China CNC routers - 3020 oldschool vs. 3018 GRBL control

[Kjelt]

So:
After assembly, what initial tests should I do to make sure everything is allright, including accuracy, nothing bent or whatever?

Manually feel if the guides move smoothly, check the axis visually for bends.
Then place a Clock gauge against the axis turn them and see if they have acceptable run.
Then check motor parameters, move everything in the middle so you don't run to the limits.
Place paper or carton on the machine bed.
Place a pencil or something else in the z axis and lower it till it marks the paper.
Let the mchine run a specific length on the x axis for instance 50mm and check with measurement.
Repeat for Y.
Then check the alignment between x and Y by making a two point opposite corner rectangle marks where x and y are choosen so you can easily measure the corners distance as with Pythagoras law.
 
Those kind of things.

also found cheap (< 2 € !) chinese inductive sensors (shipping 2 months ) which trigger ~ at 4mm. Sounds nicer (no bounce?) and more long living - good idea, or do they have their own problems?

They detect iron better than aluminium so you might need to add a small iron piece.
The ones with a flat sensor are nice, the one where it is sticking out could be more sensitive for surroundings. But these are often used as proxys (proximity sensors).

Some ppl on YT said they don't use limit switches for Z axis, they use soft limits. Really? Why? The motor is not homed then?

Limit is not home switches 
If you have closed loop steppers and good soft limits if you gomoutside and the motor encounters a physical limit it shuts down with a driver error since the servo has a stepencoder on its axis.

[Magnum]

Get something like this
https://www.paulimot.de/messen-und-pruefen/messuhren/686/fuehlhebelmessgeraet-feintaster-0-bis-0-8-mm

and / or this
https://www.paulimot.de/messen-und-pruefen/messuhren/461/messuhr-0-bis-10-mm-fuer-messstativ?c=88

and this

https://www.paulimot.de/messen-und-pruefen/flachwinkel/384/flachwinkel-900-200-x-130-mm-rostfrei-inox?c=84

That makes it easy to check the machine.
You might need to manufacture something to fix it to the machine.

[TinkeringSteve]

"Limit is not home switches "

Well. Where I encountered limit & homing switches before, namely in an automation to assemble some parts we built at my old employer's place, one of the limit switches of each linear motor was also used for homing, so in my mind it was almost the same (no it was not I who designed the overall setup, it was a mechatronics guy).

Magnum:
Ah, thanks. Hm, what's the difference between the first two? DE wiki says it's especially good to check runout, I don't get why, though. Apparently you can make measurement errors also, if measuring at an angle. I bought something like the first second for now.

[Magnum]

IMHO the first one is better for checking the machine. It doesn't need much force for the measurement and the measurement point is quite small. Also good to check runout, as you mentioned. But the measurement range is not as big as the second one.

[Kjelt]

Well to clarify the difference:
A hard limit switch is a last resort safety mechanism to protect the machine for damage, you need two per axis and it should be hardware based (safety)
A home sensor is the location where the machines 0 point is located, this can be arbitrary placed but usually it is a for a human logic location in the left corner. You need one per axis and it is software based.

[TinkeringSteve]

Btw, that "armored vehicle chain" with the cables inside to move alongside the whole thing: it's also plastic of course.
Would it make sense to apply a drop of silicon oil to every joint, or has those plastic parts rubbing against each other not yet resulted in problems for anyone?

Ah, status report

I finally built a table to put this monster onto ;-) So the base and the "portal" are in place, after I figured out which screws need to go where (no instructions, and some screws have same diameter and threading, but differ in other parameters )-
But it's not all done yet, still some mounting going on.
I also had trouble with getting to install LinuxCNC.
Am also entertaining the idea of getting one of those cheaper motion controllers from China. Just need to find out whether there are some with not crappy algorithms inside, I guess. They have some nice oones with ethernet support that one could mount in place of the printer port module inside the controller box. Or cheaper ones just with USB.

I have not 100% given up on LinuxCNC yet, though. But should I learn about a cheap-ish chinese (or otherwise) motion controller, I might try both more or less "at the same time" to find out which route I like better, let's see...

