Double sided boards: Plated-through Holes (PTH)

[Eriond]

I'm really exited about the ability to print double-sided boards with the new software update. This leads to the inevitable question; how to make connections between the two sides.
You guys at Voltera have already shown us the LED matrix board, where you use a technique to fill the vias with conductive ink before baking the board. That covers normal vias, but how should we tackle PTH:s? We need to figure out a way for those cases where you need a combination of an electric via and a mounting hole for Non-SMT components.

First, if you would "plate" the inner cylinder of a hole with ink, the resulting diameter would be smaller compared to the drill bit that made the hole. How much smaller? Unless Voltera already has the answer, we need some experimenting. My guess is that it would be apx. 0.2mm smaller (in case the cured ink layer stays at 0.1mm). This means that PTH:s need to be drilled 0.2 larger.

Secondly, which is the most efficient way to dispense ink along the inner cylinder of a hole, without breaking the tip of the nozzle?
I guess that you could use the probe to locate and identify all PTH:s before dispensing ink on the second layer. That way, the printer/software would exactly know where to make a small dispensing circle just over the edge of the hole. Then, gently dipping the tip of the nozzle inside the hole, it could do a second circle movement. This would then smear the ink along the walls of the hole cylinder. Finally, a quick rise movement should leave an opening in the center of the hole.

This being all in theory, there are still some preconditions that should be met, such as minimum hole diameter, etc. Well, these are just my ideas, I'd love to hear yours!

[emcniece]

Having broken 3 tips already, I am pretty scared about filling vias by hand. Having the machine do this by lowering the nozzle into the via is undoubtedly more reliable, but all it would take is a single off-center drill to totally bust the tip.

Not sure what the solution is here - it seems like we're hitting the upper limits of the V1. A semi-auto/interactive via filler mode might be possible, but that would be a pile of work.

Imaginary via-assist sequence:

1. Select "Via Assist" mode from main menu
2. Load files, align with probe, swap tip to ink
3. Start Via Assist - V1 moves to first via at high-Z
4. User lowers tip with +/- buttons, visually confirming that tip is in the correct spot
5. When user confirms that tip is properly positioned within the hole (visual inspection, noting that the tip is directly over or even below the surface plane of the board) user clicks a "Fill" button
6. V1 moves to defined Z depth and begins ink dispensing/Z-retract movement. Volume depends on user preset, perhaps configurable with +/- buttons like flow
7. V1 finishes dispensing routine, retracts tip and moves to high-Z on next via
8. Repeat steps 4-7 until done

Looots of work and testing here, and I expect the team already has a roadmap of more pressing features.

[Eriond]

That's kind of what I was imagining. I have seen the probe calibrates itself using the four microswitches before going onto the board area, but I'm not sure weather its pressure sensitive just in the z-axis or if it can be used to probe x- and y-axis as well. If not, there's room for developing a new snap-on device! Maybe a camera?

[gramana]

Has anyone tried using rivets for PTH's? I've been planning to go that route for an upcoming project, and was thinking of laying down a solder pad, drill hole, press a rivet, then bake. The solder pad may well just be overkill.

[DavidYon47]

I've used rivets in double-sided boards that I've milled. They work well, though I ended up hand-soldering each side. I suppose you could reflow one pad then hand-solder the other side, but I haven't tried it personally.

The other trick I've used is to have through-hole components (assuming your board has any) do double-duty as vias where possible. Then just hand-solder the component's lead on both sides of the board.

But yes, would so love to find an automated solution to plating vias. Rivets work but can be time-consuming. There are hits for DIY plating out there on the 'net, but most are techniques to actually plate the holes and as such are fairly involved. At that point you may as well send out for a real board.

[DavidYon47]

So I just watched the YouTube video on doing double-sided boards, including using manually-dispensed ink for vias. Looks cool.

I was wondering, what is the deal with the ink bonding to copper? I'm going to have a prototype board in my near future that has QFP parts with 0.5mm spacing, which means I'm forced to use my mill to create the boards on double-sided copper. My thought was to then use the V1 to dispense paste, and somehow reduce the via count and use rivets or something.

But if I could have the mill drill the narrow via holes and use the ink-as-via-plating trick, that would be way easier. That presumes the ink would maintain good contact with the copper, and doesn't have any issue adhering to FR1 board substrate.

Any thoughts before I dive in with something simple as a prototype?

