>> This reads like a skill issue to me.

> Not sure how this got down voted.

You’ll notice the parent comment could have gotten the point across without sounding toxic, by just dropping the “skill issue” paragraph.

See also: https://news.ycombinator.com/newsguidelines.html (the section about how to write a good comment).

No, it is not. You may think my contract manufacturers lack skill. (No comment from me.) But I have to live with them and make things work.

I try to get my CMs involved in design early. I think it is telling that whenever I give them choices, they reject Tag-Connect and pick one of the other options. Every. Single. Time.

The connector that was specifically flat-out hard rejected was the TC-2070 14-pin version. The number of pins was part of the problem. Apparently (this was a while back now so I may be misremembering) they had trouble with the density at 0.050": 6 pins gives you a lot more room on the sides to squeeze stuff in than 14 does. So they have to do it with a special premade block that comes in and hits the pads, and that block was nearly-but-not-quite unobtainium for the 14-pin version. The CM hated the Tag-Connect in general and wanted it gone, so we didn't trust them too much, but then we tried to build the fixture in house and prove them wrong... after that experience I have joined them in their hatred.

The fact of the matter is that there are many, many other good ways to do it, so it's not Tag-Connect or nothing. Castellations are right out in HVM because of the cost hit, so that rules out Edge-Connect and friends. Würth has WE-SKEDD which looks like the same general thing as Tag-Connect but I've not had cause to try it.

My favorite thing to do, if space allows, is to just put down the unshrouded surface-mount header. Cortex and ESP parts nominally use a standard 0.050" header and you can just place it down. Then don't populate it, and you've got an array of pads that are long enough to stagger test probes on to in a bed-of-nails, or for bench use it is very easy to hand-solder the header on. Plus it's surface-mount so the space below the header is available for use (often things like pull resistors or ESD diodes go nicely here). The biggest wrinkle here is the solder stencil. You do not want to have paste put on these pads if you're not soldering the header, because you badly want your test pins to hit clean ENIG finish and not flux-covered no-clean solder (doubly nasty to probe, even clean solder is bad enough). So it's harder to do a small run of 100 bench-debug boards with headers then the rest as production. You usually end up just soldering the header by hand (or having the CM do it), which is OK.

Otherwise it's traditional pogo pads all the way. This is pretty much required anyway whenever the board is too small for other methods (did I mention Tag-Connect is huge? Tag-Connect is HUGE.) and it works great as long as you were already planning on fixturing.

This is the way

Right, plated through holes, but with almost no exposed anular ring. As little as I can get away with without getting mask sprayed down into the hole.

For 100 mil pitch, 25 mil square pins: 36 mil holes, 6 mil off center, 12 mil hole to hole.

The short side of a generic pin header is the side that goes in the holes, so the long side is free to accept dupont sockets and shunts the same as if you had the same pins soldered in the pcb. So the cable is just ordinary programmer fly wires with female dupont ends. You don't even need to make an actual cable.

If space is tight on the pcb then it does use up more pcb than pads that leave the other side free. And pogo pins are going to be a lot faster for producing something in numbers.

I don't mind buying nice stuff like a fancy purpose made good-working tool for myself but I'm always making open source projects and one design goal is to require as little as possible, and as generic and universal as possible from the user. So I avoid fancy special things where possible. It's not designing for commercial production runs nor designing for one-off for myself, it's designing to a kind of a platonic ideal to strip away anything unnecessary and yet try to meet 2 opposing goals at the same time as much as possible: Don't require special tools that make things work reliably because of how fancy the tool is, and don't require the user to be a zen master craftsman that can attain a successful result with rocks and nails. Try to make the process reliable and repeatable while still only requiring basic materials and supplies. As much as possible anyway.

Pogo pins are pretty common these days and not exactly exotic or expensive any more so maybe I can start using them.

Then again, the through holes do 2 extra things besides make the connection.

With pads you need to aim/register the pins to land on the pads, and you need to hold them there. That means aiming with your eyes and holding with at least one hand, or it means adding some kind of extra registration and grabbing features to both the pcb and the cable, like extra drill holes or slots and extra plastic shapes on a special cable-end etc. Or no extra features on the board and instead a whole clamping jig that holds both the board and the pins.

