Okay, I was a bit fast...
When the patch became a bit more complex it didnt work anymore with the pulse object as described above.
I figured out the CPU spike when updating the display was actually caused by the Sleep command in the code.
I commented out all chThdSleepMilliseconds in the //update text in the K-rate code of the object.
I now use a timer/delayed pulse object to update the display to keep the update rate slow, and this seems to work, when the delay is set above 100ms.
Could maybe be of help for someone who is experimenting with the LKM board too.