Logging frequency for knock signal

Trying to work out the logging frequency required for logging knock and I am getting confused.

Let's say if an engine was running at 6000 rpm, it would be making 100 revolutions per second. Each cylinder should be firing once every two revolution meaning it is firing 50 times per second.

However, once I starting to consider the duration of knock and the length of each knock window, things start to get complicated and I am confused.

With per cylinder knock detection enabled, the detection window is usually set around 50 degree out of the 720 degree combustion cycle. If signal were polled at real time from the sensor, there is only a 1 in 14.4 (50/720) chance that I am polling data within the knock window for a specific cylinder. For the rest of the time, it would miss the window and should give me no reading yet such zero values do not appear in my log.

When the window is closed, does the ECU simply retain the most recent knock value and report it to the logger as data is polled? Also, how long does the shock generate by knock typically last for? If the event occurs at the very beginning of the window, would the signal have faded away by the time it reaches the end of the window? Would the ECU still log the highest value detected during each window or only the value at the moment it was polled?

When logging knock, people usually suggest to log at the highest rate possible but log rate seem to vary greatly between ECUs. For example, I have recently started tuning a Syvecs ECU for the first time where it log at up to 1000Hz and people still recommend to log knock at maximum rate. Other ECU I use tops out just 100Hz, and this makes me worry whether the lower log rate will significantly increase the chance of knock being missed if relying on logs only?

The ecus don't log the raw signal but an interpreted knock level which will be "held" in active memory for a certain time period depending on ecu, providing you log at a rate that will ensure a sample will fall within that latency period for the ecu you are fine. If you were trying to log actual waveforms from knock you would really want 5-10 times the wave frequency so something in the order of 100khz which no ecu I know of will do.

Thank you Slides. No worries I am definitely not trying to log the waveform here. Just the knock signal strength.

Yeah the held in memory makes sense as that's the only way I could explain to myself how logging works.