Knock Detection
 

I'm trying to tune spark using my electronic det cans. I'm listening to a bosch sensor bolted to the engine mount. In low RPM I can pick out knock clearly, but in midrange RPM it sounds like there's knock everywhere in boost.

I keep cranking down the timing and adding fuel to no effect.

Here are some recordings I made from the sensor. Can someone tell me whether it's really detonation I'm hearing, or just some normal 1.6L miata noise?

http://soundcloud.com/tenpennygrin/s...ock-detection/

It's been band-pass filtered, centered at 7400Hz.

Thanks!

That doesn't really sound like what i associate with knock when I am listening, but then again, I'm not an expert. Mine did dissipate as I removed timing. Mine sounded like the clicking of a stun gun.

Can you post an image of your timing map or msq? What's your afr look like when you recorded these?

I've examined knock using Spectrograph before.
Pls. make your files downloadable.

All of the files have knock. I played around using Goldwave and Spectrograph.

The knock almost sounds like a referee's whistle when played in real time.
If you slow down the playback by 10-20x, the knock has a characteristic "descending bell" tone. You can see it in the spectrograph. The rate that the tone descends is proportional to RPM. The combustion chamber's resonant frequency decreases as the piston moves down and the volume increases.

The low RPM knock file shows strongest frequencies in the range 12800 to 13500 Hz. The tone descends from the higher, to the lower frequency.
The other files did not have these frequencies. I saw this high frequency too when I did knock recordings at low frequency.

Your mid-high RPM recordings show knock in the frequency range 5720 to 7450 Hz. The tone descends from the higher, to the lower frequency.
If I use goldwave's bandpass function with the default 5th order with the above frequency range, you can see that a threshold of -55 dB works reasonably well at detecting the knock, in all the files. A simpler, 2nd order filter has difficulty discriminating between very light or no knock, and light knock (below).

The 2nd file has light knock at around the 4.5 sec mark. The earlier peak at 3 sec is not knock, or is very light knock. Any detection algorithm shouldn't be false triggered by that one, and should detect the one at 4.5s. The first strong knock is at 13.5 s.

A simple bandpass followed by an amplitude detector algorithm has difficulty detecting light knock. A DSP that pattern matches the "descending bell" tone will be much, much better.

Is this waht it sounds like to everyone using electronic det cans? If so, then I think that I am hearing something else in my det cans. Is it possible that I could be hearing pre-ignition? Mine sounds like the tapping stun gun makes as it fires. It does get less pronounced as I reduce the timing and then dissappears.

I'm wondering if i should try a colder plug for kicks.

dirtythyrtyix,

Do you have the raw unfiltered files?
What filter did you use, a 2 pole bandpass analog filter?