Variations in copper thickness on PCB messing up PCB milling.

[cj7hawk]

Hi All,

I thought I'd share this as it's done my head in over the past two years. First of all, the problem - Copper between traces not being removed cleanly, so that I cannot use a PCB after milling. It's a huge problem and it keeps coming back to haunt me, and I've tried almost literally everything to solve the problem.

I improved the locking of the bearings in my mill.
I tracked down issues between the CNC software and the stepper drivers I'm using.
I have gone through every possible combiation of autoleveller.
I migrated from normal v-tips to ball tip tools ( This one was actually worth it - big improvement in PCB quality using 2mm long 0.4mm ball-nose end mills... They make great clean PCBs and cost barely more than quality V tips. ).

I've even dropped the Z by another 2 thou, and it still doesn't cut through in those areas... Something is seriously wrong.

And still, I'll produce some perfect PCBs and some that look like the autoleveller failed.

I had assumed there were three possible areas where the fault could still lie.
1. The Milling Software is screwing up the depth calculations.
2. The PCB is moving, or the surface level is changing ( though areas of the PCB with problems tend to remain consistent no matter when milled in the process )
3. The copper thickness is varying wildly. ( I really did not expect it to be this.... )

Now, normally I figure 3 thou for normal cheap PCB should be enough... That's about 75 microns, but the copper should be around 50 micron right?

So today, I decided to eliminate the final uncertainty.

After three failed PCBs in the same place, that wouldn't mill through the copper with an 8 thou cut, I went back and did a re-auto-level of the place where it failed, since it was right at the edge. I'm still within 1 thou of the original measurement, which given I've moved a few though on the y axis to remeasure, seems reasonable.

Then I manually move the mill to the location, and manually drop 1 thou at a time until the copper is touching the bit and the probe light shows.

Then back so it goes off.

Spindle on, plunge 5 though... Cut... Copper remains. Drop another 3, Copper remains.

Pull the PCB and place under a microsope.

Copper cuts OK...



Copper doesn't cut OK.



Notice something different? Same PCB, just an inch apart.

Copper varies from 2 micron to 8+ thou thickness over the PCB. From about 50 micron to 200 micron... That's a lot of variation !

NOW I finally know - and it's mostly near the edges, so I only really noticed when I started cutting large PCBs that take the full PCB segment I install.

If anyone is having a similar issue, and can't explain it any other way, and is getting cheap PCB from Ebay to mill, do some manual experiments in the areas it's not cutting and see if your copper thickness is the same all the way through.

Regards
David

[PlainName]

Sounds like the PCB is flexing - it's not flat and your auto-level correctly detects the top of the copper, but when you push with a tool the PCB is moving down slightly.

I use a pretty crude setup and have no problems. Took a lot of dicking around to get here, but it works for me. I've explained elsewhere the parts that I think are important, but to recap:

1. Use carpet tape to stick the PCB to the mill bed. Carpet tape is thin and very consistent in thickness, with practically zero give. It ensures the PCB is flat and level (but see below).

2. Use oil to reduce cutter wear. I use 3-in-1 but any similar should be fine. It massively improves the cut edges (nice and smooth as opposed to jagged and bitty), makes the tool last lots longer, and captures the fine FR4/copper powder that's being churned out. You don't need a lot, and the small size of the cutter tip means there's no spray.

3. Use a 90-degree cutter. Sounds a terrible idea since the slightest change in depth will obviously make a bigger difference to width than with a fine 30-degree one. All I can say is it works for me - doesn't break or flex and you don't have a stupidly deep V just to get a decent cut width. I splashed on a set of LPJK cutters (part # 101010-4) which are usually blindingly expensive but I got cheap on Ebay. Anything similar should be fine. Also got their 60 degree cutter, but the 90 seems to produce more consistent results.

4. Slow feed, high RPM. The diameter of the cutters is very small so you need very high RPM to produce a decent cutting speed. You probably already know that but it can be non-obvious and easily overlooked. My drill runs at around 15krpm so the feed is low.

