CNC3018 upgrade it or replace it?

[elektryk]

Hi,
I'm playing a bit with borrowed CNC3018, which I use for acrylic engraving.
The first tests were done with diode laser, but it was not good for transparent acrylic, now I've switched to RS775 spindle, which was also included, but I've got some problems:
1. after some time ER11 chuck fell off from the spindle,
2. couplings beetween stepper motors loosen,
3. and the worst, despite adjustments, it was not possible to obtain an even Z depth.

As a solution for 2 first problems I used thread locker and for the third problem I'm already working on MDF wasteboard and I want to start milling on the entire surface to make it plain...


About the mods, there are some kits to expand x/y axis and better BLDC spindles, but does it really make sense, or maybe it'd be better to look for something other?
I've seen for example TTC450 machine quite cheap, but i has also rather small spindle and TTC6050, but this one is slightly more expensive.

[Ian.M]

A 3018 CNC with smooth rods and linear bearings is what it is.  It has very limited XY rigidity due to the smooth rods deflecting under load and their relatively close spacing in the gantry and Z axis assembly.  The spindle also deflects vertically which causes problems when drilling or plunge cutting in harder materials.  There also tends to be significant backlash in all axes.   

Yes you could upgrade it, but fitting properly supported linear rails to X and Z axes to gain enough rigidity to make a more powerful spindle useful would be a lot of work and expensive.   

Increasing the table size to 3040 specs with a Y axis extension kit without fitting linear rails under it will only make your Z depth problems worse as now the table is two sections bolted together, and it is difficult to get the joint perfectly flat, and the longer smooth rods will bow more under load.  To some extent this can be mitigated by using the old side frame rails to brace under the table and keep the two aluminium table sections flat  The extra Y axis mass will probably require GRBL's max. Y acceleration and speed to be reduced.

Before trying to cut the spoilboard level you need to get the frame properly squared and the spindle rough trammed.  See https://www.sainsmart.com/blogs/news/mastering-cnc-router-tramming-a-comprehensive-guide-1 to get the basic idea.  However it is worth checking squareness and tramming is as close as you can get it with a sheet of thin card taped down on the table, and a tramming bar initially fitted with a smooth ended probe, and when that's dragging uniformly, a short stub of pencil at the end of it.  Then you should be ready to cut the spoil board surface to get it properly flat. 

If you want real rigidity to make milling non-ferrous metal easy and milling iron and mild steel practical you need a milling machine with heavy castings for the frame and ways, and either expensive heavy duty linear rails or cast iron or hardened steel dovetail ways with adjustable gibs.  Then there are the backlash issues which will be considerable unless it is fitted with high spec. ball screws.

[elektryk]

Thanks, I read it carefully, I haven't heard about tramming bar previously, nice idea indeed.

[MadTux]

CNC3018 is just much too large for a junk router, unless you only do soft wood, foam or use it as laser machine only, with near to none cutting forces.

Softness/stiffness increases with something like 3th exponent as machine size increases, i.e. double work envelope in each axis and keep construction the same, you get something like 10x more wobbly machine.

Much better get a small aluminum CNC1310 for really small stuff or a much heavier router, if you need 3018 size.
3018 size with worse construction/less material than a CNC1310 just results in a bad machine, for anything that even needs a little cutting force.

When you have a little more stability, you can care about putting all axis perpendicular to each other, tramming spindle and so on, but it's just useless, if you dial everything in and then it just deflects during cutting.

[elektryk]

unless you only do soft wood, foam or use it as laser machine only, with near to none cutting forces.

I did some tests and it can cut 3mm acrylic sheet within 1 pass with 1-flute 2mm cutter at 600mm/min feed, I don't know, if it is too much for this toy, but it seems to work.
So cut through its quite ok, the only problem is with engraving, in extreme cases, after setting 0.5mm depth on one end of x axis, the cutter does not touch the material on the other end.

I also measured the spindle and it has 10000RPM without load at 24V, isn't it too low for this 2mm cutter?

[Ian.M]

So cut through its quite ok, the only problem is with engraving, in extreme cases, after setting 0.5mm depth on one end of x axis, the cutter does not touch the material on the other end.

That would imply the gantry or possibly front and back frame are well off square, with one side higher than the other.   Another possibility is the sliders on the underside of the 300x180mm aluminum table extrusion are slightly different heights tilting the whole table.   Start by determining how much out of level the table is with respect to the frame and gantry.  You can do this by connecting the table and a sharp pointed tool e.g. an engraving D bit in the spindle to the controller's Z-probe (A5) and Gnd, and using a G38.2 probe command on the four corners of the table, noting the Z height of each.  Ideally map the whole table in a grid, so you can check it isn't warped or bent.   Some G-code  senders have an auto-leveling function that can be  used to make and save a probe heightmap automatically, saving you much effort.   

