Machsupport Forum
Mach Discussion => General Mach Discussion => Topic started by: wdp67 on December 11, 2007, 10:45:34 PM
-
I am having more trouble. It seems that from day to day things change on my mini mill. I cut a part one day, and if I cut the same part the next day and do not change any settings, it comes out a different size, what am I doing wrong?? I set the steps in the auto setup, using a dial indicator, but then when I run a program, seems to just fall apart. Motors are not making any funny noise like they are losing steps or anything, but I can not for the life of me get any kind of consistancy.
Thanks
Walt
-
I take it this is why you are doing the home switches !!!
When you say "it comes out a different size" what are we talking about here - thousanths of an inch, or a factor or two, or what.
When you say "if I cut the same part next day" - do you mean 24 hours later, or just cut two consecutive parts.
The number of steps per unit set for your motors do not make any difference - even if they were wrong, they would be wrong all the time and the parts would come out the same size (albeit the wrong size). When you say "they just fall apart" do you mean the figures in the motor tuning alter, or that the motors just seem to do their own thing.
You only need to set the steps per unit once - finish - not every time you start the machine.
I don't know whether you are using steppers, or motors with some sort of positional feedback.
If they are steppers, then they are accurate and (I would have thought) that the only problem you have is a mechanical one, where somehow, something is slipping, either the motor IS missing steps (in which case things would be consistantly smaller) or a belt or gear is slipping. If you motors have a positional feedback, then maybe this is faulty.
Just make sure your acceleration and speeds are not too high so that you do not loose steps on acceleration, or overun when stopping.
The only other thing is "backlash" - have you got this set up. If you haven't, then your results may be indeterminate, depending on which side of the backlash you happen to be when you start.
The main thing with a CNC machine is consistancy - yes, that is what you are after, but you must be consistant as well - in that on a run of parts, YOU must do exactly the same positioning moves, so the machine is ALWAYS in the same position when it starts.
The ideal, I suppose, is to run the machine to the homes, put the work in a jig (so it is in exactly the right position) and set off from there, which, I assume, is what you are thinking.
Stick with it !!!
-
There is a real chance that you are losing steps and not actually hearing it. Often times it can happen under a heavy cutting load. Other than that, your steps per unit & backlash settings could be off.
-
Thanks for the replies!! What I am saying is I did check my steps per unit, but everything checks out ok. In set step per unit I tell it to move a set amount,and it does everytime within a .001. I am happy with that, I think. Then I go to cut a part. Mde a fence that fits in one of my table slots, then I cut the fence square to the x travel. I clamp my part to the table, which in this case is a peice of 1/8" plexi glass making sure it is square against the fence. I am using a 1/8 end mill to cut the outside profile of the part. In my drawing I have offset the outside of the part by .0625 and used that to create my tool path. I cut out the peice using 2 passes. When it is all said and done the peice is generally about .010-015 bigger than it is supposed to be and that seems to vary from day to day, or I should say evening to evening. One day I cut a few out and they were all .010 bigger, the next night I cut some more and they were .015 bigger.
Hopefully I am explaining myself clearer.
Thanks again for the help!!
Walt
-
Walt, it definately sounds like you are losing a few steps. It can happen for a number of reasons. One common reason is if your rapids are set too high. While you won't miss steps at jog speeds or feed speeds, often on a G00 move as when the tool returns to the origin for the next cut, it's quite possible for you to miss steps. Two things to look at, your motor tuning for one, try backing down the velocity. Secondly, what is often overlooked is the Rapids Speed setting in your post processor. Don't know what CAD/CAM you are using but double check your post processor and back the rapids down.
I had a similar issue a long time ago, and I could replicate it by using a 2" travel dial indicator. I could G00 2" and it would be off 9 times out of 10. Drove me nuts as I was thinking it was backlash but it was missed steps.
If the problem was consistant, i.e. always a few thou short or over of the required move then it would be leadscrew error. That can be tweaked in the Units setting for the axis.
Likewise backlash can show up in a similar manner. Backlash is tweaked with a dial indicator and jog stepping a thou in +/- directions until you get a thou movement consistantly when you change directions.
-
Hi Walt
I started a thread called "is this a bug in Mach 3" as i was having a similar sort of problem. It's been deleted.
I was consistently gaining steps which it seems that you have the same problem.
I have now resolved the issue and it was down to one of three things.
