Mechanical stability of limesdr altera chip

[oe6rke]

hi together,

to support the limemicro Team (still in first contact Situation for more than a week) i did some further Analysis which turns out the applied force in this matter.

Here the Analysis i’ve done and sent to limemicro:
…
i did some further analysis regarding the forces on the BGA during this incident which are as follows:

In this scenario the EPS Casing (closed with no cabling, rounded and attached board) and the BGA Altera Cyclone IV EP4CE40F23C8N are included.
The specifications are:
BGA: http://www.mouser.com/ds/2/612/cyiv-51001-1099062.pdf
Casing: http://www.produktinfo.conrad.com/datenblaetter/500000-524999/523130-da-01-de-Euro_Kunststoffgehaeuse_gelb.pdf

Because of used BGA type 23 there are 484 pins /balls to hold the BGA on the print.

The calculated path is the drop of 1 m height on a solid surface (aka free fall).

According the gravitational law the change of potential energy plus the impulse energy during the crash will apply to the overall case and as well to the bga.
Taken from weight measurements the weights are (referrer to attachments):
Total weight case + board + bga + internal cabling = 185 g
Weight of BGA + cement glue + heat sink = 10 g

The following formula’s are used:

Free fall

Weight

Difference of potential energy to the object = Epot = m * g * h
m = mass in kg
g = gravitational constant 9,81 N/kg
h = height

Impulse to the falling object = p = m * v
m = mass in kg
v = 4 m/s² assumed. Ideal free fall would be 9,81 m/s² after recent distance

Transformed impulse energy at crash point = F = p / t
p = impulse from above
t = contact time object with surface. 0,3sec is there a common value.

This means for the total (case 185 g):
Epot full = 0.185kg * 9.81 * 1 = 1.814 Joule or N

p = m * v = 0.185 kg * 4 m/s² = 0.74 kgm/s²
F = p / t = 0.74 / 0.3 = 0.246 N
Total Force applied to the case = 1.814 + 0.246 = 2.06 N

In assumption that the total energy got concentrated onto the BGA only (no force on board, other components, no thermal energy on the casing) the applied energy onto the BGA is 2,06N.
Ideal force per pin should be = 2.06 / 484 = 0.004356 J or N (this because the board worked and got clearance from your qa)

If we look the scenario just on the bga itself with the heat sink the values are (10 g):

Epot BGA with sink = 0,010 * 9.81 * 1= 0.0981 Joule or N

p = m * v = 0.010 kg * 4 m/s² = 0.040 kgm/s²
F = p / t = 0.04 / 0.3 = 0.13 N
Total Force applied to the BGA = 0.0981 + 0.13 = 0.23 N
Ideal force per pin = 0.23 / 484 = 0.0004752 J or N

So speaking of the worst case (unlikely case because bga had no chance being the center of the force) of 2.06N the applied weight equivalent would have been
Fg = m * g or M = Fg / g = 2.06 / 9.81 = 0,2 kg.
I see more this equivalent of 23 g pulling equivalent weight applied to the bga during crash (Force 0,23N).
Both of them are no extensive forces!

Because of this figures I see a clear quality issue for soldering the bga. Furthermore it makes no sense that the other chip didn’t rip of either.
I added the calculations to give to chance to follow the analysis.
…

Anyone who find an error in my calculation please post!

Attachments:

[hardcore]

Just a few other things.
PCB design is rarely Newtonian maths.

A few things you have not taken into account.

1. board curvature, these “blue pcb” if made in China are not always laminated correctly as regards to small area bonding (they have a higher warp rate as well)

2. Reflow oven, sometimes to speed up production , a faster reflow is performed with some heavy fans at the end of the oven to do a rapid cool, this causes the PCB to “curve” and does not allow the Lead free solder enough time to “relax” into the curvature, weakening pads that don’t have a buried via.

3. an impact on the edge of a pcb will not follow the rules above, especially you have the pcb on soft pegs, the bending of the pcd due to the shock of the PCB weight (NOT the chip+heatsink) can cause the pcb to bend and “ping” the heavy chip off.

Why?
because it is a fixed solid block that will not track the bending of the PCB (which is why perhaps the item on the SOLID aluminium back plane faired better ,the clamping force prevented the PCB from bending.)

(you have assumed it is a 10g impact), not a weight of the PCB impact & board bend. (I dropped a $1,000 pcb and the damned xilinx FPGA chip have pinged off & pulled unused pads, i have some pictures some place.)

Would recommend some “pulltests” on a couple of bare pcb, screen print the PCB and stick some brass pegs on the FPGA pads and reflow, then pull test the pads we see in the picture as coming off. (something the PCB assembly shop can do, the PCB shop may have a % of bad boards, it’s what we used to use to test processes)

Check the pcb layout, maybe add some “dummy” track sections on pads that failed.(if it is a design problem)

but most of all “reroute” the PCB and add some mounting holes for fixing the heatsink to the PCB!!!
becasue a heatsink+fan+vibration + lead free solder has been seen to fail due to the vibrations from the fan.

I think the above may explain why other chips did not “ping” off but the FPGA did.

[andrewback]

@hardcore

if made in China

They’re not. They are made in Taiwan by an ODM we have used for many years, for larger, more complex boards also, and without problem.

@hardcore

but most of all “reroute” the PCB and add some mounting holes for fixing the heatsink to the PCB!!!

There are no plans to re-spin the board and we would recommend attaching the fan to the case.