Once  you know how far vertically out of square the frame and gantry are with respect to the  table, you can determine how and where to adjust them.   Your frame should have been initially tightened on a very flat surface^*, *without* its feet fitted, and the gantry installed on the same surface to ensure it was initially sitting level.  If you have had to move the  gantry mounts to square it up or center the area of  the table Y axis the spindle can reach, and did so without the frame directly on the good flat surface, that may have been when you introduced the tilt.   

Once you know your frame is flat, you can loosen the gantry and on the  flat surface,  temporarily shim under one of its sides or under the opposite frame side to correct the tilt, then re-tighten the gantry mounting screws to lock the correction in.  If there isn't enough adjustment range to square the gantry, you may need to add permanent shims between the table and the sliding bearings on one of the Y axis smooth rods.
  N.B. you will probably need to come back and tilt the gantry front to back to tram the spindle, so note what temporary shims you used.

Once you've got it as close to square as you can, rough tram the spindle, then fly-cut the surface of a spoilboard, retram the  spindle and any remaining Z height inconsistencies should be gone.   If you ever remove the spoilboard, mark it and the table so you can reinstall it in the same position and orientation, and ideally avoid the need to resurface it.

* Ideally an engineer's surface plate, but a polished stone kitchen worktop is likely to be flat enough for any hobby CNC!

[elektryk]

It took me about 30min with 3mm cutter, probably stepover 3mm with 3mm diameter wasn't the best idea, but sanding paper did the rest.
After some "fine tuning" one side of the gantry is raised all the way up, while the other one is lowered.

[Ian.M]

Good.  Now its basically flat and square laterally, engraving should work much better.  Now you've fly-cut the spoilboard you might as well see how it does before putting more effort into levelling it.   

However that big a lateral height offset and having used up all the slack in your gantry mounts to compensate strongly suggests there's something going on with the linear bearing mounts under the table, and it may be worth your while to shim the bearing mounts on the side that was too low, re-level the gantry on the table surface with no spoilboard then refit the spoilboard and re-fly-cut the spoilboard surface.   Ideally you'd treat yourself to a 1/4" collet so you could use a straight flat cutting router bit 10mm or 12mm diameter, with about 80% stepover to get the fly-cutting done in under half the time with a better surface finish (unless the tram is way out).  I wouldn't expect it to have enough power for much over a 12mm bit, even taking very light cuts.

You had a bit over 0.5mm out of level laterally, lets say 0.6mm, and your Y smooth rods are about 60% of the way from the center to the table edge so you would have needed about 0.36mm of shim to get 0.6mm of tilt correction edge to edge leaving you enough gantry adjustment to permit tramming.   Incidentally aluminum from regular soft drink can sides typically is about 0.12 mm thickness, so I would have started with six shims cut from such a can, three for each linear bearing mount on the low side.

Edit: It is important to determine what is causing the out-of-square condition before trying to improve the situation.  e.g. if it was due to the MDF spoilboard being slightly wedge shaped, thinner at one side, then the correct action is to fly cut it to a uniform thickness, but if the frame is out of square vertically, then the table will appear to rock laterally as it moves on the Y axis, and surfacing the spoil board will make matters worse, as you will cut a diagonally oriented saddle shape rather than a true flat surface, resulting in unstable clamping of larger flat bottomed work.

[elektryk]

Here's another mod, I replaced Atmega328 with ESP32 S3 to get second UART...

Can you guys recommend some cheap DRO/MPG?

I've seen the one from Makerbase MKS, but I'm not sure, if it will be compatible with HAL GRBL (it looks like they use their own FW).

[Doctorandus_P]

The problems with these things that they are all quite bad, and bordering on unusable. For my own standards, I would not go for less then fully supported linear bearings with square rails that are supported over their whole length and (very likely) ball spindles too. But for this sort of quality, complete machines are over EUR 1000, although you can DIY for less.

One (kit) machine I like is the Shariff DMC2 (Kit costing around USD 2500), and the "low cost EUR 500 mill" by "Vulcaman" on instructables. This is basically a copy of a Sorotec, but made in mostly wood and with thinner rails (MGN instead of the HGR that Sorotec uses.

For some low force applications (such as 3D printing, PnP or a laser cutter) using a timing belt is a good option, but I would not skimp on the linear rails. MGN12 is quite affordable from Ali, and considering less than that seems a bit silly. Even Decent to good 3D printers use these rails, but they're also strong enough for a small milling machine.

[elektryk]

Nice project, but would it still be good after resize to 4040?

[Infraviolet]

Anything where you resize a CNC to make it bigger ends up needing a major strengthening boost to the structure to keep it rigid enough. If you consider a beam (any kind of linear lement, solid, box section, T slot extrusion...) within the frame as either a cantilever beam (attched at one end) or a beam held at both ends then the amount of deflection it can suffer (due to the material itself bending even if the end points are perfectly rigidly held) when a force F is applied goes as:
L^3*F/ (3* E * I)   or  L^3*F/(48*E*I)   or   L^3*F/(192*E*I)

E is a property of the material the beam is made from, I relates to it's corss sectional area and cross sectional shape, which equation applies depends on cantilever vs simple supports both ends vs fully fixed supports both ends.