1: I was using a BOB that had opto isolated outputs and my drivers had opto isolated inputs. IE two sets of opto isolators.
2: There was noise in the BOB board.
3: I hadn't wired the drivers correctly.
The problem is resolved due to a change of BOB that no longer has opto isolated outputs to the drivers
I also changed the wiring to the drivers the original wiring had a common ground the steps and direction were connected to the positive side of the drivers they are now wired as 5V common the negative now controls the steps/dir signals
Basically what was happening was on every direction change some extra steps were added.
It happened on every axis by the same amount.
HTH
Phil_H
-
Ok I took out all the rapid moves and made them all feed moves and the results stayed the same, still .010-.15 larger than supposed to be, I will trymessing with the aceleration I guess.
Any other ideas?
Thanks
Walt
-
Is the problem consistant or random? If you set up a DI to measure your distance moved as outlined in the manual, is it consistantly .010" off or does it vary? Is the error on all axis or only one.?
Can you describe the configuration of your system. What CAD/CAM are you using to generate your code?
-
I do not wish to complicate matters, but...........
How is it possible to LOOSE steps and have the part come out BIGGER ? (Rectangular pocket or island)
This doesn't make sense to me, but it doesn't take much. ::)
-
I do not wish to complicate matters, but...........
How is it possible to LOOSE steps and have the part come out BIGGER ? (Rectangular pocket or island)
This doesn't make sense to me, but it doesn't take much. ::)
Duh, you are right!
For some reason my brain was thinking 'smaller' as it's a more common situation. Larger would be either backlash setting being incorrectly set but more than likely, the steps per unit setting needs to be tweaked. Ball screws are not perfect and have error. Use a DI and a gage block to determine the travel distance and correct the steps per unit setting. I had to do that on all 3 axis of my machine to get it tweaked in.
For example, I have 200 step/rev motors and I use 1/8 step so 200 x 8 is 1600 steps per rev. I have .200" per rev ball screws so 5 x .2" =1" of travel 5 x 1600=8000 steps per 1" of travel or 1"/8000 steps = .000125" per step.
If you have your steps per unit set at 8000 and you move 3" in Mach and the measured distance is 3.015" it means you have .005" error per inch in your ball screw.
So we take 1.005" divide it by 8000 steps and we get.0001256" per step. Divide 1" by .0001256"/step and we get 7962 steps per 1" which would be our correct steps per unit setting.
-
One other thought,
If you were to CLIMB mill around the perimeter of the rectangle, backlash could allow the part to be bigger. ie: programmed dimension + backlash.
This isn't likely though with a small cutter in plastic.
-
Two things come to mind here, an 1/8" end mill is a tiny tool to ask to not flex and walk around in 1/8" thick material, and secondly, Plexiglass has horrible moisture absorbtion rates and will expand with just a small humidity change, you didn't say how big your parts are, on a mini mill can't be very big but still the stuff is trouble to keep a size with. It can also cause semi melted chips to build up on a cut surface after the mill has passed, causing size increases. Try a larger mill bit, and make sure the swarf is sucked out or something while it's cutting.
-
Walt - I have read all these suggestions and whilst I am not confused and they are all good suggestions, lets get down to the nitty gritty.
You have the machine in front of you - what is it doing wrong. Is it cutting too small or too large - and from that, does your axis appear to miss or gains steps. That is the first step.
Having said that, and I missed it first time round, plastic is very difficult to cut. As daft as it seems, the cutters must be very sharp, and the speeds must be properly controlled. (I only say this because my local plastic supplier who uses CNC said so).
I think I would try the cut manually first, say setting the cutter at exactly the right position with the spindle running, and then jogging the cutter up the cut, using a low percentage of the jog speed to do it with. Then measure the result and see if it is accurate.
If you CAN cut it accurately, then you can progress and see why the program won't do it.
JIm
-
I am at work right now so cant play with it, but it seems like if I tell the machine to move a set amount in the mdi line it moves that amount, but once I run a parts program it seems to cut larger than what it is supposed to. It does not matter if I am cutting plastic, metal or wood. All the results are the same, it is anywhere from .010-015 larger.
Walt
-
Hi Walt,
Are you using lead screws or ball screws ? What is the backlash in the area where you are cutting the parts
Did you check the travel with your indicator in the same area of the screw that you are machining the parts with ?
Are you using backlash compensation ? If you are, could you possibly be OVER compensating ? This would make the parts oversized.
