Herewith report on follow-up actions to the anomalous behaviour experienced.

 

The rather disorganised structure of the .ini file comes straight out of
PNCconf utility, and it in fact inserts the MIN_VELOCITY globals. The actual
'bug' is that it defaults the fourth joint to ANGULAR (which is acceptable),
but by setting minimum velocities - separately for linear and angular joints
- the velocity vectors remain zero (or so I conjecture) until respective
thresholds are reached.

 

Because PCNconf by default inserts 'loadrt trivkins' and it is up to the
user to change this to 'loadrt gantrykins' with necessary parameter setup
(which as you know we did), LinuxCNC has quite different setups for the Y &
A joints of the gantry!

 

Having deleted MIN_VELOCITY parms and correcting the A-joint type to LINEAR,
the time lag (having resulted from very different velocity vector commands
to the 5i23) is eliminated.

 

Something I wish to know (which our Mesa colleagues will I am sure answer)
is the tradeoff between control loop time and number of pulse generator and
stepper motor instances invoked in the 5i23. How is the stepper control loop
cycle time affected by pwm/pdm frequencies (I guess only FPGA processor
resource sharing between 5i23 tasks)? Has maximum step rate been graphed
against number of instances? Is similar data available for other family
members? Appears to me that to optimise 5i23 performance for stepper
control, one should keep pwmgen frequency around 20kHz (accepting pwmgen
task is trivial vs stepper control). Racking becomes worse as we increase
jog speed to a step interval around 50us (and concurrent motion on all
joints).

 

All the user group support is very much appreciated in bringing us quickly
to the conclusion that the supplier of the power supply has grossly
underspec'ed his product. Yes, beefing up the stepper power supply this
weekend (having overcome the red herring!) should confirm the racking
problem is indeed 'one of those silly things'. This is strongly indicated by
following the suggestion of simply tying Y & A stepper controllers together,
which (either way round) resulted in bad stepper growl, and worsening
racking!!!

 

This exercise is greatly encouraging of the flexibility and quality of
LinuxCNC, but considering that PNCconf is targeted largely at the novice
user, it does need some revision. It would be helpful to include a switch
button to invoke gantrykins when appropriate (with pull-down selection of
coupled joints), to eliminate the mixed linear/angular definitions, and
remove those nasty minimum velocity globals! I will be happy to progress
this with the developer of PNCconf.

 

Thanks to all, we will confirm outcome of beefier stepper PSU asap!

 

Hugh Wylie

 

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to