Precission issue when milling

New Home Forum Mostly Printed CNC – MPCNC Troubleshooting – MPCNC Precission issue when milling

This topic contains 25 replies, has 5 voices, and was last updated by  Ryan 3 months, 3 weeks ago.

Viewing 26 posts - 1 through 26 (of 26 total)
  • Author
  • #98862



    I have a strange precission problem when milling.

    I’m milling a square of 30.0mm width as pocket with a 10mm island in the center.

    I’m using 6mm tool, with reasonable settings (12k rpm, Fxy 10mm/s 2.5mm depth per pass, total 5mm depth) and a peeling strategy leaving enough space to the edges, followed by a first pass along the edges (outer and inner) with finishing allowance 0.4mm and a finishing pass.

    I see /hear that the finishing pass really cuts some material – so the over-dimension could not come from tool vibrations during peeling.

    The result is as shown here:


    On the larger/outer square, I have about +/- 0.5mm deviation. As it is + (in Y) and – (in x) it could not be a tool diamteer missconfig (btw: the tool is brand new and the cutted surface looks reasonable).

    Spindle is aligned perpendicular to the surface. I also tested planing the workpiece (milling away 1mm on top), before doing the squares. Same effect.

    I have these results reproducible in different wood and also in foam XPS (which however is a bit hard to measure).


    So my first idea was that the axis position accuracy might be bad.
    I inserted a needle as tool and placed a tape measure along y.

    1st test: position accuracy exactly where I milled the squares with a step of 30mm:

    With this approach, one can easily ‘see’ deviations of about 0.1mm or even better. We would expect 0.4 – 0.5 mm, so this should be obvious. However, position accuracy is good.

    Then I tested along the complete Y axis

    127 – 697 mm = 570mm which is exactly my Y width.

    From the second image one could estimate that it might be about 0.1mm off (shorter), which is good i think.


    So now I’m helpless. Where could the miss-dimensions when milling come from?

    Any help appreciated






    The axis that is making a larger island and smaller hole (looks to be your X? 29.6mm), My first guess is your belts on that axis could be a tiny bit loose. Or can you show a picture of your zipties.

    Another test is resize that same part to double and see if the error is the same or doubled. How I see it is slightly oversized consistently could be explained by run out and other “standard” spindle/collet/tool issues. Undersized ID oversized ID is a little different suggesting a little flex.



    The axis that is making a larger island and smaller hole (looks to be your X? 29.6mm),

    Yes, it’s X. Would that only apply when milling or also in the needle test?



    Only when loaded, if the cable ties are too large they create a stiff spring, or just the belt being too loose in general. You can either make them smaller or zip tie it…hard to explain. I looked I could have sworn I had a picture of this. I can not find it. I will take another and add it to the instructions.



    Well, a poor picture.

    The best thing is to have your cable ties as short as possible. No matter what one side should be all the way tight, the other end as short as possible. If you messed up and have a short belt and need a long cable tie on one side do like I did in this picture.


    The left most front cable tie is tied to keep it from springing.



    You might also double check the actual dimensions of your bit. A 6mm bit that’s actually cutting 5.85mm would put you into a bit small for both a pocket and island.



    Hi Bill,

    this was my 1st idea, but its to small in x and to large in y, so this cant be due to but diameter wrong – not?



    Thanks Ryan for the image regarding cable ties.

    The needle test shows about 1mm less actual travel on 800mm.

    Yes, one side is long, but I replaced them by metal wires. The X was really a bit loose, so I added some tension. Needle test not much better.

    I then milled another test and a double sized test. Here are the results:

    same sized square: numbers in green: effect worse – I would not have expected that.
    Double sized: in X error gets a bit larger, but much less than doubled. Y: for the outer square I see a huge (triple) error.

    I’m clueless




    What drivers and board are you using?

    Some import 8825 drivers do not work correctly.



    Try to fix the stock as high as possible



    Its a RAMPS 1.4 with DRV8825.

    Could it be the drivers when the needle test (no milling / no force) is almost exact?
    @guffy: Do you mean Z as high as possible?



