Desk CNC 3020 (or similar) accuracy and practicality for SMD PCB prototyping?

[jeremy]

Machines like this can be OK for very small boards, but maintaining an accurate cut depth over a larger, possibly slightly warped board is likely to be impractical. You really need something that can sense and adapt to surface height

Some high-accuracy, PCB-only machines I have seen use a combination of vacuum hold down, and a "stiff skirt" around the spindle (made of delrin/HDPE?) on a strong spring to push the board flat around the cutting area as the spindle plunges. I'm sure there is a proper name for it, but I don't know what it is

[G0HZU]

Can they not adopt the foot method?

Why? The height probe is much easier to make and you typically want one anyway to set zero Z.

What if you want to mill a really thin and bendy PCB material? eg 0.020" or 0.010" or even 0.004" thick like paper (including the 2 copper layers?)

This is where the foot method really performs well. Sometimes you have to set the pressure carefully. eg for cheesy/bendy stuff like PTFE because the foot can compress the material if the pressure is too high.

[mikeselectricstuff]

Can they not adopt the foot method?

Why? The height probe is much easier to make and you typically want one anyway to set zero Z.

What if you want to mill a really thin and bendy PCB material? eg 0.020" or 0.010" or even 0.004" thick like paper (including the 2 copper layers?)

stick it down to a more rigid backing board. or use vacuum hold-down

[G0HZU]

I'm not sure I'd want to use glue for a double sided board but I guess it's an option. It does rely on the table + backing material being very flat though.

I've not used a machine with the vacuum hold down feature so I don't know how effective it is. Presumably it requires some air permeable backing material below the PCB material to allow it all to work?

I can see the advantage of the vacuum with respect to preventing bowing up of the board material. It can be a real pain keeping the material flat because the ride height of the tool when lifted has to be higher than the PCB bowing in order to prevent fouling. This is one limitation of my old T-Tech machine.

But I have milled double sided flexi PCBs where the PCB laminate is only about 0.0025" thick with it.

[mazurov]

I'm not sure I'd want to use glue for a double sided board but I guess it's an option. It does rely on the table + backing material being very flat though.

There are adhesives that can be released by gentle heating ( ~160F, that's below boiling for water asl ). To make a mounting surface flat you mill it flat prior to mounting the first board to it. That's the beauty of general purpose CNC mills, they can be used for all kind of things, like making jigs for PCB milling (including vacuum tables). They won't work as well as specialized PCB-only mills but they are often good enough - I can mill 10 mil clearance consistently and when I want smaller clearance I also want a soldermask and PTH so I just order the board.

[jeremy]

The vacuum systems are quite simple really, just a grid of holes in a flat piece of material hooked up to a vacuum on the underside. You just tape over the holes you aren't using, or I guess you could use a silicone sheet or something?

[Richard Crowley]

The vacuum systems are quite simple really, just a grid of holes in a flat piece of material hooked up to a vacuum on the underside. You just tape over the holes you aren't using, or I guess you could use a silicone sheet or something?

Indeed, constructing a vacuum hold-down table is very nearly trivial. The CNC machine itself can be used to create the surface.
Here is a video of a PDJ PilotPro creating its own vacuum table.  http://youtu.be/172rLt0oxog
I am finishing a 26x26 inch version of this machine and I will likely make myself a vacuum table for it.

And you can use something as cheap/simple/prosaic as a domestic vacuum cleaner (BrEnglish: "Hoover") to provide the vacuum.
And you can cover the unused surface with scraps of plastic or even plain paper.

[mazurov]

The vacuum systems are quite simple really, just a grid of holes in a flat piece of material hooked up to a vacuum on the underside. You just tape over the holes you aren't using, or I guess you could use a silicone sheet or something?

The difficulty is the setup. You mount a vacuum table (or any other jig for this matter) on the machine and start wondering if its surface is orthogonal to the spindle. The fastest way to make sure is to mill the top surface flat; however, since vacuum table is expensive and your machine may not necessarily be good enough to mill aluminum you spend time with indicator in the spindle measuring and shimming.

