Understanding Reference Blending WRT Analysis FPS

[Burtski]

I am trying to understand the effect of Analysis rate vs Camera Frame rate VS Reference Blending.
My motion detection Camera is running at 5 FPS ( 640X480).
I have the Analysis rate set to 2 FPS.

Is the reference image blended at 5 FPS or 2 FPS?

So something like a car moving directly at the camera does not trigger an alarm until very close to the camera.
As it gets closer the number of changing pixels get bigger. The alarm will not occur until there is large change.

But a car crossing the field of view will trigger much sooner and have a smaller amount of change.

With these settings it appears that the analysis in only happening every third frame.
But it feels like the blending is happening every frame.

This sort of makes something that happens on the analysis frame more sensitive than something that has been blended into the last two frames.

Am I missing something?
( Does that even make sense? )

Running Version 1.37.56

[iconnor]

We blend every frame.