PCB Milling Failure

[CJay]

First attempt at PCB milling today, using a Roland Modela MDX20 and it's not great.

I know what one failure is, I didn't level the bed but the tool to do so wasn't there, it will be next time so the cutting will be level across the board and the unmilled pads should be fixed..

Second fail, the cut edges are *very* rough, i suspect this may be choice of cutter and or feed speed.

Third, the geometry, the lines aren't parallel and the squares aren't square, I have *no* idea what causes this except maybe slop in the cutter mounting/head?

Fourth, scaling, it's meant to be for a minicircuits SBL-1, 0.2" between pins and it's off by a few thou over the length, not unworkable but not ideal.

So, am I barking up the wrong tree with this machine and it's not suitable for PCB milling, is it worn out, am I doing something wrong, can I fix it?

(And yes, I know the board design is pretty poor, it was done as a lashup for a test mill session)

[DBecker]

That is laughably bad.

There must be a significant amount of flex and lash in the mechanism.  Software can reduce some effects by using a consistent feed direction and speed, but it can't generally work around soft mounting and chatter.  Having both chatter and lash (free play) is probably unsolvable.

[CJay]

I hoped I was wrong about the slop in the mechanism but that may be beyond my remit to sort if it is even possible to do so.

It's a machine in a community workspace so I think it may be something that's been handed down from industry, as such it's probably worn out or in dire need of maintenance.

Next time I'm in there I'll see if i can detect play in the cutter head and see if they'll let me do some maintenance.

I'm interested to know if anyone has experience of the Roland Modela machines though, especially when used for PCB milling

[Benta]

From what I've seen, it's an engraving/modelling machine, made for working with resin, wood, etc.
It is not a milling machine, for that it's not sufficiently rigid.
Using it for PCB contouring is taking it to its limits.

That being said, PCB contouring should be possible, working at high rpm (20,000+), low cutter overhang and relatively slow feeds. Your jagged edges suggest far too low rpm and too high feed, making the cutter flex and "dig in" excessively. I also suspect that your PCB substrate is not being held tightly enough to the table.

[rstofer]

That's a very expensive machine, I would expect better.
How come the web site says FR-1 PCBs and doesn't include FR-4?  That's a problem!  It can cut phenolic but not fiberglass.
Higher speed on the cutter and lower feed rates as mentioned above.
Also mentioned and a likely problem is board mounting.  It needs to be held really tight.

[drussell]

From what I've seen, it's an engraving/modelling machine, made for working with resin, wood, etc.

Engraving machines need to be rigid to make proper engraved signs/plaques, etc. also.

My old Dahlgren Wizzard XL (1980s vintage, dual Z80 processors) is still accurate to 0.001" with no issues, anything less is pretty much pointless...

The internal CRT hasn't worked for years, I really should find a flyback for that... 
The ol' Commodore external monitor plugged into it still works just fine, though... LOL

I've used it to mill and drill housings and panels even, though...  something it was NEVER designed to do...  It's hard on it but it seems to always just take it like a champ! 

[janoc]

The non-squareness of the square pads tells me that the frame of the machine is pretty much out of whack and that it is either loose or flexing and it depends on which direction it is going. The axes are certainly not square to each other by any measure.

That needs to be sorted out before you can have any hope of using it even for just wood. It has likely been well abused and not maintained properly.

[Benta]

The non-squareness of the square pads tells me that the frame of the machine is pretty much out of whack and that it is either loose or flexing and it depends on which direction it is going.

I disagree. The odd shape of the square pads, plus the jagged edges indicate that the cutter is flexing, due to wrong speed/feed.
If the machine is out of adjustment, the geometry errors would be consistent (eg, parallelogram shape).
Play in the guides could lead to similar results, but would be more abrupt and not as consistent as these. Note that the pads look ugly, but are the same, play would be much more random.

[DBecker]

If you look at the 'square' pads, you can tell quite a bit.  The slope changes depending in which direction the cut was made, which suggests that the milling head is walking and/or the work piece is moving.  The trapezoid shape indicates lash on top of the flex.

It's common to reduce lash by always cutting in the same direction, commonly by putting the machine frame into compression or belts in tension.  Tool walking or slow compression/flexing of the mount or work piece isn't readily handled.

[CJay]

Well, I can go back tomorrow and try again, I won't have time to level the bed (the only way appears to be to mill away a sacrificial surface which takes time and a 3mm bit which is on order) so, comments on the below please?

You're going to have to excuse me if I get terminology wrong as this is new to me.

I've been thinking about how the machine works and the events of today, I now suspect it's mainly feed rate and the bit flexing as has been suggested.

The reason I think this could be the case?

The software defaults to a 20x20mm offset start position, sending the head to that position was repeatable after several resets of the PC and mill so, to my thinking, that would indicate the workpiece was fixed solidly and the guides etc. are fairly slop free and that the main problem I have is feed rate. 

