EEVblog #753 - Inventables X-Carve Milling Machine Unboxing

[EEVblog]

Unboxing of the new Inventables X-Carve Milling Machine kit.
How well will it work for PCB milling?
https://www.inventables.com/technologies/x-carve

[rs20]

It'll be interesting to see how it goes with PCBs, and especially how fiddly it is to set up. In an ideal, utopian world you want a floating head (whereby the head rests on the surface of the PCB, and a conical cutter protrudes out by a certain, controlled amount). That way you can precisely control the width of your cut by adjusting that protrusion, you can use a not-at-all fragile conical cutter, and you don't need to worry about your machine being set up perfectly parallel and the PCB being perfectly dead flat, since the head is just resting on the PCB regardless, and always digs in by the same amount.

A more useful comment: try (good quality) double-sided tape to hold the PCB down. That way, you can use the entire area, and you're not at all worried about crashing into the clamps (or, much worse, the bolts inside the clamps). Also much less fiddly than clamps, although the X-carve does have a nice distribution of threaded holes that might make clamping quite straightforward.

[kizzap]

I don't know if you have read instructions, or figured this out already, but it looks like that mains input PCB is designed to attach directly to the PSU Screw terminals...

Edit: Adding to that, it looks like that black enclosure for that mains connector is designed to attach to the PSU brick. I am not sure, but I hope that the PSU brick has a fuse in it...

[station240]

I take it from the comments in the video, the Pick N place machine isn't fully assembled yet. Make me wonder what hope most of us would have given David knows about CAD machines.

I wonder if the X-Carve could have the table made into a vacuum table, that would hold parts/PCBs down.

[Fungus]

Make me wonder what hope most of us would have

Of assembling it? Not a problem. Just like flat-pack furniture but with wires.

I'd be much more worried about whether the software side is up to the job.

[Barny]

I'd be much more worried about whether the software side is up to the job.

If the software don't work, Dave is able to use LinuxCNC.
In middle Europe this sort of mill is knewn as Käsefräse "cheese mill".
(It is only possible to mill material which isn't harder then cheese.)

The shown mill will surely be able to mill PCB's.
Wood and plastic will be able to be milled with some limitations.
Aluminium will be troublesome and steel will not be possible.

I love it to heare Dave talking about rigit parts and holding cute thin parts.
I think Dave have the same feeling hear machinefitter like me talkeing about a power supply or oszilloscope


A little hint:
A brushles motor would be a better choice than a brushed motor.
Have I seen right that the vent holes are blocked with the motor mount?

And make sure that the concentric run-out is within few 1/100mm.
If the run-out is to large, it will kill the tool realy fast.

[bookaboo]

I'd love to see something this cheap work for PCBs but I'm not hopeful. I've used both T-Tech and LPKF professional machines, they were serious kit and even they were a pain to use. Ages to set up correctly, dusty, noisy and lots of tool changing. Then as a bonus it only takes one mistake to ruin hours of noisy drilling and milling.

[VingTor]

I bought a Sable 2015 CNC engraver machine several years back. The original spindle it was delivered with at the time was extremely noisy and bad. Later a new spindle was available for it, and I upgraded to it. Now I can sit and work next to it while it is running (lets call it expected level of noise). The new spindle uses the same type of locking nut as the on in this machine it seems (ER11) which is really good in holding the drill/mill bit. It is also much more stable than the old one.

My machine did not come with any end-stop switches, but the hardware supported this so I have glued in place switches to ensure that I do not run the machine to far. Did not check this machine, but end-stop switches is in my opinion a must, especially since one can use these to home the machine to a good position, then use soft limits for the rest of the run (given that the software supports soft limits.

As mentioned above, a floating head would be nice for PCB work. I want one... My table top is not perfectly flat. Probably 100-200 microns difference across the table top. I solved this by adding a sacrificial wood surface on top of the bed and milled this flat compared to the milling head. It helped, but still there are issues. When putting a PCB on top and clamping it down you may add a small tension and curve into it, making milling a bit hard. A vacuum table top would probably solve this nicely as it will pull down on the entire surface of the PCB.

