You may want to try programming the G84 in Feed per Rev mode. I vaguely remember a problem with g84 and feed per min mode where mach could violate the max accell rate at times. It would also trip my servos here and they are 1kw drives.

FPR mode makes it much easier to program the cycle with(;-) and never had a tripping problem in FPR mode.

ALSO N42 motors are famous for mid range resonance problems. Gecko drives have a setting for them other drives may not.

Just a thought, (;-) TP

Works fine on my mill, see vid http://www.youtube.com/watch?v=deYcqlBMhsA
However possibly worth trying what Terry suggested.
Hood

I'm narrowing this problem down.  The first time I started having this problem was right after hooking up VFD which gave me ability to do semi-rigid tapping.  I spent a week chasing noise issues but couldn't find anything and realized that it has something to do with motor acceleration.  In my troubleshooting I disabled spindle control and things were working normally which is why I originally concentrated on VFD.  One thing clicked today.  It has to do with spindle spin up delay.
When spindle control is disabled mill runs fine; spindle goes down and right back up.  In automatic spindle control MACH uses spin up delay.  In this mode spindle goes down and stops, spindle starts in CCW, comes up ~0.060" waits for the amount of spin up delay, attempts to move up and this is where the stepper stalls.
So, I removed the spin up delay for CCW (tap retract) and now it works fine without stalling.  This is O.K. because of the floating holder but I am really curious as to why I have a problem if there is a spin up delay in CCW direction.  I could see this becoming a problem based on the RPM and the amount of float in the holder.

Hood, there is no need to test the g-code, I think I ruled it out.  Thanks.

Hood, there is no need to test the g-code, I think I ruled it out.  Thanks.

Too late I already did and posted a link to the video
Hood

You may want to try programming the G84 in Feed per Rev mode. I vaguely remember a problem with g84 and feed per min mode where mach could violate the max accell rate at times. It would also trip my servos here and they are 1kw drives.

FPR mode makes it much easier to program the cycle with(;-) and never had a tripping problem in FPR mode.

ALSO N42 motors are famous for mid range resonance problems. Gecko drives have a setting for them other drives may not.

Just a thought, (;-) TP

BR549 I'll research G84 acceleration violation issue you mention.  I didn't have a speed feedback when I started so I always programmed in IPM now with index and VFD an FPR approach sounds like a better way to go.  I'll give that a try.
My drive and motors are a bit dated (1999) but the drive has anti-resonance feature and the motor is set for maximum power band according to the motor curve I attached earlier in the thread. 

Hood I really appreciate your effort.  It's cool seeing other machines running the same code.

No problem
Hood

Even if you don't have the spindle speed index signal IPR works fine it just assumes that what you programmed for spindle speed is what it is.(;-)

Do not try to sync the spindle in mach mill as it is broken on the mill side and it messes up the G84 tap cycle.

To test further you can always program the tap cycle as a sub program and let it run. That would simulate the cycle but do it as straight gcode.

(;-) TP