Milling PCBs on CNC 3018?

[Chris56000]

Hi!

I've been trying to mill the PCB design attached in the zip file (a Z80 test board for my friend) using a CNC 3018 machine and the results are proving totally hopeless!

Can the PCB experts have a look at my design, the cutter and the results I'm getting from trying to mill this please?

I made a 555 single-sided PCB on this same machine a week or two ago which came out quite well (apart from drilling the holes too large for the component leads!) and I can't see where I'm going wrong!

The CNC is running GRBL 1.1f with Candle 1.1.7 and I've tried a Z-axis depth in the FlatCam software of -0.010 and it's still hopeless!

I told my friend a PCB as complex as this should really go to China but he's a skinflint and won't spend any more, having bought the CNC machine!

What are the minimum clearances, in thou/mils, I can reasonably expect the CNC 3018 to mill to, in order to set the DRC Rules in my Design Spark Software?

I've included all the Gerber and native design files as it's not in any way confidential!

Sorry for being a pain in the proverbial, but my friend's in a nasty nark over this!

Chris Williams

PS!

The [ictures shown are two pieces of double-sided PCB already spoiled trying to mill this - I wonder if my friend could have bought cheap 2 oz copper PCB stock by mistake which won't mill?

The standard offering from PCB Way/JLC, etc., is only 1 oz of copper of course!

[ataradov]

CNC 3018 is a toy. I doubt you will get any sort of performance form a hobby motor used as a spindle.

That being said, your milling looks way too deep. -0.010 in what units? Have you checked the bed for flatness at all?

[saike]

You are cutting way too deep. I did 4 passes over the same board recently with a 3018 machine while my LPKF was out of action, to get the cutting depth right, and only on the 4th pass it cleared the isolation channels, this was with 0.65 pitch components. It just comes down to experience & practice. 2oz copper in my experience doesn't mill as well as 1oz, maybe I need more practice at using it.

Edit    I fibbed, my backup machine is a 3020, a bit more heavy duty and perfectly useable for making pcb's.

[beanflying]

Make sure the board is locked well to the bed and very very flat (see below), more passes, sharp bits and most of all cut down the flex on the bit by using a ER11 collet style chuck (grubscrew chuck doesn't work well IMO). Mine has been mothballed for a few months but was doing a reasonable job on PCB's, ply and balsa.

Used in combination with one of these phone repair holders and a few chocks helped keeping PCB's flat a lot too.
https://www.aliexpress.com/item/Universal-LCD-UV-Glue-Mold-Mould-Glass-Holder-for-iPhone-Samsung-HTC-Screen-Refurbishing-Repair/32470846617.html?spm=a2g0s.9042311.0.0.27424c4db4z1vg

[bsdphk]

You absolutely want to use some kind of "height probing", also known as "levelling" when you mill PCB's with a CNC router - google "pcb cnc height probing" or "pcb cnc levelling"

I originally invented that concept (http://phk.freebsd.dk/CncPcb/), but by now there are a lot of implementations of it out there.

The fundamental problem is that the thickness of the substrate is not a tight tolerance and you clamp the PCB by the backside.

Trying to mill away 35µm of copper on top of a substrate with more than 100µm thickness tolerance only works if you height-probe the shape of the surface beforehand, and then wrap the G-code over the resulting "topographical map".

[Chris56000]

Hi!

Ah!

No, I've not implemented height levelling – can anyone enlighten me what connections are made to the  Woodpecker control PCB for this?

I thought of simply connecting the spindle–motor casing to 0V on the woodpecker board then the level input to the one corner of the workpiece – is this correct?

I was using the standard grubscrew chuck that came fitted to the spindle–motor but we did get an ER11 one in a small plastic yellow box!

I'm not sure what units FlatCam is but I ihink it's metric – I assume "–0.010" is 10 microns cut depth but if anyone can recommend a better setting please advise me!

Thank to everyone who replied!

Chris Williams

[kosine]

Buy a cheap jewellers loupe magnifier off ebay, even the sub £2 ones are OK.

The engraving bit you're using is a good choice (I use them myself), but you absolutely need to get it spinning true - i.e., with no observable runout. With the spindle running, get up really close to the bit and look at it through the loupe. If you can see any shadowing around the tip, stop the spindle and reseat the bit. Repeat as necessary until the tip appears needle sharp with no blurred edge.

You can also use the loupe to check if the PCB is totally flat. With the spindle running, just travel to the four corners, and incrementally lower the bit until the tip just touches the surface and leaves a small mark. If the machine is up to it, you should be able to get a resolution of <50um. Either shim up the low corners or compensate in software. (A discrepency of 50-100um across the PCB is usually OK, any more will require adjustment.)

For cutting depth you want to be starting at 50-100um, which is 0.002-0.004 inches. Obviously go a little deeper if necessary.

For cutting speed, the spindle should be as fast as possible, and the feed rate about 10-30mm/min. If you go too fast on these hobby machines the cutting pressure will deflect the bit and leave a poor cut. Slower is better.

