Machsupport Forum
Mach Discussion => General Mach Discussion => Topic started by: VWDan on May 25, 2009, 04:46:41 PM
-
Hello Everyone,
I'm new here and could use some help. I've been working on an X2 CNC conversion for some time now. I have no prior CNC experience. I have some machining experience so I'm comfortable using a manual mill or lathe. I'm just starting to learn Mach3 and g-code.
My X2 CNC conversion is almost operational however, I discovered a problem while testing this afternoon. While checking position repeatability I discovered the X axis is traveling .006" short of what Mach3 is reporting on the DRO.
My test setup is simple. I've clamped blocks a fixed distance apart, 2.000". The DTI has a 0.099" diameter ball end. The indicator is set to read 0" at the starting block. It travels 0.001" after contact to get to a 0" reading. I wrote some simple g-code to move the X axis from 0.000 at the starting block to the end block (2.000" away). By changing the distance little by little until the DTI reads 0.000" (at the end block) I found that distance needed to be 1.906". I put a dwell in so I can read the DTI at the end block before it returns the table to the original position. The repeatability is fine. It returns to within 0.0005" on both ends repeatedly.
Here's my breakdown.
Start = 0"
Reported distance traveled = 1.906"
+ DTI travel to 0" = 0.001"
+ indicator ball dia. 0.099"
Actual Distance traveled = 2.006"
Over travel of 0.006" per the Mach3 DRO
I've also tried the above test with blocks set 3.000" apart with the same 0.006" over travel reported by the Mach3 DRO.
By using my Shuttle Express to jog at 0.001" per click I found that the DTI doesn't move til about 0.006" of reported travel when changing directions.
So, would this be the backlash on the X axis?
Thanks for the help,
Dan
-
Yes, certainly seems like backlash. Could be a loose pulley or even a sloppy bracket or could be the ballscrew.
Hood
-
Looks like it's backlash. I'm surprised to see that much backlash in a ball screw. I've got the lower cost Nook ball screws. I've checked the table mount and the ball nut is screwed into a bracket. I guess I'll have to tear into the table and see if the bracket is secure.
-
Looks like it's backlash. I'm surprised to see that much backlash in a ball screw. I've got the lower cost Nook ball screws. I've checked the table mount and the ball nut is screwed into a bracket. I guess I'll have to tear into the table and see if the bracket is secure.
0.006" backlash in a non-pre-loaded single ballnut is not at all unusual. You can minimize, but not totally eliminate, this by loading the nut with oversize balls. To get to very low or zero backlash requires spring-loaded double nuts.
Regards,
Ray L.
-
Thanks for the input Ray. I'm going to take it apart and see if I missed something. If not, I guess I'm going to figure out how Mach3 handles backlash.
Thanks,
Dan
-
are you using ball screws? I didnt see them mentioned. You also need to make sure the thrust bearings are preloaded slightly to eliminate any play there. I have done a number of x2 conversions and most of the factory screws when new run around .002 backlash
-
Yes, I've installed Nook ball screws. I disassembled the table and found the X axis ball nut mount set screw wasn't tightened down. I'm still working on getting it properly positioned. My first try cut the backlash down to 0.003" but I still think it can be improved. I'm going to have to look into oversized balls. I can tell there's some slop in the nut.
Thanks,
Dan
-
Dan - I admire your tenacity - but rest assured, if all else fails (and it will - you must have SOME backlash in the system otherwise it would seize up) Mach copes with backlash compensation admirably. I recently changed my lathe to ball screws, and was pleased with the results, but I still use the Mach backlash compensation.
If you think about your manual work, you always automatically compensated for backlash by always coming up to a cut in the same direction. CNC does not do this, it cuts in any direction, hence the need for backlash compensation. I must admit I still, if writing my own GCode, tend to run past the job on the return, and still always approach the cut from the same direction. I have no problems and my work is accurate and repeatable.
-
Thanks for the note. Once I'm satisfied with the mechanical setup of each axis I plan to set the backlash compensation for each in Mach. I've played with it a little but haven't really figured out how it works yet (noob curiosity).
Dan
-
For a milling machine, I found Mach's backlash compensation unuseable in its current implimentation for anything but drilling holes of other non-motion type of operations becuase it literally stops and takes up the backlash at a set rate at each direction change and then begins moving again at the commanded feed rate.
-
Dan,
Like Jimpinder says you will have some backlash, just a matter of degree, and like pan..m said you need to
preload the bearing. Somtimes though, as you refine to eliminate, you may get to a point where
as a system all the things come into play, a little here and there and trying get it to "zero" is impossible or not worth the effort or even impractical.
