Shift knock?

After being tuned a number of times, my setup has changed slightly, necessitating maf scaling changes and such.

My CL maf scaling is within 1% at all times (except learning A which swings ±2% or so) - OL scaling is tuned within 2% depending on the day (I need to work on IAT compensation).

 

Anyway, I've been taking many logs, and all of my WOT pulls even from down low in 3rd or 4th show no knock whatsoever.

Randomly going WOT in any gear at any rpm doesn't cause knock.

Mashing the pedal like an idiot in 5th going up a hill shows no knock.

Shifting from 2-3, 3-4 or 4-5 going wot shows no knock.

 

The minute I try a 1-2 wot run, I run into varying amounts of fbkc with a corresponding increment(s) of knock sum following the shift into second and reapplication of pedal.

Sometimes it's just -2.

Other times it'll pull upwards of 6 degrees.

 

I've reduced timing in the affected areas and richened the fueling to try and find a cause/effect but have so far been unsuccessful.

I have considered it could be just noise, but since I never get any fbkc going wot or in any other gear, I tend to think that's not the case, although I do get random fbkc very occasionally when driving around town gently at very light load in vacuum.

 

Anybody have any ideas?

I've attached an excerpt of my most recent log showing this phenomenon if anybody wants to take a look.

This in on CA91 gas.

romraiderlog_20121020_172344.csv

What's your hardware setup? You sure produce a lot of boost even at redline.

 

Anyway, just taking a stab in the dark but your AFR looks lean after your shift when you get back on the throttle...

Running an AVO 380 (GT2871r) with a PW TMIC, 73mm intake, TGV Deletes, EL Headers, catted dp, DW850 and Walbro 255.

Boost targets are 20psi tapering to 17.5 at redline.

I noticed the lean spike after the shift too - I suppose it could be not enough tip-in.

I had just figured in the past that the spike was just residual throttle lift afr that wasnt picked up fast enough due to the wideband being in the dp but I could be wrong.

Shift knock is tough to fix... to be honest, I've given up. I've tried a few things and I've mostly failed. It doesn't happen consistently in my car but I do see it in my logs once in a while. There seems to be a general problem with rapid changes in load and/or throttle position. I see knock sum increment sometimes when I touch the throttle after idling for a few seconds, and I see it shortly after releasing the throttle at the end of a WOT pull, and when getting back to cruise throttle after coasting down after a WOT pull, etc, etc. And of course I see it sometimes when shifting as well.

 

The thing that leaps out at me in the log you posted is that your total ignition timing is negative 6 or 7 degrees when this is happening. I haven't noticed negative timing in my logs but it's possible that it happens intermittently and never happened when I actually was looking at timing during shifts.

 

I'd bet that your timing tables don't have anything negative in that area (1.6 g/rev, 4000 RPM). So I suspect that the culprit here is some sort of compensation that pulled timing during the shift. There's another negative timing value a couple rows later, too. Really weird.

 

If you are running the A2WC521N or 522N ROM, I can get you some experimental logging parameters that might shed some light on this.

I noticed the negative ignition timing as well.

Like you say, the load and rpm ranges in question do NOT command any negative timing.

I've noticed the negative timing pop up in various logs in varying conditions.

I'm running 522N and would love to test out some logging params.

This thread at RomRaider has a bunch of experimental stuff:

http://www.romraider.com/forum/viewtopic.php?f=25&t=7864&p=73895#p73895

 

The second post has tables that you can paste into EcuFlash's A2WC522N.xml file.

 

The third post has logger parameters. These need to be pasted into RomRaider's "Logger.xml" file, just above the final "</ecuparams>" element. The last set of parameter are timing compensations. They've all been zero when I logged them, except for one that is used for idle RPM control. But I probably never logged them while ripping through the gears at WOT - I just never thought of that.

 

They're all 4-byte parameters, and the ECU won't let you log a lot of those at once. You'll probably have to try 3 or 4 at at time. If you find one that is active when you're shifting, let me know which one it is, and I will add it to my list of things to investigate.

