psi) after working at 16 psi. It gives me longer WOT data logs since each gear lasts longer. At 18 psi things are happening too quickly even with 3.63 ring and pinion gears. I tweaked the 16 and 14 psi Duty Cycle rows. I decided since I was seeing 3-4% higher DC in the logs once boost settled, I might as well try changing the DC lookup table to match those numbers DESPITE not needing as high DC to hit and hold target in open loop and CL setup modes. I was surprised that the initial .5 second overshoot didn't get worse, the ~.5 second dip after the spike might have improved slightly. What still confuses me is that I thought once the PID loop started (159 kPa with the delta of 24), that DC would drop immediately to the lookup table number, then react from that DC number. NOPE, my max DC is set to 62%, as intended, while the turbo is spooling up DC sits at 62. Then as it passes 159 kPa the PID loop changes DC lower but it takes 1-3 steps to lower down to match the bias DC table number. This seems to be causing the spike and following dip.

For example. I think my setup needed 24% DC to hit and hold 183 kPa. Every time it passed delta, I would see 62 drop to 34-36, then another step or two downwards until it would hit or drop below 24% DC. This delay in setting DC to what works causes the (usally) small spike. At 350 sensitivity it seems to realize this and just before the top of the spike, DC is ramping down below 24% by a few clicks which quickly pulls the spike back but then DC is too low so it drops below target and quickly recovers. This happens within 1 second from most of the logs I checked yesterday. Increasing D term is my next step hoping to limit the spike, maybe higher I term or sensitivity to see if the oscillation duration shortens or becomse more shallow.

Going WOT at 4000 RPM or higher results in a greater spike. That 1 psi becomes 2 psi over target. Get into boost in 2nd gear and shift quickly to third going WOT immediately and that spike increases to 3-4 psi over target. The recovery is crazy fast but tuning out this single oscillation, spike and dip is proving challenging. This is also why I dropped from 16 back to 14. I was tuning at 18 psi with a 21 psi boost cut. It was far too easy to slam into boost cut, had this happen a few times where I can do a U-turn at a stop light on a 6-lane area of this county road in front of a Walmart. 1st at WOT is great, WOT into 2nd, really good, into 3rd and immediately BAM, boost cut. 16 psi works ok, haven't tested that scenario as once I got the DC table adjusted to where the PID loop was settling to hold target, I dropped target to 14 psi, maybe even 12 psi. As I stated, my memory is hazy with the data overload. 2-3 WOT runs in each log, pull over, make changes to settings, start a new log, wash/rinse/repeat 6-10 times before I'm bored and want to move to another area to avoid annoying anybody within earshot.

I'll update if I am successful but my end numbers probably won't work for other Miata, even if they had the same intake/exhaust manifolds, turbo, exhaust size, and injectors. Most of it makes sense but it's still largely trial and error until you stumble upon what works. Then cross your fingers that it remains stable with weather changes! The price of gas also isn't helping, I'm getting ~18 mpg, might improve as I lower boost target to check the DC table settings against the DC numbers where the PID loop settles.

 

This afternoon runs didn't improve, probably got worse but I had a new thought. If the spiking cannot be tuned out with lower sensitivity and correlating PID values (and proper Bias DC table), maybe the Bias table needs work. My lowest RPM column is 2550 targeting 7 psi as that was the earliest I could achieve that target in the data log of my 2k RPM WOT in 5th gear to 18 psi run. I might try changing the RPM columns as the DC numbers below ~3350 RPM might be worsening the spike when target is set to 18 psi (225 kPa) as the turbo cannot achieve that BUT, another thought just occurred to me...those lower RPM numbers aren't taken into the equation as far as I know up until you hit delta and the loop starts working. That's when it looks at the table and uses the DC numbers, right? So those lower columns at higher boost targets being set to whatever DC it takes to hit and hold those higher boost levels makes no difference for open loop or closed loop. The valve is held at that DC for open loop no matter if RPM is below where the turbo can achieve that target.

I tried larger delta and that can lessen the spike but that screws up higher gears, low RPM boost where the I term winds up too far causing a spike. I created three test tunes based on previous tunes that were working better. Sensitivity varied from 275 up to 430. Hell, I might turn sensitivity back down to 150-250 and start over tuning P after I verify (for the third time this year) the DC target table in Setup Mode (open loop). I think this time once I get it hitting and holding a lower 12 or 14 psi target I'll turn up the sensitivity until I start to see it oscillate as the pop guidance suggests. Maybe sensitivity still being set too high was causing the initial 10-20 kPa spike (gear/RPM dependent) and dip below that was taking ~1 second to recover. Although, I tried lowering from ~375 down 325. I seem to remember my car working quite well around 250, maybe 275 sensitivity. The remote tuner I hired to help me over a few bumps in the road and also helped me climb the tuning knowledge mountain (sometimes more like a brick wall)...he said most Miata seem to need higher sensitivity than I settled upon way back when.

I quit for today because it started raining, it's not worth the risk or anxiety concerning traction. This week I work every day but Sunday and Mondays are usually forced OT, sometimes Wednesday and Friday as well. I doubt I'll have enough energy to go out after dinner with temps at ~90F for the middle of the week. I work in an ancient Grumman LLV with a tiny fan that looks like it's from the 1940s or 50s, can get so hot that it's barely tolerable.

 

So, it has been quite some time since I’ve messed with CL Boost Control, and I was on MS2 instead of MS3, but your sensitivity slider and PID values are much higher than what I saw once mine was very well dialed in. See this post for an idea of where I landed.

I cannot remember the resource that I used for dialing this in, but it was pretty much a step by step process rather than trying a few different settings and just hoping for the best. If I can locate that I will share it here.