MPCNC Accuracy

I have finally finished my MPCNC. working area is about 350x450x50mm, will share some images soon.

Iv run a few test cuts in MDF and aluminum and they both turned out pretty well. I was very impressed with the surface finished on the aluminum.

Just wondering what sort of accuracy other people are getting with their MPCNC. At the moment Iv cut some simple 100mm squares in MDF, and the edge lengths range from 99.9mm to 99.5mm, I think this is very good for first tests, but I’d look to improve it. If I could get all dimensions to ± 0.2mm id be very happy.

I suspect I’m getting way too obsessed about a few tenths of a mm. Which considering this is a DIY machine made from 3d printed parts with cheap conduit for linear guides is probably very good! Hence why im wondering what others are getting, and if anyone things ±0.2mm is achiviable with a machine like this.

MDF might be hard to hold +_0.2mm tolerances so I think that you are already pretty damn good for wood. I have not done a rigorous testing, it has always cut way better than I need.

With that being said, how did you do your test cuts? Was that a full width rough cut? The ideal way is leave enough material for a finish cut. Do a bulk of the cut and then come back for a small finishing pass. Doing it this way you should be able to achieve finishes about as equal as your rails are smooth.

This is equivalent to cutting something on the table saw and getting it to dimensions with a light sanding.

Some people get bad results and place a lot of blame on the machine, CAM I feel is by far the largest factor of dimensions/tolerances. You sound like you have already achieved amazing results and I feel with time and experience will only get better!

Have a look at the FAQ’s about some tolerancing info.

Yes, as Vicious1 said, do a finish cut :slight_smile:
if you redo the 100mm squares, rough them to 100.5mm and then a second pass to 100.0mm, then any stress to the tooltip will be minimal when finishing and I’ sure you be closer to the ±0.2mm.
Even with a professional CNC machine, most cnc operators always do a finish cut, for holding tolerances and the finished surface are better.

Thanks for the advice.

The finishing cut didn’t make much of a difference, and I had a long post theorizing how i could improve the machine. Then I found two of the pulleys were really really loose! Gona try some more test now I have tightened everything!

So after tightening all the nut and bolts, a tiny adjustment to my steps per mm (200 to 199), and the use of multiple finishing passes, I managed to achieve the ±0.2mm accuracy I wanted!
In fact, on my test piece, all but one dimension is within 0.03mm! Just one long side is a whopping 0.13mm out! Iv attached a screen shot of the measured dimensions, and as promised some photos of the machine.

I added a few photos of the rollers that show a couple of mods I made. Firstly a second bracket on the back of the motor. This made a huge difference to the overall stiffness. Secondly, I added a tensioner under the roller to pull the bearings onto the rail, this again made quite a big improvement to stiffness, but at the cost of speed. The pressure on the bearings increases friction, and therefore reduces speed. At the moment I can get away with 50mm/sec travel speeds. An upgrade to nema 23 motors should improve this.

I plan on adding similar tensions to the center assembly, and I have a new design for the roller which I am yet to try, see attached CAD image.
The new design replaces the bendy tube clamp with 3 bolts that fasten the tube directly the roller, I noticed a lot of flex in this area so I hope this mod will significantly reduce this.
Currently, there are 3 nuts pressed into the small removable insert below the tube, you can see this in the section view. I don’t really like this and I’m gonna have to think of something better.

I also want to add another bearing to the roller, making it 6 in total, although I have not added this to the CAD yet. With the 6th bearing, I will have more room under the tube to redesign how the nuts are inserted/held in place.

Any thoughts and comments on the current or planned mod are more than welcome.

Overall I’m super happy with the results.
Time to start some real projects!




So after tightening up all the nuts and bolts, a tiny change to the steps per mm (200 to 199), and multiple finishing passes, I managed to achieve the ±0.2mm tolerance i was aiming for! in fact, all be one dimension on my test piece was within 0.03mm! Only one long edge was a huge 0.13mm out. I have attached a pic of the measured dimensions, they should be 25, 50, and 75mm.

Also, there are some pics of the machine, including a couple of the mods I made to rollers.
The first mod is the addition of a second bracket to the bottom of the motor, this improves overall stiffness alot. At some point i will replace these with aluminum ones and nema 23 motors.
The second is the addition of tensions to the underside of the roller, which i can use to pull the bearing onto the tube. this also improved stiffness alot, but at the expense of speed. The added load on the bearings increases rolling resistanc, and therefore reduces the maximum usable speed to about 40mm/sec. Nema 23 motos should increase this.