I didn't have much time this weekend to drive my car but managed to at least get the defs set up. Seems like timing comp 5 is active only during warm up.

 

Timing comp 2 was interesting. Don't know what triggers it but it kicks in randomly while driving around and subtracts timing, sometimes sending it into the negative.

I want to say it does things when there are big changes in wgdc while in vacuum even without big changes in pedal angle, but it doesn't always happen.

Easy to see if you load it in the old ecuedit graphing tool.

 

All of the other compensations stayed at 0 the whole time although I wasn't able to do any WOT driving to test further as I forgot my serial cable for the lc-1 and didn't want to log a wot run without it.

I've posted a log of comp 2 in action. Logs of other compensations happening will be posted as I encounter them.

comp2 sample.csv

Very interesting, thanks for posting that. I'll see if I can figure out what that parameter is all about. It correlates with big reductions in total timing, so it seems likely that it is what caused the negative timing when you went to WOT in that first log you posted.

 

Most of the other stuff I've investigated has changed smoothly over time but this is pretty weird. It seems to operate like a light-switch, being off most of the time and occasionally pulling 20 degrees when it comes on. Really odd.

 

This could take a while to really figure out, but I think I can probably find a way to simply turn it off pretty quickly. Then the question will be whether disabling it any negative side-effects - in other words, is this feature of the ECU doing anything useful? It's weird enough that I think the answer might be "no."

Notice that it doesn't register as knocksum or fbkc like the first log though.

I think there is a possibility that there is something else besides or in addition to comp 2 that affects the wot shift.

That might just be due to the higher load in the first log - you were at about 1.5 g/rev in the first log, and 0.3-0.6 in the sample. So maybe it only triggers the knock sensor at higher loads.

 

I made a couple logs from my car today while running errands and I got results much like yours. When Timing Comp 2 acts up with low load, Knock Sum doens't increment, but when it happened at 0.8 g/rev, Knock Sum went up by 2 (that happened while throttle was changing from about 5% to 17%).

 

But in my case there are a lot of rows where Timing Comp 2 is nonzero and total timing is unchanged. I have a hunch that's just due to TC#2 not being active when the total timing value was sent over the wire, and then being active when TC#2 was sent, but it's hard to say for sure.

 

Also just noticed that all of those Test Timing Comp values are in units of "RPM." That was a mistake.

 

I'll look into the code later tonight, and see if anything makes sense.

Still need to get out and drive the car and have no idea if this is even relevant, but there seems to be some development on the preemptive fbkc front:

http://www.romraider.com/forum/viewtopic.php?f=15&t=8509

 

No idea if this applies only to newer canbus 32bit roms though.

Definitely related in that it's a previously-undocumented feature of the ECU, but these are two different features. That one pulls 1.4 or 5 degrees, and it shows up in the FBKC parameter; this one pulls 20 and doesn't appear in FBKC.

 

I'll start a thread at RR about this thing as well, after I have a bit more to share.

Are you guys sure it's not noise from the shifts?

Yes, absolutely sure. That would show up in FBKC or FLKC, and Knock Sum would increase, but none of those things are happening here.

 

This is an ignition timing compensation, it just needs to be defined so that we can control it.

Well, he's not entirely off track with the noise comment, if you look back to the first log I posted, knock sum DID increment and did show FBKC when coming back on the throttle.

Of course, we also see the strange negative timing issue at the same time as well.

 

That said, I'm not sure that it's noise.

I have no heatshields on the exhaust except for the ones bolted to the tranny tunnel and the turbo heatshield.

The DP is properly bolted to the transmission, and the tmic bracket is secure. Pitch stopper and all mount bolts were properly torqued.

 

I've checked the orientation of the knock sensor, and verified that nothing in the area makes any obvious rattles.

I'd also expect to see more knock in random places if it was noise, but like I said in my first post, the only place I ever see this stuff is on a 1-2 shift.

 

Of course I won't say it's definitely not noise, but I have my doubts.

I suppose it could be my chattery-as-hell ACT clutch but that I'd expect to show up as knock to the ecu in other places too if it was the culprit.