Another way is to clamp a 1" thick piece of HDPE and make the vacuum table out of it. Everything can be machined except the horizontal hole which you would drill, tap and screw in the vacuum port. I've never done this myself but I heard that drilling/milling into the vacuum holes ( or grooves ) is not a good idea - weaker vacuum, abrasive swarf going into the pump, etc., so you spend time designing your table for the board you're making.

Both methods will pay off if you're milling a batch since changing stock is very fast. However, when I mill a one-off I can do it many times faster with carpet tape. But again, I'm no machinist. I'm not a good enough circuit designer either so I have to build my circuits to make sure they work otherwise I wouldn't bother.

[G0HZU]

Does the vacuum table add any significant noise to the overall system noise?

My old T-Tech system is very noisy already as it uses an old school Nilfisk GM80 feeding to a basic nozzle at the tool area. I run with the GM80 pushed through a window to try to minimise recirculated dust particles and also to keep the noise down a bit. But the newer LPKF machines seem much quieter so I wonder if the vacuum table noise is noticeable on the newer systems.

I'm probably 10 years out of date on milling technology. My T-Tech machines are both from the 1990s and the LPKF machine we have at work must be at least 10 years old.

[Things]

I've just used doublesided tape in the past - probably not practical if you're making a bunch of them, but for occasional hobby work it works fine.

You'll want to fit a panel of wood or plastic over the actual metal bed to act as a work surface - basically set up a program and get the mill to face off the work surface, then you know it's completely flat. It'll also save you if you stuff up the programming and try drive the tool bit into the metal bed!

[ozwolf]

The vacuum systems are quite simple really, just a grid of holes in a flat piece of material hooked up to a vacuum on the underside. You just tape over the holes you aren't using, or I guess you could use a silicone sheet or something?

Indeed, constructing a vacuum hold-down table is very nearly trivial. The CNC machine itself can be used to create the surface.
Here is a video of a PDJ PilotPro creating its own vacuum table.  http://youtu.be/172rLt0oxog
I am finishing a 26x26 inch version of this machine and I will likely make myself a vacuum table for it.

And you can use something as cheap/simple/prosaic as a domestic vacuum cleaner (BrEnglish: "Hoover") to provide the vacuum.
And you can cover the unused surface with scraps of plastic or even plain paper.

Just a consideration with the domestic type vacuum... Allow some bypass air as this is also the cooling air for the fan motor.  If you cover all of the holes, the motor will soon overheat and shutdown/burnout.

[ivaylo]

And what's with the parallel interface of all of these brand new machines? The last computer I had with such a thing was like 20 years ago. Yes, I know, get an adapter, DYI USB, etc, etc, but there must be a reason for selling brand new PC controlled hardware with an outdated interface. Thanks for enlightening me...

[mikeselectricstuff]

And what's with the parallel interface of all of these brand new machines? The last computer I had with such a thing was like 20 years ago. Yes, I know, get an adapter, DYI USB, etc, etc, but there must be a reason for selling brand new PC controlled hardware with an outdated interface. Thanks for enlightening me...

Because the Chinese are no good at designing anything - they just run old designs for as long as they can get away with it.
Even if they did design a USB based one it would probably suck ass.
Unfortunately there is no accepted standard in the CNC world for a USB interface, and Mach3, with its crude bit-bashing, works well enough most of the time, so the Chinese are happy to ship cracked versions with their kit.   
Doing motion control over USB is a somewhat nontrivial, but perfectly solvable problem, and for reasons I don't quite understand, nobody yet seems to have done a really decent job of an open source motion controller.

Seems to me that Mach3 is the Eagle (PCB) of the CNC world - they got in early with a free version that was OK for a lot of people, and became a de-facto standard because it was free and there wasn't anything better at the time.   

 

[ivaylo]

Thanks, Mike...

[Richard Crowley]

Unfortunately there is no accepted standard in the CNC world for a USB interface...