I have a couple more modes planned.
I want to add similar tensions to the center assembly to further improve stiffness, and also change how the tubes are attached to the rollers. I plan on using 3 M4 bolts with an insert inside the tube, just like how the Z axis tool mount is attached. I have seen a lot of flex around the existing tube clamps so I think this will be a good change. Last but not least I want to add another bearing to the roller, making it 6 in total.

Any thoughts or comments on the current or proposed mods would be appreciated.


Why are you using the old rollers and mounts? You modded the heck out if the to achieve what it did with the update.
Before you buy 23’s you could just get regulars 17’s the ones you have on there are tiny. What are the specs on those?

A physically smaller spindle will also increase your accuracy. Just like the nema 23s, a bigger spindle has its down sides, accuracy be the by far the largest downside, explained in the FAQ.

Awesome that you got the accuracy you need. That is easily 10x’s more accuracy most people want so I am more confident in my machine as well. You should be able make some money with accuracy like that.

I printed the rollers just before the new ones were released, and I’m not too sure about the spec on the motors. I bought them ages ago before I really understood anything about stepper motors. If/when i do get new motors i think its a no brainer to get 23. I would just regret it if i didnt.

Iv seen the new rollers and they definitely look like an improvement over the ones I have, but im not too keen on the use of a clamp to fix the tube to the roller. If they are anything like mine you can’t get any tension on the clamp because the plastic is just too flexible. Even if you beef them up and use a stiffer plastic, it is still a plastic under a bending load, and it will creep, resulting is a loss of clamping force, and unwanted slop and flex in the machine.

I’m my opinion we should avoid where possible putting any constant and significant load through plastic parts, as they will creep and change shape.

The gantry pipes don’t need any type serious clamping force, they are under very little load at any time. Zipties would be overkill let alone a full rigid plastic clamp. Every single part on the machine is under load that can be affected by creep (as can be said about most all machines), but plastic has a much higher elastic zone before you get into permanent plastic deformation, that is why I have designed the parts the way I did, flex to tension.

The 17’s you have on there look at least 3 times as weak as the ones commonly used on the mcnc (72oz/in), they look smaller than standard 3d printer steppers(42oz/in). And remember you have two on the x and y axis, so doubling the strength of 1 stepper gives you 4 times the power. Nema 23’s aren’t always stronger than 17’s but hey are always heavier, more expensive, and need different drivers. Moving mass is our enemy (much slower accelerations). It is not as simple as bigger is better. There is a real nice balance to size and weight, and accelerations.

The distance of your cutting tool from the crossing point of the x and Y gantry bars is a very large factor in rigidity as well. Each time you double the distance, horizontally or vertically you at least half the rigidity and accuracy. Do to a larger lever arm. This was originally using a flex shaft tool with the handle as close as possible to the axis and milled aluminum just fine. I think it is actually harder to mill aluminum with the dewalt because of how far away it is, the the dewalt is much cheaper.

1 Like

Let me clarify so I don’t seem like an troll. I am very impressed with what you have achieved and to make it any better is going to be very difficult. I am very interested to see how good it can get so I am looking at every aspect of this.

I am trying to nothing but encourage and help you. I am making suggestions and trying to provide my reasoning for it, I very well could be wrong.

The corner parts you are using might also be hindering you. I can’t be sure because they are not mine. With the new center that you are using the pipe center to center distance changed a bit, I have no idea if those corners got updated to reflect that. You might want to try and measure the and see. the distance is very little so it might not even matter. if it was big enough of a difference you would be forcing your rails to slightly bow in an unpredictable manner and varying distances from the center to the corner on a parabolic arc. To get my new corner parts to fit well I have to sand the little center spacer to the exact fit I want during assembly and it is pretty easy to see because it stacks up 1 piece at a time and you can see if a rail doesn’t touch.

To be honest iv not even looked at the difference in torque between 17s and 23s. I just assumed 23s would be more.
Once iv found some data sheets I guess it’s a pretty easy calc to find out which is theoretically better, Force=mass x acceleration.

Im not sure where the corners came from. Will check the dimensions in CAD. What should the tube center distance be?

Regarding the router, I just went for a cheap Makita knockoff from ebay to start with. Ideally, I will get a kress at some point, mainly because they do proper metric collets, and from what iv seen metric bits are cheaper than imperial ones in the UK. The big questions is 800w of 1050w.

Im a big believer in if it’s worth doing it worth overdoing, so im leaning towards the 1050w model. Although if the current one is working so well, why bother replacing it.
To be honest, if the whole machine is working so well why am I bothering changing anything!

