Usually what I've seen is the OEM sensor will pick up knock in a frequency range undetectable by human ears. It "lights up" and then we check to be sure it was real by adjusting octane or pulling timing or a combination of that with fuel adjustments etc... I will play with it multiple ways to be sure what I am seeing is really the "onset of knock".
Here is a video I made a long time ago, you can see an area of the map where it "always knocks" but there is no actual knock. I graze over it in the video but it is a good example of a knock sensor "lighting up" for no good reason. The engine probably has a part which resonates at the same frequency the sensor is designed to detect in that rpm range, regardless of what load, timing, fuel, or octane I used.

Once the knock becomes audible it may no longer trigger the sensor.

Many ECU have "knock tables", like the 411 LS ECU I am using. It has an enormous chart which adjusts sensitivity of the knock sensors PER cylinder, PER rpm, and so forth. You could spend hours/days/years playing with it.

I do NOT rely on them. The best thing we can do is check the plugs, and tune the engine conservatively for our daily drivers.

 

How would you determine that? You'd have to be able to know that this is true knocking and that is valvetrain noise or BOTH. I know mine does have the knock count going up and back to 0 from 255, I think. Not sure anymore.

 

This is "tuning" there is no one way to do this. My method if written down would take up pages. If you want a couple pages of tuning document I've written you can PM me for it. Here is a quick rundown of things to try when knock is detectable.

1. if the knock keep recurring in the same place on the fuel map, like the same particular cell, it might be false. Especially if you note knock when passing through that cell on a fuel-cut deceleration (when no fuel is being injected).

2. If you have knock going through a cell on acceleration with load, the first and easiest thing to do is try less timing and more fuel.

3. Next we increase octane. Try both the original timing and reduced timing with added 10+ points to the octane. I use C16 racing fuel added to 93 pump fuel to boost octane and reduce knocking.

4. If knock is scattering around randomly regardless of timing and octane, it may be an ignition fault, or improper dwell time in the coil, causing premature sparking in some or one cylinder. Or it could be a too-sensitive-sensor. Or a hot-spot in one cylinder. Or a bad lifter. Or a... See what I am saying? Need to keep looking at the plugs.

5. A hill is useful for applying additional load when there is no forced induction. A load dyno can also be used to hold the engine at a steady RPM under load. Depending on the application you may or may not be able to achieve a non-knocking steady state condition at a constant RPM with a given engine. For example that 3L engine, if I held it at 3500rpm and didn't allow it to accelerate, it would probably build enough boost there to bend every rod in the engine, with no knock. If that didn't happen it might start to melt after a while, if precautions had not been taken to cool specific parts of the engine under prolonged load (like a Boat engine would need).

6. write everything down, generate a chart with "yes knock" and "no knock" for all these different scenarios you attempt. At the end it may come down to trial and error plus intuition, as to how you run the engine or what you do next with the sensor or it's sensitivity. It would suck to "tune out" true knock, and it also doesn't pay to keep having "false knock" showing up either.