Machsupport Forum
Mach Discussion => General Mach Discussion => Topic started by: mick on August 06, 2007, 12:43:36 PM
-
hello to all,
i have a small 4 axis mill x@y recentley fitted with 16mm *5mm ball screws and angular contact bearings ,driven by 320oz steppers through a cnc4pc g11 bob and arcuro drivers,after cycling each axis 50 times 0 to 127 to 0 the error is about .150" dti.
no combination of acc @ vel seems to improve on this, gen/config/step pulse(5),dir (5).
drives have been swopped with other motors,tried sherline mode but the problem is still there.
thankfull for any suggestions, mick.
-
I am trying to envisage what a "small" mill is that needs 320 oz motor to drive it. You do not say what your maximum travel is, therefore I cannot work out the accuracy. There would appear to be no benefit in moving the axis 50 times, because that confuses the issue.
If it is a problem with backlash, depending on where you start, (say the centre) then loving left, moving right and back to centre should get rid of backlash and be accurate. Doing it 50 times would make no difference. So you can eliminate backlash (and do not need to compensate for it) by doing this. DO NOT start at one end and go up and down, because you only get one set of backlash.
You seem to be saying that you inaccuracy is 0.150 inches which seems to be a fair bit if you have ball screws etc. but test for back lash - move in one direction then the other an equal distance and see if you have any back lash. Only do it once.
If you have rid of back lash then the other problem is you travel is not correct. Test your travel (in one direction only) and see if your table moves say 1 inch at a time, using your digital calipers. You may have to adjust your number of pulses per inch on your motor configuration to get this accurate, particularly if you axis thread is metric and you are working in inches or vica versa. Having said that, if it is wrong in both directions, it cancels itself out when you do it twice, let alone fifty times.
There really is not other reason I can think of, other than then, you are missing steps driving your stepper motors, but if you are saying this 0.150 is common to all axis, it is difficult to envisage each axis missing a similar number of steps and all in one direction (because if they missed steps in both directions you could not see anything wrong.
I have a feeling it is backlash - but it seems rather a large figure to me.
-
JIMPINDER thanks for replying,
motors are 220ncm ? 1.8deg, drivers set 1600 steps per rev,320 per mm
the backlash is all but zero.
the reason i ran so many times is, after 10 cycles from 0-127-0 mm i measured a -.0011" error and wanted to know if it was
accumulating, to me it says that steps are lost in one half of the cycle,
is this correct?
-
Hi Mick,
if you reset your drivers to 400 steps/rev and retune mach3 to match, then run your tests again are the errors 4 times greater?
Graham.
-
thanks Graham,
will try that tomorrow .
mick.
-
hi,
is it 100% repeatable?
if your steps per inch were no where near correct it should still
zero on return.
if your not missing steps that is...
tony
-
good morning Graham,
have tested as you suggested,and yes the error is 4 times as big.
also ran driver test:
cpu 1597
pulse per sec 24.136
apic 4142
shortest 39.5
longest 46.6
mach ver 2.2
regards mick.
-
Tony,
thanks for your input,
mick.
-
It might be worth trying a simple, straight through BOB just to eliminate it as the problem. I had a CNC4PC C11 board that did not like my driver cards and had symptoms similar to yours. Ended up using the CNC4PC C11G card, which has worked perfectly. Let us know how you solve this one.
Ian
-
IAN,
thanks for the info, i have been thinking along those lines.
ordered a set of new chips for this board this morning , but by the sounds of things maybe its money wasted.
regards mick.
-
Hi Mick,
from the test you have done I would say that it is NOT a machine error more a driving error, try a different cable and/or breakout board.
Graham.
-
I do not like spending money until I find out where the problem lies.
I run 220CNm motors on my Warco Lathe. I have Stepmaster drivers. I do not have a breakout board, but run the drivers direct from an old printer cable about 18" long. I got the motors from ArcEuroTrade.
The motors were eight wire that I connected in series to make bi-polar motors. I run them from a 24 volt supply - from large batteries which ensures a steady voltage over a wide current range.
My only difference to your set-up is that the motors are geared down 3 to 1 to drive my axis and therefore I need 6400 steps per rev.
I have tested my set-up quite rigourously for accuracy - I am fortunate because I kept the old 1thou dials on when I converted it, so you can see that the lathe returns to where it started - or you can have a second check on distance travelled. The geared down drive was because I tried using smaller motors - which kept missing steps. With the 220 performance is fine and I upped the gearing to 3 to 1 from 5 to one.
With these settings I have found it impossible to move my axis faster than 4 inches per minute - but I am happy to trade accuracy for speed.
Without getting your ****** in a twist, I think you must be patient and keep trying to find where the loss is. Yes - from what you have said, it would seem that you are loosing steps, but in one direction only.
