Neil Baylis wrote: > I hope this isn't a really stupid question, but what's the relationship > between encoder resolution and following error? > Ideally, there should be no relationship. But, encoders are a "noisy" source of position information. First, the reading is quantized at the encoder's basic resolution. But, then the position is sampled regularly but the encoder counting hardware and/or software. So, it is DOUBLY quantized, both in position granularity and time granularity. Finally, position is only updated when the encoder passes a quadrature transition, so when moving slowly, position is unknown until it moves a whole count. Position is bounded, of course, in that you know if a count has not occurred, it has moved LESS than one count. > For example, the machine I'm building currently has a linear encoder of 250 > cycles per inch, or 1000 counts. I assume this won't allow me to achieve > 0.001 following error. And I understand FE is also affected by machine > rigidity, encoder noise, and other factors. > Right, if you only have 1000 counts/inch, then you can't maintain position much closer than a couple encoder counts. When it moves one count, you have to apply a corrective force, but you have no idea of the velocity, so the PID has to make a conservative GUESS what that velocity might be. Generally, without continuous-time feedback from a tachometer, it is hard to keep error to less than a few encoder counts. For instance, on my Bridgeport, the X and Y resolution is 25,000 counts/inch. I can comfortably keep the machine within a couple ten-thousandths of an inch under most circumstances. That is probably 4-5 encoder counts. This machine DOES have DC tachometer feedback to a velocity servo amplifier. > Is there a formula that would tell me what FE would be reasonably achievable > with my machine? > > If I install an encoder with double the resolution, will that cut my FE in > half? > It "should" enable you to reduce FE to about half. It certainly will help reduce FE in linear units. But, if you really need to control position to .001", it would be desirable to have encoder resolution of .0001" That is just an old rule-of-thumb that you want 10 X the encoder resolution relative to your smallest movement.
Jon ------------------------------------------------------------------------------ _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