    Yep, close to gantry



    There is nothing consistent about your numbers, nothing that can be fixed until those stabilize, at double the size some of your tolerances got better. I can only suspect you have a large build or your gcode is not correct.


    Again, pictures would really speed up this process, guessing take forever.



    I was also thinking it could be nothing consistent in the mechanics as the effects are opposed (in x/y and outer inner).

    I will (tonight)

    • have a look at the G code
    • position the machine when milling with individual G code from repetier as I do with the needle test (which is fine) to see if this makes a difference
    • test at Z up
    • provide some pictures (my builds is this one, however the picture still shows the first setup with the long plastic cable ties)


    Well you have a very large/heavy spindle, and a gantry cable chain on a very long lever arm connected above the Z axis, and it is a pretty large/tall build. That lever arm is working against the axis you are having the most trouble with and would explain non consistent error on only one axis.

    I think some of these numbers can get a bit better but a shorter build will help for sure.




    I made some further experiments (later more).

    First you ‘main point’:

    you have a gantry cable chain on a very long lever arm connected above the Z axis

    I was following the advices regarding not to connect the cable chain etc. on the Z axis. In fact it is NOT as far as I understand the situation. Here is a more detailed image:drag_chain
    The drag chain rail is connected to the x-steppers, the aluminum harm is connected to the middle part; by this, thecables leaving the cable chain and going to the z axis (stepper and spindle) are ‘static’ viewed from the z axis (the cable chain movesand the middle part connected aluminum arm hold them in place).

    Is this still problematic? I can not think of a better solution.


    large/heavy spindle

    Yes, you are right. I partly regret this (for the weight) but it was chosen to allow for ER-16 inserts (10mm) which I will need.

    I clearly see the effect of the heavy spindle when makin 1mm spets, the elasticity of the belts/the machine leads to some ocillating (could be reduced but I’m not sure if its worth it).



    Regarding the cable ties:

    I found it beneficial to replace them by metal wires as shown here:

    This reduces the elasticity in this region to almost 0. I made experiments and found the parallel wires best (due to beeing metal, they do not flex in this shape).


    Next will come some results of new experiments wich look much better (i need to get the numbers in excel first)







    Is this still problematic? I can not think of a better solution.

    Yes, it is above your Z axis, it is almost as bad, in this case it is so tall I would easily say it is worse. Torque multiplier.





    Yes, it is above your Z axis,

    I don’t understand this. The arm is connected to the middle part, not to the z axis. In fact I have nothing connected to the z axis except the necessary cables which however are not moving when the machine is moving and they do not induce any force on the z axis (the force enters – via the arm – into the middle part)

    How else could one get the wires force-less to the z axis?




    Just my opinion, I use wires from the Z to the Middle, then the middle to the roller in a loom, not cable chain.


    Most people do not care about 0.3mm in wood, you do so I am trying to point out all the possibilities. A bar that long is acting as a torque multiplier, any drag induced is multiplied linearly to your gantry. Is you issue happening on that axis? No pictures to reference with your cut so I can only guess. If that is not it, try cutting your entire machine down to 3″ Z, or each axis shorter in general. Like I said you have a large machine, that is tall, with a giant spindle, using cable chains, all of which can be working against you with tolerances you are wanting.



    OK, no offence meant, I appreciate your help! It was just that I did not get it.

    You say: the cable chain has drag and the arm sclaes up that drag against the mid part when moving (in this case the Y axis should be affected – right?)



    Ok, here are detailed numbers.

    Some ‘management summary’ first:

    What I did change:

    • replaced cable ties by metal wires
    • increased tension in timing belts
    • milled a larger structure with more depth (8mm) for more reliable measurements
    • used a dedicated caliper for the larger square for more reliable measurements

    What I milled:

    67x67mm outer square with 42×42 inner island in coarse MDF at 2 different positions on the workplace and 1 fine MDF (which I suppose to be harder). Followed by a finishing pass (taking 0mm). Then removed 0.5mm on the contour, followed by one or several finishing passes.

    The second image is just for illustration of the coarse mdf.



