Desktop CNC

[blueskull]

Finally the time has come! I decided to get a desktop CNC for 3 main purposes:

1. Machining small integrated heat spreaders and shielding covers for tiny modules (soft metal only, less than 1/4'', mostly less than 1/8'', mostly copper, Pb free brass, 6061 or 7075).
2. Milling PCB for quick and dirty prototypes.
3. Cutting large featured solder stencils, such as 0805+ SMD stencils.

Now I know there are a lot of options, but particularly I am looking for a desktop one that's residential area friendly. Those messy, dirty, loud shop machines are excluded automatically.
Also, I'm not a machinist, and I have no idea how to design G code properly. I would like an easy to use one that's designed for EEs, not MEs.
I need one with either bundled or open source CAM that runs properly on Windows, as I'm not going to spend $3850 on SolidWorks and more on Catia. I use SpaceClaim and FreeCAD.
Precision is crucial, <4 mils (0.1mm) of error is accepted, and anything higher will be inadequate.
I need a neat package that looks nice and works nice. I don't want to waste my time maintaining a stupid machine that vibrates or even crashes every time I use it.
Our lab has a broken MakerBot 2, and I know exactly how fucked I feel whenever I use that machine.

NO CLOUD BASED BS LIKE CARVEY OR GLOWFORGE!

Any suggestions? My budget is up to $3500, but the lower, the better. My total backyard fab budget is $5000 and I want to have extra money for a LitePlacer and some other semi-automated SMT tools.

[Koen]

easy to use [...] Precision is crucial, <4 mils (0.1mm) of error is accepted, and anything higher will be inadequate. [...] I don't want to waste my time [...] $3500, but the lower, the better.

Nope, it won't fit your requirements.

Tooling, learning, setup is at least twice the price of the machine itself. < 0.1mm is a feat in itself for cheap-light desktop machines. You will waste your time on it, there's plenty to learn and plenty of mistakes to be made before being up-to-speed. Honestly, find a fabricator around you.

[Koen]

As another EEVBlog user signature would put it : "there's lies, damn lies, statistics and cheap milling machines specifications". There's not a single hope this accuracy holds true.

It's not to keep you off of this magical world that is CNC and stuff, just trying to save you time. Ideally, find a fabricator close to you and ask him for parts from time to time. Let him worry about this.

[Koen]

Lies. Yes, one day in their control room at 25°C, they set to measure it in free air while milling nothing at all and they stumbled upon impressive numbers. Then someone hit it inadvertently with the elbow and it was out of spec. Or it was colder the next day and out of spec.

Here's an example. It's mine. It's three metric tons. It's quite big for a custom. It has plenty of bell and whistles. And if I need 1.5mil on a part, I first spend two hours re-calibrating it. [Removed as you've seen it.]

Anyway, feel free to ask questions. I'm not here to crush your hopes. But hey, if the machine can be twisted by pushing it with a finger, imagine what milling does to it.

Also keep in mind that dissatisfied owners of lightweight milling machines will rarely complain about them online, they'll sell them first.

[Mechatrommer]

Any suggestions? My budget is up to $3500, but the lower, the better.

nope... but the higher the more expensive, the better. asking for lower price is asking for every problems you mentioned above.

usually people will just do preliminary test review (rough and not real life test) to gain "likes" click and then you'll heard from them no more... if there is good compliance to specification you should already hear it by now or... the price will rose exponentially already...


for the true 0.1mm spec seekers and the keen eye, they should be already getting the answer by just watching the above "preliminary" video review... implicitly... dont be too optimistic... another option is go to the big factory who has the machine in the business and you should be a able to get a good advice and a ballpark price point from them.. fwiw...

editing reason: formatting

[sn4k3]

Milling PCB with decent quality and equal results for multiple clones is very hard and very expensive, require a lot of investment and care.
There are better solutions for that but expensive, better order online or first prototype yourself with chemicals at your home

[khs]

Accuracy is the one, stability the other.

During milling there are alway forces moving your cutter away.

It's like a powersupply with a high output resistance. As long the output is open all ist fine. But if you connect any load the voltage drops or you get oscillations.

Maybe I'm wrong but from my feeling the mounting of the X-Z axis looks not to be very robust.

The gray material looks to be plastic. Take a look at the bottom of the left side, then you can  see the screw holding the vertical X-Z axis holder.

This small piece of metal must hold all forces of the X-Z axis holder with a very disadvantageous lever.

So for milling pcb's it may work, but for milling something like aluminum I would expect heavy vibrations.

[rx8pilot]

The specs presented by Carbide 3D are ridiculously overstated - while not technically lying.

