Prior to tackling the control system of my son's machine, I completely rewired spindle, stepper and limit / home wiring since he had significant EMI issues (despite opto-isolation between VFD and PC / breakout electronics).
I put in Lapp Oelflex shielded cable throughout (spec'ed for low bending radius, although better and more costly cable is available, particularly for robotic applications), and reinforce the notes written by Gene Heskett re grounding practice. But on your issue in particular Viesturs, it is only one axis that is divergent. This is interesting, given the comments by Andy that MIN_VELOCITY should have no effect. I had hoped my hypothesis of the time lag problem explained the (deterministic) anomalous behaviour on my son's machine! I will apply halscope to this lag problem once racking has been eliminated (assuming sharing the four joints between two identical PSUs is the short-term fix!), since it appears there is something deeper perhaps specific to A axis in LinuxCNC. Let's keep in mind that hostmot2 issues velocity (not position) vector commands to the stepper driver, and if different thresholds to initiate motion for each joint, this should be recordable on halscope. Viesturs, has your client problem arisen only since LinuxCNC was deployed? Can you revert to the original control system to validate this? Perhaps it is a mechanical issue as suggested in the posts. In general, it is good to know that there is plenty headroom available from the Mesa FPGA-based PCI cards. [I had the good fortune to design around gate array technology during my younger days, and how these FPGAs have transformed the industry!] Another step I wish to follow is to look at breakout board alternatives. That currently in use uses a daughter board to map the 50-way ribbon connector to parallel port pins, so requiring parport deployment for input and output expansion. I guess the local supplier (whose hardware is also sold by www.leeuwinga.nl ) will adapt his breakout board which has rather too short charge pump retriggerable monostable period for LinuxCNC (given capacitor manufacturing tolerances). We increased the cap from 0.01 (+/-20%) to 0.47uF to provide 50ms nominal timeout (knowing that minimum should be ~5x servo period, to which thread charge_pump is linked). We can trawl the web but based on community experience, which breakout board has proven to be of sound EMC design and best suited to exploiting the flexibility of LinuxCNC (free of parport I/O expansion)? The local supplier may also enhance his product to include input and output flexibility achievable with the 5i23 or other family members, since it is otherwise quite usable (many Mach 3 references, I understand). ------------------------------------------------------------------------------ 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 Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
