On Sunday 15 July 2012 08:32:08 andy pugh did opine: > On 15 July 2012 02:56, Kasey Matejcek <someo...@lkm.bz> wrote: > > No home made > > With optical pickup > > The problem with a single-channel encoder is that if you double-count > an edge then the apparent velocity appears to suddenly increase > massively. > Quadrature counting almost entirely eliminates this problem. > If you can add a second sensor N + 1/2 pitches away from the first one > to provide a quadrature signal then you should see better results. You > will also have the ability to sense spindle direction, and that means > that rigid tapping becomes more possible (you also need an index for > that, but index is fairly easy to add)
All very true Andy. But I think I just figured out why my encoder signals that look great on a 100 mhz scope, are so noisy digitally. I may have, at the AB inputs, perfect square waves whose duty cycle is within a percent of 50/50, and whose quadrature is a near perfect 90 degrees. This should give an accurate digital image of the spindles position, right? But this is the real world, the duty cycle might be 48/52 for one cell, and 49/51 for the other. By the same mechanism, quadrature might be off 2 degrees depending on which edge. I made about 15 wheels to get it that close. But the internal calculations for velocity etc are done at the servo threads granularity of nominally 1 millisecond. So there is no chance of the figures being accurate when the count may be up to a millisecond old. Perhaps with more resolution than my 39 cycle wheel, which is 156 counts per turn or a new position every 2.3076923076923076923 degrees of rotation, it might improve. But the addition of the noise caused by the 1 millisecond granularity of the floating point error process is an apparently huge error sample to sample. Looking at the encoder output scaled to rps, with the spindle at full 1200 rpm (in low gear) s/b an rps of an even 20. With the rotational mass of that several pound chuck, there is zero chance of even a .1 rps REAL error. Its very difficult to determine without running that signal thru a lowpass block with a gain setting of .001 before feeding it to a halmeter, because even at that rps, the halmeter flickers between 17 and 23 from this (to me utterly false) digital noise! When cutting threads with the G76 canned cycle, with the spindle between 2 and 6 rps, it cuts great threads, but the racket as the z motor tries to track this noise is similar to slowly dropping a shovel full of pea gravel into a washtub. I should post a video with audio, its very rough sounding. I am not privy to how the encoder module works internally, but it seems to me that if a timestamp could be put on the base thread sample _only_ when a new edge is detected, giving the encoders FP thread an idea of how long the data has been in that state, one that could be compared to the instant time, then much of this digital error noise could be calculated away by simply using that frozen timestamp as the basis for the calculations instead of the instant system time. Only when this has been done should the elapsed time differences be applied to compensate for the actual age of the sample, which in turn should give us a pretty good idea of the spindles actual position at the instant the FP thread runs, doing it to a small fraction of a degree in spite of a 2.3 degree granularity in the encoder wheel. Food for thought. Or am I carrying coal to Newcastle here? That, and I'm still a litre low on coffee, I made it, but haven't input any yet. But I survived the night, so a hearty good morning to all! Cheers, Gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) My web page: <http://coyoteden.dyndns-free.com:85/gene> is up! There cannot be a crisis next week. My schedule is already full. -- Henry Kissinger ------------------------------------------------------------------------------ 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
