Savington

iTrader: (31)

Argh. I thought it was the other way, since P100 I0 D0 gave me wastegate-only. The other problem is that higher P values result in a very low duty cycle at peak boost, which ends up undershooting the target and never recovering.

aidandj

iTrader: (5)

What did the duty log look like with 100% P?

aidandj

iTrader: (5)

Oooooh. Also something they changed in 1.4 was to get rid of the control interval. Because your valve can only react at 26hz (~38ms) PID isn't happy because it is changing every 10ms, But your output is only changing every 38ms. So change the control interval to 38ms and see how that helps.

Savington

iTrader: (31)

Spikey, but it won't come close to overshoot. The moment the wastegate gets anything less than about 30% DC, boost falls to ~145kpa, and it takes an act of god to drag boost back up to the 160kpa target afterwards.

P100 I0 D0 Slider 400

2ndGearRubber

Dumb question, I know.


Does the third graph represent knock events during the logged period?


How would one quantify 0.5 of a knock?

aidandj

iTrader: (5)

Now its too much P

I've found messing with the slider in advance mode is a recipe for disaster. too many variables. Go back to simple if you are going to play with the slider.

Read through the thread I linked on msextra. Ken had some tuning tips for advanced PID.

aidandj

iTrader: (5)

Its %knock. So if its above a certain percent you get a knock event.

2ndGearRubber

So, if I understand the graph correctly:

Cyl 1 (white trace) has a max recorded knock% of 8. Max on the graph for cyl. 1 is 9.2%


I would infer that 9.2% is the computer estimated point where a knock event would occur? Or am I shitting up Savs thread with eggplant?

codrus

No, 9.2% is simply the max value for the displayed chart. By default, MegaLogViewer auto-scales all of the values it is displaying. This is actually kind of annoying when you're trying to compare multiple values that are using the same units, because the relative heights of the charts don't mean what you think they should. In this chart, peak knock on 2/3 is 30% higher than on 1/4, even though they're both hitting max at the same height on the chart. In this case, that's probably because the knock sensor is located between 2 and 3 so it's more sensitive to noise in those cylinders.

Note that the percentage is somewhat arbitrary, because it's affected by the gain and other parameters that you configure in the knock sensor settings. You also need to configure the threshold above which it should take certain actions (and then configure the actions too).

--Ian

Savington

iTrader: (31)

It's logging noise on the knock sensor as assigned to each cylinder based on the windowing algorithm. My theory is that by looking at the noise on all four cylinders, you can more easily pick out knock events in a single cylinder.

Here's an example of that, running a bit too much timing on this 10:1 bottom end at ~170kpa. Totally inaudible, but when one cylinder registers a "noise" level 4x the others, that's knock to me:



Pulling 3deg across the board eliminated those little spikes. I'm not spending any time on the timing map right now, since I'm changing things so rapidly.

Savington

iTrader: (31)

Why does the PID control keep plunging the duty cycle to the minimum well before my target boost? This is with virtually no PID authority (slider=16, P100 I0 D0).

codrus

What does your initial value table look like?

--Ian

Savington

iTrader: (31)

Pretty simple:



Target table is 140kpa under 50%TPS, 160kpa above 50%TPS

aidandj

iTrader: (5)

Look at the small angle of the boost duty.

You get to the initial value, and then the PID algorithm sees that you aren't at boost target and it drops it to min.

Its weird. Its like it's acting backwards. And thinks that 0% duty is closed wastegate.

aidandj

iTrader: (5)

I would post this over on MSextra. That's weird.

shuiend

iTrader: (24)

Did you upgrade to 1.4 yet?

At the megameet a few of us were working on getting EBC working correctly on a MSLabs MS3 and found that the "output polarity" had to be switched to inverted to work correctly. Ken and James the MS devs were completely baffled by that as it should not be that way. We had 2 cars running EFR 6258's using the BW EBC. One was a rx7 with a ford swap running a home built MS3x, this one ran had EBC working perfectly. The other was a Chris Almerretes miata with the MSLabs MS3. It was Chris's car that needed output polarity switched. Ken and James were figuring it had to be something with how Reverant was building his units. Unfortunately Chris left before we got a chance to pull the MS out of his car so they could take a look at it and try to figure it out.

aidandj

iTrader: (5)

That matches what the log did exactly.

Savington

iTrader: (31)

I did a couple pulls early on with polarity inverted and it just overboosted, but that may have been from way too much authority. I will try it with very low authority today.

aidandj

iTrader: (5)

What lars said explains your log perfectly. And explains why I was really stumped about why it dropped to minimum when trying to spool.

codrus

Huh. I have a Reverant MSLabs MS3 and I have output polarity set to normal, not inverted. Mine's an early unit though, so it's possible something changed after that.

--Ian