The above is about running a job, but your mill must be set up well first. Relating to PCBs, the bed should be dead square to the Z axis (so sticking the PCB down works and you don't have to faff with compensating depths on a grid). My preferance was to bolt down a sheet of perspex and flycut it to ensure it is flat and square. It also acts as a sacrificial buffer when you drill holes in the PCB. With a bit of practice you should be able to mill out the PCB leaving just the carpet tape holding the parts together, but going too far and making grooves in the perspex is a non-issue.

Regarding your ball-end cutters, those are probably OK and similar to the 90-degree ones when it comes to cut depth and width. But they probably wear out fast without the oil, so you'd get a decent job and then with the same setup the next would have problems. Presumably they are carbide? HSS ones would likely be killed pretty quick by the FR4.

[cj7hawk]

I thought the PCB was flexing also as a possible cause, but that is not the case. Well, it may contribute to the bit skipping slightly when pushing through thick copper, but it's only a contributer at best.

I forgot to mention the tool is a ball mill of diameter 0.4mm - And in the secon picture you can see a neet hemispherical hole where I let the bit sit for a while before lifting, and it's cut a hemisphere clean out of the copper - which means the copper there is about 0.2mm thick, or around 8 thou!  In other places, it's only 1 thou to 1.5 thou... It's definitely a copper thickness issue. The second picture were cuts I made by hand by measuring the thickness to the surface then lowering 5 thou and then 8 thou and cutting.

Also, the problem occurs only in small patches - eg, I'll get a section of PCB about 2cm x 2cm that cuts less and less, then doesn't cut through. But either side of it, the cuts are absolutely predictable right to the edge, so the mechanical stress pushing the PCB isn't much different between the two.

I have been watching people using oil, and wondered about it. I get pretty good dry results though and can leave the anti-oxidation coating on the copper in place, which also makes it solder pretty easily. It's just these small patches where suddenly I can't remove copper for some reason. Originally, I was suspecting some cross-over between Z and X axis due to the way it manifests, and can be somewhat random, but it didn't make sense for it to be so consistent between cuts - eg, a batch of PCBs all had the same place cut thin, regardless of what I printed, or where it placed the probe data.

I eliminated that possibility by using a probe file and converting the autoleveller output to fixed numbers rather than formulaic ones.

I feed about 10" per minute as any faster than that, and I can't control the inertia in the cuts so well, and I get momentum from the motor on sharp edges causing issues, but I can cut clean tracks up to about 20" per minute at 12000 RPM, which is the limit of my modified mill before the cuts lose quality. ( Has a brushless air cooled - the older kind ).

I've had good results with 90 degree cutters before, and also with 10 degree cutters before, but a long-time run with a big batch of bad PCBs from Ebay caused me the most headaches over the past two years and I couldn't understand why I was getting perfect results before and not later. I make tracks as fine as 10 thou regularly with 15 thou clearance, so a 90 degree cutter may not be a good idea. But a 0.4 diameter ball mill of length 2mm is great because it cuts the tracks, mills the holes then cuts out the PCB in a single operation. I usually use bakelite PCBs because they are less wear on the cutter than fiberglass, but have been cutting a lot of fiberglass out lately with such a tool ( which sometimes breaks tips at 10"/min, so I need to slow it down a bit or make shallower cuts when cutting out and milling holes... ).

I might dig out some of the 90 cutters now you've made me think about it, since they would make a good base for testing the new PCB I ordered... If I cut 10 thou through, I can check copper thickness all over the PCB quite quickly as a test run.  If the copper on the new batch is bad, that should show it up a lot more easily than my other test.

The ball cutters are a mix. I tried a softer tool ( one for cutting copper )  yesterday, and a few hours on FR4 and it did well, but was struggling a bit towards the end, and I cut shallow since I had more confidence in the PCB, so the problems were near the very edge (within a few millimeters) where the PCBs appeared to be guillotined and had some slight warpage... Mostly it was OK and clean though.