For holding down PCBs I would probably recommend to first start with standard sized copper clad PCB boards. Add a sacrificial wooden top, mill a pocked fitting the copper clad exactly, and standardize on this setup. Milling the pocked will fix any table top leveling issues. PCB will be at correct 90degrees angle each time. It makes alignment for double sided PCBs easier. One could in addition modify this setup to add vacuum to hold the PCB down. I have not done this yet, but it is on the todo list.

Another "issue" with my machine is that it did not come with any means for clearing chips from the milling process. It is also on my todo list to add a small air blower next to the milling/drilling bit to clear away dust and similar. Lubrication/cooling can also be added in case of milling metals or other material requiring that.

On the electronics and software side there seems to be big differences. I use a machine with LinuxCNC and controls the machine directly through the stepper drivers by the parallel port. The machine have to do all the realtime motor control (hence the special linux distro). I have also played a bit with arduino control. I used GRBL, which I assume this machine is also using a variant of. Actually worked quite a bit on a GUI with openGL support for GRBL, works but not finished. That said, this is more or less just about streaming G-code to a machine, so the more important is probably how the G-code is generated. For this I tried a lot of different tools, but finally found that bamcam, http://www.cambam.info/ , was the most stable one for my use (and free for the 40 first projects). Also not to expensive if i want to buy it. Takes some time to learn to use it, but very capable software when one learns to use it.

So, what are my experiences with milling PCBs?
I only consider it for prototyping use or through hole projects or projects with large surface mount devices (0805 and bigger). Trying to do super thin traces and isolation is a pain. If the table top is not perfectly flat it will be bad. If the PCB curves a bit while strapped down it will be bad. Due to this I prefer through hole projects or SMD pars where the isolation tracks can be fairly wide. Using a flat 0.8mm endmill is no problem. Bought a couple of 0.2mm endmills and broke them both in 10minutes. V-type milling bits works great in some cases, but you do need to consider the fact that they are V-shaped when calculating cutting depth and width of traces and isolation. Flat end-mills are much easier to use as it is easier to see how the final results will be in the CAM software.

I still consider my CNC a useful machine, but it takes time to learn and know its capabilities and how to setup what you are trying to make. Given how much I use the machine I'm not sure I ever will save money compared to just ordering the PCBs online instead, but since I like to make stuff, I consider this as part of my hobby and dont mind this fact One nice thing is that drilling perfect holes in the PCB becomes very easy.

An enclosure for the machine is a must in case you do not want to have a mess to clean up (or dust to breathe in).

[PlainName]

I have a Sable too, and whilst I took some while (and bust bits) to get to where I am, now PCBs just run off without a problem. A few points you might find of interest:

I solved this by adding a sacrificial wood surface

I use a half-inch thick slab of perspex, milled flat. The advantage over wood is that it is more stable and doesn't break up once you've drilled it full of holes (or over-milled into it).

For holding down PCBs

I use carpet tape. Thin, constant thickness, sticky enough to hold warped PCBs flat but you can still pull it off again. Reasonably cheap too.

did not come with any means for clearing chips

I have a small vacuum cleaner (a DeWalt DC500 is ideal) hooked to some tubing pointing at the business bit. Catches most stuff. But for PCBs, I now just smear 3-in-1 oil over the surface of the PCB. That has two effects: firstly it lubricates the cutter so that lasts a lot longer and cuts better, and it stops the copper/board dust from going anywhere. Doesn't spray either (even at 20krpm).

end-stop switches is in my opinion a must, especially since one can use these to home the machine to a good position

I have a larger mill with end stops, and use those to zero the axis before any job. however, for PCBs the actual origin is very flexible (so long as it all fits on the PCB it doesn't really matter where on the blank it is), so I manually position the bit over the lower left corner of the blank and call that X0,Y0. The limits of the job aren't a surprise (I can see them in Mach3, and can manually move the mill  head to each corner to confirm it's going to fit), and given the flexible origin I've never missed having limit switches. IMO, you don't really need them on this for PCBs. YMMV

Bought a couple of 0.2mm endmills and broke them

Yeah, did exactly the same

My solution is to splash (via Ebay) on a set of LPKF cutters. These are 4mil to 8mil on a taper (can't remember the angle, but it is steep). They are sharp and robust, and SSOP devices are not a problem.