Aren't there Arduino (or clone)-based controllers that take G-Code on a SD card and drive the motors as a stand-alone unit?  (i.e. Rep-Rap)  Presumably open-source.

They are now promotoing USB-to-parallel external adapters to keep the old design running on modern computers.

Seems to me that Mach3 is the Eagle (PCB) of the CNC world - they got in early with a free version that was OK for a lot of people, and became a de-facto standard because it was free and there wasn't anything better at the time.

Rumor is that Newfangled Solutions is developing a completely new generation ("Mach4"?) and will raise the price an order of magnitude and drop Mach3 (US$175) Will an alternative/replacement emerge that will take up the market it for hobby-level users?  Mach3 seems remarkably fiddly and amateurish. I can't see how they have managed to sell it for THIS long.

[Things]

I've heard most USB->parallel interfaces don't provide all the signals required to drive CNC machines, just things like printers. Just something to watch out for.

They could probably make a USB interface OK - but don't even think about the software!

[bookaboo]

+1 to what Mike said, previously I've used LPKF and T-Tech milling machines and sometimes they work beautifully so you are can go from concept to PCB in a single day. However they often needed a lot of attention to get the right results, particularly on larger boards.  Even the slightest warp in your FR4 can destroy your results and leave you with multiple efforts. Broken mill bits are another pain, you sit listening to the machine drone (plus vacuum and air compressor) for  hours while you do other things only to realise the mill bit broke 10% of the way in. Bear in mind these were EU and USA built specifically for PCB work.

Perhaps things have improved in the 5-6 years since I gave up on milling PCBs but personally I think you are much better off just paying for one of the fast turnaround fabrication services. When you consider the cost of your own time, the machine costs, material and of course bits it probably works out cheaper to get them done externally. You also get the added benefit of soldermask and silkscreen. Unless you are in the situation where you constantly need to be turning around designs on 2-3 day lead times I don't really see the benefit of this method any more.

[dr.diesel]

Speaking of, whatever happened to that nylon cube, Kickstarter mill thing that made a few headlines a year or so ago?  I know it was small, had some hype, surprised nobody here ended up with one to show us?

[sacherjj]

I'm using a Probotix Fireball X90 setup for all kinds of general CNC prototyping with machining plastics and also PCB prototyping.  Luckily most of the quick prototypes are small enough in area that I mount as flat as I can and set height in the center.

I've really been happy with the quality of bits and service from Think and Tinker precise bits.

If I had the input capability for a touch probe for PCB CNC, I would do it immediately.  However, it means replacing the stock controller board that came with the Probotix kit.  It isn't causing enough pain right now to cause me to drop a couple hundred on changing the setup.

For setting Z height, I use a 0.250 brass rod.  Lower the cutting tip with a very slow Z jog.  Then I raise it until the bar just passes under the cutter.  Then I set Z at 0.250.  This gives very accurate and fast setup.

Now as far as vacuum hold down, I see two problems:

1 - Some PCB material is bowed enough that vacuum might not pull it flush.
2 - Drilling, Routing of larger holes, and edge cut out will all erode the surface such that hold down force is no longer present.

I am using CopperCAM to convert Gerbers and drills to G-code.  For the most part, I've been happy with it for non-slotted designs.  I've started using multiple holes to emulate slots where they are needed.  This is a little annoying, as I can't use the exact same file for real fab.  I make multiple pads over the existing slots and just ignore the slotted NCDrill file for CNC milling.

CopperCAM allows you to add drill locations for layer alignment.  If these are positioned so that board flit remains aligned, then you are set.  I use a T-slot table and have some threaded rod that goes into them.  I then turned down the 5/16" rod from the T-nut to a smaller diameter on the lathe and threaded (to 8-32 I think).  If I position my holes to line up with my T-slots, then I can pull the board down to a flat reference with small washer and hex nuts into this threaded rod.  By turning the rod through to T-nut to the base first, the position is fixed.   While CopperCAM allow up to 4 layers, I've had no problem indexing dual layer boards.

