PCB card edge manufacturing


I have created a board that has a 0.8mm pitch 10+10 pin dual sided card edge connector (fingers). The PCB was designed with pads and board outline exactly to the specs of the data sheet. I have manufactured the device with the card edge fingers and the interface board that contains the connector with skillet reflow, and after getting rid of some bridges it seems successful.

Now to the questions:

  1. I chose OSHPark for the boards as they provide a gold plated finish on the pads. These devices will not be connected and disconnected very often. However since I am at the prototype stage writing firmware, there are more connects and disconnects than usual, which made me start thinking about wear. So is it common to manufacture pads for the card edge fingers with enhanced plating that make them more resistant to wear, or is the standard gold plating I am getting from OSHPark regarded as adequate for this purpose? If enhanced plating was needed, who would provide the service and how is it specified?
  2. The fingers were designed with with a cutout in the place of two pads for keying. However I could not find any keys to insert into this rather small card edge connector at Digikey or at the 3M site. Does anyone have some clever tips to how to make a key and keep it in place? (The one I made from a board piece quickly fell out.) I sort of hesitate using epoxy or other glue in the card edge connector to keep the key in place…

Board edge socket

I have some USB-A-pcb connectors made with ENIG (not hard plated) by another shop and they work - but they are not engaged/disengaged a lot (10-20 times maybe).
No problem so far.

Probably some marking on a extra layer that you provide with the gerbers. KiCAD offers the layers Cmts.User and Eco1 and Eco2 for this.
In the end the communication medium depends 100% on the fabricator of your choice and any answer you get here is moot, unless the support from one of them posts here by any chance (0.1% ?).

I got some picture in my head, but would LOVE to see your example.
Why do you need mechanical keys for the [quote=“otoien, post:1, topic:6337”]
…prototype stage…


Thanks for the reply. The physical keying is to avoid connecting power from the main board where it should not go (on the signal lines. Signal lines are on one side, power and GND on the other (multiple connectors). The prototypes are planned used for a research project (once I duplicate them in a few copies). It is too easy to connect the wrong way in a moment where one forget (already done once briefly) , thus the board was planned with a key as shown in the attached image.

The development goes in several steps, of course first there is a breadboard prototype where most of the development has been done with respect to firrmware, but then there is need to test the real thing, in a more or less final form.


You could also use a housing with a key, you appear to have some sort of circular housing.

Unfortunately, creating customised connectors is impractical unless you have a significant budget for tooling. If you can’t find a suitable off the shelf connector, then gluing in a key may be the best option.

Of course, there is one sneaky way to make the polarity redundant, although that reduces the available conductors by half.


Very good idea - my dad used it a lot (mostly a power electronics guy).
If it’s only the power that is a problem, you can still cross/short a lot of signal lines, as long as their mating doesn’t destroy anything if done the wrong way around.

Or, if possible and you got control over power and connection design - you could also use some sort of resistor as matchmaker - along the lines of power tool battery chargers and power over ethernet.
It could always test for some value there and only power up/connect if it’s matched.
Would only cost you 1 pin.

If that’s dust on that connector in the picture it’s not very big what you got there.
Can you cut (with a 32TPI hacksaw blade - 1mm wide cuts) into the mating connector where that key is supposed to be?
Then you could wedge something in there, perpendicular to that pcb going into the connector… but that is no final solution.


Thanks Bob and Joan for further suggestions.
The board for the device, a data logger, is pretty small, about 42x15mm (card edge fingers section only about 9mm wide) and there would be no room for the crossing of lines to make it resistant to incorrect connections (besides the project has come too far for that). The signal lines pretty much comes straight out from the main chip without crossing. The suggestion to key by the casing of the interface unit is a good one and one I consider for a production version. One would just have a big hole in the casing and the device capsule would fit down that throat - also much easier to make neat than cutting/filing out the slot for the card edge connector in the interface casing. The only problem could be if the logger device also were to be distributed in a naked version where the user would supply own casing.

