Knock Control - Dealing with uneven engine noise graph between cylinders

I recently switched my 1.6 NA Miata (stock engine no Poweradder whatsoever) from an ECUMaster EMU to a Link G4+ Fury.

Knock Control on the ECUMaster was fairly straight forward. I usually gather lots of WOT Run Data to get a nice scatter plot. This way i can easily point out what the knock threshold has to look like (see picture below)

With the Link G4+ cylinder 1+4 and 2+3 are paired for knock control and i get 2 knock level plots. The Problem is that the cylinders are very uneven in their graph (see picture below). While both plots look rather steep compared to the ECUMaster, cylinder 1 and/or 4 have roughly the same overall max. level, but have a much higher level in the mid to high rev range compared to cylinder 2 and/or 3.

There is no change on the engine, same knock sensor i used with the ECUMaster. I use a Bosch Wideband knock sensor. On the ECUMaster i set the frequency to 7.2kHz, and with the Link G4+ to 7kHz narrowband. Engine is running knock free with reduced timing (22°) and on european 102 octane fuel.

At this point i don't know how to set up the knock threshold with this differences between the cylinder pairs. If i reduce the gain for cylinder 1+4 i will also reduce the max level.

Should i deactivate knock control for cylinder 1+4? (which would be my least favourite) Would switching to the second harmonic (14kHz) make a difference?

Any hint or advice is much appreciated!

Hello in the the perfect world i wold be working on not having the cylinders paired up the link fury digital knock is amazing but also needs to be used to its potential,

Regards Ross

I did further testing and tried different frequency settings. Initially i followed the recommendation to use the narrow band setting for the frequency i calculated which is 7.2 kHz. The 7 kHz narowband setting gave me the problems i described in my initial post. I then tried the second harmonic 14 kHz narrowband setting which made it almost even worse (see picture).

After that i tried the 4-10 kHz wideband setting and things look much better now (see picture). The overall shape of the curve is very similar now. The max level is now a lot higher on cylinder 1+4 but that can be adjusted by altering the gain for for cylinder 1+4. I think the issue is resolved now.

Just thought i would share my findings in case someone ran into similar problems.

Greetings from Germany

Markus