Knock Logging...

[AbeFM]

I'm trying to (and have been) use the individual knock-per-cylinder (MS3Pro).

I'm logging with MSDroid, but analyzing with MegaLogViewer

To set up the per-cylinder-gains, I drove around, took the entire map and took the average-of-the-averages, divided it by the average for each channel and used that for my gain.

I revisited it that today by getting warmed up and doing a nice long, 0-throttle decel down a hill. Attached is the log, but I only used the decel part.

Code: Select all

		Avg read   cur gain	raw read    	new gain
knock 1	17.17	1.143	15.02187227	1.016593118
knock 2	16.25	1.063	15.2869238	 0.998966971
knock 3	14.39	0.944	15.24364407	1.001803237
knock 4	19.12	1.231	15.53208773	0.983198925
Avgs : 	16.7325		   15.27113197	

My question: It looks like the raw numbers were almost perfect - well within 2%, and that it's only my gains which caused the problems. Does this seem reasonable? I could set all the gains back to 1.0 (or the shown numbers if the gain in the chip will allow it), but is that right, or is this a display issue?

In simple terms, is the "Knock cyl# 1" term the corrected knock or the raw knock, and how is that used in order to trigger knock conditions in the software?

Thanks!
-Abe.

[AbeFM]

I set all my gains to 1.000 and found my data to have a wide reported spread again. Also, I found that (it seemed) was was triggering knock events when perhaps there weren't any - due to some channels not having their gain set low.

I'll need to review the logs, but I'm hoping this 'triggers' some information/insight in one of the readers.

[AbeFM]

I'm still really having issues with this feature.

When I look at the logs, the recorded values are often well below the trigger value set in TS for knock threshold.

I suspect the reported "knock in" is not quite the same thing that the "are we in knock" function is using. Is there a better way to tell the difference between what is measured, what is used, etc?

I think I'm getting either old numbers (i.e. I'm in knock retard from an event which took place, was processed, but never logged because the log point happened once the vlaues had fallen), or an average over 4 channels instead of the current knock level of the properly windowed cylinder.

Similarly, as reported before, getting the gains to make sense is tough. I would like to get the response to a clean run to be as flat as possible and I'm not sure the gains to use - if I average over a large number of samples, I don't know if those recorded cyl 1, 2, etc knocks are pre or post gain (is that software or sent to the hardware as the chip supports)?

Thanks for clearing up what you can. It's one of the best features of the nice new box but it's hard to tell if/when it's working right. I really want to get it correct.

[AbeFM]

So, every few months, I get a bug up my butt about making knock sensing work. It's a powerful feature, but I'm always in knock it seems, so the car never runs right. I don't trust it one iota, but don't want to run without it.

If someone could walk me through how to set it up, I would be VERY VERY VERY grateful. It is probably the single biggest performance/enjoyment hit on the entire car.

My issue is, looking at the logs, I'll be in knock retard when I'm still below the set threshold. Looking at individual channels shows values distributed about the reported "knock level".

I want to:
1) Pull lots of timing, perhaps run e85.
2) Do a couple "known knock free" pulls to redline
3) Somehow use that recording to create a meaningful knock threshold curve.

So far, all attempts at doing this have been utterly fruitless. I would (of course) be even more happy with a "learn knock thresholds" function, which would not rely on the logging rate but rather internally picks the highest knock and sets the threshold XX% (10? User setable?) above the highest values recorded (and perhaps even would warn if there was a sharp peak).

Otherwise, is there a way to log every measurement made (even at the expense of all other information) - just RPM and "highest knock sensor level seen since last record". The rest is a simple manual process.

I see LESS people using knock sensing now than before because they are having trouble setting it up. I'm a firm believer in it and see this as one of the most important features in the MS as a whole.

Please advise on how to do these recordings (like, is Knock Level actually what's used to determine knock retard?), etc. Similarly, setting the per-cylinder gains could be done this way.

Thank you!

[CSXRT4]

Can you post a log and msq of an instance where you had knock retard but the knock levels were not above the threshold?

I think you are on the right track with calculating the gains based on averages. Im not sure if the logged values are pre or post gain but it would be easy to find out, just set one cylinder's gain to 0 and log to see if that cylinder's knock level goes way down compared to the others. It should be that post gain knock level needs to spike above the table threshold the amount of times specified by "knock count(knocks)" before knock retard is activated. Then it waits the time specified in "retard check time" before it starts incrementing knock counts again in the event that you are still above the threshold, at which point it will retard again once knock counts have reached the amount specified in "knock count(knocks)".

I haven't been able to play with the MS knock system yet but will be in a couple months. Based on OEM subaru tuning though, the stock settings for the knock window are starting at 10deg ATDC and the window lasts for about 30deg if I recall correctly. Also you can try doubling or tripling your bandpass frequency to see if you can get a cleaner signal, ive read that second and third order harmonics usually have less noise/interference. Im assuming you did calculate the bandpass frequency setting based on your cylinder size?



To developers, it would be nice to have a knock threshold logging parameter that would log the current knock threshold based on the table output. It would make it much easier to spot occurrences where the knock level exceeds the threshold.

[AbeFM]

What's interesting is which cylinder seems noisier seems to vary - one will lead for a while, then another. This COULD be a sign of numbers getting crossed somewhere, and the gains being applied to the wrong cylinders, but I certainly don't have enough data to back that up.

It's a reasonable suggestion to move the gains - especially since that would also answer what the numbers represent - if the distance between threshold and knock retard increases - i.e. I seen timing being pulled when I'm far from the value - then I would at least know that some of it is "display only". It's sort of silly it would work that way... But I'd feel a lot better if I could talk to whomever wrote that code in the first place.

My understanding of how it should work parallels your own. This 10 ATDC + 30 degrees seems a reasonable place to start, I don't have much else to go on. I'm sure rod-ratios and all that comes into it, but it's a place to start. I'm curious if there window is the same throughout. I'll have to see if I can find some logs - after spending an extended weekend in the desert playing with a sand rail and a competing EMS... Well, I'm kinda behind on all my "real life" stuff and need to do some homework, check in at work, etc. But I'll find you some - the issue is finding a log showing it and having a corresponding MSQ where I know the limit.