dave wrote: > Tuning linuxcnc come up time and again. As far as I can tell it is still > an art. I keep hoping someone with the requisite brights will write a > usable diagnostic that gives us a bode plot and then tell us what to do > with it. I certainly can't do it. Zu hilfe! Zu hilfe! Zu hilfe! > Yes, that would be VERY nice. But, I'm not sure how good a job we can do with LinuxCNC and typical motion interfaces. Typically, you only get samples at 1 KHz, and the encoder resolution adds a lot of noise to the amplitude and phase signals. Still, a Bode plot would be really useful. What is needed is to sum the hal signal generator into the position command, and to grab a sample of the halscope data and run a code on that to extract the magnitude and phase of the fundamental response. This is a moving target, so I think actually what you want is the mag. and phase of the error vs. the stimulus. That shouldn't be all that hard to code.
But, then, you need to combine this with the current PID settings and decide what to change. First, I'm not real clear on what to do with a Bode plot in connection with ALL the parameters we have available. Well, OK, first, FFx are out, as they are not part of the control loop. That simplifies things, but we still have P, I and D. I've never had much luck with I, I still think it contaminates the near-present time with really old error measurements. So, maybe that just leaves P and I, and maybe I could figure out what to do with them based on a Bode plot. First, you want to search for mechanical resonances and any big lags (phase shifts) in the servo amps and motors. If there is a big mechanical resonance at, say, 40 Hz., then you have a problem with a 1 KHz servo rate which gives a Nyquist bandwidth of 500 Hz. If you want to keep the 1 KHz rate, you will have to cut down the gain a lot by 40 Hz so as to avoid having gain at or above 40 Hz, where the resonance causes a large phase shift. So, while a Bode plot would be a great thing to be able to generate, and I'd love to play with such a plot on my machines, I'm not completely sure what a Bode-inspired tuning protocol might look like. I have a $13K box (Schlumberger dynamic signal analyzer) that is supposed to acquire the data for Bode plots. I tried to use it on my home-made analog servo amplifiers. I cut the servo loop and added a 100 Ohm resistor, and connected the drive/sense terminals across that. It excited the system with a sine wave, and at the lowest frequencies, I could see the table rock a small amount. But, the data I got made no sense at all, and may not have been repeatable, either. The unit may have some problems, so that may explain the useless data. (I rescued this unit from the local scrapyard and fixed the display, but haven't gone through it for any other problems.) Generating Bode plots from arbitrary machines would be good. The Schlumberger analyzer is analog in/out only, being able to inject a perturbation digitally into the PID and observe the output from the encoder would make it possible to test many more types of motion systems. Jon ------------------------------------------------------------------------------ LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial Remotely access PCs and mobile devices and provide instant support Improve your efficiency, and focus on delivering more value-add services Discover what IT Professionals Know. Rescue delivers http://p.sf.net/sfu/logmein_12329d2d _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