That is kind of what I am getting at, with this kind of machine, overdoing it can cause a much larger problem than under doing it.

I am with you, I would not change a thing, the numbers you are getting are far beyond anything I have ever specified with the exception of some tiny molds that were going to be edm’ed ($$$$$).

I have given my opinion on accuracy in the faq’s.


I have some input as well!

I have been building my mpcnc for the last week when the hardware was received from your shop.
The workspace is 600x580mm, and so far i havent mounted a spindle yet, ive had a looot of tests with my pen :slight_smile:

I made a simple piece to draw, just a square 120x120mm with a circle in the center with a 100mm diameter, just so its easy to measure, and to see if eg. the circle wouldn’t fully draw a closed circle.

I drew this model 20 times, and with measuring the 4 sides of the square, and the diameter in two places of the circle, i had a possible result of 6 out of 6.
So 20x6 = 120 measurements.
The “failure rate” was 34,1% with the maximum alternation being ±2mm.
18,3% ±1mm, 15,8% ±2mm, 65,9% perfection.

-But again, its drawing on paper, and even tho i taped it on the workspace the paper might still have had some bulges here and there, and the worksurface isnt perfectly smooth, so it might be a poor way of testing the accuracy, atleast for me. Tommorrow, ill try an find a nice plate, and just clamp it on, so that it eliminates the “worksurface-roughness” as a possible factor.

My build is the standard build, all hardware is from the shop, i printed the plastic myself with no problems whatsoever, the conduit is 25mm with an outer wall of amazing 5mm!
Theres 3 things i changed, or added… The toolmount is screwed into the Z-axis conduit, i drilled the holes, and cut rivets in it, so i could screw the toolmount directly on the conduit, with the outer walls being so thick, it would be impossible to have a nut trap in there :smiley:
And then i put a thin piece of aluminum across from the Y-axis gantrys, to support a cablechain with the cables from the Z-axis motor.

I feel that my gantrys are a bit loose on the conduit, like one side could be a millimeter further down the rail than the other one , and im afraid that one might always be a tat being the other half of the pair. They are turning the same amount and at the same times, but still, if im moving them into 0,0,0. It might be dodging them a bit, and then the piece would end messed up anyway.

All legs are the same height, everything is square. Im just afraid, when i get to the milling part, that when the middle assembly is around the middle moving around, one gantry will be forced a bit further down the rail, than the opposite side. Ofc, by turning down the moving speed of the system, would remove more material slower while moving, and therefore ensure that theres isnt too much resistance between the spindle and the piece, when moving.

Happy easter!

Theres 3 things i changed, or added.. The toolmount is screwed into the Z-axis conduit, i drilled the holes, and cut rivets in it, so i could screw the toolmount directly on the conduit, with the outer walls being so thick, it would be impossible to have a nut trap in there ? And then i put a thin piece of aluminum across from the Y-axis gantrys, to support a cablechain with the cables from the Z-axis motor.

Not clear on what you did with the tool mount. There is a parametric customizable nut trap linked in the directions and parts list I believe.
Anything on the Z motor mount is a bad idea.

I feel that my gantrys are a bit loose on the conduit, like one side could be a millimeter further down the rail than the other one , and im afraid that one might always be a tat being the other half of the pair. They are turning the same amount and at the same times, but still, if im moving them into 0,0,0. It might be dodging them a bit, and then the piece would end messed up anyway.

Loose, or not square? Either is bad. loose should not actually be possible, not square is. 1mm is not bad depending on where that measurement is taken. You can see by the thread you posted this in extreme accuracy is easily achievable even with older parts. I suggest using it a bit and then rebuilding the center part and triple checking for square at each step. Making a square and a circle with a pen should be easily obtainable and you shouldn’t really move on to milling foam until you get that. It could easily be the pen, the mount, the surface, the pressure of the tip, , the speed, the build, the CAM, etc, etc…

If you continue to have issues start a new thread with lots of pictures and details, videos so we can help you out.

Hi Teddy.
Not sure how you are attaching the pen to the tool holder, but I really struggled to keep it fixed in one place until I made my own pen clamp thing.
Then I ran some simple square and circle tests on graph paper. Is super easy to see if you are square or not on graph paper. I found I had to add a small correction to the firmware to make it square. 0.5mm over 100mm if I remember correctly.

