Tag-Connect Edge-Connect

[jmw]

Has anyone used this hands-on? While expensive, it looks promising by offering a locking connection while also being low profile like their no-leg connectors. OTOH, lack of actual in-situ product photographs, sparse info on datasheet (covering only one of the three models), and no stocking distributors are setting off my vaporware alarms.

http://www.tag-connect.com/EdgeConnectIntro

[ataradov]

If I understand the pictures correctly, this will require castellated vias, increasing the price of the board if they will be there only for the purpose of programming.

This may be fine for programming modules. But then again, there are those Chinese board mounted loops that can hold the module.

Edit: Yes, the text explicitly calls out castellated vias and Wi-Fi and BLE modules.

[rx8pilot]

I have been using their original design for years and really like it. So easy and my customers have not yet figured out what it is which discourages hacking.

This one could be very interesting, but the castellated vias. Wondering if the low-cost PCB houses will support them or at least how much extra it will cost.

[donotdespisethesnake]

Obviously, it makes sense for people already using castellated connectors. Our house rules don't allow copper at the board edge, so any sort of edge connectors are out.

You can read more about it in their patent application. https://patents.google.com/patent/US20180261940A1/en?assignee=tag-connect&oq=tag-connect I'm struggling to see what the inventive step is, people have been using edge connectors for decades. I am sure there is prior art.

[KL27x]

  people have been using edge connectors for decades.

It seems to me that most people doing this are making their own equipment, for in house use. I don't imagine this being a huge $$ patent or commercial thing. In house, I flash my chips before the assembly, and I don't have to clip/unclip connectors to every pcb to do it. I don't touch the chips.

Other have the chips preflashed with a bootloader, and they gang program panels of pcb with mousebites.

This seems like it's mostly good for JTAG debugging/testing for pcb ... but without the need for larger number of test points/jigs.

So... I don't see a huge commercial market. Most this stuff is done in China, and they aren't paying for this when they make stuff at least as better.

[bson]

Looks great for designs that need the full JTAG, or want it through an opening in a rear panel.  But I can find no info on what's at the other end - RJ plug?  100mil IDC?

I've asked them about it.

[Gribo]

Their other end depends on your debugger/JTAG device. I don't think it is a good design. Automated fixtures favor bottom side pogo pins. Side or top connections add costs.

[jmw]

Looks great for designs that need the full JTAG, or want it through an opening in a rear panel.  But I can find no info on what's at the other end - RJ plug?  100mil IDC?

I've asked them about it.

They told me it's a female .1" IDC connector, with same pin assignment as their other designs.

[donotdespisethesnake]

I have a XMC100 dev board from Infineon which appears to use castellations with index hole to program the debug chip. I wonder it they are using "edge-connect technology"?

[jmw]

Not the exact same thing - if you look at the footprint there are unplated holes to either side of the castellations for keying and also an unplated hole in the middle to hold the connector in place. Maybe they're doing the same idea though.

[Dubbie]

Someone in this forum told me a few years back about tag connect. I use them on all my boards. They are great!
I hadn't seen edge connect. will have to check it out.

[bson]

They told me it's a female .1" IDC connector, with same pin assignment as their other designs.

Yup, me too.  That should work nicely.

[bson]

Here's my current project using a TC-2030-NL with an MSP430F5510PT.  I can easily see using EC for projects in the future! 

[donotdespisethesnake]

Not the exact same thing - if you look at the footprint there are unplated holes to either side of the castellations for keying and also an unplated hole in the middle to hold the connector in place. Maybe they're doing the same idea though.

Oh yeah Infineon use a plated hole for keying and to hold the connector which makes it COMPLETELY different 

[julianhigginson]

I think it looks good. Very nearly picked it for a project that's currently out being built, but the normal/old tag connect suited the mechanical details of the product better this time, so it got used instead.

[mrpackethead]

Its cheap enough that i'm tempted.  I've bene using the their products for a while and they have been really good for saving time.    This is a little bit of a solution looking for a problem for me though.

[drussell]

A re-invention of the card-edge connector, only smaller for the SMT age. 

[SiliconWizard]

I've tested the TC2030 - not bad but not ideal either. Still bigger than I'd like for some small boards.

This edge connector? Why not, but getting castellated edges can add significant cost to a PCB in my experience. They say it's cheap - well, depends on the manufacturer. Some chinese ones can do that almost for no added cost, but if you're using an higher-end one, it can be expensive because of yield and quality issues. Also depends on the number of layers.

Still looking for the ideal JTAG connector here.

