Hi Katemonster,
See this post could be of help.
https://www.cnczone.com/forums/pcb-m...44316-cnc.html
and scroll to post #15.
Cheers,
Hanspeter.
Hi all.
I hope this thread fits under this forum. The machine I'm using isn't really a DIY, I started with a Hermes V3400 engraver and have been hacking it apart to fit my needs, I began by replacing the electronics with an Arduino and a Raspberry Pi. I also changed out the steppers and limit switches to make it run more smoothly. Now I'm looking to find out what else I need to modify to make the machine capable of milling out PCBs with high accuracy.
Ok with that out of the way. I've attached an image of attempt number 1. Disregarding the uneven cut depth (I think the PCB moved after autoleveling), it seems like I'm having wicked backlash/inaccuracy on my Y axis. I am here because I suspect the design of the machine to be the culprit. The X axis screw goes through the center of the gantry. The Y axis, however, drives the gantry through a lead screw in the left side of the machine, in the housing. It relies on a steel rod in the right side of the housing that the gantry floats on, to keep the gantry straight. This seems like it could cause serious accuracy issues. My thought here was to make a whole new bed for the machine and have it driven by the Y axis. What do yall think?
EDIT: I just noticed that the image appears to be rotated 90 degrees counter clockwise. When looking at the image, the Y acid movement is left to right.
Similar Threads:
Last edited by katemonster; 10-26-2019 at 02:46 PM.
Hi Katemonster,
See this post could be of help.
https://www.cnczone.com/forums/pcb-m...44316-cnc.html
and scroll to post #15.
Cheers,
Hanspeter.
You do seem to have some serious mechanical issues. I don't think the design of the machine is the problem though - it did (presumably) work originally, so you should be able to make it work for engraving PCBs.
Can you feel any backlash or deflection if you try to push the spindle by hand?
Hi,
if you wish to have PCB features as small as 0.2mm then the backlash/accuracy needs to be about 10 times better than that, that is about 0.02mm.
That requires a nice machine and near zero backlash is required.
Craig
The good thing is that a PCB milling machine does not need to be particularly rigid, fast or powerful. It does not even need to be particularly "nice". All it needs is low backlash and low runout. I can consistently mill 0.25mm (10mil) traces with my toy CNC machine made of plastic. That means I can use chips as small as SSOP (0.65mm pin pitch).
If both axis are running on lead screws, just ensure it has anti-backlash nut installed and working.
I'm assuming your CNC is Hermes V3400 engraver.
How do you hold PCB on the bed?
double sided adhesive tape should help on a square and milled bed.
PCB's don't lie perfectly flat, except you have an auto leveling tool/software to do this for you.
Is your CNC controller Arduino or Raspberry Pi?
This was last year, I guess you must have resolve the issues.
Citizen of Dreams I am very curious about this plastic CNC of yours that can achieve this kind accuracy. I always struggled with pcb milling and I have an all metal chinese type CNC. Every new pcb is a headache. I have anti backlash nuts in the x and y axis but I've never had consistent results. Can you share a bit more details about your milling settings, engraving bits and your machine setup?
Hi,
I can second the low backlash/low runout requirement.
While my machine is made out of cast iron and steel, rigid enough for metal milling, making PCBs does not even come close to testing its rigidity.
When I designed and built it I bought and used three second hand ground C5 ballscrews by Karoda. Even second hand they were expensive but
I've never has cause to regret the money I spent, and even after six years use I still find myself thinking 'damn these things are good!'.
Likewise I bought a German made highspeed spindle and have never had cause to regret that purchase either despite the premium I paid at the time.
The backlash is near zero, certainly less than I can measure, but the 'lost motion' due to flex in the couplers is 4um.The spindle runout is 3um by spec
and less than I can measure. I can make boards down to 0.4mm pitch, even if I struggle to solder such chips. In fact I look on 0.5mm pitch as being
a challenge to solder but find that 0.65mm pitch is entirely possible and repeatable.
Low, or even zero backlash, and very low runout make PCB making easy, even enjoyable. I've been designing and making PCBs for sale in the way of
a business for over a year. I would make a new PCB at least daily, and sometimes two or three a day.
Craig
With the exceedingly cheap price of having them made with all kinds of features, cheap, is it really worth it?
PCBWAY cost me $2.00 for 10x 100mm x 100mm boards, doubles died with vias.
CNC, Mechatronics Integration and Custom Machine Design
“Logic will get you from A to B. Imagination will take you everywhere.”
Albert E.
Hi,
The reason I mechanically etch PCBs is not price.....as you've pointed out getting PCBs made conventionally is very cheap, but its not fast.With the exceedingly cheap price of having them made with all kinds of features, cheap, is it really worth
The PCB company may turn them around within a couple of days but then you have to wait two-three weeks postage from China.
A customer can approach me in the morning for a new tachometer PCB for his classic XYZ car, I can design the circuit and PCB , make
the PCB and populate it that afternoon, and charge the guy $350 for a custom made PCB. In the vast majority of cases that is the first and last
PCB you will ever make of that exact design, having multiple boards does not make me any more money Tomorrow it will be for a ZYX car, and
you can reuse the electronic design within limits, but the PCB will be different, say round rather than rectangular or something.
I have a catalogue of about half a dozen designs that I do frequently, say 20-30 a year. For those designs conventional PCB manufacture is
worth while.
If you really want to save costs then the PCB is only a small part of the cost. I populate all boards, most surface mount, by hand. If I want to save
that time and therefore cost then I would have to engage a manufacturer with a pick-and-place machine, but that's only effective for
production runs of 50 or more. I don't have the capital to invest in that manner.
Even if I did invest in conventional PCBs and machine populated boards, the customer might see a saving.......but would I make any more money....
probably not, and certainly not enough to justify the investment.
The whole point of customers coming to me to repair/remanufacture instruments is because there is no 'new made solution'....if they want their car
to pass a 'Warrant of Fitness' (New Zealand Road Safety Inspection) for which they require a speedo, and they can't buy new or secondhand
then my remanufactured solution comes into play. Before someone asks....it's not a legal (NZ) solution to use GPS for a speedo......who knows when the Yanks
decide to turn it off.
Craig
The first most important thing I did to achieve consistency was replacing the original spindle with a better one (a 500W BLDC motor). This is still not a "real" spindle, but it has low runout and good balance.
The second most important thing was buying a copy of Autoleveller. It is not possible to maintain constant depth of cut by simply mounting the PCB flat. You need to use either a depth limiter (as found in commercial PCB prototyping machines), or surface probing (the easiest way for a hobby machine).
I also replaced the original M8 all-threads with T8 trapezoidal screws. That was mostly to improve the rapid speeds; I did not have any backlash problems with the old screws either.
The tools I used are regular Chinese 0.1mm 30 degrees carbide cutters, either pyramid-shaped or half circle. 50 cents a piece on eBay.
The best speeds and feeds is something you need to find for your particular machine. At the moment I have it set to 4000RPM and 100mm/min. And yes, I know it's slow.
That's pretty much it, the rest is just trying, gaining experience, figuring out what works and what doesn't.
Joeaverage pretty much nailed it down. Making PCBs at home is not about money, it's about turnaround time. If you need a small one-off board and you can have it in your hands within an hour or two - that's an advantage. If you are designing a prototype and you can try several versions of the board in one day - that's an advantage.
And the feeling you get when you watch your ideas turn into copper traces - that's simply priceless.
Hi,
I also use Autloeveller, and at the time I paid 20GBP as shareware/donation, but note that it's now free. I'd be lost without Autoleveller.
I use double sided tape to hold the PCB blank down with two 1.5mm pins as alignment dowels. I find that the double sided tape does not tend to pull any bows
or warps into the blank whereas clamps, bolts or screws tended to do so. I use 10mm polycarbonate as a spoil board and the double sided tape adheres well without
being overly difficult to remove. As CoD commented, it's what works for me.
My spindle does 24000rpm and I set the feed rate at 600mm/min. Long continuous runs at 600mm/min cause a borderline overheat of my steppers so am not tempted
to go faster. My new build mill has 750W AC servos on each axis so will push the speeds up BIGTIME when its finished. I'm thinking that 3000-5000mm/min will be
doable......can't wait to try it out!
I use, mostly, Kyocera-Tycom 0.2mm 45 degree engraving tools, approx $4.00USD each, for fine trace boards and use Kyocera-Tycom 0.5mm two flute endmills, approx $4.00USD each,
for through hole boards and SMT boards down to 0.6mm between pins, so SOIC and bigger.
Craig
If you are on Windows, you should try ioSender. I like it a lot better than bCNC, UGS and other senders. It has excellent leveling support. I use a 3D probe (cheap one) and typically probe a 1 cm grid over the PCB. It produces a height map and will apply it to the loaded GCode. Then you can set your Z and route away.
Probably the biggest issue I have is securing the PCB material before probing. I wind up having to use double sided tape on the entire bottom of the board. Even a small untaped area can have flex which will confuse even a relatively sensitive 3D probe.
On ioSender, it is a very good G Code Sender. Probably the best out there. It was originally written to support grblHAL (32 bit version of Grbl) but works with the original 8-bit Grbl and has a lot of great new features. In height mapping, you can save and reload the height map though I always probe the actual PCB material each time because FR4 is somewhat variable in it's thickness. It has lots of great probing options - the author is adding rotation to allow aligning GCode to the measured part. Visualization is excellent and looks to be realtime. It visually shows the state of all the inputs (limits, controls). The Grbl settings section is really good - no more encoding bit fields. If you are running it with a grblHAL controller, it shows the settings organized by function (not just a linear list). Windows only but I hear reports of it running just fine in a Windows VM on Linux.
Obligatory "easier to get them made " discussion point. While I have made PCBs via toner transfer, photo mask and routing, I send the vast majority of mine off to a Chinese fab. With DHL delivery, I can get 5 boards made in about a week for less than $20 (with delivery to USA). I also use OSHPark for small boards - about 10 days. $5 per square inch for 3 boards. Though, I will still route the occasional PCB for proto work - typically to test an idea or try a new chip where turnaround is important. It is a bit of a time sink. I can't imagine going back to chemical etching even though a quick dunk of a routed PCB in an etch tank cleans it up nicely.
Have been using OpenCNCPilot for four years and never looked back. https://github.com/martin2250/OpenCNCPilot
I can mill 0.3mm tracks on a warped PCB as OpenCNCPilot is an auto-leveling open source software, paid £20 for Autloeveller that never worked for me.
My CNC is the usual Chinese CNC3020Z this CNC comes with ballscrew leads, by the way I use GRBL compatible G-Code Sender rather than the
Chinese stepper motor driver boards with Toshiba TB6560A that came with CNC milling machine. The stepper drivers on GRBL boards are DRV8825, and
GRBL boards run on Arduino UNO or Nano.
Take a peek at users gallery https://github.com/martin2250/OpenCNCPilot/issues/47
Cheers,
Hanspeter.