This could also be an analog oscillation problem.
That theory would be supported if you get predictable spindle speed using a battery (e.g. 9V gives you 90% speed, 1.5V gives you 15% speed).
If that appears to be the case, you could try one of the "snubber" solutions.
See https://centroidcncforum.com/viewtopic. ... 98&p=46127
VFD control voltage
Moderator: cnckeith