    The results (short version):

    • x/y results are comparable (x:y deviation after a finishing pass is around 0.1mm which I think is great!)
    • I have a persistent total deviation towards structures being smaller than intended (mill forced away from the workpiece due to elasticity?) of around 0.5mm in coarse mdf and 0.6 or even 0.7 in fine (harder) mdf, which in principle makes sense.
    • The deviation of the larger square is larger than for the smaller one – which is not clear to me
    • The results are reproducible at different positions on the working area (which is great)
    • I have the absolute deviations even after repeating the finishing pass 10x.

    Here are the numbers






    I measured the milling bit, it actually is 5.88mm instead of 6mm (due to wear?).

    So we could correct reduce the absolute errors by 0.12mm.

    What I do not get is that even 10 subsequent finishing passes leave an error of 0.6 for the larger square. The needle test – force free positioning gives me an incredible precision for that size of the machine which is better 0.1mm @ 540mm in y (almost 0 deviation for 70mm) and 0.8mm in x along 800mm (better than 0.1mm for 70mm).

    Is this due to the fact that the bit is pushed away from the workpiece due to elasticity and can only cut away a negligible amount? In fact, the numbers from 1. to 10. finishing pass actually get slightly better.


    I need to do some experiments in hard foam (XPS) which should result in much less force and therefore in better results; I’m not sure if I can measure the foam with the necessary precision; however it would be interesting to see if the numbers get even better in foam.

    edit: I did use ‘conventional milling because estlcam states it should be used in soft machines. However, the tutorial seems to prefer climb milling. do you think climb milling will help or be disastrous?


    Why did I build that large machine: my main intent is to cut surfboard foam kernels – for which even 1mm ‘off’ is negligible – so the machine is great for that I think.

    Every now and then I make some airfoil moulds. I could make them in XPS as positives and cast the actual mould from that (for which the machine should work ok). However, milling the mould directly in MDF would save me the positive to mould casting step.



    OK, no offence meant

    I did not take any of that with offense, promise.

    I measured the milling bit, it actually is 5.88mm instead of 6mm (due to wear?). So we could correct reduce the absolute errors by 0.12mm.

    Plus collet and spindle run out.

    I use climb milling most of the time.




    Here are some further numbers:

    Whats in there (all numbers corrected  for real tool diameter):

    left: last mdf numbers from yesterday for comparison

    mid: switched to climb milling, cut away 0.2mm followed by 1 finishing pass.

    Interpretation: even for the cut pass the numbers a better than after 10. pass using conventional milling. The x/y deviation is quite small -> i will use climb milling in the future

    right: new square milled with gantry all up (80mm above the level of yesterday)= most stiffness in Z i can get

    Interpretation:Numbers get a tiny bit better


    In conclusion:

    • using climb milling and gantry all up, I get an accuracy 0f about 0.3mm and x/y deviation of almost 0
    • at z=-80, I get almost comparable numbers

    which is a good result for that large machine – don’t you think so?





    This is fascinating.

    I generally think of climb milling as dangerous because it pulls the cutter into the cut, but now thinking about it more carefully there are actually two actions.  Yes it pulls the cutter into the cut in the direction of feed, but it pushes the cutter away from the part lateral to the feed.  Conventional milling is reversed: it pushes the cutter away from the cut in the direction of feed, but also pushes the cutter into the part, lateral to the feed.

    And this would also have big implications for slotting.  You might expect one side of the slot, the climb milling side, to be accurate to dimension, while the other side, the conventional milling side, could be off due to the lateral force on the cutter.

    Maybe this is something that veteran machinists know without thinking about it, but for me this is an Aha moment.



    Try thinking of it in terms of cutting metal. Starting thin to thick can actually work harden the material by rubbing before it bites or even ramping the cutter away until it has enough lateral force to bite, to me I used to think of it as “easing in” that is wrong. Taking a good bite and easing out is a better cut. Slotting, too complicated o sum up in a few words.


    Climb Milling vs. Conventional Milling

Viewing 26 posts - 1 through 26 (of 26 total)

You must be logged in to reply to this topic.