Note how they all feature L^3. Length cubed. Even a small addition to length makes a beam far more deflection prone. Which substantially reduces the N/um rigidity of the machine. Which limits how hard a workpiece material it can reliably cut in to.

As a general rule, if you're sticking to some sort of weight limit for a machine, a big machine where that structural weight has to be wrapped around a big working volume will be floppy, the same structure but enclosing a smaller working volume will be more rigid.

Making a milling/routing CNC machine bigger means making the structural members much stiffer.




The Shariff DMC2 is very interesting to me, one of the inspirations for my own design which I'm working on. And thanks too, Doctorandus_P, for that Vulcaman design, looks a bit floppy compared to the DMC but will be interesting to read about and see if some techniques involved in the design are of use in other designs too.

[negativ3]

TouchDRO.

https://www.touchdro.com/

[elektryk]

I just discovered that Openbuilds Control has mobile (webpage) jog widget, so now I've got DRO/MPG substitute without the need of additional hardware.

By the way, I've seen many builds use a standard 800W wood router as a spindle, those routers use non standard collets and there are also version with ER11, but those are about 4x more expensive, is it the only difference?
I'm thinking about this cheaper version and additional ER11 adapter.

[Infraviolet]

More from general suspicion than absolute knowledge, I'd warn against that sort of other-collet-to-ER-11 adapter. That sort of thing will make your spindle shaft longer (so might limit your available Z depth unless you can mount the router higher up) and will almost certainly increase the runout (how far away from truly rotating about the exact expected centre point of the shaft the tool ends up being) and vibrations in the tool fitted in the end.

I have bought myself, but cannot actually use until I've finished building the rest of my CNC (mine is a small work area but will be for milling aluminium, not surface enrgavings in acrylic), a Genmmitsu 710W "router" which already includes an ER-11 collet, price was about £85 (not that much more than the £50 or so for generic routers with a fixed tool shank size, and less than the >£150 for branded router tools with a fixed tool shank size). The marketing material and some online discussions suggested that, because it has a proper ER-11 which can take tools from the very tiny right up to >6mm, these are quite good for light and medium-ish cuts in aluminium with end mills of 4mm and less, but I've yet to find out for sure.

One has to buy a full collet set separately, the router only comes with one collet of the 6.35mm size, but ER-11 is standardised so plenty of choices for where to buy.

I don't see many circumstances where a router with some "weird" collet which can only take tools of a fixed shank size, plus an adapter from that size down to ER-11, don't add up to a price almost as high as a router which has a proper ER-11 collet to start with. Any price difference between them is going to be small compared to the amount you have to spend on other CNC elements anyway.

[phil from seattle]

A little late to this party but a 3018 has so many compromises that upgrading it is going to cost more than getting a decent machine to start with. The spindle, even in the "pro" version is really poor.  Upgrading to something like a 500W spindle is going to tax the gantry rails so you'd need to upgrade them. Because the plastic gantry risers are specific to the rails, you would have to upgrade them to something like 8mm aluminum or thicker.  At this point, you are well into more cost than the original machine.  You would probably want to upgrade the screws to some sort of ball screw.  With all that weight on the Z axis, you will want to upgrade the stepper motors to something a bit less anemic. Which will probably force you to toss the controller/drivers and get something more substantial.  And, you will want to run your steppers at a higher voltage so toss the PSU.

At this point, what's left of the original machine?

[elektryk]

I have bought myself, but cannot actually use until I've finished building the rest of my CNC (mine is a small work area but will be for milling aluminium, not surface enrgavings in acrylic), a Genmmitsu 710W "router" which already includes an ER-11 collet

It looks like the one from Genmitsu has ER11 directly on the shaft, but I've also seen routers, which have separate ER11 chuck from the factory (see photo), so still not the best solution...

A little late to this party but a 3018 has so many compromises that upgrading it is going to cost more than getting a decent machine to start with.

I'll probably leave the 3018 as is and maybe look for something better.
Wood router itself probably won't be a real waste of money, because I can still use it for manual works (not only CNC).

At this point, what's left of the original machine?

Some screws maybe, but indeed I think it is still a good toy for learning.


A bit beyond budget, but what do you think about this new model?
https://www.foxalien.com/products/2025-new-cnc-router-machine-xe-ultra?variant=44423628652681
This time something that has factory linear rails and ball screws.
I was also looking for PrintNC, but it will probably cost 2-3x more.

There is also total cheapskate option, TTC450 + linear rail upgrade. I've seen that some yt guys did that...