[emcniece]

@DavidYon47 it would be really interesting to see an ink-to-copper bond test. My guess is that you would have to make the copper trace a bit wider than the ink trace, overlap the ink and copper traces, then flood the overlap with solder.

@Eriond the camera idea is very interesting... what do you think about a small laser-based range finder? If the board can be scanned and precision mapped with a new attachment (camera or laser) then it might actually be feasible to have the V1 fill manually-drilled vias

Hot damn, the things I'd give for code and internal documentation access

[jazozaya]

Hi @DavidYon47,

The ink adheres pretty well to just about everything, we've tried printing on glass, ceramic and polyimide with no problems. I would just recommend making sure the board is clear of dust before printing on it

Ages ago we talked about SMD breakout boards for very fine pitch components. We mentioned it in a kickstarter update during our campaign. These breakout boards would allow us to bypass the resolution limitations of the V-One. That idea has been sitting in the backburner for some time, but now that the software is in semi-decent shape we can look at it again.

We should have some lying around somewhere - I'll see if I can dig them up.

[DavidYon47]

I'm having some success with milled FR1 copper boards and the baked-ink-as-via technique. My first attempt failed because I didn't quite ensure that the holes were completely filled (resulting in double-digit ohms for resistance for some vias), and I also ran into a problem where wiping off the excess ink leads to shorts.

The difference is that on a blank FR4 board, the ink can be safely wiped off onto bare substrate. With milled copper, you end up with a small trough around the trace(s) coming off the via pad, and it's easy for excess ink to get wiped into there. At that point it's impossible to get it back out, and once it's baked you now have a short between the trace and surrounding un-milled copper.

Solution? Only mill the holes as a first pass, then bake in the vias before doing anything else. That gives you holes surrounded by a flat expanse of copper, and wiping it off is easier, with almost no downside to some excess (as it will just get milled off just like the copper does). So I also more carefully ensured that the via holes were fully loaded with ink, and it's now baking.

I'm a little concerned all that copper will act as a thermal and adversely affect the baking, but definitely worth trying anyway. I'll post an update once I finish this 2nd test board.

[DavidYon47]

Well, not so much. The problem seems to be reliably filling the hole and also having it come up and over to cover the via pad a little bit. I was hoping that if wiped flush then the ink would at least come into contact with the inside border of the pad (and maybe a little bit underneath), but that doesn't seem to happen.

Plus it's very hard to tell when you'd really filled the via. So when baked you can end up with a bit of a divet in the via hole and there's no continuity whatsoever. I even tried manually dispensing some solder paste into those botched holes and reflowing via an air gun, but that doesn't work very reliably either.

When it works, it works well. But at the moment all the workflows I've tried tend to degenerate into a complete time sink, and has a 50% success rate.

The technique seems to work well on V1-only boards because you're filling a "cup", where the bottom is cured ink, and then on the 2nd pass the printer will dispense a complete pad over the via, allowing more tolerance for via holes that may not be completely filled. But on milled copper you don't have any of those advantages.

My next approach is tried-and-true 30-mil rivets, with the hope that maybe I can convince the V1 to dispense solder over the the via pads, which will reflow the top of the rivet to the board. I can either manually dispense paste to the bottom pad and use a hot-air gun afterward, or if I'm lucky I could manually dispense and reflow it all at once.

The latter will probably be too tricky so I'll probably end up going with a second reflow with an air station.

Was certainly an educational experience. There's probably not a lot of people trying to mix fabrication technologies as I am, but if anyone else solves this puzzle, I'm all ears, It's a promising way to solve the DIY PCB nightmare that is vias.

[DavidYon47]

Oooooo..... very cool. I tried my first Gerber export from Eagle for the purpose of solder paste. Per my previous post I was hoping to have the V1 dispense at least the top layer for the rivets. Of course, with the default CAM file that doesn't happen. So I went to the Top Paste section, and selected layer 18 "Vias". After processing the job and loading the resulting Gerber into the Solder option, the Vias show up!

There will be the problem that solder paste gets dispensed into the hole, but who knows, maybe that will cause the rivet to push enough paste to the bottom such that I don't need to manually dispense. (yeah, that's probably too good to be true)

I'd give it a whirl but I don't have the parts gathered for pick-place/reflow. So this experiment will have to wait given the the time.