I think your problem may be mechanical more so than electrical.
Wish You Good Luck
-
I am using the stock lead screws. I have tried with both backlash enabled and backlash off, I honestly can not remember the results of the backlash off. When I get home tonight I will try again with backlash off and post the results here tonight. Thanks for all your guys help I really appreciate it. Hopefully I will get it resolved. The parts that I am making are not critical so it really doesnt matter, but I want to get it fixed so I can cut accuratly!
Walt
-
Hey all
I played around with the machine tonight a little bit and may have found some or all of the problem I think. When the y axis switches direction, watching a dial indicator, before it actually starts to go the other direction it moves .005 further in the previous direction. Does that make sense? It even does that if I do it manually. I can see where my parts would be .010 bigger in the y direction. I am not sure what to do to try to fix this though. It is a cnc'd harbor freight mini mill. Any ideas?
Walt
-
just a thought but what's the (theoretical) resolution of your machine in:
steps?
and
microsteps?
-
I am not sure I can tell you, It is a harbor freight mini mill using a xylotex control box I think that is shipped 1/8 step and I have not changed it. I am using the stock lead screws which move the table .0625 per turn.
-
Walt - Mach3 cannot do this. If you reverse direction, then Mach3 reverses direction immediately. What sort of motors are on the mill, are they servos with some sort of reaout to locate them. They certainly can't be steppers doing such an odd move. If they were steppers, they might not reverse straight away, but your DRO's have no feedback from steppers and would count down straight away.
If this is the case then Mach 3 is putting out the pulses, and the reverse sugnal, but your machine is not immediately picking up the reverse pulse - there is some delay in interprting it - as you see about 5 thous worth. I cannot think what to do. Have you a breakout board that is delaying these pulses (although if it delayed one, it sould delay them all). I don't see how you can measure the timing of the pulses without some sophisticated measurement devices.
The only thing I can suggest is slow down your traverse speeds and see if this is reflected in a lower than 5 thou readout when you change direction. It is certainly odd - anybody else have a thought.
Jim.
-
I agree, I am sure that it is a mechanical problem, because like I said it does it when I switch directions manualy also. I just can not fugure out what to do about it.
Thanks
Walt
-
If it is a mechanical problem then it would not show in the DRO's.
If you tell Mach to move 1.000" then the DRO will show 1.000", the only time this will be different is if you are using compensation of some kind.
Backlash compensation is applied on the start of the move and is not shown on the DRO's
Even on a servo system Mach has no idea if the mechanics are moving. It just counts pulses out.
What is in your start up line on the config page?
Graham.
-
I never said it showed up in the DRO, I said the parts actually measure .010-.015 bigger than they are supposed to. At least thats what I thought I was talking about. :)
Walt
-
If I understand this correctly you move the table in say the Y+ direction and the DTI shows 1.000" you then stop and reverse the direction to move the table in the Y- direction and before it moves in the Y- direction the DTI shows 1.005" before reversing?
This happens on both stepper driven and manual driven lead screw?
This would indicate slop in the table slides to me, you'll never get any sort of consistant result if there is movement in the slides.
If it isn't the slides then I'm at a loss, I can imagine turning the lead screw manually with the handle could move the table a little but the stepper is attached to the table so it shouldn't be able to.
Steve
-
Yea that is right. My mill is cnc'd as per fignoggle plans so I still have the manual capability. How would I fix the slop in the slide?
Thanks
Walt
-
I have to add, you are asking a $400 Harbor Freight mini mill, with the original Chinese leadscrews, NOT ballscrews, to be accurate within .005". That machine doesen't even have tapered gibs, it has setscrews that you load the parallel gib with, very flexible setup as the table loads go through these tiny setscrews. Maybe you need to map the leadscrews with Mach's mapping feature. Or spend the money and buy ballscrews for it.
DaveA.
-
I have a question on Mach3. I don't know if it was ever addressed here or not, but let me try and explain my situation.
I have being running my CNC router for more than 2 years now.
I bought a dial gauge the other day and last night I put it on to check the setup.
I was very very close with my steps, but then I notice the following. I have seen it also while cutting. I programmed the piece to be cut 2mm deep but Mach3 shows 1.985 mm for instance.
I then did the following, ONLY on Mach3.
On the TAB called MDI I enter a value...say Y6.5 to move my Y axis 6.5 mm.
