Potentiometer bug?

I noticed a strange behaviour with the G0 and G1 speeds when they are set outside the GUI with a potentiometer.

Lets say that the potentiometer is at 50% of its travel which equals to a speed of 5000mm/s

Scenario A: If I slowly rev the potentiometer from its starting position to the starting point == 0% then the GUI reads the value as expected and the feed rates are close to 0mm/s

Scenario B: If I rev the potentiometer from its starting position(50%) to the starting point == 0% but with a very fast movement then the GUI reads the value incorrectly. The feed rates are high somewhere between 5% and 45%. It depends how fast you rev it out.

I have the potentiometer wired to the 24v line and scaled so it gives 100% at 24v and 0% at 0v.

In both scenarios if I read the values at the potentiometer pin it is close to 0 which implies that the GUI reads/calculates the feed speeds somehow incorrectly. Wrong interval?

Even if the pooling interval is set to high (it lags), The GUI should read the value of the potentiometer at 0v and set the feeds accordingly.

What am I missing?

I should also add that the "progress bar" - percentage of the feed rate in the GUI is shown correctly - 0% but the machine still moves with like 20% or more of the max feed rate. To correct this problem, I have to back up the potentiometer a bit or just move it slowly around its end. Then linuxcnc's feedrate corresponds to the GUI "progress bar".

This may not be a problem with gmoccapy.

When milling, you can see the feed rate on motion.current-vel

When you turn the potentiometer to 0% with a fast move, the motion.current-vel still stays over value 10
Then when you back up the potentiometer a little towards the 100%, the motion.current-vel initially drops and then starts to increase.

This will not happen if you slowly turn the potentiometer towards 0%. But slowly is not always an option.

Where lies the problem, does the axis GUI also have this phenomena?

This is very interesting. I have pot ORs on my machine as well and not noticed this issue when running Gmoccapy or any of the other UIs. I tend to use feed and max vel ORs instead of e-stop, and I easily get scared with the 300ipm rapids on this machine so frequently spin them all the way to 0% very quickly.

What version of LinuxCNC and what interface card are you using?
What HAL pins do you have the pots connected to, halui.feed-override.counts and halui.rapid-override.counts?

A classic car has recently forced my mills into a corner, but if there is enough room to operate the pallet jack I will pull one out enough to run and see if I can duplicate this.