A-Axis improper movements

I am experiencing A-axis improper movements during a long run with version LinuxCNC 2.7.15. My system is Debian 9 stretch. A-axis scale is set correctly and well tested.

In my testing I can see that it does have something to do with the long run. The long run takes 3.93 hours to complete. If I increase the feed speed for a faster test run, it runs complete and accurate. What is most curious is that when it fails, it overruns the expected feed. What I mean is that each movement is supposed to be 7.5°, but when it fails, it runs about 12° instead. 

I have attached my Gcode file, hal, and ini files. 

Does anyone have any ideas on this problem?

Thank you in advance


 

I forgot to mention, the LinuxCNC software shows the correct axis positioning for all the axis's while the physical A-axis(Rotary table) is well out of position. This indicates to me that the software feed to the para port is in error.

I have successfully installed Debian 10 Buster. Further testing reveals that LinuxCNC version 2.8.4-1-gb7824717b as the same problem.

Extensive testing shows that the a-axis runs better when codes G64 p0.0001 and G61 are not used. At the g1 a-180 movement, it ran to exactly +5 degrees over setting to table at 185 while the LinuxCNC display showed it being at -180.

Scaling has been set and tested multiple times. It ran perfectly for the first 27 increments, and suddenly it jumped +5° of it where it should have been. I should add, even though the table was at 185°, and Linux showed 180° on the display, when I ran g1 a0 f1000 it ran back to 0° perfectly and the Linux display showed it at 0°. Very strange!

I see those settings in hal. I will have to read up and see what they do. This is how they look now:
setp stepgen.3.position-scale [JOINT_3]SCALE
setp stepgen.3.steplen 1
setp stepgen.3.stepspace 1
setp stepgen.3.dirhold 145000
setp stepgen.3.dirsetup 145000
setp stepgen.3.maxaccel [JOINT_3]STEPGEN_MAXACCEL
net apos-cmd joint.3.motor-pos-cmd => stepgen.3.position-cmd
net apos-fb stepgen.3.position-fb => joint.3.motor-pos-fb
net astep <= stepgen.3.step
net adir <= stepgen.3.dir
net aenable joint.3.amp-enable-out => stepgen.3.enable

You have steplen and stepspace specified as "1" but I'm not sure if this makes sense
unless "reset" mode is being used.

(and reset mode is not being used, though it might be a good idea
to use reset mode with your low base thread rate)

The hal and ini files that I loaded were the original stepconf with presets settings. stepspace = 0 is for reset mode. I did try that with reset mode, but no joy.

After many days reading and testing various settings, I finally got something very close. I made one disturbing discovery through that process. My rotary table that I have been using for over 12 years has a scale error. The etched scale at 140° jumps to 150° on the next mark, but is only 5° movement. This throws the rest of the scale off by 5°.

It still overruns one total revolution by 1°, but it is appears that is equally spaced over of complete 360° rotation. I was able to get working parts made with this setting.

I want to thank you all for your help. I now know way more than I thought I would ever need to know about hal.