[jwr]

I'm a big fan of the Tag-Connect no-legs footprint (see the "Choosing a debug/programming connector" article I wrote at https://partsbox.io/blog/choosing-a-debug-programming-connector-2019.html) — but these look useful only in very specific cases, where either you already have a castellated edge, or you are extremely space-constrained.

That said, I might consider it in the future. See for example the picture (in the article linked above) of a board where I had to place components on the other side of the Tag-Connect footprint, because of space constraints. A castellated edge might have been possible on that board (it is not always the case, mechanical design might interfere) and could potentially be a better solution.

[ataradov]

I had a chance to play with them at a couple trade shows now. They are very bad. The people in the booth have to give you a lecture on how not to break them before giving you a try.

It is extremely easy to break something, it just does not feel good and solid. Every time you try to insert or disconnect it, it feels like you are about to break the PCB.

[rx8pilot]

I had a chance to play with them at a couple trade shows now. They are very bad. The people in the booth have to give you a lecture on how not to break them before giving you a try.

It is extremely easy to break something, it just does not feel good and solid. Every time you try to insert or disconnect it, it feels like you are about to break the PCB.

Practical insight. Glad to hear from someone that has actually seen and handled it.

[dgtl]

I got a sample on my desk and havent found a reason to use it.
1) In addition to SWD/JTAG, I use a debug serial port as well. Solving half of the problem is not enough.
2) I have no problems soldering my own SMD 1.27mm/2.0mm headers on my development boards. In schematics, these are set as no-load and only devel boards get  them soldered on.
3) Too flimsy to use for manufacturing
4) Most of the modern microcontrollers and application processors have built-in bootloaders. In most cases, it can be arranged, that an already present external port (mostly USB, sometimes serial or others) can be used for programming in manufacturing, if it's a small run and constructing test jig is not justified. For larger runs, the test jig handles connections.

I've been keeping my eyes open for no-solder debug connectors. So far, incorporating a card edge connector would make the most sense, but it does not have large enough advantage to start using it and create additional adapters and cables. For now, I've used the same serial port footprint for over 10 years and a big number of boards; the same JTAG for all Cortex-M projects and so on. Keeping things uniform is a good thing as well.

[andyturk]

The people in the booth have to give you a lecture on how not to break them before giving you a try.

Yep. Even after their instructions, I managed to break their demo. 

I'll skip the edge connect, but I use the TC- series frequently.

[bson]

Good to see some hands on discussion.  The problem I see is the form factor doesn't have any advantage over the TC2030, but requires castellation which is a bit of a complication.  To be honest I think an edge mounted 50mil 10/14/20 pin header is preferable even though it does require populating a header, for large boards or boards where surface access is limited (like in a cage).

[OwO]

Why use overpriced tag-connect crap when you can just:

[julianhigginson]

I like that peg idea! it's pretty cool. What do you use for the pogo pin spacing? (I guess it's that light blue plastic bit ? what's that?)

But I'm not really sure this is a suitable solution if you're a development contractor - sitting down to build these on contracting rate time, and needing to explain to other people (client maybe, or any other subcontractors they engaged) how to use it and how to build their own.. hourly rate for all of that and you could buy a big box of tag-connect cables, I think. :-)

Also, since you have holes in the PCB already for this approach, you could probably do almost as well with a plain long pin header run through the holes and pushed at an angle... use the weight of the PCB to hold it in place.

[IDEngineer]

We use a LOT of Tag Connect's traditional connectors in production here. Very happy with them. Clever idea, well executed.

We looked very seriously at using that EC style, I even had a nice conversation with the owner/inventor at the company, but ultimately abandoned it. We sent their spec sheet to several PCB shops that we use and every one warned of cost and QC concerns. My interpretation of that was they had just given themselves an "out" if things went wrong down the road... "We warned you, but you insisted, and now here we are."

What I'd like to see (but haven't shared with them yet) is an inline version of their standard connector. Instead of two rows of pins and contacts, just a single row. This would give you a long, narrow form factor that would fit better in a lot of PCB layouts than their present, more squarish footprint.

[SiliconWizard]

What do you use for the pogo pin spacing? (I guess it's that light blue plastic bit ? what's that?)

It's probably just based on something like that:
https://www.ebay.com/itm/Spring-Loaded-Pogo-Pins-Receptacles-for-Jtag/250546684662
https://www.ebay.com/itm/Spring-Loaded-Pogo-Pin-Adapter-for-Jtag-cable-pitch-0-1-2-54mm/260430819053

you can find many variants of those on eBay, Aliexpress, etc. I have a few and that's handy.

