laser ablation of PCB resist

[(*steve*)]

Armed with my normal settings I headed off to do another board. On arrival I noticed I had left the painted PCB at home.

After finding another piece of single sided PCB and some paint, I decided to just try it.



So here it is before being etched. Because this was new paint (and not long dried) I tried a number of different power settings.

It looks real ugly.



And here it is held up to the light after I stopped etching it. It's terrible!

However note the fine resolution in some of the pre-etched boards. This paint seems to give me the opportunity for better results

Spoiler

My next post will show you how good. But I have to go out now...

 

[(*steve*)]

As promised... An update on what I did yesterday.

First some technical details for anyone who might want to repeat this. I was using an LG500 40W laser cutter/engraver. Details here. The paint used was "Fiddly Bits" High Gloss Enamel Grey Primer as seen here. This seemed to go on thicker than the "aluminium" colour I used on the previous test. Only a single coat was applied (three coats of "aluminium" were used in the previous test).

The previous test showed very well how hard it is to control the ablation of the paint when the laser speeds up and slows down while drawing complex shapes. In order to keep the laser speed constant, I went back to engraving -- essentially rasterizing the image.

I examined three variables: cutting speed, scan gap, and number of passes. All tests were done at 100% power (40W).

Each board is almost exactly 15mm x 25mm

The 8 tests were:

1. 100mm/s, 0.1mm gap, 1 pass
2. 100mm/s, 0.1mm gap, 2 passes
3. 200mm/s, 0.2mm gap, 1 pass
4. 200mm/s, 0.2mm gap, 2 passes
5. 400mm/s, 0.2mm gap, 1 pass
6. 400mm/s, 0.2mm gap, 2 passes
7. 400mm/s, 0.3mm gap, 1 pass
8. 400mm/s, 0.3mm gap, 2 passes

After cutting, the board looked like this:



In this orientation, the tests go

1 2 3 4
5 6 7 8

The time taken to cut these boards on the laser cutter were (approximately):

1. 60 seconds
2. 120 seconds
3. 30 seconds
4. 60 seconds
5. 20 seconds
6. 40 seconds
7. 15 seconds
8. 30 seconds

You may notice the subtle difference in appearance between the single and double pass tests. This turns out to be significant and from here on in I'll only bother to show the two pass tests.

A closer look at tests 2, 4, 6, and 8 follow: (edit: 8 is correct now)



If you take a look at these you can clearly see the effects on the change in scan gap, but the differences due to speed are a little more subtle. Because the scanning was bidirectional, you will notice that in the higher speed tests the alternate lines are not fully aligned. The scanner can do single direction engraving which would fix this at the cost of reducing the effective cutting speed.

The major difference with the change in cutting speed is not visible from these images. The relative effective power when doubling the scan speed is half, so if test 1 (100mm/s) is assumed to be at unity power, test 2(200mm/s) is at half relative power, and tests 3 and 4 (400mm/s) are at 1/4 relative power.

Test 6 (the third of the tests above) has a couple of visible blemishes. These were caused because in the previous test the "aluminium" paint proved to be extremely easily scratched and some etchant had come into contact with this part of the board. The paint used in this test exhibited the same delicate nature. Anyone using this paint needs to be aware of this and treat the painted surface with care.

Etching was performed in hot ferric chloride with both agitation and gentle brushing. Test 2 completed very quickly (followed by 1), and whilst far more vigorous agitation and brushing were applied to the other tests, etching was abandoned because it was simply taking too long.

Here is the board after etching:



In this orientation, the tests go

5 1
6 2
7 3
8 4

I may have done a little more etching after this photo, but nothing changed significantly.

When held up to the light, it looked like this:



In this orientation we're back to:

1 2 3 4
5 6 7 8

It's pretty clear which the successful test was! (test 2: 100% power, 100mm/s, 0.1mm scan gap, 2 passes).

After removing the paint with acetone and another wipe with isopropyl alcohol, the board looked like this::



It's very easy to see the cutting depth (or lack of it). This is in the same orientation as above.

And here are tests 2, 4, 6, and 8 again:



Test 2 is clearly successful.

