Distinguishing Knock From Noise

I've got my Haltech Elite 1500 K20 setup working pretty well. At the advice of a local tuner, I replaced the stock 1-wire knock sensor with a 2-wire Bosch Motorsport sensor hoping for less noise. Unfortunately, even though I installed it with nice properly grounded twisted/shielded Raychem cable, there's still quite a bit of noise. This engine has Drag Cartel 2.2 cams, dual valve springs, etc. so it's rather noisy even with the valves adjusted perfectly.

Under WOT conditions I don't see signs of anything that seems to fit with a pictographic description of knock:

Under lower load transient conditions I see some "noise" (I assume at least, I don't hear knock) that is above the baseline by maybe 4-5dB:

In terms of amplitude (percentage, absolute, etc.) is knock a little louder or is it a lot louder (like double?). I want to set the detection up, and don't want it picking up false knock, but I also don't want to desensitize it so far that it misses real knocks (even subtle spark knock at low load) if possible.

Is there a rule of thumb on determining the difference between noise and knock? Is there any conventional wisdom that would guide you toward using a second order frequency (for example 13333Hz instead of 6666Hz, which seems to be appropriate for my engine's 86mm bore)?

--Matt

Absolutely try 2nd order frequencies. That is almost always better for me in setting up knock detection.

I suggest you setup a test to actually induce knock. This is easiest on a load-controlled dyno, where you can hold a fixed RPM and vary the load. You just advance the ignition at your chosen RPM (usually something easy in the 2000-3000 rpm range), only at high load. Now you can increase the load for just a moment, see the knock, and back off the throttle.

Your Knk 1 Signal does not look ilke what I would expect, unless the ECU has some feature to show just peak values. Normally a knock event will be 50% or more higher than your typical peak when running without knock.

Looks like the Haltech has some tools to help you set the frequency. So check out this webinar which details how it's setup. Particularly start at 15:46 where Andre demo's this on the dyno.

https://www.hpacademy.com/previous-webinars/057-elite-knock-control-haltech-elite-2500/

Thank you. Yeah, I'd love to use the knock spectrogram! As a person without a dyno or a helper it is really tough, though. I've not managed to get results from it 😅. The data definitely looks weird, coming from KPro where the signal was much more graph-like. The stock sensor produced the same sort of graph, too, so the Haltech logic must just look like that.

Inducing it in the map and looking at the data seems like something I can do though. 1500-2000rpm should be low enough to avoid too much damage too. 🤞

Where are you located? Everybody that says "there are no dynos near me" is surprised with I find several dynos for them.

Just use your left foot to keep the RPMs in the 2000-2500 range and squeeze the throttle all the way down. Your brakes are much more powerful than the engine and you can go full throttle and hold the car back with the brakes.

Oh, there are several dynos near me (Cleveland, OH area) just no competent tuners to avoid me needing to do it his 🤣, and me too shy to try to coerce them to let me buy time to use them myself. I've bought "dyno tunes" from both capable shops in my area and only received halfway decent fuel and timing at WOT from both for way more money than you'd think for that level of service.

Ok, I'm familiar with doing that to fill in my fuel maps. I should be able to bump some squares up 10° and hit them tomorrow.

Well, it was scary but I did make my engine knock this morning using that process. I wasn't able to use the second order frequency because NDS won't let me pick frequencies past 12kHz, but the sensor definitely picks up real knocks how it is setup. In the range I reviewed, they were an easy 20dB above the typical noise level.

Glad to know it's working, but I didn't like having to test that at all!