Welcome to LinuxCNC.org

The range is 1 to 100, try with another values to see what happens.
HAL connection example:
net x-torque-ff pid.x.commandDD => hm2_5i25.0.7i77.0.1.analogout1
or
net x-torque-ff pid.x.errorD => hm2_5i25.0.7i77.0.1.analogout1
[AXIS_0]Torque_Feed-forward_OUTPUT_SCALE must be in INI file and in this case must be analogout1 not analogout0

Manual says: loadrt pid .... debug=1. I have never used that, so check to see what happens.
pid.N.errorD float ro (only if debug=1) Derivative of error. This is the value that is multiplied by Dgain to produce the Derivative term of the output.
pid.N.commandDD float ro (only if debug=1) Second derivative of command. This is the value that is multiplied by FF2 to produce the second order feed-forward term of the output.

Connect the hal oscilloscope to check to have values only on the accelerating/decelerating phase of the move. Must be 0 during the constant velocity phase of the trajectory. If this is not OK, we have to choose another output parameter of the PID or correct them doing some mathematics.

andypugh wrote:
OK .... I get the subtle differences between the two.

It’s a steep learning curve, when starting from scratch, even using Linux for the first time.
I confess that I haven’t entirely grasped the concept of “net this” and “setp that” in its entirety.
I’m sure it will all just click at some point.

So, as my drives respond (supposedly) by increasing the servo rotational speed, hence velocity of the axis proportionally with the output of PID-X, which relies on velocity from the encoder, for feedback – then as far as my HAL config is concerned, that constitutes velocity mode.
(Bit of a eureka moment for me there!)

"The range is 1 to 100, try with another values to see what happens"

I arrived at “1” after trying 1-20 in steps of 1 and 20-100 in step of 2 – with servo on and off the machine - so this isn’t for the lack of trying!

I’ll try some of the examples you ve kindly come up with and see what happens in halscope to start with.
Thankfully the param for the torque ref on the drive is scalable so i can start with small amounts and see what happens.
fingers crossed ...

I’ll report findings.

ROG wrote:
Absolutely. (though if the feedback to your HAL PID is really the encoder velocity, then there is a problem

Typically a PID commands something one level removed from the desired quantity. So you command velocity and feed-back position, or command valve PWM and feed-back oil pressure, or command valve position and feed-back coolant temperature. (the last two are not entirely random examples that I am baking in southern italy to test at the moment)

Update

Ok ... in ini - I’ve got

[AXIS_0]
Torque_Feed-forward_OUTPUT_SCALE = -100
Torque_Feed-forward_ OUTPUT_MIN_LIMIT -10.0
Torque_Feed-forward_ OUTPUT_MAX_LIMIT 10.0

In HAL I’ve got

loadrt pid names=pid.x,pid.z,pid.s Debug =1

setp hm2_5i25.0.7i77.0.1.analogout3-scalemax [AXIS_0] Torque_Feed-forward_OUTPUT_SCALE
setp hm2_5i25.0.7i77.0.1.analogout3-minlim [AXIS_0] Torque_Feed-forward_ OUTPUT_MIN_LIMIT
setp hm2_5i25.0.7i77.0.1.analogout3-maxlim [AXIS_0] Torque_Feed-forward_ OUTPUT_MAX_LIMIT

net x-torque-ff pid.x.commandDD => hm2_5i25.0.7i77.0.1.analogout3

Basically, I get a square wave “pulse” at the beginning of the move, and the end of the move, pulse width is the same as during the acceleration phase.
analogout3 is connected to the torque ref input on the drive and toque feed forward is enabled.

If I plot pid.x.commandDD in Halscope, it has a value of 350 (regardless of scale)
If I plot hm2_5i25.0.7i77.0.1.analogout3 is showing a value of 350 (regardless of scale)

Although the scaling appears to be working, as I had to add the minus sign as the output was reversed. So despite this scaling appearing to work, the value of pid.x.commandDD & analogout3 is either zero or 350 in Halscope, forming a square wave for the duration of the acceleration period.

Any ideas why 350 seems to be a constant and where this figure comes from? (Just thought .. acceleration is set to 350 in ini)
Any thoughts why analogout3 plots with a value of 350 when it should be scaled?

I haven’t had chance to assess whether this is going to cure the problem yet but it is definitely adding some grunt to the acceleration phase (amoujt of grunt changing with scaling).

I’ll do some more tinkering tonight and upload some plots of the results.

May you plot pid.x.error, pid.x.output, encoder velocity, axis.0.axis-vel-com, pid.x.commandDD and pid.N.errorD during the acceleration phase ?
Try to use the same scale in the graph of requested velocity and feedback velocity, is easier to see and analyse. The goal is to match the real velocity to the requested velocity. As example use www.linuxcnc.org/index.php/english/compo...mit=6&start=12#20671