If you've got this far and wondered why test 1 doesn't have the centres of holes marked, it's because I cut the wrong file :-( It was identical except for those "pilot" holes, so I used it.

Things to look at next:

• It's pretty clear that multiple passes are the way to go.
  
• The second pass seems to remove a large amount of what is left behind in the first pass.
  
• It's also pretty clear that a fast, high speed initial pass can remove most of the paint, and that a second slower pass could be used to remove the material left behind.
• I don't think there's much to be gained from a scan gap smaller than 0.1mm
• I think there may be a lot to be gained by moving to the large (120W) laser cutter.
• Building the device from boards 1 and 2 so I can use them

 

[(*steve*)]

It's possibly also worth pointing out some of the other details involved in the preparation of the files for cutting and engraving.

The steps are basically:

1. Draw the PCB
2. Output milling files
3. Do modifications to the output file as required
4. Cut/engrave on the laser

However there are some differences if you go for cutting vs engraving.

I assume you are familiar with your software and your laser cutter. If not, become so FIRST.

Cutting:

1. Draw the PCB (same in all cases) but minimum clearance needs to be sufficient for the cutting path generated in the following step to cut between pads I used 0.5mm for an initial tool width of 0.45mm. I also recommend you place a trace around the outside of your board. This allows easy editing later to produce a rectangular (or whatever shaped) board.
   
2. One or more output files are required. The miminum is 1, which should produce fine cuts separating pads and tracks. The problem is the potential for solder bridges and inadequate distance between high voltage tracks, so you may wish to do multiple passes each with progressively larger tool width. If your software supports it, one of the milling files (I suggest the first) should mark the centres of each drill hole. I used 0.45mm for the first cut, and additional spacings 0.55mm thereafter. 0.45 for the first cut seemed about right.
   
3. If you have produced multiple HPGL plot files you will need to append them together and then edit them to remove any initialization sequences from all but the first file. For other file types you're on your own. When asked to mark the centre of each drill hole, my software just does a "pen down" followed by a "pen up" at the point. This does not produce a cut on my laser cutter so I have written a program to convert such single points to small squares. If you have placed a track around the outside of your board, it is now time to remove any milling paths (except maybe one) outside of it.
4. Load the job into the laser cutter and cut. You will need to be very careful with the corner power to ensure that the cut width does not get larger in corners and around bends.

Engraving:

1. Draw the PCB (same in all cases) but minimum clearance needs to be sufficient for the laser to turn on and off. (and larger than the scan gap). I used 0.25mm, but I think I could use less. I also recommend you place a trace around the outside of your board. This allows easy editing later to produce a rectangular (or whatever shaped) board and it may be required to get the correct engraving (i.e. not a mirror image of what you require).
2. Produce a single output file. Mark hole centres if possible.
3. The output file may need to be edited as above to handle the hole centres. Note that in this case the size of the square generated needs to exceed the scan gap. You will also probably need to remove the single cut around the outside of the bounding line on your board to ensure you get a positive rather than a negative image.
4. Engrave on the laser cutter. Choose a power, speed and scan gap appropriately.

 

[(*steve*)]

Doing another board just to prove it's no fluke



Yeah, some are a little under-etched.



But they look pretty good.

And another test to try to get traces between pins of small outline chips (in this case SO-8.



The successful test (2 passes at 100% power with 0.1mm scan) is pretty marginal.

 

[(*steve*)]

I will give more details later, but acting on a tip about the use of a different type of paint, I have the following etched board:



The paint edges are far cleaner leading to thinner lines being more practical. In addition this only requires a single pass at a faster speed and lower power.

 

[(*steve*)]

After doing a bit more work, I have managed to get reliable ans repeatable fine detail



The three pads in the top left corner are a SOT-23. On the far right are pads for three SOT-363devices and the thing that looks like a spider on the bottom centre us for a uSMD-14. The really small set of 8 pads are for a 4x0402 resistor package. It's a little larger than an 0803 but has 8 connections.

I'm very very happy at how well this has turned out.

I will give details later...

 

[Arouse1973]

Wow very good Steve.
Adam

 

[(*steve*)]

Now for some specific details of how I achieved this and how you can replicate it with your very own laser cutter (everyone has one, right?)