The idea of making a slot in the card edge receptacle for a key is interesting and one I need to look into, especially for the interfaces I am to use myself where the presence of a chip programming connector in the interface unit close to the card edge receptacle makes the big hole approach less feasible or at least less neat (programming connector not needed on production samples). Another idea spurred by your suggestion: If I can find a small enough drill bit I could drill a hole across the card edge receptacle instead of making a slot and insert/glue in a piece of string as a key.


Hehe, always many ways to solve a problem :slight_smile:

Why not design using an edge connector, with a built in key ?

Other comments:
Edge connectors these days are not cheap, and are trailing edge/low volume products.

  • with a bonus irony, that more pins cost less than fewer pins.

eg PCI 36 way edge connectors do have a key, and show ~20c/500|+

OR Can you not design in a proper connector, cheaper than trying to key something manually ?


Hm… have you considered M12 or M16 connectors? M12 comes with a max of 12 pins… they are coded and depending on pin count can carry up to 45W per pin.
The hole in your housing must be roughly 12-14mm wide already to accommodate the 9mm pcb edge connection.
Especially as this is for a datalogger in a housing.


Because at that point you may still encounter cases of quick cycling between the firmware development system and the hardware product, without giving full attention to what you are doing as you grab and plug the hardware. For example, how many times have I told the administration that I was detained by “unscheduled development tests” after spending half a day recovering from a battery inserted wrong-way around?



Hehe, for that reason I did chose a CR2032 cell holder that doesn’t cause problems when I (or other users) insert the battery the wrong way around :wink:

My dad once told me a story about something similar.
The former product had all sorts of connectors on it’s main board… he didn’t like it and standardized most of them, to have better scale, less inventory… streamlining essentially.
Then a prototype got assembled - not by him - and naturally didn’t work.
Turned out the person doing the assembly did plug in wherever stuff ‘fit’ - *doh*, right?
Luckily for him (and probably my dad) he anticipated that and also standardized the pinout for those connectors.
In the end the signals were all wrong and nothing worked, but at least it didn’t blow up :slight_smile:


Thanks for further comments. The card edge solution was chosen as the device needed to have minimal ferromagnetic properties; a connector like the one shown above would be out of the question, besides being way too big with respect to depth. There is only about 11mm from the front edge to the pads of the main chip. I will keep it in mind though, it could be useful one day for a different project.

I looked for a keyed version of the card edge connector, but none found, only for those with higher pin numbers.

Yes, if something has the possibility to be connected the wrong way it will happen. :dizzy_face: . It is just surprising that not more stuff is smoking.


What is the card edge connector you are using?


3M SPD08-020-RB-TR, Digikey part no. 3M10682CT-ND.


That’s a nice compact connector, but doesn’t leave much room to work with! Normally orientation would be enforced by the enclosure, but I don’t think you have room to do that.

Possibly I would glue an extra piece into one end of the slot, and have a notch at the end of the PCB, but that would mean redoing the PCB.

Also, that doesn’t work without the housing to provide guidance.


If you are Ok with that price level part, the Samtec MEC1 series have a key, on 16 way.
(I think you use 16 contacts)


That is a nice 3D rendering bob, although it does not solve the problem of keying the “naked” device.

Thanks for the reference PCB_Wiz, it looks interesting, although the “8-position” version (part no. MEC1-108-02-F-D-A) only has 7x2 connectors. As I use multiple pins for power and GND in the present design I could possibly work around that. With 1mm pitch and fairly thick body I am not sure if it would fit inside the throat of my capsule (the circuit board has an asymmetric location due to the battery), but I ordered some free samples so we will see. It could anyway be useful for a different version of the logger not using a capsule.


No, I couldn’t think of a way to do that without extending the overall width. I don’t think I can improve on the idea of drilling across the connector and inserting something like a pin.

For info, the drawing is created in OpenSCAD, it takes a few line of code. I downloaded the STEP file for the connector from digikey, then exported an STL for OpenSCAD.


For what it’s worth, I think the simplest solution would be to leave the slot in your board and drill a hole through the mating connector into which you could insert a plastic or steel pin.

But I’m not going to render a 3D model to illustrate it. :wink:


Oh, I modelled that as well, here ya go :slight_smile:


Somehow I thought you would. :slight_smile: Very nice, but I think the pin/slot need to move one position to the left. :wink: