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

[python50]

Hello, I am considering getting a cheap desktop CNC mill for PCB prototyping. Specifically a cheap Chinese CNC 3020 which seems to come with everything needed to mill. I have etched many boards and I have gotten them professionally made before, but I have never milled a PCBs. I always have enjoyed being able to mill out designs but I don't know what the limitations would be for milled PCBs.

• What does it take/is if feasible to mill fine SMD footprints like a 0.5mm pitch TQFP100 on a CNC mill?
• I notice the machine I am looking at has a max spindle speed of 8000RPM, would higher speeds be of much utility for PCBs?
• Is a different avenue for prototyping recommend instead? To me it seems that it a mill could pay for itself in a few large prototypes given its accuracy is not an issue.
• Do these Chinese machines work reliably over the long run or will I constantly have to tweak the mechanicals to make it preform satisfactorily?

Thanks,
Jason White

[Frost]

• What does it take/is if feasible to mill fine SMD footprints like a 0.5mm pitch TQFP100 on a CNC mill?

To mill SMD footprints with an extreamly small pitch, you need very high RPMs
to use a very small, we call it "Stichel", it's a mill cutter primary for engraving
with only one sharp 15 degree V-shaped tooth.

For this small cutters you normaly use a inverter driven high speed spindle
with up to 25000RPM and more.

• I notice the machine I am looking at has a max spindle speed of 8000RPM, would higher speeds be of much utility for PCBs?

Definitely yes!
The smaller the drill diameter the higher the needed RPM,
same for the trace spacing with the Stichel.

• or will I constantly have to tweak the mechanicals to make it preform satisfactorily?

I never ever saw a chinese (except taiwanese) tool or machine, within the machine building
segment, that worked satisfactorily ex works without the need of massive tweaking and rework

For PCBs with "normal parts" this mill could work for hobby purposes,
but for SMD parts with small footprints like TQFP100, I am not sure.

[python50]

To mill SMD footprints with an extreamly small pitch, you need very high RPMs.

Is the spindle RPM the primary limitation with machines or would more precise stepper motors and etc. likely be needed as well?

For this small cutters you normaly use a inverter driven high speed spindle
with up to 25000RPM and more.

Would it be possible to simply run the machine at a greatly reduced federate to compensate for low spindle speed? Or might that have some undesired consequences?

For PCBs with "normal parts" this mill could work for hobby purposes, but for SMD parts with small footprints like TQFP100, I am not sure.

In your opinion would a SO (50mil pitch) package be feasible, or is that getting into the painful territory for milling?

[orangecat]

There's no problem milling the smallest SMD footprints. Here's a video milling a TQFN-48 (0.5mm between pin centres). Using a generic chinese CNC3040:

[jeremy]

Well done, that looks amazing!

[ecat]

Impressive.

But. It is hard to know if anomalies are by design or due or errors. Looking at the introductory still image the stepping on the central horizontal channel may be backlash in the Z axis and the variation in the shape of the IC pads backlash in the X and/or Y axis.

In common with the vast majority of CnC videos showing the final result is mostly pointless, what we need to see are a selection of squares and circles cut and then measured to better than 0.1mm.

Two other points before the OP spends their cash: The machine in the video uses the 24,000 RPM water cooled spindle and these machines are available in both trapezoidal screw and ball screw configurations.

High rpm is good for PCBs, the other consideration wrt spindles is run-out, the eccentricity of the tool when rotating.

Ball screws are more expensive than trapezoidal but they should be more accurate along their full length, present lower backlash and maintain the lower backlash for longer. You can tweak trapezoidal setups to provide good results, then tweak them again after a couple of hours, rinse and repeat.

[Fred27]

I used to mill PCBs on my converted Proxxon MF70, but have switched to photo etching. Whilst I got reasonable results, I found that the the depth of cut meant that applying dry film solder mask was harder.

If the router is anything like the Chinese laser cutters (and i've read that it is) then it probably won't be smooth sailing out of the box.

Whilst these are both negatives, I would still say "go for it". It will be ideal for drilling and routing out etched boards. I'm sure you'll enjoy trying out PCB milling even if you don't completely switch to it. If you don't mind tinkering with the machine to get it running well, then they're great value. Consider it as a kit of parts rather than a polished commercial product. I perhaps wouldn't recommend one to the average guy on the street, but if you're on this forum then it could suit you.

CNC machining is another fun new hobby. You're just as likely to use it to make quadcopter frames, front panels or project enclosures as you are to make PCBs.

[orangecat]

Machine uses ball screws on all 3 axis (I wouldn't bother using a machine with t-screws on any axis if TSSOP, 0.625mm or 0.5mm pin pitch, is required). The 800 watt water-cooled spindle is also a good choice, air cooled spindles tend to much noisier and they may have more runout. Any significant runout will break the endmills or ruin the board or both.

There might be some backlash in Z, or it might just be the board slightly changing vertical position as the stresses in the board are altered as copper is removed, or as the tape holding the PCB down or simply due to temperature variation from start to finish. It's very small, something like 0.02mm -- not even important for cutting boards.

The board looks almost exactly like the simulation:

[jaxbird]

Machine uses ball screws on all 3 axis (I wouldn't bother using a machine with t-screws on any axis if TSSOP, 0.625mm or 0.5mm pin pitch, is required).
...