I cannot think that this will be in the middle of travel, but is probably at the start or finish.
Make up a table with different settings on it. Try a different set of motor speeds and do short - say 1 to 6 inches movement, and check them with digital calipers - in both directions. Start with slow movement and slow acceleration, and see if you were ever accurate.
Gradually increase the speed of travel, and finally the acceleration.
I left my pulse length at 5 - I tried putting in 10 but it didn't seem to take it.
This may sound a bit complicated - but with a little thought you can do it all in an evening.
I have to be careful with my set up - which works perfectly in the mid area of my lathe (which, or course, I used to use a lot) but at the extreme ends, I can loose steps, where the lathe (not the steppers) sticks a fraction.
There must be an answer.
Jim
-
Try changing the step active low state ... It is worth a try :)
-
GRAHAM,BRIAN ,
thanks for the suggestions ,will have a go.
regards mick.
-
thank you gents,
my problem is solved thanks to Ians suggested bob change.
after fitting the new board this morning, i ran x & y axies through 50 cycles each , after which the d.t.i read .0005 error.
happiness ! thanks mick.
-
Mick,
Pleased to hear you got it sorted.
Ian
-
I am having a problem loosing steps on one axis. When I run a 35000 line project I loose aprox. .025 inches on my Y axis, the X axis comes back exactly to 0. I switched the X with the Y and the problem switched axes too. I am using a gecko G540, xylotex 270 oz. in. motors, running 20 in per min. with the acceleration of 20 percent. I can run it at well over twice that speed and acceleration without stalling. The kernel speed is 25000 the step and direction pulse width is 2. This board has a charge pump that I believe is working. I noticed a post saying he changed the pulse width and fixed his problem so I changed mine to 5 on both and am running the project now; it takes 2.5 hours. If that does not work does any one have any ideas. It is 28000 pulses per inch. This project covers an area of .75 inch by 4 inches.
Thanks silveron.
-
Silveron,
Don't know how manny moves your making with the y axis, how good all the mechanical stuff is, etc.
How much backlash do you have and where is it coming from?
How linear are the screws?
Mechanical errors can add up over manny moves and result in what your seeing.
Just some thoughts,
RICH
-
Thanks Rich
I am only loosing position on one axis.
When I switch the axis X with Y the problem switches axes also. If it were a mechanical problem the error should stay with that axis.
I believe this would rule out noise between the driver and the motors being the problem also. I should mention that I had a TB6560
driver a few weeks ago the problem was similar but worse, much worse, it lost a little position on one axis and a lot in the other.
That is why I bought the Gecko G540.
I have not been counting the moves because after running this program multiple times which would amount to well over a hundred thousand lines and ten hours of running the good axis looses nothing but the other one looses aproximately .025 per 35000 line run. When I then switch the motor plugs X for Y the problem switches axes and the axis that was loosing steps now works perfectly even over multiple runs.
Thanks: silveron
-
That is a lot of moves! Is the error accumulative? or does it vary up and down. If accumulative, then it could be a rounding up (or down) error in the maths associated with the G code, but finding it would be time consuming. You would have to look for square roots and trig functions or something similar.
Ian
-
Have you got the active state of the Step pin set correctly? What have you got set for the Pulse Width?
Hood
-
Thanks for your thoughts Ian.
I have run the G code that is called Cross.tap that comes on the Mach 3 cd. I used it instead of the Roadrunner file because it is larger.
The problem is the same, and yes it is cumultive. That file is a little over 13000 lines after 7 runs the total amounted to approximately .030 inches. So far this seems to be between the driver and the G code. Does that sound correct? I am going to change parallel cable.
Thanks: silveron
-
The ideal would be to count PC pulses to the driver, driver pulses to the BOB, BOB pulses to the motor.
When compared they should be darn close. Not the kind of electronic device you will find even in pro shops.
The one my friend brought over once was custom made for his workplace and measured to a few counts / parts per million.
Doing that confirms the electronic end of things and leaves no doubt as to what the culprit is.
Downstream of the motor a calibrated device is used to compare the mechanical motion and that is compared to Mach's DRO and maybe even another independant reference such as an axis readout. With that you have comparison of electronic to mechanical motion.
The mechanical motion of each axis can be isolated and checked and related.
Now we meticulously did the above once...in fact it was done to my lathe before using it to find a fix for the lathe threading cycle.
In general the pulses the pc sent were way within the pulses the motors received. Different motors provided for different linear motion / accuracy. Belt tension could and would affect linear axis motion. Same for timing gears. All screws are different and forward and reverse motion of the screw may be different, especialy if they are rolled. There is always some lost motion due to thrust bearings, manny times
overlooked and just lumped in with all the rest of so called backlash, but it is real and just a component of the whole. Mach's DRO readout,
at least on my pc, was superb.
Sorry, more later, dinner is served......
RICH