Good find! After running that program, I'm seeing similar spikes. That error is a subtle one, and I think I know what's causing it. There are rare cases where a segment hits its target almost exactly, so the split time is small or non-existent. All of the acceleration spikes in my logs correspond to this case, so I need to take another look at what happens when we hit the target exactly.
As for possible realtime errors, my answer for now is to simply run the loop slower until we can do some profiling and optimization. My code as it's written now is a bit expensive to run a 5 or 10 kHz, but we shouldn't need to run that fast unless you have a very fast and very accurate machine. I've also been working on how to correlate desired accuracy and velocity to the update rate. For example, if you want segments that are 0.001" long with a peak velocity of 120ipm, then you need a ~12kHz update rate to be able to hit every segment (with blend arcs) at that speed. This is of course the worst-case scenario, as other limits such as cornering acceleration and machine accuracy will be the limiting factors. -Rob On Sun, Dec 1, 2013 at 2:46 PM, sam sokolik <[email protected]> wrote: > ok - I still think there is an issue - maybe with really short line > segments? this program causes a huge jump in position and accelleration.. > > http://pastebin.com/97YFUnsM > > (around line 1064) > > That is with strait G64. If I do a G64P.010 - it runs much further > through the program - almost 3/4 or so before it fails > > (whole program) > http://electronicsam.com/images/KandT/testing/stellabee1.ngc > > no realtime errors on any runs... > sam > > > > On 12/01/2013 11:33 AM, sam sokolik wrote: > > wow - more testing - so far - it looks like the constraint violations > > are from the realtime issuses.. running 1khz on this computer is not > > showing the errors I was seeing > > > > It really is quite awesome. The performance improvment over the current > > tp is pretty substantial.. > > > > Still testing > > sam > > > > > > On 12/01/2013 10:45 AM, sam sokolik wrote: > >> ok - those blends make sense (other than I don't know why they are > >> parabolic) as those segments are longer.. > >> > >> sam > >> On 12/01/2013 10:12 AM, sam sokolik wrote: > >>> ok - a little more testing (just a little) > >>> > >>> I wonder if what I am seeing is caused by realtime errors.. I was > >>> running 5khz thread which is on the edge on this computer. when I went > >>> to 1khz it obeyed constraints.. > >>> > >>> There are a few blends(large) that look odd to me - you can see it here > >>> > >>> http://imagebin.org/279832 > >>> > >>> sam > >>> > >>> ps - if the previous failure was caused by missed realtime - I wonder > if > >>> that needs to be figured into things. (if a realtime target is missed > - > >>> the machine should not go wonky) > >>> > >>> sam > >>> > >>> > >>> > >>> > >>> On 11/30/2013 04:52 PM, Robert Ellenberg wrote: > >>>> I'm looking into it now, but I think it's an optimization issue. I'm > >>>> exporting position, velocity, and acceleration data through sampler > now, so > >>>> it's easier to identify small blips in velocity. With smoothing > enabled, > >>>> there shouldn't be any significant blips in velocity on the spiral > example. > >>>> > >>>> > >>>> On Sat, Nov 30, 2013 at 5:24 PM, sam sokolik <[email protected]> > wrote: > >>>> > >>>>> I pulled it yesterday. mid day or so.. I will try to get a halscope > >>>>> tonight (and check to see if I have the latest git) > >>>>> > >>>>> what are your thoughts on the spiral velocity? > >>>>> > >>>>> sam > >>>>> > >>>>> > >>>>> > >>>>> On 11/30/2013 01:41 PM, Robert Ellenberg wrote: > >>>>>> That's an interesting result. I just ran 3D Chips myself with my > test > >>>>>> config (tests/trajectory-planner/circular-arcs) and it didn't have > that > >>>>>> hiccup. However, I had an issue a few commits back that looked like > that. > >>>>>> It happened when a segment switched from parabolic blending to arc > >>>>>> blending. I fixed it by adding the split-cycle check during > parabolic > >>>>>> blends, so that we shouldn't get to the end of a segment without > knowing > >>>>>> exactly when to stop. Are you running the latest code? I pushed an > update > >>>>>> recently, so that might make a difference. Also, could you upload a > >>>>>> screenshot of the spiral code HAL scope? > >>>>>> > >>>>>> There's another possible cause: I found a possible error with one > of my > >>>>>> formulas in the tpCreateBlendArc at lines 768-774. It will be a > simple > >>>>> fix > >>>>>> I think, so I should have updates by tonight. Thanks for the > feedback! > >>>>>> > >>>>> > ------------------------------------------------------------------------------ > >>>>>> Rapidly troubleshoot problems before they affect