PCB Isolation Milling Software for a fine feature test project

[rx8pilot]

I am planning to do some PCB milling tests soon and a bit frustrated with the options to generate the G-Code I will need.

The test is essentially trying to figure out just how fine of features I can pull off with a full size, very accurate mill. I am planning to build a special vacuum fixture to keep the board very flat and very rigid. I also spent some $$ on very high-end cutting tools. Curious if I will be able to get good results. The goal is to do tests for high-speed, impedance sensitive structures as well as some power electronics applications to test sections of a design. I will also use it to fabricate pogo pin test fixtures.

Anyway....the best looking option I found was FlatCam but I struggled for hours trying to get it working on my machine. There are a number of Python dependencies that I clearly do not have right. It will not even start up. I also used Carbide3D software which is web based. It is a bit simplistic and will require a lot of hand editing, but at least I get the basic tool path out of it. It would be nice to be able to place tooling pin holes and manually choose a starting offset.

Any ideas on which software I should try?

[TomS_]

If it's a decent python project it should have a "requirements" file that lists all of the dependencies, including any pinned versions.

In that case it should be a simple matter of "pip install -r requirements" to sort all of that out.

Make sure you use the appropriate python version too. Python 2 and 3 are largely the same but there are some differences in the way they handle certain things.

[rx8pilot]

I am away from my computer now but from memory it recommended installing WinPython and then adding RTree and Shapely. It requires PyQt4, but I can't get that installed with PIP or the WinPython manager. Shapely errors out too.

Tried on Windows 7 and 10 with the same result. Python is not a strong skill for me either. I have only used it for relatively simple automation tasks.

Hoping there is another Gerber to G-Code option that isn't such a problem.

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

[saike]

There is the software used by https://cirqoid.com/ at http://cirqwizard.org/.

I tried it ages ago and can't remember much about it but I think it only does small (Eurocard?) size boards.
There is also https://github.com/pcb2gcode/pcb2gcode/wiki/Manual  if you don't mind command line stuff.

I notice you are going along the route of buying expensive cutter bits, I have been there and I get better results from the <$1.00   Far East 'D' shape bits on Ebay.  0.2mm  20 degree angle seem to be best for a fine cut and sturdy tool tip.

[rx8pilot]

The cutting tools are a big part of the experiment. Over the past few years, I have read a lot of articles and gone to a number of trade shows.

The equipment range in PCB milling is from a few hundred $ kickstarter stuff to $80,000 LPKF. The reviews range from 'total crap' to 'amazing' at each end of the range.

As the features become more difficult, the increases rapidly. I am interested in learning just how far I can push it with a very precise machine and high-end cutting tools. Once I find those limits, I will try the same tests with Alibaba tools to see what the difference is.

The precision on the z axis is very important when the goal is to have predictable trace geometry for RF and high-speed circuits. My machine is very precise, but the challenge will be getting the material truly flat. Double sided tape is not particularly accurate so i will machine a special fixture to get the best flatness I can.



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

[saike]

Keep us posted, I will be interested to see how it goes.
I regularly produce boards for 0.5mm pitch components with only a few minutes setup time.

Here is something interesting for anybody into milling who hasn't seen it

[Koen]

The video is so beautiful it's soothing.

[rx8pilot]

Keep us posted, I will be interested to see how it goes.
I regularly produce boards for 0.5mm pitch components with only a few minutes setup time.

What machine, tools, and software do you use?
That demo video looks great, I saw it a while back and my first thought was that it was idealistic relative to the rather lightweight machine construction.


I tried the Cirquid software - nope. Not even in the ballpark. Moving on.

Discovered that Flatcam needs Python 2.7 and PyQt4 among other dependencies. My machine had 3.4 and PyQt4 is not compatible. Sure wish it was an .EXE or installable package, but when it is free and open source - hard to be a complainer.

[ajb]

I hope you'll be sharing your results, it would be really interesting to see what you find.

