> Man I hate intermittents.... Say it -loud-!
Although it does not pay a prophet to be too specific, especially about the future, I think this story has / will have a happy ending. Yesterday evening I swapped in a generic PCI parallel-port card (NetMOS chipset, if it matters), aimed the stepper output at it by tweaking the port address in standard_pinout.hal, and had some initial success. All this runs with yesterday's emc2 head. Since then, my test programs have run perfectly, with three exceptions that I (think I) understand. However, I also think the answer isn't as simple as "bad hardware". I modified flowsnake.ngc to plot symmetrically around the origin with a specific radius. The core of the routine is still Ken's original code: http://pastebin.ca/394977 That's run 7 times with no errors: X, Y, and Z are all correctly aligned at the beginning & end of the program. Two runs (after I fiddled a bit with my starting point) left the X axis off by 10 mils, but it turned out that flowsnake stuck out an itsy beyond the right edge of the X-axis travel, just enough to knock the stepper off. So that's a mechanical problem of my very own and, whew, easily solved. I also ran my cam program twice with no errors. One other time the Z axis was off by 8 mils, but I'm not convinced I had the knob zeroed at the start. I think those may be the first three times it's ever run correctly all the way through to the end, so this is a definite, um, step in the right direction. I ran the programs at their default speed and at 200% overrride, both of which worked fine. Of course, my pokey motor acceleration limits flowsnake to well under its top speed, but I'll experiment with that setting another time. As nearly as I can tell, there were three layers of problems in this particular onion, each interacting with the rest. The hard -thunk- and runaway axis behavior after about half a motor turn was (most likely) due to my incorrect (grossly large) FERROR settings. Easily fixed (after you folks pointed it out) and not seen since. The occasional wrong-way, high-speed runaway with no windup was (probably) due to the old stepgen's integer error accumulation that produced an occasional "backward" step at the very start of the motion. Depending on various mystical factors, that errant step could trigger a motor resonance that would spin faster than normal in any direction; in my case it was generally downward. That problem vanished with the new stepgen code. Which left the slow axes slippage during repeated motions. That problem Went Away when I swapped parallel port cards, so it sounds like bad system board hardware. However, I've used three different Dell desktops with this mill and, looking at some early (and rather simple) projects done with those other PCs, I see some of the same slippage. Sooooo... could it be that Dell system board parallel ports have a very small error rate when hammered by EMC? You folks have already established that laptop parallel ports (and system boards) aren't suitable for real-time work due to SMI and other below-the-OS events. Maybe Dell desktops have a similar issue, one that creates a glitch rate of just a few parts per million? I'm not sure I believe that, but ... Anyhow, subject to my actually making some chips, it looks like we may have finally hammered this one into the ground. In the process, we've flushed out a handful of other errors that would have gotten fixed eventually; I know my setup here is a -lot- better than it was. Thanks to all of you! Incidentally, as a side benefit of spending way too much time in the basement, I figured out how to stick HP plotter pens into the Sherline spindle. Turn 'em down to 10.27 mm OD along the full length, punch a suitably snug hole in a random plastic cap that slip-fits on the spindle threads, then lower a drawbar through the spindle to put a little pressure on the tip. Tape some paper to a sheet of plastic, tape the plastic to the tooling plate, and away you go. More (I hope good) news later... -- Ed ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
