Custom Flex PCB connector thermal expansion

[OwenH]

Hello everyone,

I have an application using a custom designed spring/pogo pin connector made in-house from machined PEEK. The connector has a series of surface mount spring pins pressed into the peek body and then the connector assembly is surface mount soldered to a high current flex pcb. The whole assembly is fixed with 2mm bolts into an aluminium housing leaving a 1mm clearance in the area below the flexpcb area.

When the first prototype parts were reflowed to the flex pcb I realised I had not properly considered the thermal expansion of the PEEK material in the connector. The first parts had warped like a banana as the part cooled and the solder solidified. The Peek had expanded more than the flex pcb and the solder solidified before the peek shrunk back to it initial dimension.

I was able to successfully solder the parts with a machined aluminium fixture on a my hotplate and I don't get any warping. By doing this the body of the connector doesn't get as hot as the flex pcb on the hot plate and there is no noticeable warping. I am however concerned that there may be a long term issue with thermal expansion overtime.

I have been running the parts at their design limit using my programmable load letting them warm up then putting them in the freezer at about -15c then removing from the freezer and running at full load until the parts are back up to 50c. I have done about 100 cycles with no change in the resistance of the connection. I know that I would need thousands of cycles to do a proper stress test but I won't be able to do this myself.

So my main question is how tolerant of thermal expansion is a flex pcb? My intuition would be that the flex pcb is more forgiving than fr4 but I don't have any experience. How flexible would the Polyamide flex material be laterally?

I have attached some photos to help show how the parts are constructed.

Thank you for your help!

Owen

[ajb]

I remember you posting CAD screenshots of this a while ago, looks like the final assembly came out nice 

How long is the PEEK section?  Guessing 30-35mm from the photos?  What's your anticipated temperature range?

I'm no mechanical engineer, but if you assume that the FPC is constrained to remain flat, then you could assume that the PEEK is perfectly rigid, and look at the modulus of elasticity of the FPC to approximate the maximum stress induced in the FPC for a given strain (really a limiting value, since the PEEK will take up some of that strain in reality).  If the PEEK reaches the same temperature as the FPC during soldering, the strain can be calculated from the difference in CTE between the FPC and PEEK at any temperature T₁ as (T_solder - T₁) * ΔCTE * Length.  But presumably the PEEK is not getting nearly as hot during soldering as the FPC, so you'd need to know or approximate the PEEK's temperature at solidification and calculate the dimensional change separately between that and the FPC to determine the strain.  Also, the FPC isn't homogenous because the copper layer isn't solid across the length, and you have those metal reinforcements, with some sort of adhesive bonding them to the FPC, so there will be shear in that adhesive interface and...it gets a bit complicated.  But you can still do some approximations via simplifying assumptions, like assuming the PEEK is perfectly rigid, and that the FPC *is* homogenous, and ignoring those reinforcements.  Once you approximate the strain, you can determine the approximate stress from the modulus of the FPC, and finally compare that to the yield point of the FPC and see how close to failure it is.  That would be for tensile failure of the FPC, in compression, I think your failure mode would be the FPC buckling away from the reinforcement, probably in one of the gaps between the copper sections.  You could approximate the magnitude of that buckling from the strain. 

Some rough CTEs for the materials in question just to get an idea:

PEEK 55 ppm/°C
Kapton 20 ppm/°C (value from Dupont's Kapton FPC datasheet, other forms are much higher)
Copper 17 ppm/°C

So there's a ΔCTE of about 35 ppm/°C between the PEEK and the FPC.

[jbb]

Those connectors look great!

That existing assembly might be totally fine…

The sandwich side view does look like a bit of a mechanical stress concentrator. Possible mitigations:
- have you considered double-ended pogo pins? That way no soldering required and you might even be able to get rid of the PEEK parts. But contact resistance would be higher
- could you place some slits in the flex PCB between pogo pins and maybe even some serpentines to soak up the mechanical strain?

Reference: https://www.hemeixinpcb.com/rigid-flex-pcb/multilayer-flexible-pcb.html
(Note that was just 1st good picture, I’ve never used Hemeixin)

[OwenH]

The connector body is 47mm long, with the area where the flex PCB is soldered being about 32mm long. At an ambient temperature of around 20°C, I measured around 45°C at 20 amps using a thermal camera. The device needs to be able to handle a maximum peak current of 20 amps. It’s difficult to differentiate the different temperatures of the PEEK connector and the PCB with my thermal camera. The flex PCB has an aluminum stiffener and is thermally coupled with a thermal pad to the aluminum enclosure. I imagine that the flex PCB would be getting hotter than the PEEK considering that the brass spring pins should be conducting the heat into the flex PCB nicely, but due to the assembly, it’s difficult to measure.

I hadn’t really thought about the failure mode being buckling of the flex PCB. I had figured that if it failed, it would be cracking at the solder joint.

I also thought about creating slits between the pins in the flex PCB, but I’m not sure how much of a difference it will make with a solid aluminum stiffener below it. I suppose it may depend on the flexibility of the adhesive. I had also thought about double-sided pogo’s, but the current capacity is too low and the resistance is too high.

[jbb]

…
I also thought about creating slits between the pins in the flex PCB, but I’m not sure how much of a difference it will make with a solid aluminum stiffener below it. I suppose it may depend on the flexibility of the adhesive.
…

Ah. If you’re getting an aluminium stiffener as part of the FPC, you probably can’t (easily) get the polyimide FPC material split where the stiffener isn’t.

I too would expect the solder joints to fail, (especially if there’s vibration), so I suggested slitting the flex to allow a bit of springiness in order to reduce the stress on the solder joints.