Buffer Underflow Problems

[Hullhorst]

Hi Tom,

If G code is 'heavy small' , on certain parts of models in produce there is no use if I increase the LookForwardTime. System go to Underflow . [ i.e. from 2.5 to 5 or even 10 sec ].
I am getting fast in a position to decrease the overall speed to 1500 mm/min even the CNC play well up to 5800 mm/min [ 55 inch/min respectively 228 inch/min ] . Based on USB 2.0, there is any setting in Windows [ services, priority so on ] or C [ timing factors, canceling process so on ] to increase the data transfer ? [ KMotion interface work without Graphic Plot and main processor run decent on 2.8 GHz ] - all task in background where stopped, Network interface is disabled, mouse is wired, kbd is wired - just in case you may ask -> all but Kmotion is stopped in Windows services / USB cable is 1m long, screened and protected with Ferrite rings on both ends].

I am not in a crisis , I do operate same manner since first Kflop board [ and this one is the third with his third CNC ] , I am willing to cooperate to test even odd solutions.
I type now because last clients demands went a zone where G code will be even double or triple as size. Somehow I want to keep my timing under control.
[ to have a example, I may add a sample from a program that is now on run and a sample from one I want to run , accuracy go to 0.003 '' ( 0.08 mm ) ]

Sorry to bother, Thank you,
Vio

[TomKerekes]

Hi Vio,

A couple of things to check:

Too Small of a Facet Angle in the Trajectory Planner Settings can cause excessive segments to be sent and buffered. What is your setting? Try setting to 2 degrees or larger.

You might check your USB rate for your Computer. There is a USB Test in the SimpleFormsCS.exe example. Push the Test USB. My (fairly fast i7) computer shows:

PC->KFLOP N=100000 Int32, Time=2.276 sec, 390KBytes/sec
KFLOP->PC N=100000 Int32, Time=2.047 sec, 433KBytes/sec

If you post some GCode and your settings we can look into it further.

Regards

[Hullhorst]

Hi Tom,

Your advice is appreciated.
I check-it-up by small steps, until my models surface was smooth applying effect from : (G code generator + Facet Angle ), therefore I arrived to 0.3 deg.
Below that value was small or no significant change in surface [ I check all surfaces with magnifying glasses and I check again after manual finish is apply ].
Today I run the model with 2 deg Facet angle. ( Beginning of the program was Facet Angle : 0.3, speed stack somewhere 435 mm/min ]
To be done :
How do I send a sample from my program ?
To be checked :
USB Test.

Sincerely,

Vio

[TomKerekes]

Hi Vio,

To attach files use the "Attachments" Tab below the Reply box. Then select "Add Files".

Also include what Version you are running and what type of system this is?

Regards

[Hullhorst]

Hello Tom,

Sorry -> I was expecting a large .nc file [ 4Mb ] to be refused as attach. This is why I asked for an uploaded method.
My usual .nc files stay from 2 to 16 Mb.
As I say before, I operate the program - after 75% of his volume, on 2 deg adjustment.
Compared surfaces - identical surfaces both as model, angle and direction of mill travel - show differences in accuracy.
Speed, however , in the '2 deg' adjustment was possible up to 1250 mm/min [ rough 3 times faster ].

To be done :

speed test on USB.
I will answer on questions and give several details on next post.


Thank you,

Vio

[TomKerekes]

Something else you might try is Coordinated Motion Smoothing. What is yours set at in your Init C Program? You might experiment with different values? Larger values should provide more smoothness but slight corner rounding. The following code will set the smoothing Tau value to 1 millisecond:

Code: Select all

#include "KMotionDef.h"
main()
{
	double Tau = 0.001; // seconds for Low Pass Filter Time Constant
	KLP = exp(-TIMEBASE/Tau);
	printf("Tau=%f KLP=%f\n",Tau,KLP);
}

Here is an example plot of motion through a ~25mm radius, at ~8000mm/min, 0.5 degree facet angles, with and without smoothing:



This is described here.

[Hullhorst]

Hi Tom,

For practical reasons I must take all information and check-them step-by-step.
USB test : File mentioned -> SimpleformsCS.exe start.
Error on run.
Error signature :

EventType : clr20r3 P1 : simpleformscs.exe P2 : 1.0.0.0
P3 : 534a05e2 P4 : kmotion_dotnet P5 : 1.0.1.0 P6 : 531ca6f7
P7 : 3b P8 : e1 P9 : kmotion_dotnet.dmexception
Test done with Error.
System details.

PC : Dell Optiplex Proc ; Pentium E5800 RAM : 2Gb , Video : Intel G41
OS : Win XP SP2 - licensed
Kmotion 4.32 -> for some reasons 4.33 work bad with sudden stops and I took a step back ( is not your fault, I just did not took time to fix )
Trajectory adjustment :
Break angle : 85, Corner 5 E-7 , Colinear 5 E-7 , Facet : 1.5 ->

Next reply will contain more data upon the tests.

Thank you for your deep sharing,

Vio

[TomKerekes]

Hi Vio,

Please use our latest Test Version if possible. Links are here.

USB test : File mentioned -> SimpleformsCS.exe start.
Error on run.

The SimpleFormsCS.exe uses .NET 4.5 and will not run on Windows XP.

Break angle : 85, Corner 5 E-7 , Colinear 5 E-7 , Facet : 1.5 ->

Those tolerances are very small (13nm!). Please use something more reasonable like 5e-4.

Regards

[Hullhorst]

Hello Tom,

I made adjustment upon Colinear and Corner Tolerance as suggested.

Tau factor, however, is not implemented in my system.
Regarding link and descriptions : I feel out-dated

Next days I will focus on my main revenue paid-job->automation.
Therefore I will refresh the info after next tests .

Please consider : maybe some users can accept online debug/filters adjustments/calibrations from your side, as paid services.
A TeamViewer session is easy to implement and 50 $ or more/hour with 50% payed in advance will keep people more focused on CNC-s job.
At least C problems or Step response will be no more an unsolved issue. [ I think ]

Thank you,

Vio

[Hullhorst]

Hello Tom,

Lat week, 4 days , I had an extensive test with 3 different 3D models from 150 sq inch to 5 sq inch surface.
All test performed with 3500-4000 mm/min and various Colinear and Corner Tolerance not bigger than 2-10e-3 and not smaller than 5- 10e-4.
So far USB comm went smooth , therefore I presume I will keep adjustments in this area. ( mills from 4 to 0.5 mm diam )
Regarding implementation of Tau, was not done yet. I keep in mind that is will give even more play-ground for the USB speed and smoother corners.

I had another error trying to use test speed but I don't have in handy the error file, maybe I will explain that later on.
( like explained below, I don't feel Panic now )

There is a lot to learn in order to achieve a deep nice use of this product and I am glad this way. I will go forward and implement some planning for this learn.
Thank you for your previous detailed answers.
I am considering : Emergency of this subject is level 1/5 for me.

Thank you again, best wishes Tom.

Vio