--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Troy,
>
> It is very complicated.  I wish I understood it.  When running KMotionCNC there are basically 3 Threads.  Two in KMotionCNC.exe (one doing the GUI and another doing GCode Interpretation and Trajectory Planning).  One in KMotionServer.exe doing the USB IO.  Then there is Windows Kernel mode for IO like USB, Disk, and Display.  Then there are probably 50 other Windows background things going on.  If everything is balanced properly the main bottleneck should be the USB data rate of 12 MBits/sec.
>
>
> TK
>
> From: tmday7
>
>
>  
> Wow, i didnt relize that multi processing is jumping back and forth between tasks like that. Interesting.... So one should be able to force more CPU time to Kmotion and such?
> Troy
>
> --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> >
> > Hi Albert,
> >  
> > There is now a preliminary release called V4.30z that has he Trajectory Planner Bug Fix.  Here is a plot of the before/after result:
> >  
> > http://dynomotion.com/Software/Patch/TrajectoryPlannerBug4Resolution.pdf
> >  
> > Regarding NtSetTimerResolution(): I'm not sure.  I did a lot of experimentation and on my computer (i7 W7 32) i doesn't make any difference.  My system is quite fast.  I get download rates of 3315 segments/sec.  I updated the TestUSBRate.ngc GCode test a bit.  The main thing is that if the test is run so that the Feed Rate is limiting the speed then there is half as many segments as when the max acceleration and curvature is limiting the speed.  This is because in the second case each segment has two parts to download a constant velocity and an acceleration portion.
> >  
> > Anyway I think we should to do some more investigation before adding the NtSetTimerResolution().  That is an undocumented Windows call that has a global effect on Windows,  One of the main goals of KMotionCNC is to run standard Windows without relying on any special tricks.
> >  
> > A User can always run a separate program to change the Windows setting if he/she wishes.  Maybe we should supply one.   I think there is a Windows Registry setting as well:
> >  
> > HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\PriorityControl\Win32PrioritySeparation  2  ??
> >  
> > http://www.jhouseconsulting.com/2008/05/13/processor-scheduling-20
> >
> > Regards
> > TK
> >  
> >  
> >  
> >  
> >
> > From: albertplatek <albertplatek86@>
> > To: DynoMotion@yahoogroups.com
> > Sent: Thursday, June 28, 2012 9:57 AM
> > Subject: [DynoMotion] Re: Trajectory Planner Bug 4 (uncommon cases)
> >
> >
> >  
> >
> >
> > Hi Tom,
> >
> > Thank you for very precise description of the problem, if there will occur other bug I will try isolate it as you did. But not sure :)
> > Your understanding is good for me. ( I made lot of tests, and it seem that it is it)
> >
> > (btw. does it make sense to add NtSetTimerResolution() to KMotionCNC app?
> > I know that it help not on all systems but more performace is always good)
> > (it seems that it can use more battery power on laptops but it does not matter for cnc systems)
> >
> > here simple c++ code (devcpp) of it:
> >
> > #include <windows.h>
> > #include <stdio.h>
> >
> > HINSTANCE hDLL;
> > typedef void (*PFN)(unsigned long,bool,unsigned long*);
> > unsigned long ActualResolution;
> > extern "C" { _declspec (dllexport) void NtSetTimerResolution(); }
> >
> > int main(int argc, char* argv[])
> > {
> > hDLL = LoadLibrary("NTDLL.dll");
> > PFN FuncPtr = (PFN)GetProcAddress(hDLL,"NtSetTimerResolution");
> > FuncPtr(0,TRUE,&ActualResolution); //set smallest possible timer res...
> >
> > printf("Press any key to restore normal timer frequency.\n");
> > getchar(); //just pause
> > return 0;
> > }
> >
> > Thank you for your time!
> >
> > Best Regards
> > Albert Platek
> >
> > --- In mailto:DynoMotion%40yahoogroups.com, Tom Kerekes <tk@> wrote:
> > >
> > > Hi Albert,
> > >  
> > > Update I think I have a full understanding of the bug.  See description here
> > >  
> > > Hoping to have a fix soon.
> > >  
> > > Thanks
> > > TK
> > >
> > > From: Tom Kerekes <tk@>
> > > To: "mailto:DynoMotion%40yahoogroups.com" <mailto:DynoMotion%40yahoogroups.com>
> > > Sent: Monday, June 18, 2012 2:42 PM
> > > Subject: Re: [DynoMotion] Re: Trajectory Planner Bug 4 (uncommon cases)
> > >
> > >
> > > Hi Albert,
> > >  
> > > Absolutely.  Please report all bugs.  You are providing a great service.
> > >  
> > > I'm thinking it may have to do with that one GCode Block which is a segment only ~0.3um long!  But that should not be a problem as it should be removed by the collinear tolerance.
> > >  
> > > Regards
> > > TK
> > >
> > > From: albertplatek <albertplatek86@>
> > > To: mailto:DynoMotion%40yahoogroups.com
> > > Sent: Monday, June 18, 2012 1:28 PM
> > > Subject: [DynoMotion] Re: Trajectory Planner Bug 4 (uncommon cases)
> > >
> > >
> > >  
> > > Hi Tom,
> > >
> > > Thank you for replay, but dont hurry with this it happen really rare.
