Opinions sought on cheap desktop PCB milling machine

[Jonathan]

I also use that cheap PCB from rapid, but tend to need the bigger sheets.

Have you considered toner transfer, where you print the mirrored artwork off using a laser printer onto the correct type of paper, place it over the bare board, iron it on then etch as normal? I've used it quite a lot and you can do very accurate boards with a bit care. The best paper is the type that sticky labels peel off as the toner doesn't stick very well to it. The finest I did was an 80-pin TQFP as a test. PCB was fine but I couldn't solder it!

Anyway, clearly milling PCBs is superior. Whilst selecting the track outlines and offsetting them to create the toolpaths works, if you're trying to make fine board it can be difficult to ensure pieces don't overlap. So instead I convert the tracks to the minimum required lines to isolate each connection.

From this:
(drawn with funny pads/track widths to make conversion easier)


To this:


Unless you're doing fancy RF stuff the odd shape tracks shouldn't be a problem. It's also less cutting, so clearly tools last longer and it doesn't take as long to cut. The latter isn't really an advantage unless you're making several of the same as it takes a while to redraw it.

As long as you don't skimp on the stepper motor and drivers there's no reason to expect long term problems with missed steps. This is particularly important since you seem to have chosen a relatively heavy machine bed with dovetail slides, not rolling balls, so the total force required by the stepper motors (F=uC+ma) will be high compared to a moving gantry router if you want to get decent acceleration and speed, which you do for PCBs since they have a lot of short moves and changes of direction.

[HankMcSpank]

Re what you've done to minimize thecut lines...there's a free app out there called visolate - it's basically a midline voronoi region generator ...it takes a gerber as an input file & creates an NC file. It's poorly supported & the documentation sucks, but it does work ok (you just need java loaded on your PC with JAva's 3D libraries loaded too, it's explained a little better here...

http://blog.makezine.com/2010/07/26/...sing-visolate/


Terrible stray capacitance though so forget it for anything over 1khz or so!

(I'm also in email contact with a chap who after a bit of communication to/fro, is soon about to lauunch a web portal where you upload a dxf & it'll generate the voronoi midlines & give you the file back with the midlines done for you)

Re worring about offsets.....in cambam you can import a gerber, fine tune the offset & see the cut width onscreen...the trick is to get the offset so that the midlines of ajacent pins of your smallest IC pitch so they almost share the same path. Difficult to explain, but look here...



the IC pads are black, the tool cut width is turquoise & the midline for the toolpath is the dark blue thin line in the middle ....it's important to set your offset to get these midlines to almost kiss, but not go over, else you'll be unintentionally cutting into the neigbouring pad.

[saxonhawthorn]

http://blog.makezine.com/2010/07/26/...sing-visolate/

Terrible stray capacitance though so forget it for anything over 1khz or so!

Re worring about offsets.....in cambam you can import a gerber, fine tune the offset & see the cut width onscreen...

Well I think it's a safe bet that I'll be going quite a bit over 1 kHz, and I can do without yet more stray capacitance problems, so perhaps I'll give VISOLATE a miss.

But I take the point, and your pic is helpful. CamBam has probably got to go on the shopping list.

Ian

[saxonhawthorn]

Thank you Jonathan. Yes, I'll make sure the steppers and drivers are beefy enough. I haven't done any CNC yet, but I used to plot PCB artwork on a Roland plotter and I know what you mean about all the stops and starts and changes of direction. Dovetails have more friction than ball bearings, but unless there's a similarly solid XY table out there with ball bearings, my guess is that it's easier to compensate for high friction by beefing up the steppers, than it is to compensate for a less-than-rigid (aluminium) ball bearing table of the kind I have seen being offered by the oriental suppliers, some of which look a bit flimsy to me. But I confess to a total lack of experience.

I hadn't even got as far as worrying about minimising toolpaths, and I'm going to have to print off your two examples and lay them side-by-side to ponder them. Obviously there are things to learn about this CNC business! Is there a good book out there which deals with subjects like toolpath optimisation?

Ian

[John S]

I wish someone had told me how hard it was to do circuit boards before I ran the last batch of 20 off this weekend without any glitches.

[HankMcSpank]

I wish someone had told me how hard it was to do circuit boards before I ran the last batch of 20 off this weekend without any glitches.

Yeah, that happened to me when I first started pcb milling too .....but next I had to make a pcb with something more than just two very large through hole components on it!

[John S]

That's about 80 x 80 mm.

[Jonathan]

That's a large track width, all through hole and no tracks between the pins - I'd have been amazed if anyone had problems with that. A lot of SMT IC's have a quarter the pin spacing of DIL, which is when it gets more difficult.

If it's a circuit which is running at relatively high frequencies I do engrave it in the 'normal' way or etch it.

I cut plenty of PCBs on the Roland PNC2300A (same as one John S has) at school with a V-cutter. It's a very weak machine, yet did what I needed at the time (up to SOIC). So you don't need a massively rigid machine for PCBs, since the cutting force is minuscule, however if you're happy with the travel on the table and expect to use it for other things then clearly rigidity is good.

[HankMcSpank]

Exactly....anything through hole like that, feels like groping "Dawn French" after Kylie ....Dawn might have a welcoming large warm bosom to nestle in but not many want to go that way nowadays.

(nice board though)

[ecat]

Curiously enough I've spent two weeks trying to coax 0.4mm tracks and 0.3mm isolation out of my little X1 - quite the experience for the reasons mentioned previously in this thread. I agree with the statement, if the machine is going to do the drilling it may as well do the milling, especially for double sided boards, and wanted to see if the X1 could hold the resolution before I spent any real money or even bought a 'real' CNC machine.

The results are looking quite promising atm, I think the next issue to address is spindle speed - 2,000 rpm does not make for a happy experience. I'm reluctant to add any additional weight to the X1's already over stretched arrangement but I reckon I can drop a 50mm spindle directly into the existing spindle hole. This looked just the job, ESX16 collet, 28,000rpm http://www.ebay.co.uk/itm/2307559193...84.m1423.l2649 but worried about its age and condition I was out bid :( If any of you fine gentlemen know of a suitable substitute I'd be most grateful.

Looking at 'real' CNC machines for the first time in quite a while I see the UK hobby market remains in a dismal state and the UK DIY/kit market practically non-existent.
Strike CNC routers looks promising, if expensive, I'm not convinced by CNCDudze and I refuse to roll with Dice. Reading the Chinese threads make the machines sound a bit of a lottery, even with a little reinforcement and upgraded linears there is still the question of the ballscrews and I've not seen any actual measurements. I took a look at the Zen Toolworks machine, could be ok in a light breeze ;) and spent an entertaining afternoon reading the XZero threads on the zone!

Anyway, it's been a while. Nice to see some of the old faces again and good to see some new ones. Time for more coffee and low fat procrastination.

... Oh, John, when they asked for a PCB with a bus I don't think they meant the #47 to New Street Road End ;) Joking aside, nice board John, smooth, good definition and more consistent isolation than I've managed to date.