[troy]

Hey (again) David,

Can you share your circuit or a stripped down version of it here or send it to support@voltera.io. No promises I can achieve anything better, but I'd like to at least understand what you're up against.

[DavidYon47]

Sure, it's just a simple ATMEGA328 throw-away project that I have been using to do proof-of-concept with various workflows. Enough small parts to be a bit of a challenge, but nothing that bogged me down circuit-wise. I basically started with Sparkfun's Arduino Pro Mini (5V/16Mhz), trimmed off some stuff, and added four LED's.

When I was trying to dispense solder automatically on the vias (for the eyelets) I ran into the problem of the Z-probe diving into a via as if it were a pad. That was 0.6.5, and I know there have been improvements since then. Don't know if 0.7 fixes that, but for the moment I manually dispensed paste on both sides of the vias.

Keep in mind (per my earlier message) that I tricked the CAM job into including layer 18 in the Top Paste section, so this is a slightly altered situation from what others were reporting. If using ink for vias is infeasible, it would be nice to be able to dispense paste on the via pads without risking a probe jam, but I don't know if a feature on the board being both a hole as well as a solder pad would still confuse 0.7 into attempting to probe that feature.

Anyway, here's the circuit in question:

Arduino-Pro-Mini-v14.zip (36.7 KB)

[DavidYon47]

If it helps, here are pics of the board. Short of having to jumper a broken trace, it worked the first time. Notice the things that look like plated vias are actually eyelets/rivets which were reflowed with solder paste underneath their flange.

Even though I milled this (and with the ATMEGA's 0.8mm pitch I could have probably used the Voltera), the solder-paste dispensing was was an enormous help.

[jazozaya]

Hi @DavidYon47,

Thanks for the pictures.

Did you have any problems with the paste adhering to the copper? Or any shorts in the Atmega? Did you also reflow with the V-One ?

Also - to clarify, did you insert the rivets after milling but before the solder paste ?

Do you have a recommended supplier for them? I wouldn't mind checking them out

Cheers!

[DavidYon47]

Yes, I do find that the paste has some trouble adhering and ends up being something of a dot in the center of the pad rather than being a flat layer covering the pad. There was one pad that needed a manual shot of paste from the syringe. There were no shorts anywhere.

But compared to the KS8999 I'm planning on using, the Atmega is downright quaint. I do have worries about the paste on 0.5mm copper pads, which will not only be present on the controller IC, but also the SMD HDMI connector that's required for this prototype. Those pads will be pretty small if I end up doing manual fixups from the syringe, so the paste dispensing pass has to be pretty reliable.

Would using something like 500 grit to slightly roughen the copper before dispensing help at all? Or is it simply a materials problem?

I milled everything first, then used automatic dispensing on the normal pads. I then manually dispensed paste onto the via pads. Once that was done, and the board was still clamped, I inserted the rivets. That was tricky because they require some insertion force and I inadvertently wiped off some paste off nearby pads. I then placed the other components.

The other (minor) issue is that the rivets protrude slightly below the board, which meant they wouldn't fully insert when the board was clamped. So once I placed everything, I needed to hand-hold the board while pushing the rivets the rest of the way through so that the flange would bottom-out against the paste and pad.

At that point, I no longer have a board that sits flat on the heater plate. It might have reflowed ok on the plate, but I opted to use my DIY oven instead.

I'm using the 30mil eyelets from T-Tech, which aren't cheap but are very high quality:

http://t-techtools.com/store/index.php?main_page=product_info&cPath=9&products_id=40

They are also the only supplier I know of that makes eyelets in that size.

[mewertow]

I've seen issues with poor adhesion to copper if paste in the nozzle has begun to dry out - this typically happens if the dispenser is left out for extended periods without a cap.

The fix is pretty straightforward. Try to dispense some solder paste manually, and observe how it flows out of the nozzle tip. If the paste has begun to dry out, it'll flow slowly, and you should start to see a slight colour change as the dry paste is cleared from the nozzle. Once the dry paste is cleared, the adhesion will be much better.

[DavidYon47]

Interesting, I did end up just changing out a nozzle because I was getting almost no flow at first. Didn't know I could just flush it out.

But even with a fresh nozzle, I did have adhesion issues. That said, I very well could have had the flow rate and Z-height not calibrated optimally (although I played with those values 4-5 times before dispensing the rest of the board).