What about converting the PCB design to DXF and doing CAM from that?  I'm not sure how nice of a workflow that would be in a CAM package meant for solid models, but perhaps there is more general purpose 2D/2.5D engraving CAM software you could use.  There are gerber to DXF converters, and your PCB software of choice may be able to output directly to DXF as well.

[rx8pilot]

I am making a video for my YT channel Factory400.

Win or lose, you will see the results.

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

[phil from seattle]

This one is on my project list.  I have a mill and would like to understand how small it can go so count me as interested in your results. My goal is to be able to whip out a quick prototype or test PCB and not have to wait a week or so.  If .4 mm is the best, that's not going to do it for me as that's just a bit better than 16 mils. 10 mils/.25 mm is probably the min acceptable. Maybe I could live with .3 mm.  I would only use it for quick test kind of stuff as milled PCBs lack through holes, silk screen, and soldermask. The point of it is to shorten my design cycle time for testing new (to me) parts. What I find a bit more compelling is making stencils which has similar size/accuracy requirements.

[saike]

What machine, tools, and software do you use?

I have a hotchpotch desktop machine made from bits and pieces I have built over the years with cheapo ballscrews driven by stepper motors and round linear rails but everything is rigid and it has an accuracy way in excess of anything needed to produce a fine pitch pcb, confirmed by doing several runs at increasing depth over a board full of smd components with no loss of any features. The drawback with it is that it is relatively slow because the spindle is heavy and the accel/decel rate has to be kept to a modest level.
I use either pcb2gcode or Flatcam with this machine. 

I also have an ancient LPKF C20 I picked up for a few hundred dollars years ago. This has a floating head that compensates for non-flatness of the pcb material as long as the tool bit doesn't leave a burr on the edge of the cut which happens with a lot of the $20+ bits I have tried, that is why I prefer the 'D' shape bits which produce a much cleaner cut.
This machine uses proprietry Circuitcam and Boardmaster  software, both overcomplicated and 90% of what they do could be discarded and wouldn't be missed by me. Pages and pages of stuff in the manuals about aperture sizes, tool size overlaps and the ability to add features to the pcb design as well.  The advantage of this machine is the lightweight construction makes it fast, a big advantage when you have a lot of boards to make.

[rx8pilot]

I have a hotchpotch desktop machine made from bits and pieces I have built over the years with cheapo ballscrews driven by stepper motors and round linear rails but everything is rigid and it has an accuracy way in excess of anything needed to produce a fine pitch pcb, confirmed by doing several runs at increasing depth over a board full of smd components with no loss of any features. The drawback with it is that it is relatively slow because the spindle is heavy and the accel/decel rate has to be kept to a modest level.
I use either pcb2gcode or Flatcam with this machine. 

I am still trying to configure Python for Flatcam - nearly ready to give up.

Looking for a supplier of the cheap tools still. My fancy micro grain carbide tools arrive on Tuesday next week from Harvey. The routing end mills are called compression cutters and have some really complex geometry to squeeze the laminate together instead of spreading/splitting it apart. Is it worth it? We shall find out! 

Can I get 4mil traces? Can I get 4mil spacing? Does it require luck or is it a function of machine and skill?

[george4657]

Flatcam is the best pcb milling program I have found.

I just did a download into windows 10 using https://bitbucket.org/jpcgt/flatcam/downloads/FlatCAM-Win32-8.5-Install.exe

This  loaded the files into C:\Program Files (x86)\FlatCAM\   including python27 files and PyQt4 files and setup the program to run.
 84 File(s)    104,978,661 bytes
I have attached dir listing if this helps

The program just ran after this. I don't remember having to do anything else.

George

[saike]

6 mil traces with 6 mil clearance is the smallest I do with 1 oz. FR4  and a cut depth around 2-3 mil set empirically by test cuts. If the cut is any deeper than 1 or 2 mil into the laminate then it is way too deep.