My next upgrade is to not use just scraps from the same board underneath, but machine a few slightly varying heights of plastic to tune the board flatness.  I'm looking for some flat head 5/16" nylon bolts to make counter sunk mounted sacrificial board that I then machine flat.  This with a common reference pattern on it to screw the PCB to might be the best solution.  But I've been to busy to tweak the setup lately.

Here is a picture of my setup (pardon the dust, I was milling a gift plaque for my wife out of Oak):



I included the good old DS1052 for size reference.

The guys at Think and Tinker are very knowledgeable on runout of common routers.  We have their precision collets for 0.125 and 0.250 bits.  This is required to machine with very small bits.  I did some prototypes of clips in spring steel using a 0.010" bit from them.  I had to manually drip fluid at the bit, but it worked great.  I used a router speed controller to get the speed down.  This has been replaced with a Super PID closed feedback speed control that lets me run a router down to 5k RPM without issues.  This lets me machine plastic with a 0.25" bit with no melting issues.

[sacherjj]

Perhaps things have improved in the 5-6 years since I gave up on milling PCBs but personally I think you are much better off just paying for one of the fast turnaround fabrication services. When you consider the cost of your own time, the machine costs, material and of course bits it probably works out cheaper to get them done externally. You also get the added benefit of soldermask and silkscreen. Unless you are in the situation where you constantly need to be turning around designs on 2-3 day lead times I don't really see the benefit of this method any more.

I think this is true with almost any non-trivial design.  I've got to prototype some fairly simple power PCBs with 2 and 3 ounce copper.  We started looking at the prototyping costs for some boards that were just glorified wiring and it makes sense.  The whole board might take 20 minutes to cut, so I can baby sit.

What I've found really useful is cutting out boards that have no traces on them.  We are brainstorming a few ways to solve some interface and case design issues and 3D isn't as good as holding them.  So I've been routing out the edge and mounting holes for jacks.  Then I solder jacks directly to the copper clad.  It is a real quick way to give our boss a real world feel for a board in the case prototype and physically plugging and unplugging cables.

Another great use for this thing is cutting the opening for one off or short run projects in hobby boxes.  I've even taken to engraving what would otherwise be a stuck on P-Touch label or something.  It gives results that look really sharp.  4 holes around the clean opening for the LCD.  Then machine a plexiglass cover to go over it and bolt in place.  It really looks like a production piece.

[mikeselectricstuff]

Rumor is that Newfangled Solutions is developing a completely new generation ("Mach4"?) and will raise the price an order of magnitude and drop Mach3 (US$175) Will an alternative/replacement emerge that will take up the market it for hobby-level users?  Mach3 seems remarkably fiddly and amateurish. I can't see how they have managed to sell it for THIS long.

They really missed the boat on this - in the parallel port era they owned the market, if they'd done a reasonably priced USB to parallel interface to run all the existing parallel port devices (which could also have been a dongle for their SW) they could have totally cleaned up.  There is other stuff out there now so their monopoly is gone.
I bought a smoothieboard a while ago but not tried it yet.
What I think is still missing is a standalone G-code to motion controller with sufficient inbuilt UI to do the stuff like zeroing and feed rate control - that's the only thing I ever do with Mach3's UI.
Maybe something that interfaced directly with these cheap handheld MPGs http://www.ebay.co.uk/itm/CNC-3-Axis-4-axis-USB-HandWheel-MPG-pendant-for-Mach3-engraving-Router-Mill-/111464851845?pt=UK_BOI_Industrial_Automation_Control_ET&hash=item19f3d2a985

 
 

[G0HZU]

+1 to what Mike said, previously I've used LPKF and T-Tech milling machines and sometimes they work beautifully so you are can go from concept to PCB in a single day. However they often needed a lot of attention to get the right results, particularly on larger boards.  Even the slightest warp in your FR4 can destroy your results and leave you with multiple efforts. Broken mill bits are another pain, you sit listening to the machine drone (plus vacuum and air compressor) for  hours while you do other things only to realise the mill bit broke 10% of the way in. Bear in mind these were EU and USA built specifically for PCB work.