Nice work 

But you don't need ball screw to do 0.5mm pitch. I have a small cnc with acme screw and anti backlash spring nuts, no problems doing 8-10 mill traces. (got some pics under the PCB manufacturing main subject)

I think most important is having a spindle with minimum runout and some sharp bits.

I don't go as deep as your pictures show, just enough to remove the copper.

[python50]

Thanks everyone for your input. Especially after seeing that video on vimo, I think I'm going to get a CNC 3020 and try it out.

[mazurov]

The picture shows SO and 0603 milled on this machine -> http://microcarve.com/ (MDF frame, Delrin nuts, 2 start turned ACME screws) with Bosch Colt spindle and 60 deg bit from http://www.precisebits.com/ .  The gear is pretty basic but the machine needs to be squared and have very little backlash (freshly cut Delrin nut works very well for about a year ). Out of the box, any mill would require squaring, so choose the design which allows for that. If you have never squared a mill you may want to pick a model which is popular. Ball screws will help with backlash but not squaring.

I can mill 0.65 mm on 1oz and 0.5 mm on 0.5 oz; however, results are less consistent and boards are harder to stuff.

[maliqua]

I did something similar to what you have in mind I've been using a desktop cnc from mydiycnc.com with the laser cutter for traces and rotary tool for holes


i found it very difficult (technically i haven't succeeded) using the rotary tool for an entire pcb.



** i should note that i still get better results with the toner transfer method then etching in FeCl

[eurofox]

I just wonder what kind of software you use, I know the generic Mac3 but the output from PCB software is not compatible with Mac3, is there other dedicated sofware available or tools to convert it? 

[mazurov]

I use pcb-gcode ulp for Eagle. It outputs GCode usable on many different platforms, including Mach3.

[mikeselectricstuff]

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

[Precipice]

You really need something that can sense and adapt to surface height

Does such a thing exist? I know it's a pain trying to keep plasma cutters hovering at the right height.
Can you scan, remember and cut, hoping nothing's shifted in the meantime?
(I'm trying to think of a scheme for measuring depth. Maybe spotting capacitance change as you repeatedly lift and lower the tool into the copper? Something optical, watching the width of cut? This seems hard, but I've probably missed something obvious...)

[mikeselectricstuff]

measure-then-cut would probably work. Capacitance may work quite well, as you're sensing a flat metal plate at close proximity.

[G0HZU]

My T-Tech 7000 machine auto adjusts for surface height by using a simple sensing foot. This allows it to achieve good results up to a max PCB area of 13 inches by 19 inches.

It usually works well although it can cause problems if the foot falls into a wide area that has already been milled. But usually this is avoidable.

The foot technology has evolved a bit since my machine was made though...

http://t-techtools.com/store/index.php?main_page=product_info&cPath=14_19&products_id=288

[G0HZU]

In case anyone wonders how the sense foot works, the spindle assy has a foot or 'heel' that the whole assy rides on and a fine rotary vernier control sets the tracking height of the tool tip vs the foot. So even if the board is warped the foot will press it flat and the tool will faithfully track against the foot and keep a constant cutting depth.

If the cut depth isn't quite right, it's easy to just give the vernier a fraction of a turn to increase or decrease the cutting depth by a fraction of a thou of an inch as it mills the board.

[G0HZU]

(I'm trying to think of a scheme for measuring depth. Maybe spotting capacitance change as you repeatedly lift and lower the tool into the copper? Something optical, watching the width of cut? This seems hard, but I've probably missed something obvious...)

There's always the 'dambusters' method I suppose ( )  but the sensing foot method has been around for decades and it is reasonably effective...

[mazurov]

You really need something that can sense and adapt to surface height

Does such a thing exist?

There is an add-on to pcbgcode called autoleveler; it's a script which probes the surface of a PCB, stepping and probing every 5mm in each direction, and then uses the data to adjust Z during milling. Standard tool height probe input is used, you need it to clip one wire to the (stopped) spindle with cutter installed, another wire to the foil, run the script, remove wires when it's done probing, mill. I never used it myself but saw ppl raving about it on the forums.

[G0HZU]

Can they not adopt the foot method? I've not really looked at the mechanics of mine for many years but the foot pressure is set by an adjustable solenoid. It is done like this to preserve tool life. The ideal scenario is for the foot to only just lift off the PCB when it has finished a cut. So it has minimal vertical travel before hitting the PCB again on the next cut. It took me a while to set up the best foot pressure on my machine. The foot isn't used during drilling or routing.

It can't be that complicated to replicate. I think some newer systems use a horseshoe shaped foot. Mine is the older generation and is round like the ball of your foot.

[mazurov]

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.

[Richard Crowley]

Would it be practical to use a vacuum hold-down for the PC board material?  That method is used successfully for much larger pieces, so it seems like it would work with most any size PC board.

[mazurov]

Would it be practical to use a vacuum hold-down for the PC board material?  That method is used successfully for much larger pieces, so it seems like it would work with most any size PC board.

Depending on what is considered practical. If you have it already, then yes. If you don't, you can either buy one (expensive) or make one (depends on how good you are with mechanical design, I'm not). On the other hand, carpet tape is not that expensive and works just fine for one-offs.

I have a project in the future when I'll need to mill a bunch of the same boards ( 5mm 65mA SMT LEDs run cold being mounted on a PCB with almost all copper left in place on the both sides.  I'm making a grow light out of them. See attached ). Since I need about 50 of them I'll attempt to make a vacuum table, good just for this board.