[rx8pilot]

Pretty excited to try that link in the morning. Fingers crossed.

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

[rx8pilot]

Flatcam is the best pcb milling program I have found.

I just did a download into windows 10 using https://bitbucket.org/jpcgt/flatcam/downloads/FlatCAM-Win32-8.5-Install.exe

This  loaded the files into C:\Program Files (x86)\FlatCAM\   including python27 files and PyQt4 files and setup the program to run.
 84 File(s)    104,978,661 bytes
I have attached dir listing if this helps

The program just ran after this. I don't remember having to do anything else.

George

Thanks George!!

That was extremely easy. Not sure how/why I missed that. I had only seen the raw *.py project files. It is up and working now and only took about 30 seconds. Brilliant.

Now I am going to order some FR4 in various flavors - copper weight and dielectric thicknesses. I have some of the typical Chinese 'carbide' drills on order as well to see how well they perform at <1/10 the cost.

So far, I have not seen any easy way to plate through holes so this test will only focus on designs that do not need them - or perhaps only a handful that can be manually soldered.

[IconicPCB]

I setup PCB prototyping service ten odd years ago.
Looked at LPKSF and a few other brands and decided to roll my own.

Machine has been runing very happily to date without major issues on linuxCNC .

I also use Graphicode gerber CAM software which allows full gerber wditing and processing including tool path extraction and processing ( G code output).

I whole hartedly recomend PRECISE BITS cutting tools in particular theor triple fluted cutters.

0.010"  ( 0.254mm ) clearances in 0ne ounce copper are a breeze. I have even processed two ounce boards with the same feature size, a lot more demanding on tools but well within the tool's capability.

Compared to LPKF tools Precise bits are far better value for money in my experience. One thing though the smallest of LPKF tools is probably the more capable of the two brand's offerings .

Japanese MITS equipment is also well worth the consideration. In 2007 i had the opportunity to visit their factory and meet the people.

If You are interested in RF structures what is the upper limit on frequency of operation?

A professor, a doctor of sciences at Belgrade University had done some research work edge effects of milled structure profiles withthe view to minimising influences of edge deformation on structure RF  performance. I was interested in his work at the time.

[saike]

Looked at LPKSF and a few other brands and decided to roll my own.

What spindle speeds do you use? The fastest I have is 20,000 rpm and I have always wondered if more speed would have any advantage.

[rx8pilot]

I whole hartedly recomend PRECISE BITS cutting tools in particular theor triple fluted cutters.

Excellent...they have a good selection, good prices, and they are only 2 day ground shipping from me. Thank you for the referral.

If You are interested in RF structures what is the upper limit on frequency of operation?

As high as 20Ghz with some of the high-end substrates is the long term goal. I am hoping to have some freedom to experiment and measure fairly freely to facilitate the considerable learning curve.

Looked at LPKSF and a few other brands and decided to roll my own.

What spindle speeds do you use? The fastest I have is 20,000 rpm and I have always wondered if more speed would have any advantage.

My machine tops out at 15,000 rpm which is pretty quick for regular CNC milling but dead slow for PCB isolation milling. 80k - 100k RPM is a more optimal speed to maximize carbide tools. In my case, I will probably stick with the limitations of 15k as I figure out if the process is worth the effort. If I find myself doing this regularly, I would consider getting a Air Turbine Tools spindle that does 100k RPM. I have other machined parts that would also benefit from that kind of speed.

A lower cost option would be an electric spindle that can do 30k RPM or so but not so sure the effort would be worth the benefit for only a 2x RPM boost.

@IconicPCB - do you use some sort of hood / vacuum to deal with the resulting dust?

[IconicPCB]

Glad to spread the good word.

20 GHz is OK.. I looked at the requirement for high frquequency copper structuring and settled on renishaw linear scales with a 5 micron reader head.