I don't agree with Simpson on the usability, but that's my opinion. I also still put approach moves in or do prep moves
as Jimpinder mentioned even if backlash is used. Practicaly speaking, the system is only as good as the sum of all it's components. Practicaly speaking, if the backlash is small, the tolerance issue goes away for a lot of work.
Apply 50-50-50 rule for starting backlash settings. 50% of max axis velocity - 50% for backlash speed % - 0.050 for Shuttle Accel setting, and then tweak for your system.
RICH
-
Simpson - Yes, but how else could it take up backlash. It senses the change of direction,stops and applies the backlash compensation, then sets off from the correct start point. The cutting tool does not move during the backlash application - it cannot do, it is not moving the cutter, it is taking up the slack in the drive train.
-
jimpinder,
It is not the concept, it is the implimentaion in MACH that is troublesome.
When I tried it, the compensation event was not incorporated into the CV stream nor did it obey the commanded feed rate. The feature behaves as if is not integral to the pre-calculated tool path, but rather an add-on which is called like a subroutine triggered by a change of direction.
To use a example, say you are cutting aluminum at 7,000RPM with 5/16" three flure end mill with a .006" per rev feed rate and CV is active and no coolant. With backlash on, every change of direction results in a stop, taking up the backlash at a crawl, then another stop and then the tool path resumes.
The reult is at best a bad finish and at worst a ruined end mill.
I recall there is some setting for the speed at which MACH does the comp move, but MACH comps at the same slow crawl no mater where that setting is. If you are machining at a slow crawl, then you might not notice the problem, but if are cutting at anything approaching a production rate, or are cutting any type of hardenable material, you really can't use the feature.
If the comp was well integrated, i.e. the move was predicted during the tool path generation and incorporated into the CV at the same feed rate as the rest of the tool path, then it would be useful for continuous motion machining. As I said, it is still usefull for drilling hole patterns and I'll add it may be useful to to those who are already machining soft materials at very low feeds.
-
Yes - I can understand that at the speeds you are talking about, Mach2 implementation of backlash could be troublesome. I can see that,however, the axis must come to a complete stop before backlash can be applied, since, as I said , whether the steps are backlash steps or movement steps, the first shall we say 600 of these (1/1000th of an inch on my old lathe) will not produce any movement on the axis, they merely take up the backlash.
There is a speed box on the backlash page, and you could only shove this up to 100%, but yes, on reflection, I think the backlash pulses will be applied seperately. It may not be the backlash speed that is the problem.At 7000 rpm and a 0.006 per rev rate, your feedrate is 42 inches per minute. I don't know how many pulses per inch your motors are tuned to , so I can't calculate the pulse rate, but it might be that your setting of feed per rev (whilst it gives a good cutting speed) may be giving you the slow backlash rate, since, even set at 100%, it will still keep to the speed at which it is set (i.e. it doesn't revert to your maximum G0 rate).
I don't know what you are cutting, or what the reason is for choosing a feed rate per rev, but perhaps if you used feed rate per min, and backlash at 100% the results might be more acceptable.
-
jimpinder,
First, let me compliment you on your ability to have an objective discussion about a Mach shortcoming. This is how software gets improved.
I have only a little toy mini mill to play with and it is not very rigid so I need to change the spindle speed occationally if it threatens to chatter, and I want the feed rate to follow the spindle speed, hence feed per rev. Also, milling toolbits have a certain optimal chip load per tooth for various materials and it's just a lot easier to calculate the feed per rev without considering the spindle speed which can vary greatly depending on a lot of factors.
Lathes are different animals and are better suited to a travel speed type of feed fate.
I disagree that the machine 'must' stop to take up the slack. I see no reason that the comp move cannot be incorporated into the CV stream, but the velocity set to G0 speed for that 'move'. It appears now that Mach behaves as if it is in exact stop mode when the comp takes place. It has been a while since I investigated backlash comp, but my recollection is that there was a definate stop, if not a pause, before and after the comp move. Now granted the pause may be only an impression cause by the significant delta in feedrate between the cut and the comp move, but it definately behaves as if there is a stop in there.
If, in fact, the feature is triggered and a sub of some kind is called, then at least that processing time could be eliminated by incorporating the comp move into the CV stream when the toolpath is calculated by MACH.
My solution was to install oversize balls in the leadscrew nut for near zero backlash. Mo better.