> > > (i mean that below 30degrees, or maybe low pass smoothing at 30deg
> > > attenuate high peaks of accel and I ddint noticed it before)
> > >
> > > If you dont mind I will report more and more new bugs here in future and I will try isolate them as I can.
> > > I really like KFLOP, and I think that reporting it is always good idea.
> > >
> > > Thanks :)
> > >
> > > Albert Platek
> > >
> > > --- In mailto:DynoMotion%40yahoogroups.com, Tom Kerekes <tk@> wrote:
> > > >
> > > > Hi Albert,
> > > > ÃÆ'‚ 
> > > > BTW I'm able to re-produce the problem exactly as you describe, but haven't been able to understand it yet.
> > > > ÃÆ'‚ 
> > > > Regards
> > > > TK
> > > >
> > > > From: albertplatek <albertplatek86@>
> > > > To: mailto:DynoMotion%40yahoogroups.com
> > > > Sent: Thursday, June 14, 2012 3:09 PM
> > > > Subject: [DynoMotion] Re: Trajectory Planner Bug 4 (uncommon cases)
> > > >
> > > >
> > > > ÃÆ'‚ 
> > > > Hi Tom,
> > > >
> > > > Yes it is repeatable exactly at
> > > > 40deg BreakAngle
> > > > (value a bit lower 39 fix problem and little higher 41 also fix problem)
> > > > Collinear Tolerance 0.001
> > > >
> > > > It seem that exactly at 40deg motion is not blended and pass without
> > > > slowing down.
> > > > Yes 40deg is high BreakAngle but...Can simillar situation happen
> > > > at lower breakangles?
> > > >
> > > > Please check it in free time, it is rare bug, I noticed it after
> > > > many days of work and on big job... hopefully Tau 0.008 helped
> > > > and motors didnt stall but sound was still noticable.
> > > > Without low pass smoothing you can capture crazy high acceleration peak.
> > > >
> > > > Best Regards
> > > > Albert Platek
> > > >
> > > > --- In mailto:DynoMotion%40yahoogroups.com, Tom Kerekes <tk@> wrote:
> > > > >
> > > > > Hi Albert,
> > > > > ÃÆ'Æ'‚ÃÆ'‚ 
> > > > > Sorry I haven't yet had time to try to duplicate it or track it down.
> > > > > ÃÆ'Æ'‚ÃÆ'‚ 
> > > > > Is the problem severe, obvious, and repeatable?
> > > > > ÃÆ'Æ'‚ÃÆ'‚ 
> > > > > Regards
> > > > > TK
> > > > > ÃÆ'Æ'‚ÃÆ'‚ 
> > > > >
> > > > > From: albertplatek <albertplatek86@>
> > > > > To: mailto:DynoMotion%40yahoogroups.com
> > > > > Sent: Thursday, June 14, 2012 1:25 PM
> > > > > Subject: [DynoMotion] Re: Trajectory Planner Bug 4 (uncommon cases)
> > > > >
> > > > >
> > > > > ÃÆ'Æ'‚ÃÆ'‚ 
> > > > > Hi Tom,
> > > > >
> > > > > Can you duplicate this error?
> > > > > Can this situation occur below BreakAngle 30 ?
> > > > >
> > > > > Best Regards
> > > > > Albert Platek
> > > > >
> > > > > --- In mailto:DynoMotion%40yahoogroups.com, Albert PÃÆ'Æ'‚ÃÆ'‚³atek <albertplatek86@> wrote:
> > > > > >
> > > > > > Hi Tom,
> > > > > >
> > > > > > I think that this problem is not critical, but I decided to report this
> > > > > > small error.
> > > > > > (high peak of acceleration)
> > > > > >
> > > > > > Error at BreakAngle 40 and CollinearTolerance 0.001in
> > > > > >
> > > > > > BreakAngle value 39 or 41 fix the problem.
> > > > > > Or smaller CollinearTolerance 0.0008in also fix problem.
> > > > > >
> > > > > > G90 G21 G49 F5000
> > > > > > N1819760 G01 X108.0000 Y23.8750 Z-28.0016
> > > > > > N1819761 G01 X99.1716 Y23.8750 Z-28.0016
> > > > > > N1819765 G01 X98.6688 Y23.8750 Z-27.5766
> > > > > > N1819768 G01 X98.0547 Y23.8750 Z-27.2674
> > > > > > N1819769 G01 X98.0543 Y23.8750 Z-27.2672
> > > > > > N1819770 G01 X97.8112 Y23.8750 Z-27.1428
> > > > > > N1819771 G01 X97.7085 Y23.8750 Z-27.0942
> > > > > > N1819772 G01 X97.4058 Y23.8750 Z-26.9568
> > > > > > N1819773 G01 X90.0000 Y23.8750 Z-26.9568
> > > > > > M2
> > > > > >
> > > > > > Best Regards
> > > > > > Albert Platek
> > > > > >
> > > > >
> > > >
> > >
> >
>