[Benta]

(the only way appears to be to mill away a sacrificial surface which takes time and a 3mm bit which is on order)

Do NOT open this can of worms, unless you want a surface that looks like steel wool. Don't even think about it before you have your feed/speed rates worked out! Jeez!

What you need to do is provide data instead of lamenting.

Cutter type? Size? Length? and, and, and
Speed?
Feed?
Spindle bearings OK?
Play?
Pictures, especially of the cutting head.

We're all groping in the dark, because we have no information!

[G0HZU]

Wow that looks bad... At work we had a new T-Tech 7000S machine back in the 1990s that we foolishly let all the engineers play with. It took maybe a year for it to get ruined through ignorance and clumsiness resulting in several bed scores and knackered spindle/bearings with obvious out of true in the spindle. Even when it was like this I think it could still outperform what you have there. Maybe if it was also used with a faulty/worn tool it would produce the churning effects at the edges too.

But your board even has wandering on the squares that have only scored the surface of the copper.

Here at home I have a 23 year old T-Tech 7000S and its a similar model to the one we had at work. It is all original and see below for a recent closeup of a simple SMD board on FR4. This was done using worn tools that were salvaged at work when they reach the end of life in terms of distance milled. Had I used new tools it would have looked better. The traces are typically 0.012" and the SMD part in the centre is 0805. Your machine isn't designed for PCB milling so won't be able to match this but I'd expect it to get reasonably close.

[Jay_Diddy_B]

Hi,

It looks very bad. I have an LPKF c60 Protomat and I get much better results than that.

I use tools from Think & Tinker Ltd

http://www.precisebits.com/applications/pcbtools.htm

For milling traces is use EM2E8-0625-90V, this is a 90 degree carbide cutter. The spindle speed is around 30,000 rpm.

I use regular FR4 material. The LPKF is equipped with a Nilfisk hepa vacuum.

Here is a board that I milled on the c60:



Good luck!!

Regards,

Jay_Diddy_B

[brainwash]

I don't have on-hand experience but agree with most of what has been said already.
First off you need to have the right cutter, right spindle speed, right feed rate and right depth of cut. For PCB the standard (AFAIK) cutter is a carbide one. They need to run fast (>>20k RPM) and feed relatively fast and have absolutely no flex.

You can try feeding them slower but they will burn out quicker. If you feed them fast at low RPM they will flex and break instantly into shards.

If you insist on using the same cutter, I would suggest trying to do it in two or three passes, at maximum RPM. For small diameter drills you need to spin them really fast. Slow down the feed rate to 1/10 of what you have now and work your way up. I would suggest redesigning the CAD so you have rectangles with different feed rates and see which one provides the best compromise.

Regarding the machine, it seems adequate for the job, I've seen much worse machines perform better.
Work piece leveling is a must, level the bed first and then clamp the board as good as you can. Invest in a dial/digital indicator, if you don't have one, they run for 20$ including the magnetic base.

Last but not least, invest some time into some machining books or watch some YouTube channels. oxtoolco, Stefan Gotteswinter and Robrenz do precision work and there's always something to learn from any of their videos. Even 4h invested in learning will save hours of frustration and $$$ in broken tools.

[Jay_Diddy_B]

Hi,
Here is another picture of a different board.




It shows an area for a SMD part. The lead pitch is 0.5mm or 0.0197 inches.

The board was cut on my LPKF c60.

After machining, a light sanding with 600 grit wet and dry paper.

I think that the cutter is too deep on your original pictures. it may be too sharp an angle. Can you post a link to the bit that you used?

Regards,

Jay_Diddy_B

[drussell]

Spindle bearings OK?

+1 for spindle bearings being worn or something else being loose somewhere causing the obvious sloppiness....

[CJay]

OK, lots of useful info and some rather enviable boards from Jay Diddy B, suitably impressed.

I have to admit to not putting too much thought into this because I was assured by the people who run the group that they and other members have successfully made PCBs by following a user created guide sheet for PCB milling, so, from that I got:

1. Cutter, 0.4mm

2. Feed rate, 4mm/s

3. Cut depth 0.045mm

4. Guide recommends 1.25" of the cutter is left out of the spindle, I think it's flexing, will reduce this.

Pictures later.

Spindle speed of the machine is 'only' 6500RPM, could that be used with a carbide bit if the feed rate was set appropriately low?

PCB material is SRBP, not tough to cut.

After reading the replies here, examining the board made yesterday and thinking about it all:

The distortion in the squares is repeatable and varies reliably depending on feed direction, bottom row was left to right, 2nd row was right to left, 3rd L-R etc... I think this suggests the guides and frame are OK

Play in the spindle, I *think* it's OK, I couldn't detect any by hand but I need a dial guage to check properly and I don't have one, I can borrow one next week though.

I am coming to the conclusion that feed rate and a crappy worn out cutter are the main problems so today I'm going to try lower feed rates, down to impractically low rates if that helps make square squares and then decide, based on advice here, if I should order some carbide bits or if I should give up.

