Request: More Knock Input Threshold Points

[AbeFM]

I've been trying to get my knock sensing more discriminating. As there are several places in the rev-range where the background noises of the engine are quite loud, with a number of valleys of silence right nearby, I'm finding it difficult to create a descriptive filter using the 10 Knock Input Threshold points provided.

Recently, in a quest to fix this issue once-and-for-all, I went to the top of a local hill, revved the motor fairly high, and did a slow decell, with the throttle fully closed. I was hoping this would be representative of the background levels in the motor and be knock free.



I did a full 30 seconds to make sure I wasn't getting short fluctuations and to make it easy to read.

As you can see, the trough between 5,500 and 6,100 RPM is pretty deep, and as this is where I start to make real (for a miata) power, I would hate to have a high knock threshold in this region. But if I use up 8 or 9 points getting the first few peaks captured, I have precious little left over for the low end where I have the most knock issues.

I would like to see more points on this graph. Is that possible? 16 at a minimum, 20-30 ideally. It may be overkill, but even a few more would be a big help. I never know the background was so complicated till I carefully mapped it out, but now that I have I realize where many of the myths of "silent knock" come from - simply having a quiet spot in the motor where significant knock doesn't show up due to unrealistically high thresholds.

[WestfieldMX5]

7724 is indeed fairly high, you're braver then I am ;)

Are these results repeatable is the big question.

[tpsretard2]

think you need to drop your sensitivity a bit, maybe even your listening window.

[AbeFM]

You mean unilaterally turn down the gain, across all cylinders? That I could, though the peak at 6,100 is very pronounced in every test I do. This particular test showed more background in other places in the rev range than I am used to, nonetheless, it is obvious that it would be hard to draw a meaningful threshold across all that.

I will try using different settings and see what I get. Does anyone else have a similar calibration log to share?

[AbeFM]

Oh, perhaps I should put this on hold until I can verify something (though comments are still welcome):

I read that piston slap can be a big issue on decel and even on light load. Makes some sense, perhaps I should just pull 15* timing everywhere and do a pull.

@tpsretard - how do you set up knock windowing? I saw one place which says they look from 32* AFTER firing, for the next 44. That seems... odd to me, even if it's valid at peak torque, it's too slow at idle.

[masterx81]

An engine under load make different noise than in decel. It depends on map, in fact some oem "fine tune" the sensitivity with map. Do a full pull in long gear free of detonation (confirmed by headphones), then trace the maximum noise profile

[AbeFM]

Since this has gotten crazy-derailed anyway - does anyone have good advice on knock windowing? I've always had a HUGE peak at EXACTLY 6128 (or something like that, only ten or 20 rpm long), I'd love to get rid of.

[S.Bretz]

I would adjust the window to a little earlier than 32 after firing the plug. If you have high boost and lower timing (say 20btdc) that means that is will not start to look until 12 atdc, which will more than likely miss the det hit will definitely miss any hits from pre-ignition. I would think you would want to start monitoring about 10-15 btdc and go out 44 degrees from there.

If I'm wrong, some one please correct me so this guy can get is stuff right.

[AbeFM]

Does it make sense to use degrees verses milliseconds? It would seem for the lowest speeds (1000-2,500 rpm) that you would want to look even earlier.

Anyway, thanks for the advice, ANY filtering would get my noise down. 4 sets of 44* is cutting out nearly 3/4 of the time which has to help.

I'm definitely interested in hearing more.