Saturday, July 15, 2017

Using PCB rivets vias for homemade double sided PCB

You may remember in a former article the super tiny rivets I bought for double sided prototyped printed circuit boards (these but check those, they are way cheaper!). I realize that I never documented how I used them. So here it is.

The "raison d'être" of these costly rivets is specifically to help doing the vias that route the signals from one side to the other side of the board. The industry does plated holes with a chemical process that deposits a layer of conductive metal on the inside walls. It both conducts the signal but it also really helps making the board sturdier, as the copper rings (well square, here), are bonded together. Without it, the risk to tear apart one of them is much more important. Actually, a few makers do it also but it is both difficult, lengthy, messy and specially risky with a lot of nasty chemicals... not for me since I mill my PCB ;)

So in general, making double sided PCBs at home means there is no such plated holes. So we usually solder a thin wire through the hole on both sides. It is time consuming and it stresses the copper pads a lot, especially when the damn wire falls down when both sides melt again when you just want to solder the other side ;)

Top and bottom layers: making vias by means of PCB rivets on a double-sided milled PCB.
The milling job is quite bad here: the traces were not cut properly on the top layer (left side, check why here),
and the layers were mis-aligned after the board was flipped over to mill the top side... I need more practice.
Also, there are better and expensive tools to fix the rivets, but this prototype worked fine in the end anyway ;)
Now, rivets still help to make more robust single sided PCBs, most notably for the connector holes that will be subject to mechanical stress. Rivets will protect and keep the small copper "rings" tightly bounded to the PCB support material, so they will be less likely torn away.

Most notably, dirt cheap bakelite boards give no second chance in this respect, while fiberglass (FR4) endure more abuse and re-soldering. I guess the bonding is much stronger with the latter.

Which PCB rivets to buy?

I recommend 0.8mm outer diameter (0.6mm ID), because there their heads leave just enough room to route a trace between them. Hence in EAGLE CAD, I make sure to configure my vias to be 0.8mm accordingly, so they fit tightly.

You probably cannot design your PCB with the "regular" smaller vias (~0.4mm) unless you can find smaller rivets. The smallest I found were tiny 0.6mm (OD) which are already harder to secure I guess (I would try to use a pin?). Larger rivets will impact on your design as you may not be able to pack as much vias on a given surface as when using wires, or professional PCB etching.

But even when you send them to a board manufacturer later you still benefited from the valuable home-made prototyping. And tweaking a design afterwards is easy. Making it work in the first place is harder!

So there are different sizes from 0.6mm to 1.9 outer diameter. I did not find one with an inside diameter of 1.2mm: they would make life much easier for headers and connectors that will not fit in the ones I bought. Sadly indeed, unless you go really big (1.9 outer diameter!), the rivets may not help you with headers.
Source: the seller I bought from in Greece ($17+S&H for 200 of them, huh),
but there are others, and cheaper on aliexpress.
Now, they are expensive for sure, but they really are worth their price IMHO.

Routing through-hole pins on a double-sided milled PCB

The best thing to do is upfront: route all the traces so that through-hole components can be soldered only/mainly on the under side of the board. This is a general issue, not specifically related to rivets, though they can help a lot here.

Notably, do not use a component leg to route the signal from one side to the other, because there is no plate hole, and it will be hard to solder them on the top side of the PCB though you will have to for the signal itself (this really is error-prone by the way!)...

So better think about it during the design itself, and possibly use explicit, deported vias for the signals. I draw a quick example below (blue in bottom, red is top):
Here is a design tip to help soldering an ICSP header plug. When your PCB holes are not plated, which is
the case with homemade PCB, you want to avoid having to solder legs on the bottom and on the top layers.
The latter is obviously difficult because the component body is in the way.
The problem arises also when a through-hole leg is used to route signals from one plane to another.
So better design your layout with this in mind, possibly by adding explicit vias to do the job.
You bought vias? Use them instead of through-hole component legs, it will be easier and prettier in the end :)

Now if you cannot do so (or if you forget to do so as I did on the right), proceed with extreme mechanical caution!  Double-row headers such as the ICSP header above are hellish in this regard.

A risky business! I secured the pins with
epoxy on both sides before I gently
slide the plastic bracket back in place :/
What I end up doing is to first loosen the plastic bracket a bit by sliding the pins through it beforehand. Then I move it flush to the top of the pins, and I solder the bottom side.

When the pins are locked at two points, shearing is less likely to occur, so I still recommend to solder both sides for a more robust connection when possible. Use a thin soldering tip, making sure not to use too much solder on the top layer. Adding epoxy is probably a good thing also, especially for such a connector that will be used a lot. It will also help dissipate the large stress when the plastic bracket is sled in place.

Now, this will eat one or two millimeters from the protruding header pins at the end, and the connection will then be less than stellar...

One day I will proceed in two steps for a better looking result: first I solder the bottom side, I may solder the top but with very little solder to stay as flush as possible. Then I bury the bottom solder points in epoxy. Only once it is cured, I then add a little dab of epoxy also on the top between the legs, and then I slide the plastic bracket back to the PCB surface before the epoxy has cured. I suspect I would get better results, and still get a solid connector.

Using rivets large enough (here 1.0mm ID) will certainly help securing header pins on the PCB,
but look how close the heads are already ! No way a trace can be router between them.

Populating and securing the rivets

Funny game: populating the board with 0.8mm OD rivets (in 0.8mm drilled holes).
The board is halfway on the ruler here, because the rivets are protruding from below.
The metal ruler is quite important to keep the rivet head close and flush to the
(other) side of the PCB, before the tails are pushed down and enlarged.
Do not use a soft surface to push on the rivets from the other side!

Here it is. Huuh. Use a conic, pointy tool that has an "appropriate" diameter and shape.
This one is not wide enough I think, as can be seen in the first picture of the post.
But, hey, it works well and faster for prototyping (better than using soldered wires on both sides),
and it also really helps pinching the copper rings/pads to the PCB, which is very desirable!

I often (try!) to solder the rivet heads and tails nonetheless as I am unsure about the connection
(even though I never had an issue, the copper rivets really are flawless!).
My soldering iron was not properly set here, and I "flattened" the tin before is solidified here,
because I want the result to be as flush as possible, as required by the radio module I solder on the surface.

Using rivets proved another advantage: there are no protruding soldering points like when wire is used  instead for the vias. I still used a piece of kapton to make sure the serrated radio module would not make contact with a rivet head.
On the bottom left corner, you can see the epoxied ICSP connector pins. I really want this connector to be solid,
as it will be used a lot, and copper traces do not stick very well on these really crappy PCB clads.

Here we are, I hope it will help some of you.

Bonus: epoxy curing
reaction produces heat :)