[I have been a machinist for about 12 years or so, now doing mainly electronics - power electronics and high-speed digital]

The specs they print are seemingly based on the stepper motors and pitch of the lead screw. When designing a motion platform, these numbers are interesting as they represent (to some extent) the best case scenario. In reality, the real world errors injected into the system guarantee that you will not achieve anything close to that resolution. The real test on a machine is to measure the actual movements in 3D, axis reversals, acceleration/deceleration, interpolation,etc. Usually this is done with a ball bar rig and it exposes the all the cumulative errors in the system. This is where the rubber meets the road.

Major red flags:
Lead screw: They are shit for this type of work. Ball screws are the only practical way to drive the linear motion.
Linear bearings: They are also shit for this type of machine as they are never rigid, the bearings will wobble. Linear guides on a ground/shimmed surface are the only practical way to support the linear elements.
Wood frame: Seriously, I cannot even comment.
10k RPM is slow for this type of work. Not unusable, but slow.

@blueskull think of much effort you put into a design from the block diagram to the schematic, to the PCB layout, to the final validation. The block diagram and first conceptual schematic are simple diagrams that show a concept. After you start looking into the implementation of the control system, parasitics, step response, stability, efficiency, etc, etc.....it gets real complicated real quick. The generic MOSFET symbol in the schematic can become a week-long search for the right balance of specs. The PCB layout becomes a maze of parasitic inductance and capacitance that works against you.

CNC is no different, the block diagram looks easy while reality is a giant challenge. Even (and perhaps especially) small parts like heat spreaders and PCB's are not a trivial challenge. For you stated goals, I have a few comments:

PCB Milling: The challenge is largely in how you hold and locate the PCB blank. It MUST be flat-flat-flat to get a reasonable result. This is not easy and a machine like the Carbide 3D has no vacuum system or another way to keep thin parts flat. Tiny variations in flatness have a huge impact on what you can accomplish. If your goal is to make high-current half bridges with big MOSFETS - no problem. As the features get smaller, the challenge increases exponentially. I have machined one PCB design and I did it on a $150k 5 axis machine. The challenge was entirely focused on keeping the PCB flat. Ultimately, using a vacuum plate (another $3k) and a fixture to align it for both sides and drill it (5-6hrs to work that out). Not easy even with serious tools.

HEAT SPREADERS: This is probably not a huge challenge for something like a Carbide 3D. Work holding is important as always and your tolerances will not be very tight. The programming of the G-code can be trickier than it may seem. I designed and machined a heat spreader yesterday and it took the majority of the day. I have an $80k Haas Mini Mil with another $40k in tools/software/work holding. Not exactly the challenge of my life, but not trivial at all. The tiny chips you will generate will get all over the motion system and any sensitive electronics. be prepared to clean constantly or fabricate some shields.

STENCILS: This may be the hardest one of the three. Milling very thin material is hard - not for the machine but for the machinist. How do you hold it down while a helical tool is trying to pull it up? This is why stencils are cut with non-contact lasers. Mechanically cutting a stencil with any mill or any quality will be hard to do regardless of the size of the designed features.

If you get a low-cost desktop CNC - be prepared to spend many hours fiddling with a steep learning curve and a fiddly machine. Many times it is very difficult to tell if your problem is a limitation of the machine, your skills, your tools, or the process in general. What appears to be a quick PCB, heat spreader, etc....can go on for days.

Yesterday's project.....

[khs]

If there are $150K not around at the moment maybe a cheap alternative (about €0.5K - in Germany).

https://www.proxxon.com/en/micromot/27112.php

For small parts maybe not bad.

[rx8pilot]

To ensure there is no confusion on my comment....I am not saying you need a $150k machine to mill a PCB. The point I was making is that it was what I had available at that moment. It was clearly overkill in terms of technical capability and it was still hard to get a good result. The end thought is that for PCB milling, the hardest thing to do is holding the PCB flat. No matter what machine you have, that is a universal challenge. There are some that have attempted to do a 3D topo map of the PCB and adjust the z level to compensate. That technique is not only difficult in concept, the reality is that the PCB will vibrate if not supported well.

A plate that is very flat relative to the X/Y plane is crucial and the PCB needs to be supported across the whole surface. Only a vacuum plate or adhesive can do that (or at least the top two in terms of practical reality). The vacuum plate is very easy to release, adhesives not so much.

[ChunkyPastaSauce]

Tormach 440 will probably do it. Basic machine + controller is little under 6k. Its over your budget a bit but might be worth considering and can handle a lot more other smaller projects than just pcb engraving. Has 10k spindle. For work holding, you can also use thin double sided tape or just make a cheap vacuum table (it's just pcb engraving, doesn't need to be super accurate with ground surfaces nor the ability to rigidly handle large loads).  Also I know you said no cloud....but maybe also take a look at fusion 360... the CAM processor in that would have been mega expensive a few years ago... its free for hobby and businesses making less than 100k...and its integrated with a decent CAD package.