[bson]

We use a LOT of Tag Connect's traditional connectors in production here. Very happy with them. Clever idea, well executed.

I created an adapter that goes from the 10-pin .050" Cortex-M connector to RJ12 for TC2030 - for SWD/SWV only.  I've been thinking of extending this to a small box that detects VCC on the target, lights a LED on the enclosure, and either after a brief timeout to ensure a good connection (like maybe 500-750ms), or when a button is pushed, issues a text message on a USB VCP connection to let a USB host know it's plugged in.  The host can then run a programming command, or grab a complete system state dump for post-mortem.  This would permit using the legless TC2030-NL without the little gripper for programming - just push it in and hold it there, wait for the adapter box LED to come on, the host to begin programming, verification, test, and done.  Disconnect and power off; ship it!  Something to program small volumes, maybe in the 100s, that just about anyone can handle.  Glitched or had a problem?  Just do it again.

[IDEngineer]

I created an adapter that goes from the 10-pin .050" Cortex-M connector to RJ12 for TC2030 - for SWD/SWV only.  I've been thinking of extending this to a small box that detects VCC on the target, lights a LED on the enclosure, and either after a brief timeout to ensure a good connection (like maybe 500-750ms), or when a button is pushed, issues a text message on a USB VCP connection to let a USB host know it's plugged in.  The host can then run a programming command, or grab a complete system state dump for post-mortem.  This would permit using the legless TC2030-NL without the little gripper for programming - just push it in and hold it there, wait for the adapter box LED to come on, the host to begin programming, verification, test, and done.  Disconnect and power off; ship it!  Something to program small volumes, maybe in the 100s, that just about anyone can handle.

We toured a manufacturing facility in Shenzhen China that had built something like that in-house. They panelized multiple PCB's, and after stuffing but before they separated them they put them into this fixture that had X-Y control with a programming connector on it. The operator pushed a single button, and the X-Y moved the connector around the panel to each of the footprints. When it lowered and made contact, it detected the voltage on the pins and triggered the programmer. It moved around, picked up all boards on that panel, and then signaled it was done. It felt like a bit more automation than was necessary (given the low labor costs there) but they did a nice job of it... definitely homebrew but worked great.

[KL27x]

^I custom programmed a CNC machine to do that. i programmed thousands of boards with it. I gave the firmware a "wookie wiggle" option, for the time I accidentally ordered boards with paste layer on the ICSP pads, so that the serrated pins would dig through the flux residue and/or oxide layer.

When it lowered and made contact, it detected the voltage on the pins and triggered the programmer.

I just programmed mine to start flashing after lowering the Z axis. And it moves to the next position after receiving a programming verify, so on error it stops and waits for human intervention. I actually made my own controller from scratch. Just a micro and transistors to move the steppers in a grid/array, adjustable for number of rows and columns, X and Y travel, an input for programming verify, and an output to trigger the programmer. There is a start button and a clear/home/abort.

But ultimately, if just doing ICSP (not board/assembly diagnostics) I found it is much easier to flash the bare chips. Except for stupid QFP's, it is much easier and compact to handle chips than panels of boards. This means you can load a crapton of chips at a time, vs loading a CNC machine with panels of say 12-30 boards, every 5 or 10 minutes, all day. And there is less tolerance stacking to bother about if you can design a machine mechanism to program bare chips vs boards. You don't need optical correction or whatnot. I flash my own DIP gullwing down to SOP and even QFN chips. The latter is a bugger to get dialed in, initially, but once that is over, I have never had to fidget with the mechanism, again. It requires some extensive setup time to add a new chip or even pinout to the system, but it's good for medium volume use without paying 10 cents per for the privilege and allowing bulk purchases of blank chips vs trying to figure out how many flashed chips to order to get the best price without getting stuck with too many leftover... or discovering a bug or otherwise requiring a firmware update the day your flashed chips arrive. I can just reflash the flashed chips.

This is convenient for me to the point I have ordered boards completely assembled except for the micros. And I hand solder just the flashed micros to the board. This is obviously a bit more work than just flashing assembled boards, but it avoids customs fee of importing microcontrollers from China.

i still put the ICSP header on the PCB, in case of firmware tweaks post-assembly. But I have retired the CNC machine method of programming boards. Too fidgety due to pcb tolerances unless you make the pads rather large. I've pretty much settled on 0.05" pitch pads (same as Tag Connect), and I like to keep it that way for now.

[PCB.Wiz]

Why use overpriced tag-connect crap when you can just:

.. use a clothes peg (and pogo pins).
Yes, nice example