The paint used to coat the boards is critical. Different paints yield vastly different results. I am currently using "White Knight High Temp Pot Belly Black self-priming heat resistant paint". It claims to be suitable for up to 300C. The board is prepared by cleaning with acetone then isopropyl alcohol. A thin coat is applied and allowed to dry. After it becomes dry to the touch (about 2 hours) the paint surface is VERY delicate. Leave it for 8 hours or so if the painted surface is going to come into contact with anything. After is fully dries, the surface is far harder and can be handled with far less caution. A single 300g can looks to be sufficient for about 1 square metre of board!

There are a number of other high temperature paints, but there are only so many variables you can look at! Other paints may be even better! (I used the highest temperature paint I could find -- which was also the most expensive, costing about the same as a 300mm square piece of unetched double sided fibreglass board.

The laser cutter I use is a 40W CO2 laser.

Originally I was using the laser to "cut" the outlines, but because this results in continuous changes in the speed of the laser, the cut width varies and so fine detail is not possible.

"Engraving" allows the use of a constant speed, but requires that the laser be turned on and off very quickly. Electronics is faster than mechanics!

From observation of what happens when power is reduced, it appears that the laser turns on much more slowly at low powers than it does at high powers. This change in speed of the laser is more than enough to compensate for a faster cutting speed. Thus operating the laser at 100% power and 800mm/s yields better performance than 50% power and 400mm/s.

The next task is to determine the effective beam width at your chosen power/speed. This will vary with the type of paint you're using, the focus of the laser, and the power/speed.

Simply engrave a small area, setting the gap between passes to be 0.1mm (say) and then increasing or decreasing this until you see only very fine lines of paint between the passes. Etch these and evaluate the result. Ideally you want to pick a width that leaves the thinnest of unetched copper traces, preferably non-continuous. This distance will be what you use when creating the milling files. It will be your initial "tool width".

In my case the "tool width" is somewhere between 0.1mm and 0.15mm. For safety I could choose 0.15mm, but I have used 0.1mm and plan on avoiding details less than about 0.2mm wide (he says, even after the board above has 0.15mm traces).

Note that changing power/speed will cause this to vary. You may want to go through several iterations of this using different speed and power settings until you discover the "sweet spot" for your laser.

I have arrived at 800mm/s 100% power, and 0.1mm. I think there is scope to tune this even further, but at some point you've got to realise that if you can make boards you have no current hope of hand assembling that you've gone far enough!

Beware that if you tune your parameters too close to the limit that a slight change in focus could ruin your day. In my case, the beam width is slightly smaller than 0.15mm and the cuts are very reliable. However the laser is used by many people and I want this to work even if the focus isn't as good as it should be.

The next thing you need to do is determine the distance between laser passes which will give sufficient detail for the pads you're using. Note that your laser may allow unidirectional or bidirectional engraving. For finest detail you'll probably want to use unidirectional engraving, however this will double the time required. The distance between passes will be smaller than your tool width. This will ensure that none of those small traces remain between cuts.

In my case I found that for 0.4mm or larger traces and 0603 or larger pad sizes that 0.1mm between passes was fine. This could also be done bidirectionally. For the finer BGA and 0201 pads, 0.05mm between passes was required and it has to be done unidirectionally.

The difference between these settings is 28 seconds for my test board vs 1 minute 45 seconds. When you pay for the laser by time, this can make a real difference!

In order to test this properly you need to design a board that has your finest detail positioned both horizontally and vertically. If you're the type of person that placed things at angles on the board, try them too. It is also important to have at least some traces at 45 degree angles.

Once you determine what your practical resolution is, you should go back to your PCB tool and determine what the minimum clearance between copper areas should be. This is done in the design rules, and often this allows the tool to check your layout for traces or pads that are too close together. In my case, I found that 0.2mm was appropriate.

Note that there is some interaction between the clearance, tool width, and scan gap. There needs to be sufficient difference between the two that the laser can turn on and off reliably when traversing across these gaps, and at least one engraving pass will be made between structures when the etching path is between them.

I have some unsuccessful tests that can demonstrate how things can go wrong...

Oh, another thing is that your initial tests can be done on MDF. The advantage with MDF is that it's cheap, and that you can see differences in cutting depth.