All valid points and i think for general purpose FR4 boards (especially digital circuits with lots of PCB traces) it is false economy to use one of these machines today.

But I use mine for RF research/design using various exotic PCB materials and the boards are often very simple. The key to success is to be milling small and simple boards on non standard dielectric on a regular basis and also to use it for other purposes like front panel design and making laminated tools etc. Also, there's no substitute for experience and this helps enormously because there are so many things that can go wrong during the whole process that can destroy a board. I've learned a lot in the time I've been using these machines and it's very rare that I get a dud board and I don't have many tool failures.

I'm amazed that modern machines use parallel interfaces. T-Tech moved over to RS-232 about 20 yrs ago. I know an ex colleague who has one of the early T-Tech machines with a parallel interface and it is controlled via an ISA card that has to be fitted in a contemporary PC running MS-DOS. eg a slow 286 or 386 or the system will be erratic in operation otherwise.

Mine are both controlled via RS-232 and I have a dedicated (old) Win98SE laptop that runs the Isopro software really reliably.

[mc]

The main reason people still use parallel ports for CNC machines is they're cheap, and for your typical home CNC they are more than adequate. However parallel ports are becoming rarer.
Any half decent commercial machine will use a dedicated motion controller, and if PC driven, usually either via USB or Ethernet. Serial has pretty much died out, as it simply doesn't have the bandwidth to handle complex countouring, especially at modern machine speeds.

I actually find the history of Mach pretty interesting, and think that Art Fenerty, the man behind Master/Mach, is a genius. He achieved something that many said couldn't be achieved on a windows operating system, which was a highly accurate pulse generator. And as far as anybody knows, it's something that has never been repeated. I know Art has said that the driver is essentially a virus that latches onto the windows kernal and hijacks processor time, which is why it will never likely work on 64bit systems, as 64bit kernals are encrypted to improve security and the encryption gets changed with most updates.
Art developed Mach to the point of adding plug-in support, so it would work with external motion controllers, however Mach3 was originally designed to only use the PP, so Mach3 has been plaqued with many bugs due to interdependability issues since, which have only become more apparent as external motion controller/plug-in developers keep pushing the limits.

At some point Art sold the company to Brian Barker, who spent a few years ironing out bugs, before finally biting the bullet and starting a complete re-write, which has only over the past few months been released for testing.
Mach 4 is essentially just a framework, with various plugins to provide the GUI, trajectory planner, and IO. It also removes alot of the limitations that Mach3 imposed, such as limited IO numbers, and the 10Hz update rate for plug-ins/macros. Feedback so far has been good, however there is increased competition, but personally I think Mach will still be very popular, as it's far more adaptable than the other options without requiring a great deal of programming knowledge.
The parallel port plugin for Mach 4, Art wrote under the working name of Darwin, as he feels it wil be the last evolution of the printer port for motion control purposes, as external motion controllers are getting cheapers and more common place (I've personally got 3 machines that all use external motion controllers).

Art also wrote a test version of Mach with a high order trajectory planner to provide an S-curve motion profile, under the name Quantum. I read a post about what's involved in a high order planner, and my head hurt reading the description, let alone the amount of coding and thought that would need to go into writing one!
One of Art's great attributes is he is willing to share his knowledge about such things, and explain it in a way in which most people will understand.

Art's latest project is Gearotic Motion. I had a quick play with it when it was still pretty new, and it just blew me away with what it was capable of designing. For anybody interested in gears, especially more bizarre yet functional ones, check out the Users Pics, and Your Videos boards over on the forum - http://gearotic.com/ESW/FavIcons/


For those who mentioned using a USB to parallel adapter, it won't work. Motion control relies on highly accurate pulsing, which standard USB to parallel adapters are just not capable off. The nearest thing you can get is CNCDrive's UC100, which looks just like a USB to parallel adapter, however is actually an external motion controller, shrunk to fit in a very small enclosure with only a single USB and single DB25 connector. It's essentially designed as an easy way to provide USB connectivity to a parallel port machine, without the need for rewiring/retrofitting.


