John that same counting cleverness needs to be in or merged with the differential for helical gears to allow it all to remain in gear and have a fine feed and to move the cutter head back to be able to measure and take another shaving.
http://www.collection.archivist.info/hobbing.html having used the hal method I did get a slight error http://www.collection.archivist.info/searchv13.php?searchstr=speedo pics at bottom show a slight helix I did add encoders to the machine to see if there was any slip/step loss/other bugs but did not find the error, being small when sliced into gears it remains as a feature. Dave Caroline On 30/07/2015, John Kasunich <jmkasun...@fastmail.fm> wrote: > > > On Wed, Jul 29, 2015, at 09:35 PM, richsh...@comcast.net wrote: >> I am adapting a version of the "little hobber" and want to use linux-CNC >> to do the ratio division between the hob and the work spindles. >> Additionally, a third axis that is actually the feed needs to be >> incorporated. Any thoughts. > > The "encoder-ratio" HAL component was invented with this specific use in > mind. > http://linuxcnc.org/docs/html/man/man9/encoder_ratio.9.html > > It is a software based encoder counter (like the regular software encoder > counter), but it does the math differently to avoid some problems that can > happen if you use a more "normal" approach. > > The normal approach would be to use either software or hardware encoder > counters to measure the position of each axis, then scale the positions > based on the tooth ratios, and eventually compare them to determine how to > drive the slave axis. The problem is that if you run long enough you start > running out of accuracy. Imagine that your hobber has been running for > hours. You might find yourself subtracting 100,000,000.03 revolutions of > the master from 100,000,000.05 revolutions of the slave. If the math > doesn't have enough significant digits, the 0.02 revolution difference gets > lost in the noise. > > The encoder ratio component uses a different approach to the math, and will > never run out of accuracy no matter how long it goes. > > -- > John Kasunich > jmkasun...@fastmail.fm > > ------------------------------------------------------------------------------ > _______________________________________________ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users > ------------------------------------------------------------------------------ _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users