 I can't repair or service the machine as it's not mine so further time and any money spent would be a waste if I can't get better results from tinkering with the things I can control.

[Siwastaja]

Spindle speed of the machine is 'only' 6500RPM, could that be used with a carbide bit if the feed rate was set appropriately low?

This is way too low, no chance of it ever working. You can achieve 20-25k rpm easily using sub-$100 Dremels and similar.

[CJay]

OK, so it's a bust then.

Thaks for all the help. Will pursue other methods for prototyping, back to the chemical tanks probably.

[Siwastaja]

Both opto-chemical process and milling can produce good results, and both require skill, practice, specialized workplace and some investment in tools & materials.

IMHO, however, milling is often way overrated (treated as a simple, turn-key solution, what it isn't) while the chemical process is way underrated. For some odd reason, it seems that it's hard to discuss in civil (non-aggressive) ways; I guess people might have issues with the concept of needing "chemicals", and they like the concept of fully automated, one-step manufacturing so much that they tend to sometimes get religious about it. (edit: this is not in response to anyone in this thread.)

I've done a lot of PCBs by etching, and at 8/8 mils tracks/clearance, the copper quality is almost on par with commercial manufacturers, it's just lacking plated-through holes, high-quality soldermask and of course, multilayer.

On a limited budget, chemical process will get you way better copper resolution per money. A good etch tank and UV exposure unit are at least an order of magnitude easier and cheaper to build than a good CNC mill.

OTOH, for drilling the holes, you most likely still need some kind of CNC machine anyway, although a lot simpler machine with mechanical play that would never produce acceptable milling results, will suffice for drilling holes at exact locations, especially as you can compensate for mechanical play in software by approaching each hole from the same direction. Many of the "cheapo" DIY CNC kits are almost able to do high-quality CNC milling, but not quite; whereas they are more than enough for CNC hole drilling. For hole drilling, the base doesn't need to be leveled and the workpiece mounting is trivial as there are no sideway forces.

Since the chemical process requires so many seemingly tedious steps that each take some practice to perform correctly, not to mention the need for separate CNC drilling step, I understand very well why PCB CNC milling looks so appealing. But as usual in engineering, it's completely different to imagine a technical need or concept, and to provide a really working implementation that matches the expectations. PCB milling is often unsatisfactory on hobbyist budgets. For professional sector, some machines perform better, others require a separate operator to be hired in practice due to maintenance-requiring or hard-to-use design; I've heard from several professional sources that the LPKF machines have been especially painful and disappointing in this regard as they didn't provide the turn-key simplicity they were bought for, although, this might have changed with the company feedback and design of new models.

[brainwash]

OK, so it's a bust then.

Thaks for all the help. Will pursue other methods for prototyping, back to the chemical tanks probably.

I personally wouldn't give up so easily, there's a lot to be learned from failures. If you can spare a few cutters (I assume cheap and/or HSS) then go with max spindle rate and lower depth-of-cut, keep the feed rate the same. Your tracks are both skewed = high feed rate vs. DOC and chewed
 = high DOC and/or worn cutter. If you see older videos of semi-commercial CNC PCB router machines, they do at least two passes.
I would advise against Dremel for CNC - some clones/variants use bushings instead of bearings and even with bearings they have a lot of endplay, as opposed to a true pantograph setup (like your dentist uses).
If nothing else, seek help from a machinist near your location. But bear in mind that commercials setups use mostly chemical etching.

[CJay]

I've all the equipment for UV processing and chemical etching, I was just hoping I could prototype a few by milling and then perhaps outsource them to a Chinese fab.

Seems it's nowhere near as simple as I was lead to believe it was 'point and shoot' by the guys who run the place.

I have already booked lasercutter time so i'm not going to give up on on the mill completely, it's a handy machine even if it's not useful for PCBs.

I do think it wouldn't be the end of the world if I bought a couple of cutters to play with so if i can find some suitable i might add them to an order.

[colorado.rob]

I don't get the need for doing pcb etching or milling these days.  You can get a properly fabricated pcb for next to nothing with quick turn-around.  Hell, I just got 20 4-layer pcbs (5/5 & ENIG) prototype boards for under $100 in a week.  A 2 layer 6/6 HASL run would turn around quicker and cost significantly less.

[tooki]

…because you need it in a few hours, not a few days?

[Siwastaja]

Maybe, just maybe, there are some reasons for it, and maybe you could try to spend a little bit of your mental resources first trying to think about it for a second, or maybe search for the gazillion previous discussions, instead of requesting that we need to spell it out to you, each and every time.

People who have desperate need to fill up any DIY PCB fabrication related topics with their questioning about its point are the cancer of this forum. Bravo, this was the first and it took 21 posts before it happened. Maybe the situation is improving? I remember when it was everyday business that any DIY PCB fab related question was buried under dozens of echo chamber posts about how ridiculous it is.

Hint: One of the reasons is totally obvious and in plain sight for anyone not a complete idiot; but there are several subtle and deeply meaningful additional reasons, too, why we do (or did) it.