-
I will agree that behaps the backlash could be seamlessly integrated into the move, I don't know how it "senses" the changes that may be needed, nor do I understand if the number of "pulses" for a given line are calculated all at once, or whatever, to allow the requisite "backlash pulses" to be tacked on.
However it is done, if you have backlash, the axis will appear to stop moving, until the necessary number of pulses to reverse the gear train have gone through the motor.
The only other thing that comes to mind is that, where two axis are moving together, either diagonally or in an arc, one axis (usually the one with less distance to move) takes it's pulses at a slower rate to keep the line (or arc) true. In this case, backlash MUST be applied first otherwise the movement of one axis relevant to the other(s) will not be accurate .
My machine is a combined lathe and mill, a medium sized Warco, and although new(ish) it suffered badly from backlash. I think I have that contained now, and I can turn out accurate stull. The next problem is 140 steel railway wheels to re-turn to a better profile (My CNC man did them wrong)
-
Simpson36,
"I disagree that the machine 'must' stop to take up the slack. "
Hmmm........
There is a mechanical side to all this.
If an axis didn't decelerate or stop before changing direction ...what would happen?
Ever put your car in reverse at 5 mph? How about 50 mph?
Equal and opposite reactions / forces on the drive train...... not good!
Can't change the laws of physics.
RICH
-
Rich, your 'direction change' analogy and physics reference do nothing to explain Machs real world behavior.
It would be useful if had an explanation for what Mach is actually doing:
If I run with CV on and make a circle, the machine runs smoothly past the direction changes . . . . therefore . . . obviously long pauses at direction changes are not intrinsic and unavoidable.
If I use backlash comp, even at tiny numbers like .002 there is a definite pause at each quadrant.
Cutting at 40ipm, the .002 movement should take .003 seconds, i.e. imperceptible. That actual pause is many, many times that and is very noticeable . . . . thefore . . . . obviously Mach is occupied doing something that takes far longer than .003 seconds while the machine sits waiting for that operation to complete.
In addition, CV even without backlash comp, does not seem to be able to reach thru A axis movements, but that's another discussion.
I'm not saying Mach is bad, but it does what it does and facts are facts. Mach has some room to improve in some of its functions.
-
"It would be useful if had an explanation for what Mach is actually doing"
Better than an expanation, is to run things in slow and fast moves along with changes to the
"50-50-50 rule for starting backlash settings. 50% of max axis velocity - 50% for backlash speed % - 0.050 for Shuttle Accel setting". That will give you the explaination on how it behaves depending on what is influencing it. I fooled with backlash for hours. Nothing i can explain will give that experience. I don't know the "inner" workings of Mach but rather relate to how it behaves. And it does what it does and anything can be improved. You may want to post a question on the thread on Mach ver 4 and see if they are planning on doing anything differently with backlash and provide specific suggestions there.
BTW - Some controlling programs provide option of implementing the backlash over a quadrant of the arc travel or the backlash at the typical 0-90-180-270 circle point.
RICH
-
You may want to post a question on the thread on Mach ver 4 and see if they are planning on doing anything differently with backlash and provide specific suggestions there.
RICH
Brian will be looking at the backlash code once Turn and Plasma have been done and that is after mill has been done ;)
Hood
-
I fooled with backlash for hours. Nothing i can explain will give that experience.
I hear you. I only played with it long enough to determine that although it has a comp speed setting, that setting is ignored. Actually I'm not too keen on the whole concept of backlash compensation, so it was never a big deal for me anyway, I just commented on what I found when I played with it.
Frankly, I'm more and more impressed with Mach the more I dig into the macro capability, and I haven't even touched on the brain facility yet. Mach looks to me like a WIP and it doesn't cost $5,000 so in my opinion, it is an impressive piece of work all things considered.
I was not impressed with the single index threading as I could not see how it would ever work and I said so on the forum and quickly found that pointing out Machs shortcomings is not popular around here ;) However, now there is quite an effort going on to improve threading with many counts per rev, so I was obviously not alone in my assessment. The CV has some issues, the tool comp is a bit bizarre and as we discussed, the backlash comp is out to lunch basically.
In the plus column is that these issues (at least those important to me) are being worked over for the new version of Mach which is very exciting.
-
Brian will be looking at the backlash code once Turn and Plasma have been done and that is after mill has been done ;)
Hood
Sounds reasonable to me . . . compensating for backlash in software, in my opinion anyway, should not be a high priority item, considering there are much better ways to eat that particular elephant.