(I must admit I find the idea of a whole PC with real-time linux as "the motion controller" a bit weird, it is theoretically very flexible, but also from some other angles, inefficient solution to the problem, probably with worse performance than using a fast microcontroller with bare metal firmware on it ...
But I'll give it a chance I guess)

[Kjelt]

No need to lube the chain AFAIK.
Do check the cables for in the chain though, better buy quality chainflex quality cables.
Good brands are:
- Igus chainflex cables shielded ofcourse
- Lapp Ollflex
Check which you need, nr of wires, gauge etc. They are made for these applications.

[TinkeringSteve]

Ah. I guess if they are not quality flexible cables, they will go bad after some movements, eh?

I think I should look for some aquarium pump or something for the water cooled spindle. (yes I also bought a weaker one with air cooling, but in case I should need the 1,5 kW one at some point)...
I can't get over the fact that they want you tu put this pump into a bucket and then run 230VAC into it. Double insulated my butt, I do not want to get used to this as looking normal, a wall outlet cable going into a bucket of water, lol!

[Kjelt]

Well these cables are designed for this purpose to last >6 million bends and only cost about one € more per meter.
I can't see much difference besides that it is more flexible (smaller bends possible) and when stripping the outside plastick jacket came off real easy, it looked like the inner cables which are quit sturdy wrapped together and shield could move inside the jacket.

[TinkeringSteve]

Yeah, at some point, I'll probably buy shielded cables. At the latest when these here fail
For now, I will install the limit switches, connecting them as a current loop requiring some mA flow for a closed switch, which hopefully means the controller won't be confused by PWM crosstalk from the crappy cables

Different topic:
Early in the thread, Magnus strongly recommended getting a sturdier aluminum base T-slot plate, and I at least agree that the one that came with the chinese contraption seems somewhat flimsy.

The moderately priced suggested replacement at vakuumtisch.de has a wider slot interval than the chinese one, though (50 vs 30 mm).
Which means I could not use that nice big bag of "clamps" anymore, flat pieces of cast aluminum with one screw in a threaded hole, and another one in a wider slit, with a wingnut attached to it, such that you can fix the thing at a 45° angle onto the slot plate and thus easily attach square shaped objects of "any" dimensions to the plate.

Are there similar utilities available for bigger plates, or is this maybe not the best way to fix something for milling? But *something* gets mounted into the slots of such a plate, eh? I just don't know what. I guess I could fix a machining vise that way, but that's only for smaller work pieces.

[TinkeringSteve]

I also just realized that that aluminum slot table has exactly the 60 x 40 dimensions, not "made for 'cnc6040'"  I overlooked that at first.
Since 60x40 is the working area of the CNC, the real plate size is of course bigger, and I don't think anyone has such plates with those rather peculiar dimensions (77 x 48).
I am guessing that something custom made is prohibitively expensive, nonetheless I will inquire as to the price of just that, and report back here for anyone who cares. (presumably someone in the position I am in now, in 2 or so years, stumbling upon this thread )

[Kjelt]

You don't need the tslot plate to exactly match the machines dimensions unless you need to work on a workpiece that is the same size minus 3cm on the sides for clamping. Most machines have a way small tplate. If you do need it you better bought a bigger machine 

Clamping, look at the same vacuumtish site they have a lot of clamping tools, i will order there also soon 
If you need extra clamps like a box full with different nuts, bolts etc you have to look at ebay, ali-x or a local shop that has cheap clamping assortments boxes.
In our country such a cheap shop is hbm-machines.com

[TSL]

Milling PCBs is something I've done for quite a while now. I have a Carving-CNC 6040 that I added home/limit switches to it and added an output to turn the water pump on and off with the VFD. The water pump is required with my 800w water cooled spindle that does up to 24,000 rpm.

Half the trouble is engineering your PCB to make it suitable for milling on such things. The other half is making the mill as rigid as possible and the mill bed as flat as possible.

You have to pay attention to speeds and feeds + acceleration since any sharp movements can quite easily break a 0.5mm end mill or snap the tip off a 0.1mm v bit.

I use both end mills and v bits for routing but prefer the end mills since they give a superior finish. I also use a bit called a "chip breaker" that is designed for cutting through PCB.

I author my PCBs in Diptrace, output the gerbers and import to CopperCam which creates the mill Gcode and cut profile for the board. The drill output comes directly form Diptrace as a Mach3 compatible file. Mach3 is the software that controls my mill.