God, quick replies! :smiley:

There is a parametric customization nut trap available in the assembly tab, i just felt this was a lot easier, also for the future. Also i just wanted to continue with the build, and felt it was a different, but good alternative. Actually i started with wanting to fit the universal toolmount, with the slidable exchange, but with further thinking, im sure it would give me stability issues later on, so i went with the original one instead. Ill take some pictures tommorrow when im at work again, then ill show you everything :slight_smile:

I know, the only thing mounted on the z-mount is the rest for the cable chain, all the weight from the chain is supported with aluminium piece attached on the Y gantrys.

And when i mean loose, then becoming not square, because one gantry is further down the rail, than the other. Ill show you tommorrow :slight_smile:

Hi Andy!
The pen holder is screwed in the toolmount, Vicious1 shows on the thingiverse file, like here
And then the pen is mounted to the mount with zipties, i was so amazed how tight it was fastened on it, and with the two thin pieces of plastic, acting as a spring, so smart! :smiley:

Graph paper, that was actually a good idea! And i will look if my settings needs to be adjusted! I will try it out! Thanks!

So im back!

You can see what i mean with the gantry’s here
And i thought of a good example to explain what i could refer to as a situation, like a pendulum. One end is “fixed” while the other end can move. It cant move a lot, but it can move.
So i start thinking of Andy’s gantrys, with the two screws in the bottom to like clamp on the conduit a bit, maybe that would make it a bit more ridged.

I also took some graph paper, and made a smaller square and circle, 30x30mm with the circle being 25mm diameter, the sections on the paper is 5mm. I put two sheets, diagonally on the worksurface, and clamped a plate on the worksurface aswell, its a lot smoother than the worksurface itself.
And you can easily see a big difference from end to end…

And the settings youre talking off Andy, where is it exactly you changed it? Im a Donkey, sorry.

Check out the attached pictures.

I tried some more tests, a new piece, again a square and circle, both 20mm.

The only thing i have done is change the pen, and edit the drawing. Theres two clips.
The first one was wierd, the circle and square are crossing eachothers borders, when they should align. The second one, is basically perfect. And the only thing i did between these two tries, were exchanging the paper, and its in the other end of the worksurface.

Now, i wanted to try something else.
As it is now, i design a lot of rackplates for… well racks, for touring and IT systems and so on. And then i normally send my drawings to manufacturing. And they have always been perfect. But its costly, and it takes forever to receive the finished plates.
The drawing i wanted to try is one i actually ordered a few months back, so i know how everything should look.

So i loaded the drawing into repetier. Attached are two pictures.
In the first picture on the top, is the drawings from the mpcnc, and the bottom one is a print from the computer. But ofc, i know, it may never get as precise as a printer. But my circles arent fully drawn, and those that are drawn is more or less oval. Same as the small square, the mpcnc starts drawing a bit on the left, and doesnt end up in the same spot it started in.

What is shown in the video is absolutely normal, and you have a very large build so it will be more so. Before you start a cut you should be squaring the gantry before you hit go. measure each roller to a corner block, make sure its pair is the same distance. Then hit go. Have you been doing this?

How far are you compressing the pen mount spring?
How fast are you moving? regular and rapids.

Could very easily be drag from your drag chain. 3D printed ones are much less smooth than real ones, but I don’t like either on the z axis. you can see this by lightly touching the top of your z motor and watch the pen tip, or run your code in mid air and see if the tip rocks back and forth as it changes direction on that axis.

Hello! Im back!
Its been a busy week, so i havent had much time, until today.

I just think, if its normal, why do the accuracy vary this much from test to test.

I actually found out that i have been a moron, my mounting plate had a small curve to one corner, so that might have been one of the main reasons i had problem. Ive changed the plate and mounted everything again, and squared everything again.

But, i still have the problem where my pair of gantrys arent equally far down the counduit, so its not perpendicular…
Its still only 1-2mm but thats still a lot when working on a project.

I have like most, 0,0,0 as the center of the drilling head, between projects, no mattter where i move the cutting head, the center of the drilling head will still be 0,0,0.

When i move in between projects, should i meassure between gantrys and corners everytime?

Its like, when i move one gantry back so a pair is similar, the other set of gantrys move. I can force them all to be perfect by hand, and then power the motors with repetier, just to lock them. Then they work as they should. Its just frustrating having to do that everytime.
I have tried tightening and loosening the tension bolt A and C. It has helped, but not enough.

The pen mount spring isnt compressed a lot.
If i put a forceful finger on the z axis, ofc it moves a tiny bit, theres not loosenes causing any rocking back and forth there either. It feel veeeeeery rigid! : D