Thanks mactec54! I'm at a point where only experience will help (of which I have none) since the thing "kind of" works but I suspect something is wrong. Here I'll share few observations and you gurus tell me what you think:
Yes the servo is NOT connected to any load and all my testing is done on the bench.
I did play with the gain settings per mactec54 recommendation. I can make it worse but not better. At its worst the shaft spins even at idle
I've observed uneven rotation speed at very low motor RPM (2-3 RPM) where one can see with a naked eye that the shaft speed isn't constant throughout the rotation cycle. At high speeds it seems good but there isn't really a way to detect any abnormalities at high speed.
So at the low speeds there is definitely an area of the rotation where the shaft slows down. It is about 30-40 degrees wide. It's as if there is a mechanical resistance which slows down the shaft. The location of this problematic area is constant in the rotation cycle whether CW or CCW. Could this be a bad winding for one of the phases? Rotating the shaft by hand doesn't confirm the mechanical resistance hypothesis.
At slow speeds, where the encoder stream is slow enough to see a near-real-time display on SigmaWin I do see that "speed feedback" value does slow down consistent with one's visual observations of the shaft.
So in summary; is this all normal? If not what's a probable cause?