Messages

1

General Mach Discussion / Re: The dreaded random stall - servo mill running Mach2 with Geckos

« on: June 10, 2006, 12:01:52 PM »

Thanks Brian!

I am running Geckos.  I presume the daming you are talking about is one of those 3 little plastic set screws in the Gecko drive?  I will investigate how to damp and see how that works. 

In terms of vibration/frequency - I did pretzel one of my X-axis screws which has never been quite the same.  This does produce some nasty vibration at certain feeds, depending where along the axis the gantry is.  It is on the list of things to retro fit, but perhaps this is contributing.

- did the car analagy make any sense is is that a red herring?

Thanks again
AJ

2

General Mach Discussion / The dreaded random stall - servo mill running Mach2 with Geckos

« on: June 09, 2006, 01:55:22 AM »

Please forgive the length of this post.... I've learned a lot on the forums these past few nights... - but want to know if what I think is the solution makes any sense to those out there with much more comprehensive understanding of all the variables involved.

Well its been a long couple of nights troubleshooting everthing under the sun.  I built my first CNC 3 yrs ago - 3 axis, slaved X using 600 oz/in servos, 200 count US encoders, Gecko Drives, and Mach 2 ( 4'x'8'x18" ).  It has worked wonderfully well until 48 hours ago.... the first time its been used in its new shop.

A bit of history... I moved house this spring and since this move I haven't had reliable performance on programs that ran perfectly before -albeit some still run fine.  I managed to move the machine without having to rewire a thing, so short of mechanical damage (binding for alignment in its new situ) its as if it never moved.  Subsequently I have tested all my axi and they all move easily and require no where near the torque the servos can put out.

To make matters worse, the 1.3 Mhz laptop I've been using up to this point went on the fritz and subsequently was wiped clean and reformatted with Win XP Pro, MS Office, and Norton Antivirusbut nothing else.  I reinstalled Mach2 because its what I know and luckily salvaged my xlm setup file so setup was again identical to the good old days.

The program I'm troubleshooting against cuts a 3D half ellipse (concave shape) by contouring down through Z Steps.  It takes about 3 hours to run but I am finding my Y axis stalls 10 minutes or so in, running at 75 ipm.  The fault LED lights up on the Y as you hear the machine 'hiccup' - sometimes the X axi drivers fault in quick succession... but never first.  (I have the drivers set up to automatically reset a fault).  It is seemingly random but always occurs within moments (< 1 sec) of where the X axis changes direction, coinciding with Y axis maximum velocity (ie. the sharpest curve in the ellipse).  The ellipse is made up of straight line moves. 


So here's all the things I've tried so far

1) 5 us pulse width, 5 us preload
2) Follow Art's Optimization Guide and stripping all background activities out of my laptop
3) enhanced pulsing turned on
4) running at 45 / 35 / 25 Mhz (no difference) Diagnostics show 47Mhz consistently.
5) CV option on (versus exact stop); and I checked the No angular discrimination box
6) continually halfing my acceleration/velocity settings ( accel 4.0 down to 2.0)
7) Backlash has always been off.
ran the program with all axi free from load (removed couplings to gantry, etc etc)... still stalls

all of these above had seemingly no effect.

9) running the program at reduced feedrates of 50% (diminshed random stalls, but did not eliminate)
10) running in the dark with the fluorescent lights off (I heard these cause noise??)... yes I am desparate and am looking for anything at this point.  And at this point I could almost convince myself to try anything to make it work.

So at the verge of giving up I tried one more thing... increasing the acceleration.  I moved it way up (24) and the machine responded fine to joggin albeit I was afraid my mechanical craftsmanship would be the next thing I'd have to fix.  I then backed it off to 15  (originally 4, as low as 2 during my tests) and voila... the program ran perfectly twice.

And now after my long winded history... a question for the wise -  Does this make any sense?  I have convinced myself that perhaps it does.  If the geckos fault at 128 count lead/lag... wouldn't a system with faster acceleration help a lagging/leading access get back in sync with the other axi?  Something akin to a pack of racing cars trying to stay in formation... the more responsive the car's performance, the quicker it could get back with the pack should it fall ahead/behind, and more importantly the more quickly it could accomplish this, the less likelihood that it would fall completely away (ie. a 128 count fault)

I'm hoping I wake up tomorrow and find this apparent solution is permanent.  If folks out there think otherwise and have any other thoughts.. please let me know.

Thanks

AJ