I've now retrofitted three CNC machines, with a fourth being planned, and have looked at most hobbyist level motion controllers and software. All have limitations, however Mach remains my personal choice as it provides a standard interface which is relatively easy to customise. I have one machine running Denford's VRMilling5, purely because it had been upgraded to the USB controller which provides pretty good performance, however it's just not got the features or easy adaptability of Mach, which is where most other options suffer aswell.
I know I fall into the more experienced CNC builder category, so I know what I want from the hardware and software.
Whereas somebody new, mostly decides on price, and often ends up with less features, which is fine until you realise those features can come in very handy, or the junk you've bought is hopeless (TB6xx drives being the best example!).


And finally, the issue of non-level boards, check out http://www.autoleveller.co.uk/
I've not personally used it, however feedback seems good. Although it requires probing, you just need a single available input connected to a couple suitable bits of wire with crocodile clips. One clip on the board, one on your spindle/cutter, and you've got a suitable probe.

[G0HZU]

Any half decent commercial machine will use a dedicated motion controller, and if PC driven, usually either via USB or Ethernet. Serial has pretty much died out, as it simply doesn't have the bandwidth to handle complex countouring, especially at modern machine speeds.

To clarify my current setup, both of my old T-Tech 7000S machies are arranged as PC >> RS-232 >> T-Tech Control Box >> big custom T-Tech parallel cable >> Milling table

The milling table houses the various motors and the spindle motor and Z solenoid and all the power and control to it comes from the parallel interface from the external T-Tech control box.

I don't think it is a slow machine but then I don't know what modern machines are capable of. The modern? machine in the autoleveller link looked really really slow and clunky compared to my oldschool T-Tech machine.

I also had a look at a few youtube demos of the Mach 3 SW and I wouldn't want to be using that for PCB milling. It's a very versatile interface with lots of low level user control/config but it looks to be very limiting when it comes to PCB milling. I'm comparing it to my ancient copy of Isopro.

Isopro isn't 'great' SW by any means but it does seem to be well tailored to PCB milling with my 7000S. I use LPKF SW to produce the isolation/rubout files and import these into isopro to do the milling. This is because the isolation/rubout algorithms built into Isopro are very crude compared to LPKF.

[mc]

G0HZU, I should maybe of been a bit clearer and mentioned modern machines.

For 2D work, which is pretty much what PCB cutting is, then serial is more than capable, as the majority of commands will be relativelt large linear or arc moves created from a DXF or similar 2axis file in only 2 axis, with the 3rd axis only occasionally being moved to set depth.

The limitation of serial comes when you start creating G-code (or whatever machine code your machine uses) from 3D stl or similar files, which result in lots of relativelt small moves often in all 3 axis.

Also, when I say parallel, I mean something that actually uses the parallel IO standard. By the sounds of it, your machine is using parallel connectors simply for connecting power/signals. The RS232 is what is doing the actual communication, with the control box containing some form of motion controller that converts the serial signals to/from power/sensors.

Most of the Mach3 videos you'll see, are machines that have been converted using steppers on a pretty tight budget, and for various reasons, aren't really capable of high speeds. Certainly your typical 3040/6040 machines will shake themselfs apart if pushed too hard/fast, whereas I'm guessing your machine is well designed and tuned for it's intended purpose. Most hobbiests will be quite happy with sub 1m/min speeds, whereas that's verging on a crawl for commercial machines. The last commercial vertical milling machine I looked at, was capable of 6.5m/min cut rate with over 16m/min rapids. (yes I'd love one, however i can't justify one just yet!)

Quite simply, Mach3, and all other motion controllers/software, are only as good as the G-code it's fed and the machine it's controlling. Here's a vid showing move testing of a Chiron mill that has been retroffited with Mach3 and an external motion controller -
It's also worth checking the tool change vids, as the unique feature of this mill is the super fast toolchange with the spindle still turning.