On Mon, 2012-12-10 at 21:09 -0600, Jon Elson wrote: > Andreas Strecker wrote: > > Hello > > The only way i know to get rid of the Problem is to measure the Timepoint > > when an edge is created. This can be done by a Software Interrupt and if > > this is too slow one needs a so called Timer Capture Input which stores the > > timer value when the edge occurs in a Hardware Register. By comparing the > > last edge before the Controlloop with the last edge from the previous > > Controlloop one gets the exact speed. v=s/t. s is a discrete natural > > number, > > so to get exact measurements the time has to be exactly measured. > > Hopes this helps. > > > Not sure exactly what you are commenting about. But, if it is about my > comments > on the accuracy of sample time, then software recording of the time of a > sampled > edge doesn't help, as there can be a time gap between any two instructions > due to interrupts or other CPU disturbances, like DMA. Yes, v=s/t, but > if there > is significant jitter in the t, then accuracy is lost. You only have to > look at > these things in real life to see HOW BAD it can get! Take a case where > there > are 1500 encoder counts/second, and it is being sampled at 1000 > times/second. > On alternate sample intervals you get one or two counts. Is the > instantaneous > velocity changing by 50%? Of course not. Using a time stamp and a higher > sampling rate, you can smooth out these variations a bit, but you still get > some jitter that can't be reduced. It is good enough to make a servo loop > work. I have serious doubts you can extract gain vs. phase info over > a wide range of frequencies with a white noise stimulus to create a Bode > plot in one measurement. > > Jon
Indeed that is bad but maybe not an ideal example either. Just a good example of the pathological possibilities. So we either need encoders with higher counts or we need to tune and higher speeds or utilize both options. We could slow down the clock but that seems like a losing proposition. No reason to go back many years. ;-) Or we decide that bode is not the way to go and try another approach. All of this means nothing if we can't come up with a usable scheme for acquiring and analyzing data. Just an evening rant. Dave > ------------------------------------------------------------------------------ > 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 ------------------------------------------------------------------------------ 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
