Babystep bug w/ incorrect motion

nhof

Hello, I am running into a strange issue with the Z babystepping. It appears that the babystep is moving an incorrect distance, then the wrong direction.

Firmware version is 2.01 RTOS.
PanelDue version is 1.21.3 on a V2 board with the 7" screen.

Test setup: print running perpetually back and forth at Z0.20 to simulate a first layer. Babystepped using the menu from the PanelDue print screen. Jog moves are quick and I waited for the system to settle before taking readings to allow for any ramp of the babystepping.

Babystep is the value commanded from the babystep menu on the print screen (and displayed on the print screen babystep menu)
Z actual is the actual movement of the Z axis during the print. This was taken using a mitutoyo dial indicator clamped onto the extruder.
Z on control screen is the Z axis position readout from the control screen of the PanelDue.

babystep	Z actual	Z on ctrl screen
0.00	0.20	0.20
-0.10	0.15	0.20
-0.20	0.10	0.20
-0.30	0.15	0.30
-0.40	0.20	0.40

From these readings, there are a couple things to note:

Actual motion of gantry is only 1/2 of commanded babystep. for a commanded babystep of -0.10, only -0.05 of motion occurs.
Upon "zero crossing" (where the negative babystep value is higher than the first layer height), the Z axis moves the opposite of the commanded direction.
Upon "zero crossing", the axis position readout on the control screen goes up, even while the babystep commanded value continues to drop.

I have already ruled out a motor or steps/mm issue as the system will respond quickly and accurately to normal jog moves in Z, and will move the correct distance when changing layers normally, etc. Hopefully this is something obvious I am missing but I haven't figured it out yet.. Other ideas would be appreciated.

Here is a quick chart I have drawn showing the behaviour if it helps to visualize.

dc42

I've added this to my list to investigate, but please re-test with firmware 2.02RC2.

It's odd that the displayed height is changing - it should remain constant at 0.20mm.