With a bit of practice you should be able to cut tracks running between 0.1" IC pins, and SMD boards using 0805 components.

[Chris56000]

Hi!

Member kosine, thank you for your helpful reply!

I'm just wondering if Flat Cam's Z Dimension might have been in "inches" in which case no wonder the machine was making a gruesome noise and a mess!

I'll do some research and set up auto–levelling on the machine as well

Chris Williams

[ataradov]

Probing and leveling over the board surface will be critical for large boards.

But before you do that, just run the same test program multiple times. Let it hover over the surface on the first pass, and sneak on the depth that is actually cutting something. Use as small of a step as you can for that. This will let you separate the capabilities of the machine versus the material used.

[kosine]

I did wonder if your machine was running in inches, definitely something to check.

A few other tips I forgot to mention:

May be obvious, but when referencing the Z for cutting, I try to find an unused area at the centre of the board and touch-off there. Any discrepency in the flatness of the board is then averaged out - one side will be slightly higher, the other slightly lower.

If I need to cut to the edges of the board and can't find room for the clamps, I stick the PCB to a larger piece of acrylic with double-sided tape, then clamp around the acrylic. Has the benefit that I can drill through-holes and router out a shaped board without going into the table. A kettle of hot water softens the adhesive for easy removal, and it doesn't affect the flatness of the board as much as you might expect.

Because the depth of cut is critical, I resist the temptation to cut the whole layout in one go. Instead I cut track-by-track and recheck the Z periodically. If the board is not sufficiently flat I can adjust the depth of cut to compensate somewhat.

If I need to do a double-sided board, I drill some through-holes at the corners. When I flip the board I drop into those with the same drill (not spinning, obviously!) to get the X-Y reference correct. Up close with a loupe I can usually eyeball it to within 50um, which is good enough for the boards I cut.

When you start to get some results please post a few pictures. I'm sure we'd all be keen to see them. Attached is one of mine using the techniques I've mentioned.

[KL27x]

V groove bits like high speed. They would work better at over 100K rpm, if you could get it.

When I use a manual router table, I use a square end mill angled at 45 degrees to cut copper clad. Loads faster and smoother, can take much deeper cuts, and with 100x less worn out bits. If you have another 2 axis on the machine, it might be feasible to CNC mill PCB like this?

[Chris56000]

Hi!

I've now got the auto-levelling to work, and I'm getting a much better cut-depth, I've set the cut depth to -0.095mm in two passes as recommended in a Unicycle Forum which had one of the most well written Chinese CNC pcb tutorials I've ever seen, using a 2418 machine:-

https://forum.electricunicycle.org/topic/11205-milling-pcbs-with-cheap-chinese-desktop-cnc-router/

However, there's now a strange scaling problem I've now got that I can't work out - I've tried milling one of the example PCBs that comes with Design Spark, which measures 42.7mm long by 29.5mm migh, but for some reason the machine is only milling everything about one-third the size it should be - the PCB's coming out 13mm long by 9mm high!

I've tried Candle 1.1.7., gRBL Control, CNCPilot and GRBL Plotter as mentioned in the tutorial linked above but I'm stumped as to what's wrong!

Could something have gone wrong when I upgraded the firmware from GRBL 0.9J to 1.1F?

Chris Williams

[foxcolin.com]

I have managed to get successful milling on a smd to breadboard board I designed in Eagle to prototype an audio amplifier, seen in the attachment on this post.

I used https://carbide3d.com/apps/rapidpcb/ to convert the brd file into a file that Candle could handle. There was a lot of tweaking involved, as follows:

-Changing the trace depth from .2mm to .18mm (as that was the specification of the board I ordered),
-Doing a height map (as differences in thickness in the board is very important with a V-bit, as deeper means thicker passes that can make traces smaller)
-Changing tool bit size from .1 20 degree V bit to .2 20 degree V bit in carbide3d (included with the cnc-3018 pro is 10 .1 20 degree V bits)
-Start above the board, run the milling program, then slowly move down until there is contact from the V bit to the board

I plan to make a youtube video going through this since I have seen many people have issues trying to do this. If I don't reply within an hour or two don't hesitate to email me but I really want to keep the replies on this board to help out others who want to try milling with this setup.

[ataradov]

Please crop and resize your image! There is absolutely no need for this gigantic image.

This is pretty horrible quality for a board this small.

[foxcolin.com]

I am aware of the lackluster quality, I still have a LOT of kinks to work out with this setup. The resin-like finish is due to a clear coat of nail polish I put on to stop oxidation, and the sketchy solder job is due to the lack of proper smd soldering equipment and inexperience of working with such small chips.

I also did a small board to show that if this process and work with something as tiny as an smd (in this case a lmv358 with a 8-SOIC footprint) then it can work for many other things, like arduino shields and whatnot.

The quality will improve in the future, this is just the first working one I was able to get.