Since these are holes that pins go in to, you don't need any other form of registration to aim the pins at the pads. The pins go in the holes.

And they hold onto the pins themselves, so you don't need any other form of retention.

It's just like plugging a plug into a socket where the socket provides all that naturally.

I have one board that needs two different connections like that, one for jtag and one for power and to temporarily close a jumper to write-enable the cpld. So a 4-pin and a 6-pin, 2 different cables in 2 different places. The entire board is slightly smaller than a DIP-28 so no room for any real connectors. You just stick the cables in and two different cables hold themselves with zero hands while you operate the flashing software. The wires are all plain dupont wires stuck on the pins, no solder, and 2 of the pins just have a generic jumper on them. It's completely basic and not-special and works perfect.

I have another board that needs 28 pins in a small space. For that one I used 2.0mm pitch pins in straight rows not staggered, but with the holes only 1.7mm apart. In that case it's the long side of the pins that goes into the holes, and the short side is soldered into a programming adapter pcb that goes into a programmer. The pins are in 2 sets of 2x7. Each set of 2x7 has 2 straight rows of 0.72mm holes 1.7mm apart. What happens there is, as the pins start to lean over, the top of the pins hit the opposite side of the hole on the top of the pcb, and don't want to go any further. The pins wedge solid and make 4 points of contact, 2 on bottom and 2 on top, and the board won't go any further even though the pins only just poke out the top and there is still almost 2mm of travel left. So you have a lot of remaining travel to just push a little more if you get a bad connection. It works great and no special parts anywhere.

Oooh that's a great trick! Thanks, I'll try that on my next board.

Fully agree with the "it depends".

FWIW, my slightly more than anecdotal evidence is that 1.6mm is the default thickness at JLCPCB and PCBWay, and they will subtly encourage you to switch to that width if you don't want to pay more or wait longer for them to fill a sheet.

Every IC prototype board (eg Proto Advantage) and all of those Adafruit breakout boards are 1.6mm.

For these reasons, 1.6mm sure seems like a default if there is one. Is there some JEDEC standard or similar which proves me officially wrong? Happy to learn if so.

A good solder sucker is your friend.

I have the SS-02, and I like it - I had one of the cheap blue ones first, but the pliable rubber tip really makes a difference. If you’re soldering smd by hand, it’s more than worth the $20

The answer to almost every question in soldering is 'more flux'. Solder wick has flux in the center of the braid, but it's hard to get it into tight places like structural through-holes. Adding your own liquid/paste flux will make the wick much more effective.

Melt the solder and thwack the board on something hard? So the board stops but the molten solder doesn't.

Sometimes though you just have to pile on solder and flux because the via is small enough that surface tension and heat dissipation means its never coming out

A desoldering pump (manual model, $10 or so for a decent one) is very suitable for removing solder from through-holes, if that is the main issue.

I often add solder to make it easier for the wick to get everything. If the original assy was Lead-Free, using low temp solder (I can has lead? As a treat?) may make a difference here as well. Flux pen on the solder wick also seems to help especially if your wick is kinda crusty.

To add to the sister comments, you can quite easily remove such bridges by adding flux and then touching each individual pad with a fine tipped soldering iron. It sometimes takes a few tries, but eventually the solder that’s touching the solder mask will either be wicked onto the iron or move onto one of the neighboring pads. (The trick is to touch just the pads with the iron, and not to try to attack the solder bridge itself.)

Yes, the surface tension of melted solder pulls the solder to just the pad areas (assuming you don’t have far too much)

Make sure there is soldermask between the pads. This makes soldering much easier!

(If your foundry can't fabricate it, then make the pads thinner until they can fabricate the soldermask.)

Using with a little flux while tinning usually prevents the pads from bridging

I have soldered the 12-pin, power-only USB-C connectors. The real breakthrough came though when I tried a hotplate rather than soldering iron for the USB-C connector.

You cannot do that because of how the connector flips over.

(Believe me, I have tried to make it work.)