trajectory acceleration is half the value in config file

Hi -
     I'm interested in finding a good acceleration value for my three-axis router, for fast part program execution without harsh motion.  The first thing to do is to push acceleration to the point where the actual velocity profile can't achieve the same sloppe as the desired velocity profile.  My machine uses a Motenc-Lite PCI motion control interface board from Vital Systems, so the motion parameters are in linuxcnc/configs/motenc/motenc.ini.
     In the TRAJ section of motenc.ini, I set MAX_ACCELERATION to 40, and I also set it to 40 in the AXIS_0 section.
     Then I did a .5 inch X+ move at 1 inch per second (G1 X.5 F60.) and recorded the desired and actual positions of the X axis at each servo heartbeat (every millisecond).  I used SciLab to differentiate the positions to obtain velocities.  I plotted the velocities.  It took about 50 milliseconds for the velocity profile to reach speed, so the acceleration was 20 in/sec/sec ... or half the value I had entered in motenc.ini.
     I wondered if I had made a mistake.  I repeated the experiment, making sure to re-start LOinuxCNC so it would have to read the motenc.ini file.  I got the same result.  I wondered if for some reason LinuxCNC was ignoring the motenc.ini file and using a 20 in/sec/sec acceleration value instead.  To test this last possibility, I double the value for MAX_ACCELERATION two places in motenc.ini and re-ran the experiment.
     This time, it looks like indeed the system is reading the motenc.ini file and using the MAX_ACCELERATION value.  Once again it is half the value in the file.
     I'll attempt to attach a .jpg showing the plot.
     The smooth trace is desired velocity.  The jagged trace is actual velocity.  Differentiating real world data always produces a jagged trace.  These are actually pretty good plots.
     You can see the system reaches speed in about 25 milliseconds, which is consistent with an accel value of 40.
     40 in/sec/sec x .025 sec = 1 in/sec.
     Maybe somebody else has seen this same thing.

In my torture testing  >36m/min rapids and 5 to 8 m/sec/sec accel
Steppers loose steps when they get hot.
I found that by using Lam Technologies drivers, I could use their boost setting and a custom component to reduce current when at constant velocity to say 30-50% of peak.  This let the motors cool and substantially increase performance as they are only working flat out on accelleration and decelleration. Cutting at 2 m/min, that meant that the high current was only aplied for about 30 ms at the beginning and end of a segment.

By just looking at your desired vs actual profile, you are not looking at the performance impaact of heat. So its likely that your finely tuned configuration will fail once it does some real work. THe result won't be pretty Trust me!

Thank you for the reply.
Pushing the motors too hard is a topic but not really what I'm interested in right now. I'm more concerned about the how actual acceleration is half the value specified in the config file.
I found a forum post from 2019 by dm17ry that is almost identical, finding actual acceleration (in some cases) is exactly half the config value. I don't see any resolution or explanation in that earlier case.

Is it the same if you max-out the feed override?
It is possible that the TP is leaving some overhead to allow for changes to FO in real time.

Maybe posting your ini file  and LinuxCNC version would help.

I don't see this with a quick test:
(LinuxCNC master 30 IPS max 2000 IPS/S max)


 
 

Link to the mentioned thread by dm17ry:
forum.linuxcnc.org/49-basic-configuratio...eration-value#128957