I am using a "Teensy" development board in a project and would like to attach it to my main PCB including the surface mount pads for the usb data lines.

Backside Of Teensy

Are there any reasons I shouldn't just design up a footprint of all surface mount pads and install it during my regular reflow assembly step.

I can see the potential for some kind of thermal stress issue but on the other hand see many castleated radio boards mounted in this fashion.

My only other alternate solution is to persist with through hole design and use pogo pins to mate with the required surface mount parts.

Edit: To be clear my question is more focused on if it is possible to take an arbitrary non-castellated board and solder it surface mount in a reliable fashion.

  • \$\begingroup\$ A picture or sketch to show what you mean would help. \$\endgroup\$ Apr 11, 2017 at 3:04
  • 3
    \$\begingroup\$ If you're going to be designing a board anyway, you might as well put the entire circuit on your board. \$\endgroup\$
    – Matt Young
    Apr 11, 2017 at 3:16
  • \$\begingroup\$ Those modules are used mainly when it's much cheaper than self manufacturer, or when licensing and testing of a self design should be avoided. A microcontroller board like the arduino or teensy is so small that integrating the actual parts would be simpler. The teensy is barely more than the MCU by itself. \$\endgroup\$
    – Passerby
    Apr 11, 2017 at 3:25
  • 1
    \$\begingroup\$ At this stage it is a low volume 10-20 units. At which point the price of the individual components and the time (I am hand assembling these boards) becomes vastly larger than the cost of a teensy. Integration of the parts is something we have considered but decided to persue at a later time. \$\endgroup\$
    – Hugoagogo
    Apr 11, 2017 at 3:25
  • \$\begingroup\$ Will all the components on the Teensy survive a second pass through a reflow oven? Like the crystal - I guess you don't care about the USB socket. \$\endgroup\$ Apr 11, 2017 at 8:17

2 Answers 2


I suppose what you suggest should work, but I've never seen a definitive statement to that effect by someone who has actually studied this. My logic is that castellations are half of a thru-hole pad. If half of a pad can hold the board well enough, adding a second half should hold it even better.

However, this is not a area where I would trust simple logic like this. I would do one of:

  1. Use headers. These are soldered to thru hole pads on both ends. That is using the headers and both boards as intended, so really has to work. This is the "correct" way to connect something rigidly with thru-hole pads to another board.

  2. Replicate the whole circuit on the new board. If you're designing a board from scratch to hold a module, you can incorporate the module's circuit on the new board directly.

    This is only valid when you know the circuit, are legally allowed to replicate it, and the module doesn't have certifications you are relying on. For example, if you are incorporating a WiFi module into your circuit, its FCC certification would not apply to your replicated circuit.

  • 1
    \$\begingroup\$ Using headers has the added benefit of being able to place components under the daughter/module board. \$\endgroup\$
    – m.Alin
    Apr 13, 2017 at 12:20
  • \$\begingroup\$ Following this logic, do you think a reasonable solution in this case where I have some hidden SMD pads would be to solder the board down using reflow with only paste applied to the SMD pads and then to solder in pins through both boards to provide mechanical strength. \$\endgroup\$
    – Hugoagogo
    Apr 13, 2017 at 15:12
  • \$\begingroup\$ @Hugo: Soldering thru-hole pins to SMD pads doesn't make sense. Also, after solder has flowed, it may be difficult to get a header pin thru a hole. \$\endgroup\$ Apr 13, 2017 at 15:33
  • \$\begingroup\$ I might not have been clear but I was meaning if a footprint that has a mix of through hole and surface mount was used. \$\endgroup\$
    – Hugoagogo
    Apr 13, 2017 at 15:35

Don't do it. I have a board like this and yield is awful.

There are two reasons for this. First is that these are blind pads - because of constraints on the other side of the board, there's no space to put holes and un-masked areas to drain excess solder. When boards are hand-paste-stencilled control of paste quantity isn't perfect and some pads end up with too much paste. If there's nowhere for the paste to drain to then each pad has a different-height solder blob, the board floats and some pads don't make contact. If pads do have somewhere for paste to drain, it can drain instead of joining the pads.

The other problem exacerbated by too much paste is that any degree of non-planarity can mean some pads in the middle of the board aren't touching.

Also, reworking pads in the middle like this is difficult. You can't inspect to check they are correctly soldered, and it's difficult to get heat or solder in there to fix a bit connection. If you add a hole to allow reworking to conduct heat into the joint, it may drain solder away and cause a bad connection.

QFNs and similar get away with it because they're so small, so planarity is less of a problem. BGAs get away with it because just the right amount of solder is pre-installed. Castellations avoid the problems because the open half of the castellation acts as a solder buffer for the other half.

I know the specific question is old, but in the general case I'd suggest looking for another way.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.