I then check the output of CopperCam in NC Corrector which is a Gcode editor and virtual mill. This allows me to see what is going to happen in the milling process and allows me to make adjustments to the Gcode if needed.

Once I have the final output, if its the first prototype, I "cut air" This is where you set Z to 10mm or so above the bed and run the code to see what the mill is going to do without actually cutting anything. This some times exposes nasty moves that might break a bit or shows up moves that should be faster than they are.

I secure a piece of MDF or aged pine to the CNC bed for use as a spoil board, clamp that tightly and then mill a pocket into the wood using a finishing profile. This makes for a almost perfectly flat surface on which to mount the PCB and gives something for the drills to go into when you do your drill run. You don't want holes all over your mill bed, do you ?

I mount the PCB to the wood using double sided Bear tape. This is only about 0.1mm thick and as sticky as hell. I use a roller or block to press down on the PCB to make sure its adhered to the spoil board.

The process is then mill, drill, cut

Pictures of my rig attached

[TSL]

My preferred tooling is from Kyocera, they come with depth rings so that tool changes are all accurate once you have set initial Z.

Here's the box off drills and bits I use

The larges drill in that box is 1/8", the smallest mill bit there is 0.4mm

[TSL]

Whether your building your mill or not, one of the first things you should do after you have tuned the motors to the software, is run a standard test pattern.

The most common is called circle, diamond square. More info here...

https://timf.anansi-web.com/wp/circle-diamond-square/

This test pattern will enable you to check the accuracy of the cut and it will expose any non linearity in your mill quite quickly.

It will also expose over acceleration since the arc cuts stress the drives nicely and they will miss step if your acceleration is to high.

I've attached the pattern for your use.

[TSL]

Here's a few pics of my design process and the resultant boards.

You will notice the boards are different from what is on the screen which was version 1.0.

While I was initially going to have a two part board I decided to make it a single board with a cut out for the FET and mounting holes to secure the heat spreader.

So the boards are actually v2.3

On the screens, from left to right, is Diptrace, NC Corrector and CopperCam

The boards are prototypes for an LDMOS FET amp at 1296 Mhz.

If the FR4 behaves as I expect and they work as I expect then I'll redo the boards on Rogers 4003, test that and maybe then I'll send to fab.

[TinkeringSteve]

Hey TSL, or shall I say alien cook?
Thanks for your info!
I'll digest that on the weekend.

For now my plan is to try it with LinuxCNC first, and if that doesn't do it for me, well to (successfully) use Mach3 I'd have to buy a motion controller first.
The PC I put aside for this tests well on the LinuxCNC real-time test, although the speed is still poor (like 30 kHz) and with still huge jitter compared to even cheaper embedded motion controllers. But some people are happy with it, so let's see.

I haven't really tried anything but the spindle motor yet, as I have not yet installed the limit sensors, and was busy building the table, and designing a noise reducing enclosure which I now also need to build :-D
And still waiting for the replacement water pump for the spindle, as they sent me a broken one, so for now I can't let this run for all too long I guess.

[TinkeringSteve]

Btw,

I recently made that other thread, about grounding the spindle motor to protective earth, I guess Kjelt's DIY machine is all so massive that it's earthed by itself
Anyway I'm curious, what does the ground connection for the AC motor look like on your machines?
(my china CNC had none, so I need to retrofit one. The other thread, with more generic title, has some suggestions, but I'm still curious what your setup looks like)

[Kjelt]

I also have a China spindle. My CNC machine frame will get its own litze wire to PG. PG from the mains cable as well to the controlsystemboxes.
Since the VFD cause lots of EMC noise I keep those wites seperate from the motors and control cables and will let them drop down from the ceiling with the dusthose and mist cooling hose.

The spindle cable is 4x1,0 shielded cable. three are motor cables, one PG.
The shield is connected to PG in the control box and in the connector will be connected to the PG wire to return to the box. There will be no connection to the spindle of the PG in order to prevent groundloops.
The spindle is itself bolted to the Z axis with 8 bolts and that should be PG from the chassis but I have to check that.
My machine is still in the build phase currently it looks like this, the stand is from iron it was welded by a friend of mine. The feet have been replaced by heavy duty feet.

[TSL]

My spindle has a 4 pin connector in the top and a 3 core cable , one of which is earth. I took the top off the control unit to check and its wired back to mains earth.
I've never had electrical noise issues with mine.