The main issue I have with fine tracks (actually, fine gaps) is allowing for runout - there is an absolute minimum gap you can cut with a particular cutter, and trying to cut a thinner gap (like the exact cutter diameter) is doomed to failure. The trick is to find out what that minimum is and live with it, then you can cut any track width you like (subject to there being room for gap-track-gap).

[MrAl]

Hi,

Is this a machine that might be able to mill out small PC boards?
That sounds very interesting, i'd like to see if this works and how long it takes to make say a 4x6 inch board (say 100x150mm board) or even just a 2x3 inch board (50x75mm about).

Also, what is the selling price, approximately?

Thanks.

[EEVblog]

I take it from the comments in the video, the Pick N place machine isn't fully assembled yet. Make me wonder what hope most of us would have given David knows about CAD machines.

Finding time has been impossible. David2 has had exams, only starting again this week. And well, I have a 1 months old...

[mux]

Compared to something like an LPKF, this seems considerably flimsier. I wouldn't be surprised if it gets out of alignment or runs out after one or two PCB milling attempts. Not sure why Dave is acting so excited by it; is it just that he got this for free? I mean, it's an awesome gift, don't get me wrong, but I know what will happen with it:

It will be used three or four times, then you get so sick of repairing it, babysitting the damn thing and going to and from the bunker that you just go back to ordering much better quality, cheaper PCBs from china

The 2-3 years I've had the pleasure of working with prototyping mills/routers, it went exactly the same way. They are fine as long as they work, but they have a myriad of failure modes and little things to adjust. You're bound to spend at least an hour of babysitting/setup/repair/ordering new parts/etc per part you're milling.

One thing that I did find 3-axis proto cutters extremely useful for is mechanical parts. I've seen people make things that are just not possible (with enough structural strength) with a 3D printer; think little cogs, anti-backlash nuts, spring holders, pretty faceplates. Another great application for routers is vinyl and polycarbonate cutting for making housings, logos and stick-on text stickers.

[DaveW]

Compared to something like an LPKF, this seems considerably flimsier. I wouldn't be surprised if it gets out of alignment or runs out after one or two PCB milling attempts. Not sure why Dave is acting so excited by it; is it just that he got this for free? I mean, it's an awesome gift, don't get me wrong, but I know what will happen with it:

It will be used three or four times, then you get so sick of repairing it, babysitting the damn thing and going to and from the bunker that you just go back to ordering much better quality, cheaper PCBs from china

The 2-3 years I've had the pleasure of working with prototyping mills/routers, it went exactly the same way. They are fine as long as they work, but they have a myriad of failure modes and little things to adjust. You're bound to spend at least an hour of babysitting/setup/repair/ordering new parts/etc per part you're milling.

One thing that I did find 3-axis proto cutters extremely useful for is mechanical parts. I've seen people make things that are just not possible (with enough structural strength) with a 3D printer; think little cogs, anti-backlash nuts, spring holders, pretty faceplates. Another great application for routers is vinyl and polycarbonate cutting for making housings, logos and stick-on text stickers.

I've got a Shapeoko, which is the older, flimsier version of this. Although it struggles with aluminium it'll cut wood all day, and milling PCBs isn't exactly heavy duty work. After the initial fiddling with belts it really isn't too bad, got around 0.3mm accuracy generally.
It's useful during the initial stages of prototyping when you need to make changes quickly, but as you say, a Chinese PCB will likely be far better than any homemade one. Mostly use mine for producing enclosures and jigs; and only takes a few minutes of setup before a part is being made.

[VingTor]

@dunkemhigh:
Thanks for your inputs. I will try some of it next time I'm milling.

Could maybe be fun if some of us how have milling machines milled the same project and compared the results. Could compare stuff like CAM processing time (entire process), machine setup time, actual milling time, and the final result (high res foto). I guess it should be something that is not too easy, but also within capabilities of most machines in this case. Just an idea.

Edit: finally one should compare total time spent compared compared to cost of cheap chine PCBs. Was it wort using all this time making the PCB (not considering how fun it was )

[PlainName]

Was it wort using all this time making the PCB

Unlikely. My last PCB, a week or two ago, was a simple RS232-TTL converter with DC-DC converter, LED, links. Not quite single sided, but with only two top layer tracks I made it as double sided and used wire for the top (making sure the tracks were short and straight to make that bit simple). Simple, but it would be a drag to wire up on stripboard.