[rx8pilot]

+1

The Tormach 440 is perfect for the stated goals and more. A bit over budget as said, but that is life in the real world.

Don't forget, shipping, tools, tool holders, vises, metrology.... which you need no matter what machine is chosen.

Short and misplld from my mobile......

[rx8pilot]

The only reason I don't own a Tormach is because they cannot do rigid tapping..... a small issue in the context of this thread.

Short and misplld from my mobile......

[Magnum]

I wanted a desktop CNC for the same things, too. Milling PCBs and working with plastics and aluminium. I bought a Stepcraft 420 desktop unit. The unit is ok for plastics, even aluminium if you have a lot of time. But it uses lead screws, and after some hours of use you are above 0.1mm precision. So from my experience you'll need the following:

PCBs: A good desktop unit will do, as the milling forces are very low. To get the precision you need ball screws and profile linear rails or supported linear guides.
To hold the PCB down use a vacuum table and vacuum pump. If the PCBs are not perfectly flat, there is software available to compensate that. Works great.

Vacuum table:
https://www.vakuumtisch.de/index.php?page=categorie&cat=124

Vacuum pump:
https://www.ebay.de/itm/vidaXL-Alu-Vakuumpumpe-100-L-min-2-stufig-Unterdruckpumpe-Pumpe-Klimaanlagen/401471059658?_trkparms=aid%3D555019%26algo%3DPL.BANDIT%26ao%3D1%26asc%3D20161121134545%26meid%3D9ab812cebfe647cd84bc4780b3c7a76d%26pid%3D100505%26rk%3D1%26rkt%3D1%26&_trksid=p2045573.c100505.m3226

Linear rails:
https://www.dold-mechatronik.de/Linearwagen-ARC-20-FN-Flanschmodell

Linear guides:
https://www.dold-mechatronik.de/SET-4x-Linearlager-TBR20UU-2x-Supported-Rail-TBS20-1200mm

Metal works: Buy a heavy unit (>100kg). You won't get happy with a lighter unit. I bought this one to build my own desktop CNC:
https://www.paulimot.de/fraesmaschinen/mit-vario-antrieb/193/bohr/fraesmaschine-f207-v-mit-frequenzgesteuertem-deutschen-motor?c=8
It is a Chinese unit, modified from the importer for better quality.

And expect to build a housing around the CNC, as it makes some noise and quite some dirt.

Sorry for the german links, just that you know what to look for.

[rx8pilot]

PCNC440 was my first idea, but the price deterred me a lot. It's not just $5000. I need a full enclosure, which is stupidly priced at $1500, then the price goes $6500, and that's just the bare minimum to get it to work in a residential area. Throwing in some other very basic options like coolant system, bit set and chip tray, plus tax and shipping, it quickly becomes more than $8000.

And it does not stop there....the expenses never seem to stop.

I would like to keep my wallet happy, though I don't mind slower machining speed, anyway I only need some very few proto boards, less than 20 pcs per year.

For the aluminum machining, I may need to use it as a productive tool, but 0.2mm accuracy is all what I need.

If it can't do stencil, then it's fine. I have access to a laser cutter that can do polyimide.

Accuracy is the expensive part - even if you don't need raw power to take big cuts or machine tough materials. The accuracy of .2mm will only get you very course features, but the biggest problem is the Z accuracy has to be FAR better than that to mill PCB's. I understand the financial restriction for sure, but with that in mind it would be a bummer to spend $3500 at get very little productivity from that.

After researching all the available desktop CNC's for my home shop, I simply decided that if I cannot get a decent (Tormach or better), it was not worth it at all. My very first milling machine (many years ago now) I prioritized price and figured I could work around it's limitations. I was wrong. it became a fight on two fronts - my own learning curve combined with the limitations of the machine itself. In the end I got next to nothing out of it, spent a ton of time, and sold it. After more than a decade of professional experience, I can say that I still would not be making anything on that original machine. It just sucked to bad regardless of skill level.

[Dubbie]

I built a machine from scratch to do PCBs. It took me a couple of years and cost a lot in parts. I have never cut a single usable pcb with it for the reasons that rx8pilot outlines. However I cut small steel and aluminium parts regularly. For small parts (under 60mm) I can usually hit about 0.02mm accuracy. However the machine has high quality linear rails and high quality ground ballscrews on all axis. It is messy and tooling is expensive. To be honest, I think you should find a local shop that will cut you parts. It will take a lot of parts before you are behind doing it yourself financially wise.