[Kjelt]

My spindle has a 4 pin connector in the top and a 3 core cable , one of which is earth. I took the top off the control unit to check and its wired back to mains earth.
I've never had electrical noise issues with mine.

Is the frame of your machine seperately grounded, if no than that flimsy cable is your PG of the whole machine, just that you know it 

[TinkeringSteve]

Whoa! Ground loops, I hadn't even thought about that in this context. Something I still need to look at.

About the separation of VFD and control cables... by control cables you really mean cables that have control voltages, not big currents yet to the ...steppers I was going to say, wait, you have servos, right? Or at least some sort of closed loop system, so that's different from my simpler setup.
I have only cables to the steppers, where all the pulsed current is flowing, I would assume that they will not be impressed by some 400Hz 3-phase AC crosstalk running alongside?

Thanks for the photos.  That's quite a massive beast you have there I see, that won't bend so easily.

EDIT:
For the record, I *can* hear my VFD in a medium wave radio, but it's not as bad as I would have expected.

[Kjelt]

Ah ok, in your case no, the motorpower cables are pretty noisy themselves as long as you use also a shielded cable for them and they are good flexible quality (we discussed that already) than you don't have to worry about them.
For my situation I have easy servo steppers that have an encoder cable, those need to be pretty good shielded from noise or the drivers might go berserk.
I have also the homing sensorcable which is an open collector cable so also not that prone to noise and I have a lighting behind the spindle that is also not a problem.

Grounding is something to look at. Pretty much create one star ground in the controlbox and run all shields from the shielded cables to that point and don't connect the shields on the machine side. That is pretty much the generic advise given in the cnc community. Ofcourse grounding the machine itself properly would be advisible esp when high voltages like the spindle might make contact to the chassis somewhere.
Also if you have a VFD use a proper mains filter, or the rest of your house or neighbours might suffer, the radio is the least of your problems 
Then some people have problems with the VFD and the Earth Leakage Circuit Breaker, there is a whole topic about that on this forum. I would advise first see how it goes, if it runs keep it like that, if problems arise take a look at that topic.

[TSL]

Is the frame of your machine seperately grounded, if no than that flimsy cable is your PG of the whole machine, just that you know it 

You know I never really thought of that and I should know better!! I’ll see about putting a nice heavy ground back to the control box

[TinkeringSteve]

Ok, some reporting back here. As I now had some time to play with this thing again.

One thing people who also buy this thing (my "exact model"  is the CNC 6040 by "CNCEST" as sold by chinese ebay sellers, that name is on a sticker on the device) -

the longest (Y) axis' ballscrew is wobbling around.

I have not read about this as a problem anywhere else, where people mention the "by design" shortcomings of these chinese CNC routers, so probably "my" manufacturer just screwed this one up.

To demonstrate:
Running - note the motor coupling to the right:


Stopping after some steps, so you can better see how the distance between ballscrew and the steel angle thingy changes, depending on rotation angle of the motor:


Here my crude attemt at measuring the deviation, roughly 35/100 mm total, at the edge of the coupling anyway:


Looking at the ballscrew itself, it looks like a lot more than just 0.35mm, though.
So this will probably break the bearings if I use it that way, I wonder how they managed to get it that bad.
The frame of the whole thing seems relatively sturdy as such. Then again, the gantry and all the stuff on it are pretty heavy. Maybe I was less lucky than others when moving the whole thing around on the desk, and the whole setup warped somewhere?
Or they just screwed up, tired at end of working day or so

Now, I am currently struggling with the ebay seller to make him understand the problem, and he's playing dumb, of course He already did that when I showed him a photo of the ingenious re-design of his water pump - the brass connector piece for the output hose was solid, i.e. no hole for water to come through, and never mind the broken off pieces of plastic from the side of the pump - I guess more water comes in, as compensation for the "tight" output!

I reminded him that paypal protection is still on for a number of weeks, suggesting he makes better suggestions than so far.
Let's see how that turns out...

From what I gathered by a post in another forum about the proper setup procedure for such a ballscrew and bearings, it sounds like I could never possibly do such a thing myself, as bulky, complicated, expensive precision parts are required.

So the only real option is that I send the base with the Y-axis parts back for repair, eh? (apart from returning the whole thing).