I made it at 1am on Sunday morning, having decided I needed it during Saturday, and then used it in anger Sunday afternoon. There is no way any Chinese PCB house can come close to that, not any UK place come to that. In such a case, money is an irrelevance. It also didn't cost me anything notionally, since I had all the parts already and the PCB was an offcut from something else.

OTOH, I am also prototyping something for which this board is a dev tool. That one, I don't even think about a DIY approach - it will go to itead, and the only question is how much to pay for shipping.

I think the problem with discussing this stuff is that people take an either/or stance. You either send everything to wherever, or you make everything yourself. That's stupid, and in reality you use the best tool for that particular job. Suppose your PCB mill turns out to be useless for SMT parts - it is not the end of the world (though you might regret spending the dosh). Instead, you know the limits of the tool and then use it appropriately. If you're doing a board like the one I did, you might choose to use PTH parts instead of SMT. There is no shame in PTH! If you need something the tool can't manage, use a different tool.

As it happens, I chose to use a SSOP MAX3232 because that's what I had to hand, and then PTH caps and resistors because that gives me a better chance of routing single-sided. Thus I end up with parts on both sides of the board, and mix of SMT and PTH, and a couple of wire links. I don't care - it perfectly does the job it was meant to do and I had it within a few hours rather than a few weeks.

It's worth bearing in mind, though, that I would never have purchased a mill to make this board. I could do it only because the mill already existed, and I think that's a big hurdle to get over. This is one of those things where there is no single compelling reason or application, so you could go forever without actually getting round to acquiring one. But if you have one to hand, there are many things where it turns out to be useful (I got a panel manufacturing job - pays very little but helps with IR35 - only because I already had a mill and could do it).

I think this is why Dave getting one is quite important, assuming that it's not totally pants. It is like a hammer wielder acquiring a screwdriver - all of a sudden you can fit frames to windows without breaking the glass, and things you previously never thought to do become perfectly feasible. I predict that there will be a sea change in how these are perceived once Dave finds a real use for one

[sain]

I too have a mill and I've been using it to make boards recently.

Its pretty easy to set up. I used a bit of scrap wood which I routed a pocket in as described above, and I'm using carpet tape to hold the boards down.  The scrap wood was bolted to the bed prior to the pocket being routed in it.

The process is simple.  stick some carpet tape to your boards back edge and then put the board in the pocket, making sure your flush against at least two of the square edges in the pocket (or have plenty of room on the board).  Next put the engraving bit in the spindle and touch off on the top surface of the board to set the depth to zero.  I've got a simple LED that I connect to the spindle and the board.  When it lights up your touching.  Turn the spindle on, route the board. 

If its a through hole, or has mounting holes I have a tool change step to swap to a drill bit after the engraving is done. I normally use a 0.8mm bit for that step, which is fine for most components.

I find its great for prototyping.  If I want to prototype on a board I can churn one out fairly quickly.  If I'm not bothering to put PCB laquer on it then about half an hour turn around time.  Thinking about it there isn't really any reason why the boards couldn't be pre-laquered, but I imagine you might get oxidisation around the edges of the tracks.  Anybody got any thoughts on that subject?

I'm curious what software you guys are using to generate the g-code?  I'm currently using the pcb-gcode plugin for Eagle, but I'm thinking about moving away from Eagle and I'd appreciate some suggestions.

 

[free_electron]

If the software don't work, Dave is able to use LinuxCNC.

see, to me this is unacceptable. if i were to buy this product i'd expect it to work.

this attitude of , oh well if the software for the machine is crap you can use xyz, or abc.  This is a common ailment that plagues all these open source / open hardware based products.
they are all contraptions pulled together of various bits and pieces and NO BODY takes responsability for the end user experience. it is expected to 'tinker'.

there is a differenc ebetween people that want to "make machines" and "make things with machines".

the same misery exists with 3d printers.

i'll be keeping an eye on this 3d mill. it looks well made. i am hoping the software will live up to it. my ideal workflow would be : solidworks -> milling software in one click.