[rx8pilot]

Curious if Shariar(sp?) from the Signal Path has seen any success. I am guessing that he would be making various RF structures which is a great application for CNC milling. It seemed that he was struggling in his video.

Maybe I should design and make a little vacuum table for all those hoping to CNC their proto boards. After spending the $$ on a machine, it may be worth it to get the $500 vac table that drastically improves your odds of success.

[rx8pilot]

Where do you source the FR408? I would like to have some on hand for some experiments. Guessing if you are using that material, you are probably using eGaN FET's with some outrageous edge rates.

FYI: I am not trying to talk you out of anything and would love to see some success with a system like this. My best guess is that you are the type that will not give up when the going gets tough. If that is true, it is a good quality to have when machining things that are hard to machine well.

 

[rx8pilot]

I had no idea OSH Park even had FR408.

The demo video looks pretty good, but the machine looks devastatingly weak to me. Clearly they are making things with it, but I wonder how many tried it took. I saw a video where an end user seemed to have some success, but only after leveling the table which was far from flat on arrival. The Carbide looks better.

Wonder what happened to the Cirqoid product. It looked way better mechanically.

Also, Prometheus looks interesting.
https://www.zippyrobotics.com/

[mrpackethead]

Dont' forget the cost of tools, clamps, measuring equipment..  You'd be suprized how fast that adds up to at least as much as the machinery.. 

Btw,  a router on a gantry is not a Mill, and vice versa.     I have both a CNC mill and and cnc router, and they both have their place.    I make die sets in aluminum for injection moudling on my mill,  and i cut out wood and plastic on my router.

[rx8pilot]

I just sent the Zippy Robotics guy an email, and asking him whether the machine can also mill aluminum. Fingers crossed. Thank you for this recommendation.

Wow, they respond fast! Unfortunately the answer is negative. It's not designed to mill solid metal and is not stiff enough to eat into metal. If I insist, it will take many, many passes of very thin depth to do it.

Also, it doesn't have such firmware to cut solid metal, and that feature will be added after stencil milling feature, so I guess that won't happen very soon.

I think you will find this to be true with most of them. It really is a different task.

[Koen]

Example above isn't so much milling soft metals as it is grinding soft metals. You don't make chips, you make powder. It's hard on the endmills, they have to be replaced often. Also, heat won't be escaping by the chips as should be and thus, both the endmill and the piece will heat up plenty, leading to problems.

If I were you, I'd put a classified looking for someone owning a nice mill in a similar semi-pro setting. And ask them a price for machine time. You provide a nearly-done file and the material stock. They mill with a little profit. Maybe a "FabLab" would do. It's not very developed here so I wouldn't know.

Or you know, China. Nowadays, I send a step file, pay ~300 USD, wait a week and receive 4 samples by express courier. It's somewhat stupid of me I agree but I don't have much time anymore and it's much more convenient than sourcing the material myself, sourcing the endmills, any specialized tooling, preparing the machine, making a mistake or another and so-on. The mill now only serves big/heavy/expensive projects like making other machines.

[Magnum]

Here is a (german again) machine that is good for PCBs and can handle aluminium:
https://www.sorotec.de/shop/Portalfraesmaschine-CL-Line-0403.html
It has ball screws, linear rails and weights around 50kg for 300x400mm travel. The price is without electronics and stepper motors, that is another 1.500 EUR. So altogether EUR 3.800 with VAT, EUR 3.200 without VAT.
Here some videos of building these machines. You can see how rigid the construction is:



And here cutting aluminium:


Even iron:


With such a machine you have the accuracy you need. Something similar should be available in the US, just check the construction first.

[mtdoc]

I need a band saw to cut long stock into pieces, but I can also just order pre-cut from local hardware. My local hardware store offers free cutting and bending as long as the pattern is very simple.

For an amazingly capable, inexpensive and portable band saw solution I can highly recommend the following.

1. Bauer handheld 10 amp variable speed bandsaw. Yes , it's from Harbor Freight.  But this is one of their hidden gems. Read the reviews. With their ubiquitous 20% off coupons you can get it for $80. It comes in a nice case which is perfect for your portability concerns.

2. Swag portaband table. These things are awesome. They are very well made and can be broken down to a small package.

Optional - Add a good miter gauge (I am using this one) for precise cuts and a cheap foot switch.

I've been using this setup to cut stock and it works very well.  It's also nice to have the option of using the saw hand held for removing stock from large pieces (or from the junkyard, etc).

[rx8pilot]

For an amazingly capable, inexpensive and portable band saw solution I can highly recommend the following.

That looks AWESOME for my tiny shop.....thanks!