I think that will give me sufficiently high resolution to nudge 50 GHz limits.

I use an Alfred Jaeger spindle good for 100KRPM.  Alfed Jaeger now have an office in the states but their agents Centreline also supply the spindles.  Mind You spindle speed is not all it is craked up to be. A major problemwith  the spindle is dynamic balance and therefore run out of the tool tip.  Jaeger specify runout of the tool tip at the end of a 15mm long tip protruding from the collet.  The specification is in terms of runout ( micron ) at a maxim RPM.

Japanese manufacturer MITS  advises the spindle be operated at the minimum speed consistent with required throughput. Reason being good sound engineering practice.. the higher the RPM the greater the runout.  Do not push the envelope if you don't have to kind of thinking.

Additional advice on spindle speed... the spindle is stiff and it has mass ...it will vibrate ... and if Your RPM is near a resonant frequency the run-out will be even more of an issue . Try to estimate the run-out not by measuring it but by listening to the song the spindle sings as it runs up and down in RPM. Run it in a quieter window.

In addition to adjusting spindle for minimum vibration , adjust the rpm for optimal feed  or chip load.

There is no point in running the spindle at super sonic speeds when the end result might be dulling of the tool edge rather than material removal. You will wear  out a tool and not finish the job. It is better to load the tool and even cause it to fail due to cutting load rather than loose the tool edge due to excessive friction under light load.

Also make sure the CNC controller will produce acceptable deviation from desired trajectory at the programed feed rate; if need be drop spindle speed to meet tool wear criteria and acceptable trajectory performance.

A vacuum chuck is a fine idea but essentially it is not what You should be looking for.

A floating head is far better at ensuring uniform depth of cut .  Laminate is not perfectly flat nor does it have perfectly parallel surfaces. The only way to ensure uniform finish ( depth of cut ) is to map laminate surface and adjust Z axis to suit.

The next best thing is a floating head which will follow the surface variations and provide uniform finish.

One more thing ... Precise Bits offer a copper cutting compound. Ron Reed who runs the place is the fellow to talk to about the applicability of the cutting compound to Your needs.

EDIT: on the issue of swarf and dust control: find a qood quality vacuum clener and position a suction nozzle adjacent to the tool tip. The other option is suction through the body of the head. Removes debris and heat from the work space.

[rx8pilot]

This project/test is based on using my fresh new Haas Super Mini Mill. This is a machine that is not designed to do this job - it is happier taking a 10lb block of Titanium and cutting it down to an intricate 1/2 lb part. The machine is about 5,000lbs and has a 15hp spindle - obviously a bit ridiculous in the context of a PCB.

So, the idea of milling PCB's is an adaptation. The idea of a floating head is not really practical as far as I can tell. Making a vacuum plate fixture with tooling pins is a simple project relative to floating. I am not sure how flat I can expect the material to be.

[jeremy]

How do the floating heads actually work? Do they basically just use a spring to push a foot onto the pcb, and the cutter is adjusted with respect to the bottom of the foot? Or does it use an indicator to track the surface height? If it is a foot, what is it made of so that it glides over the pcb surface?

IconicPCB it’s interesting that you say a map is better, it seems like a map would be much simpler to implement too. Do you need to worry about the board shifting after you start cutting?

I’ve always had dreams of making a little CNC for 6mil/6mil PCBs, but once you look at the cost of the scales, ball screws and the spindle it becomes a little to much to stomach

[TheUnnamedNewbie]

Regarding the high frequency stuff: I don't know what the surface finish is that milling leaves behind but it might result in higher loss than expected. For projects I am working on (mind you, these are > 100 GHz things so I don't know how this scales to 20 GHz) surface roughness of the materials is a big deal and causes a lot of additional losses. You might find the same happens when the edges are rough due to milling?

I'm curious as to the performance here and would be interested to be able to compare some results.

[Dubbie]

Vacuum chuck plus z map seems like it would be the best trick.