Sometimes it does show 6.5 but but other times it differs.
The very next move say y6.66 show 6.6481.
Remember I get this only playing in Mach3...no machine attached.
When I do it on the machine and I see this difference between what I asked and what the DRO shows, I also see this difference on the dial gauge...Dial gauge shows this 6.64mm movement for the 6.66 asked.
(http://)
(http://)
-
What are your steps per unit in motor tuning?
Hood
-
Woody, Hood will be able to answer your question definitely, however in the meanwhile you might like to play with a little excel sheet that shows the relationship between various parameters. Note in particular that it is possible to specify distances that require fractional steps :).
Al
-
Thank's Al and Hood.
Previously I calculated the steps as per my pinion sizes etc and got to 54 I think. I then used the Calibration in Mach to FINETUNE it.
Firtsly I draw a block in CAD say 500 x 500 then cut it say 1mm thick. I then measured the block and got maybe 501.6mm. I then used the calibration and told Mach to move x=500. Now I know I measured 501.6mm on my block that was suppose to be 500. So when Mach asked how far did it move I entered 501.6mm. Mach then re-calculated the steps and got maybe 53.923456783......lots of digits.
I brought it back to a full number of 54 tonight and did not see the problem yet....will have to do some more testing.
-
The problem you have is your steps per unit equate to 0.0185mm per step step so really that is the minimum distance Mach can move.
Hood
-
So what do you suggest I do. I am microstepping my driver (1/16 step)currently.
Should I change it to half step and then use another value for my step setup Or am I barking up the wrong tree.
-
Really its not going to make a huge difference and I presume its a router so its probably more of a visual annoyance seeing the numbers in the DRO than of an accuracy issue. The problem withthe standard Mach screens is they are designed for Imperial so have 4 decimal positions, a metric one would be fine with 2 or possible 3 depending on the type of the machine, a router may even be fine with 1? You can easily change that by editing the screen and choosing the decimal places, for example I have 3 on my screens but my resolution is 0.000625mm per pulse.
Hood
-
Maybe you need to map the leadscrews with Mach's mapping feature.
I've been trying to find out how to map my ballscrew. How do yo access Mach's mapping feature?
-
Capt Meatballs,
Never tried or used the Mapping feature so someone else will need to chime in on it.
Your the second one in years to ask questions about it that I can remember.
Why do you want to map them?
I ask that because similar to backlash, a software fix for bad screws sugggests fixing the problem instead.
RICH
-
Well, my ballscrew is a bit erratic throughout its movement. I'll ask X to move one inch, and the mill's dro will report 1.002", then I'll ask to go to the 2" position and it will be spot on 2.000". Then the 3 inch might go 3.0035", but then the 4 might be 4.000.
My backlash is around 0.003", but this problem is a little different, it's more cyclic in nature.
My Roton ball screws are not very precise, but for my first CNC built I didn't want to go too expensive. I think I will eventually change them, but for now I'd like to train Mach 3 to recognize my ballscrew inconcistencies.
-
Sounds like what your seeing is a factor of your machines resolution. Mach is open loop, and therefore could not report back the differences even if it did move more or less than commanded. What your seeing is mach rounding the movement to the nearest step your machine is capable of making. Take for instance the 3.0035"....another step up or down of the motor, would have put you beyond the 0.0035 length, therefore, 3.0035 was simply the closest that could be achieved.
-
My question .............Is the mapping functional or actualy implemented?
To use your need a standard that can actualy give a true position over the total length for input.
I have checked ball screws opticaly using a calibrated scale and a jig transit. Longest scale I have is 48" long and only have
a 12" long scale that is calibrated to 0.0001" over it's length. A ball screw spec that states "X inches per foot" may not mean
that screw error is linear and it also dosen't mean that the reverse direction is the same, so one could have a ball screw
where the total error / more pronounced is in some foot section of the length.
The Mach counting of pulses is very accurate and before doing threading tests we actualy compared the pulse to mechanics of the
machine and the other pulses along the sending stream. Used one section of the z axis do do all the tests as it was more accurate. Practicaly
speaking, I would trust the DRO indication.
Practicaly speaking, at the end of the day or machining task, one has what they have, and the part is what it is.
Want better than what one has, pick whatever you want off the shelf but once at the register you need to pay for it!
