On 03/25/2016 02:35 PM, Sebastian Kuzminsky wrote: > On 03/25/2016 01:26 PM, Nicklas Karlsson wrote: >> As someone mentioned before a the an axis/joint may/will run away in >> case of an encoder failure. >> >> It would be possible to detect an encoder failure by comparing output >> signal with encoder signal. A velocity signal may be integrated and >> do not allow to wander to much or an encoder error will be detected. >> A torque signal would be worse. > An encoder that fails into runaway (motion erroneously detected when > none is occurring) or that fails into stop (no motion detected when > motion *is* occurring) will both result in a following error, precisely > for the reason you say above: linuxcnc compares the commanded position > with the feedback position from the encoder. > > It will not detect a following error if the failed encoder is on an axis that does not have any motion commanded at that time. This may not cause a catastrophic runaway on some systems like Mesa and Pico Systems analog velocity controlled servos with tachometers, but a slow creep toward the limits. Some other systems might have an integrator that will just keep winding up without feedback.
Old Allen-Bradley controls generated an absolute velocity analog signal from the encoder, and compared that to the absolute value of the tach signal. If the tach came out 20% above the encoder-derived value, it caused an E-stop. This was in the days of incandescent lamps in the encoders, so these errors did occur. They also used 4-sensor encoders, where they had a separate optical sensor for the A and the A-not signals. (Same for B and B-not.) These 4 signals went to Xor gates that made sure the A and A-not and B and B-not were complementary. If A and A-not were equal for more than a few us, that caused an E-stop. This detected a failed lamp. Jon ------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://pubads.g.doubleclick.net/gampad/clk?id=278785351&iu=/4140 _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users