 

[(*steve*)]

Here are some closeups (sorry the quality isn't that great) showing the top right corner of the board with cutting settings of

100mm/s 100% power, 0.2mm gap, unidirectional and 0.1mm tool width



And with the same settings, but bidirectional:



As you can see, the etch is not complete

You can also see the skew between scans that is present on the bidirectional vs unidirectional settings.

What you can see between the closely spaced pads is that the laser doesn't cut as much. I don't have any proof, but I suspect that as you decrease the resolution in one direction, the software decreases it in the other as well.

You will note that the paint that remains seems to be far wider than the actual cut. You might be wondering why I estimate a tool width of between 0.1 and 0.15mm. There are two reasons. Firstly, some undercutting occurs when etching, and secondly, you can see how the paint becomes dislodged between the cuts.

If fully etched, this would leave thin strips of copper. In fact, even where you can see copper, much of that would not etch, but the remaining strips of copper would be far narrower than they appear.

Changing the gap between passes to 0.1mm causes a dramatic improvement.



Here you can see that there is no trace of any lines in the large etched areas. You can notice small ridges in some places where the laser was turning on and didn't quite come up to sufficient power to join up with the adjacent scan.

Between the pads, there is some cutting, but the laser power was insufficient to cleanly remove the paint.

However, these settings are fine for footprints just marginally larger than the SOT-323's shown here. The traces here have a design width of 0.15mm, and from the ridges (that are 0.1mm apart) you can see that they're pretty close.

If this cut is done bidirectionally, the results are:



This adds more problems. The 0.15mm trace, while probably not compromised, is certainly on the way there.

I consider these to be good settings for 0603 or larger. Below are the 0402 pads from the same board.



Halving the scan gap again to 0.05mm (unidirectional) produces the results seen in a previous post. Here is a close-up.



Don't you hate camera shake? These are the settings I use for fine detail.

If I did this using bidirectional, I get:



So close, but no cigar.

So, what happened to my initial calculation of the tool width of 0.15mm? Well here is a cut done with the same optimal settings:



The problem is that whilst it works in the horizontal direction, it fails in the vertical direction. Reducing the tool width slightly allows the laser to more reliably fire between these pads, and whilst this affects horizontal structures, the lines cut in this direction are much cleaner, so you can live with it

 

[(*steve*)]

Long time between updates...

I'm routinely using the laser cutter to make prototype boards for myself and others. I'm kinda working on a Gerber to go plot file converter to make everything more generalised.

But most recently I've been chasing down tinning solutions.

I think I posted some results of testing a formula I found on line, but I had to compare it with something.

Here's something I bought on eBay:



I made up a small quantity of this and I was really disappointed. After leaving the boards in the solution overnight, not all of the copper had been covered. I took the boards out and cleaned them in isopropyl alcohol just in case I had left great fingerprints all over them. After leaving them in the solution all day, the results were not much better.

So I've moved up another small batch of my recipe, and even after a few minutes it looks better!



To allow me to use only a small quantity, I place the boards in a zip lock bag with the solution and exclude as much air as I can.

I'll leave this overnight again, but it already looks pretty good.



There's my first test with this solution, the bit on the right left overnight.

 

[(*steve*)]

Hmmmm, I didn't remove the paint on those very well :-(

They were done so long ago that the paint had really cured. And I couldn't find any steel wool.

 

[(*steve*)]

Some people have suggested running solder all over the board.

Maybe I'm biased, but it's not what I call quick or fun.

Here's two boards, one with suffer in the tracks, the other with chemically deposited tin. The one with chemically deposited tin was a lot easier to solder to too.



Maybe I should clean them ;-)

 

[Hopup]

I wonder how powerful the laser would need to be at minimum to still get somewhat good results?

 

[(*steve*)]

I wonder how powerful the laser would need to be at minimum to still get somewhat good results?

It's actually a maximum more than a minimum. A higher power laser makes it harder.

Something under 60W would be my preference (unless it has a really small spot size and a very fast head).

Lower power lasers will take longer. The minimum power would be where you can't it to remove the paint.

I've seen blue lasers used to expose photosensitive board -- that can be done with far lower power lasers.