[rs20]

<Warning: I'm sleepy, this may make very little sense. Consider it a stream of consciousness...>

I agree with everything free_electron says, but it's interesting to consider that everyone seems to at least slightly assume that each CNC mill should come with its own dedicated software complete with GUI, especially one that understands PCBs (like, accepting Gerbers/Excellon files, or something).

This is especially interesting when contrasted against normal printers (you know, 2D laser & inkjet printers). Just imagine if you had to use word processing or spreadsheet software written by HP or Epson. Not even HP or Epson could make something that even comes close to Microsoft Word, and here we are expecting random CNC mill startups to do better.

I get that there's a lot more complexity & fine-tuning to the G-code generation that CNC mill software does than whatever a 2D printer driver does, but that's doesn't mean that it can't live in a binary blob with a standardized API (let's call it a "printer driver") that the end user interacts with via real PCB software. It's perfectly possible for CircuitMaker or Eagle to have a "Mill..." option on the file menu(/ribbon!) next to the "Print...", but who's going to own and push that project? Altium/Eagle has no reason to and the CNC mill startups don't have the ability to.

<So yeah, I just suggested something and then pointed out how pie-in-the-sky that idea is. Sorry.>

So to be more constructive, rather than just berate the current state of CNC mill software, tell me: What does your ideal CNC mill software ecosystem look like? What should the CNC mill startup-provided software do, what input file formats does it take?

[PlainName]

everyone seems to at least slightly assume that each CNC mill should come with its own dedicated software complete with

Indeed, and I think some do. But in my case I knew what software I wanted to use to drive it, so it was a case of buying some hardware to suit. Kind of like you buy a PC to run Windows or Linux (or by a Mac isntead). Anyone with sense will wipe the pre-install on the harddisk and set up from scratch anyway

It's perfectly possible for CircuitMaker or Eagle to have a "Mill..." option

TurboCAD basically has this for 2.5D milling, as an extra cost. I guess it works (I looked at a demo) but I always end up wanting to use my own preferences. Kind of like when you write code you use your favourite editor, not the rubbish supplied with most chip specific IDEs.

What does your ideal CNC mill software ecosystem look like?

This works for me, though if I were starting again I might review some of it:

MACH3 for mill control. I use it for 2.5D milling of physical stuff on a larger mill, and PCB stuff on the Sable.

CopperCAM to convert gerber files to isolation routed g-code. It is... idiosyncratic, but between banging ones head on the keyboard it does what it's meant to to, with some nice touches

Labcenter Proteus to generate the circuit and ultimately the gerber, but anything that outputs gerber is fine.

Additionally, not strictly required but something I always use before letting MACH3 loose on anything, CutView Mill. This is a mill emulator, and when something goes wrong you write off virtual bits and fixings rather than the expensive physical ones. It is most use for normal milling (rapid moves through a clamp are a good catch), but shows the results of a PCB isolation job perfectly - can catch silly things like mirrored sides.

If I were to buy or build a new mill, I would assume I'd be using the exact same software, so whatever it is would need to be controllable my MACH3. Fortunately, pretty much everything is once you're talking to the motor controllers.

[PlainName]

I've got a simple LED that I connect to the spindle and the board.  When it lights up your touching.

I used to do that, but then switched (ho ho) to using a MACH3 script that does it for me. All I do now is position the bit within 5mm of the surface to zero on, hit the button and MACH3 advances the bit until contact, zeros the axis and backs off. There's a bit more to it than that, but that's essentially how it works.

There's another script for zeroing on holes or other circular objects: it finds the top/bottom and left/right mid-points, which is the centre. Can't recall the last time I had to manually find the touch point of anything!

[miguelvp]

I think someone mentioned this already but on the "Inventables X-Carve LIVE Build Part 2" at 2:09:08 forward, Dave 2.0 dropped a mailbag item behind the big box on the floor, he tried to catch it but couldn't and went on after he got the paper to explain the non concentric nuts.

https://youtu.be/nQH4I_p7-MI?t=7748

Tried to link the video at the right spot, but if that didn't work, look at 2:09:08

Maybe he remembered later on and picked it up, but maybe it's still down there.

[station240]

I love all the jokes about missing tools, and dave's "deep pile" filing system.