The bits I use mostly though are 62 HRC, R0.2x4Dx50L-L2 and the manufacturer claims they are a carbide tool suitable for cutting carbon steel, but I am not an expert on tool selection. L2 I find is just long enough to cut the PCB out and mill the holes.

[Doctorandus_P]

I forgot to mention the tool is a ball mill of diameter 0.4mm -

It's quite likely that this is your problem.
Ball mills do not cut well in their center, because that is a horizontal flat surface, with no room to make cutter edges, because those all come together in a point. Also, with a radius of zero, the cutting speed is also zero, (regardless of rpm).

Try some single lip cutters, or other conical cutters which do have a cutting edge at the bottom.

Something else you can try is to rotate your milling spindle. As the cutter is a ball, it can be rotated without penalty in milling geometry, but when you rotate the spindle, you start cutting on the side of the ball.

[beanflying]

These are worth a go over the more generic split point cutters. https://www.aliexpress.com/item/32995987952.html?spm=a2g0o.order_list.order_list_main.217.26201802kelOil There is the more up market spiral fluted version of these but lots of $ to get them into Oz.

You still need to 'plunge' in a little as already mentioned the middle of a cutter does nothing but rub and generate heat. Ball mills are however great for 3D contouring but not so much for removal of copper.

[djsb]

We have had good results in our LPKF S62 with bits from this company.

https://mctinfo.net/products/tools-and-drill-bits-lpkf-t-tech-mits-accuratecnc

We have used the MIL 8-20 90 degree universal cutter, and it works really well. We can't use these on our S103 though.

HTH.

[Kean]

It should not be too hard to peel some of the copper layer off the PCB at the edge and measure it with calipers or a micrometer to see if it is significantly over expected thickness.  Score around a square area with a knife and grab the edge with some sharp side cutters.

The measurement doesn't need to be high precision to confirm your hypothesis but should be done using a tool that has better than 0.1mm resolution.  You could also peel some off from a scrap piece that was further from the edge for a comparison.

[PlainName]

I don't think copper thickness is significant. The cutter will cut relative to the FR4/copper interface, so a 0.01mm deep cut will be 0.01mm deep into the FR4 regardless of how thick the copper is. The only effect it might have is in determining the cutter height to start with since that would be set by the touch automation (or manually). However, I think it would be obvious at the start if that was occurring.

[beanflying]

1 oz copper is about 0.03mm and most of the variation in height you are trying to mill into is in the FRP.

[Kean]

Oh, I agree that it is unlikely to be a copper thickness issue.  35um copper is standard and unlikely to vary much even in cheap PCB.  When I used to mill PCBs quite a lot (before access to cheap PCBs via JLCPCB et al), I found that the big difference with cheap PCB is that it is rarely flat.

Using autolevelling doesn't work well because a warped PCB will flex so easily under milling forces, and this will be exacerbated by using a ball mill (as explained above by Doctorandus_P), and why V cutters are so popular.

As I found the autolevelling process pretty poor (talking many years ago) I found the best solution was good quality PCB (I found RS supplied some good stuff back then), very thin double sided tape, and a 90 degree V cutter.  I would regularly resurface my sacrificial plastic base board to ensure it was as level as possible.

[HoadySnitch]

   Sounds pretty similar to my experience, especially with cheap phenolic
board stock from AliExpress. One track is routed and another not more than
.25 " away didn't clear the copper to the board, just scored it with 
a Candle Height map.
  So without resorting to microscopes and such, here's a  sort of
band-aid workaround with which I've saved a few boards.  To use it
however, your routed board must remain clamped to the bed of your
CNC machine unmoved. Do your ohmmeter checks or visual inspection
with the board in place.

...pardon the jerky production values.

https://www.brighteon.com/c569714f-fb7b-40b7-9bcf-1e0e630a83bd

Github link + more info there.



    Another method for phenolic board which worked surprisingly well
was just to tap two traces showing a short with a several amp 12v
source. Blasts out the small bridges with a visible spark.
    Though I should have used a charged capacitor bank for the
fr4 stock because the 'fun with electricity' method curled up some traces.