RICH
-
What your seeing is mach rounding the movement to the nearest step your machine is capable of making. Take for instance the 3.0035"....another step up or down of the motor, would have put you beyond the 0.0035 length, therefore, 3.0035 was simply the closest that could be achieved.
Thank you for your prompt answer Rich, but I respectfully disagree. Perhaps I need to offer more details to allow you to appreciate my assessment.
First is the fact that I can manually step the X axis to the any number I choose, so in this example I can step it all the way back to 3.000" after its initial commanded movement.
Second, mechanically the resolution is 0.0001" per step, but obviously not the accuracy. My screw is Roton #5932, not a particularly accurate one, but ok overall. It's a 0.200 inch per revolution which combined with my 200 steps per turn stepper motor, plus 10 microsteps per step, gives me 10'000 microsteps per inch.
So, I can physically move the table at increments of 0.0001", that's just a fact. I cannot however place the table where I want it using an X#.#### command because the spiral screw is not consistent throughout its length (my assessment).
This is where the idea of mapping the screw comes in.
Now, I've seen a few people mentioning it, but never anyone actually doing it, so I wonder at this point if it is an actual feature of Mach3, or not. If it is I would like to take advantage of it, as it would allow my somewhat inaccurate screw to achieve much better positioning, at least as accurate as my mill's DROs can measure. Who wouldn't want that?!
-
I have tested screw mapping before in a very coarse manner and it seemed to work fine. Set it up and test the results. Its straight forward. Function Cfgs, screw mapping. You jog it to a position, shows the current dro reading in the bottom dros. You measure and enter the actual distance moved in the top dros and enter a correction point. Add as many correction points as you need, save the curve for the axis and enable screw mapping for that axis.
Brett
-
Fantastic! I'll go play with it right now!
Thanks a bunch.
-
My question .............Is the mapping functional or actualy implemented?
Brett, thanks for the confirmation that it's functional ( just being lazy here).
Rich, but I respectfully disagree.
No problem Capt and if axis comp works for you that's great since I never used it.
Let us know how it works for you, now and in the future.
Just a few comments:
Accuracy or true resolution only comes from the SYSTEMS over all mechanical ability.
IE. In testing we found that timing belt tension can affect positional accuracy, verified by varifying the tension and monitoring
pulse count info at each interface of the electronics and physicaly measuring via a known stantard.
If like backlash comp, which i used, and helped me for some time.........but, when doing 3d machining it just didn't cut it.
When engraving or machining small stuff nothing beats true mechanical ability. Backlash can be problematic even when
compensated.
I am just saying that software solutions just mask the mechainical problem.
Enough of being anal about it, just spent a lot of time in the past trying to get my machines accurate.
Have a Merry Christmas,
RICH
-
I agree 100% with you Rich. I'm just trying to get the best possible setup without throwing more money at it (for now).
I'll try the mapping thing, then let you all know how it went.
Merry Christmas and/or Happy Holydays to you all.
-
Well, so far it has been a bit frustrating.
I started mapping the ballscrew and recording values as I went, which by the way were a lot better than I was expecting, but nothing showed up in the graph on the screen, just a red horizontal line. It might be because the extremes of travel were in the .5" range, while my errors were in the 0.001" range, but it looked flat lined and I couldn't find out how to adjust the graph vertical range.
Anyway, I went with that, sent the mill to 0 and enabled the X mapping correction. Unfortunately Mach3 X axis DRO read +1.#QNB, I'm not sure what it meant.
Sending X back to 9.000 resulted in an inch of travel, then the motor started seizing and rattling like it was having a heart attack, and the table stopped moving. I did this a few times then I mostly gave up on the mapping idea.
Obviously I'm doing it wrong, or mapping is not working right (it's probably the first option).
-
I'm not too concerned with the line looking flat. It should look pretty flat unless your making huge compensations I would think. I don't know how to explain it any better than in my previous post. You move to a distance and the DROs show current position. Say 1.000 for example. Then you measure and if you actually moved 1.002 you would enter that as the actual distance (1.002) and set a point. Repeat this process for the length at what ever resolution you want........ every inch for example. Get all your points set and enable screw mapping for that axis. If your screw holds better than your backlash of .003 and I would think it does....... you will be wasting your time IMO. Rich is spot on in regards to mechanical ability. Screw mapping and/or backlash comp are at best a bandaid for a situation requiring surgery to correct. If they will hold you until you see the Dr. that's great but betting your life on it would be a mistake.
Brett