Curve with noise filter set to 0. (usually I set it to 7)

The brake. Note that the load arm is very stiff. 

Id try cable tying some foam around the load cell to rule out harmonics. Not a permanent fix since it will mess up readings but if that eliminates it or moves the spike then you know its a harmonic issue.

Sorry but Im not completely understand. Do you mean to put Polyurethane foam all around load cell?

Yeah could be anything, You just want something that will absorb some vibrations to make sure thats not the issue.

Ok tomorrow I will try. 

My dyno's chassis is resting on ground by antivibration feet, could be ok or it is better ground it by rigid feet? 

Im not sure whats best all my experience has been with dynos bolted to the ground. Thinking about it logically being on feet will allow it to vibrate more.

This morning I've try with foam but no changes. I also tryed with 0.3 spike and 5 but no changes. During the test I touch the load arm but I havent feel particular vibrations between 13800 and 14000 rpm. 

Im out of ideas.

Eric,

Ref our mail correspondence. Can you confirm that the error was that you had enabled compensation for inertia and that the MOI was way too high? How did it go with the inertia compensation turned off?

Hi Jo, the compensation for inertia and compensation for fictional losses was enabled because I had just reinstall and reset the yourdyno software, but usually they are disabled. I see the test that I sent you and with these 2 parameters disabled, the noise is reduced but still remain too high.

I use a ngk bR10eg spark plug. Next days I try to route cables further away from cdi. 

Whit compensation for inertia and fictional compensation disabled, in the noise test, the noise is +-1.1Nm, and is compatible with the peak in my power curve.