your business. > Most IT > >>>>>> organizations don't have a clear picture of how application > performance > >>>>>> affects their revenue. With AppDynamics, you get 100% visibility > into > >>>>> your > >>>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > >>>>> AppDynamics Pro! > >>>>> > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > >>>>>> _______________________________________________ > >>>>>> Emc-developers mailing list > >>>>>> [email protected] > >>>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>>>> > >>>>> > ------------------------------------------------------------------------------ > >>>>> Rapidly troubleshoot problems before they affect your business. Most > IT > >>>>> organizations don't have a clear picture of how application > performance > >>>>> affects their revenue. With AppDynamics, you get 100% visibility > into your > >>>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics > >>>>> Pro! > >>>>> > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > >>>>> _______________________________________________ > >>>>> Emc-developers mailing list > >>>>> [email protected] > >>>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>>> > >>>> > ------------------------------------------------------------------------------ > >>>> Rapidly troubleshoot problems before they affect your business. Most > IT > >>>> organizations don't have a clear picture of how application > performance > >>>> affects their revenue. With AppDynamics, you get 100% visibility into > your > >>>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics Pro! > >>>> > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > >>>> _______________________________________________ > >>>> Emc-developers mailing list > >>>> [email protected] > >>>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>>> > >>> > ------------------------------------------------------------------------------ > >>> Rapidly troubleshoot problems before they affect your business. Most IT > >>> organizations don't have a clear picture of how application performance > >>> affects their revenue. With AppDynamics, you get 100% visibility into > your > >>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics Pro! > >>> > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > >>> _______________________________________________ > >>> Emc-developers mailing list > >>> [email protected] > >>> https://lists.sourceforge.net/lists/listinfo/emc-developers > >>> > >> > ------------------------------------------------------------------------------ > >> Rapidly troubleshoot problems before they affect your business. Most IT > >> organizations don't have a clear picture of how application performance > >> affects their revenue. With AppDynamics, you get 100% visibility into > your > >> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics Pro! > >> > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > >> _______________________________________________ > >> Emc-developers mailing list > >> [email protected] > >> https://lists.sourceforge.net/lists/listinfo/emc-developers > >> > > > > > ------------------------------------------------------------------------------ > > Rapidly troubleshoot problems before they affect your business. Most IT > > organizations don't have a clear picture of how application performance > > affects their revenue. With AppDynamics, you get 100% visibility into > your > > Java,.NET, & PHP application. Start your 15-day FREE TRIAL of > AppDynamics Pro! > > > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > > _______________________________________________ > > Emc-developers mailing list > > [email protected] > > https://lists.sourceforge.net/lists/listinfo/emc-developers > > > > > > ------------------------------------------------------------------------------ > Rapidly troubleshoot problems before they affect your business. Most IT > organizations don't have a clear picture of how application performance > affects their revenue. With AppDynamics, you get 100% visibility into your > Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics > Pro! > http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk > _______________________________________________ > Emc-developers mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-developers > ------------------------------------------------------------------------------ Rapidly troubleshoot problems before they affect your business. Most IT organizations don't have a clear picture of how application performance affects their revenue. With AppDynamics, you get 100% visibility into your Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro! http://pubads.g.doubleclick.net/gampad/clk?id=84349351&iu=/4140/ostg.clktrk _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