There's knock sensor activity in places in the earlier log, but in the most recent log you posted, there's 20 degrees of timing pulled with no FBKC or FLKC and no knock sum increments. That's the thing that I find most interesting about all of this.

Is it possible that the FBKC and knock sum increment is a RESULT of the undefined compensation?

Pulling that much timing following a WOT shift is bound to make EGTs climb...

There's knock sensor activity in places in the earlier log, but in the most recent log you posted, there's 20 degrees of timing pulled with no FBKC or FLKC and no knock sum increments. That's the thing that I find most interesting about all of this.

 

Speaking with Eric from torqued performance, he said that it sounds like standard overrun and that the ecu will reduce timing during overrun

Is it possible that the FBKC and knock sum increment is a RESULT of the undefined compensation?

Pulling that much timing following a WOT shift is bound to make EGTs climb...

 

Hard to say but it's definitely an interesting coincidence. The other possibility is that we'd see a MORE such knock if we turn off this compensation. Only way to know is to try it and see what happens. At low loads of course.

I did some more driving around and logged other comps simultaneously (1,2,3,6 and 7. Not 5 as that looks like a warmup-only comp)

Being consistent with the previous logs, only comp2 was ever seen as being active.

 

I luckily had a passenger staring at the laptop this time and he told me that it would sometimes activate during relaxed 1-2 shifts at low rpm (3k or so) at normal city driving speeds, and occasionally (and seemingly randomly) when simply coasting in gear off the throttle completely when slowing down approaching a traffic light (afr reading 20.33).

 

There are three logged values I see besides 0 (granted in incorrect units of "rpm"): 193, 192 and 128.

When the comp2 value is 193 or 192, it appears to be pulling timing (or the ecu is at least considering it, as it doesn't always).

When it's 128 it appears to do nothing at all. Maybe the polling rate is too slow to catch some of the events?

 

 

As for my actual fbkc events when shifting (the real purpose of this thread), I did a couple of wot 1-2 shifts and didnt see anything besides the timing comp come into play.

Maybe it's the cold weather and corresponding colder intake charge, or maybe it's the tiny bit of timing I pulled from base timing in the landing spot for shifts.

Who knows, but I'm not ready to say the perceived shift knock is completely solved just yet.

 

I'm planning on removing ALL of the remaining exhaust heatshields from under the car soon just to see what happens.

It's too bad clutches aren't and easy 10 minute swap too as I'd definitely swap my ACT clutch out if I could to eliminate that as a source of noise.

What are your minimum FLKC and FBKC load activation values set at? I had mine bumped on an etune since the newer cars have them set that way and I've seen no low speed, transitional knock.

minimum activation is the same as stock for all knock correction modes.

Maximums I've raised to cover more of the operating range as the stock values seem very narrow (I know the sensor is theoretically inaccurate above a certain rpm, but I haven't encountered any issues stemming from this alone).

Just in case anyone is wondering, I've logged my shift knock both with stock and modified (extended) knock correction thresholds.

I went on ebay and bought some random bosch knock sensor (a matching pair, actually), planning to create some sort of diy det can setup, hoping to eventually get to the bottom of this.

 

A simple combo of the knock sensor, an EV1 connector, some wire, and my Tascam DR-05.

 

For now the sensor is bolted onto the wiring harness bracket under the intercooler (at the very back of the block).

I suspect this is a poor spot and that the audio I'm getting might not be useful at all...

 

I still need to figure out what freq. high pass filter to apply, and determine what the best recording level is (without clipping), and find a better spot for the knock sensor.

 

Anyway, as I said, it needs work, but here's a quick sample of some idling and revs to around 4k rpm or so: LINK TO AUDIO

This guy says it's strictly a function of the bore diameter...

 

http://www.rhinopower.org/knock/knock.html

 

...but I have a hard time believing that the bore depth (at the time of the knock) isn't a factor. I recall reading somewhere else that the primary frequency drops during the event, giving it a very distinctive sound.

See also:

http://forums.nasioc.com/forums/showthread.php?t=2082311