[SOLVED] How to test FPGA before soldering it?

[aminpix]

I have 30 Ultrascale+ FPGAs and I took them from a board (i.e. they are used). They were soldered on a board and I removed them, reballed them and they are ready to use.
I am looking for a method to test these FPGA before sell or install them again on a new board.
Is there any method I can test them to confirm they are working and not damaged?

Is it possible to design a test board with exposed pin-pads and hold the FPGA (not solder it) and run pin read/write commands to test the FPGA?

 

[KlausST]

Hi,

you don´t say what exact BGA geometry....
So you have to look on your own for such a "BGA test socket".

Klaus

 

[aminpix]

Hi,

you don´t say what exact BGA geometry....
So you have to look on your own for such a "BGA test socket".

Klaus

such a this circuit is possible?
Do you know any similar design?
My FPGAs are xcku15p-2ffva1156i.

 

[dick_freebird]

I'd think about getting an OEM development/ eval board and trying to swap on a BGA socket. But socket body overhang might make this tough on a hard packed board and extending vertically will hurt inductance possibly making things squirrely if you're up against timing.

But I also imagine that reballing holds pin-short potential and a proper acceptance test would begin with one-to-all-others-grounded shorts test times N. That wants all very different kind of random / all direct pin access.

 

[barry]

Depends on what kind of "testing" you want to do. Shorts test is one thing. Functional testing is quite another.

But how much automation do you want to have? 30 FPGAs x 1100 pins sounds like a lot to test, but writing test scripts, designing a fixture, etc., sounds like a lot of work. Those FPGAs are not cheap, but neither is development.

 

[dpaul]

I guess what KlausST is correct, you need a custom BGA socket to be able to put in them and remove easily. I have no idea if such a product exists or not. It is a very wierd requirement.
If you have good connections with AMD FPGA FAEs, try asking them. They might be able to give you some info on what type of test socket is used internally or if there is a vendor which can supply them. You'd be lucky to find one I can say.
Try the AMD FPGA forum too...

 

[barry]

BGA test sockets are not a unicorn, they most certainly do exist. And there’s nothing “custom” about them, there are standard BGA footprints and, thus, standard sockets.
The hard part here is everything else: surrounding circuitry, test code, etc.

 

[FvM]

There are several manufacturers of BGA test sockets, e.g. Yamaichi https://www.yamaichi.de/media/Bilder/Downloads/Yamaichi-Electronics_Test-Solutions.pdf and Advanced https://www.advanced.com/products/bga-socketing-systems.
As they are more than 2000 footprint variants according to manufacturer, you'll get them most likely custom made.

 

[aminpix]

I guess what KlausST is correct, you need a custom BGA socket to be able to put in them and remove easily. I have no idea if such a product exists or not. It is a very wierd requirement.
If you have good connections with AMD FPGA FAEs, try asking them. They might be able to give you some info on what type of test socket is used internally or if there is a vendor which can supply them. You'd be lucky to find one I can say.
Try the AMD FPGA forum too...

I plan to design a custom board. In the custom board I will implementing all the power supplies and clocks and leave the BGA pins as exposed pads. To test each FPGA I would press the FPGA on its spot.

In my custom board I will group all I/O pins in teams of 4 IO members (i.e. each group contains 4 IO pins connected to each other).

To test the FPGA I would select one of the IO in each group and write a binary value on it. I would expect to read the same value in the other three I/O pins. I would repeat this for all IO groups.

It will be a low speed test (maybe few KHz).

Do you think this test is doable?

 

[barry]

"Press the FPGA on its spot"? How? You expect to be able to get hundreds of balls to make good contact by pressing them onto a PCB? What's the spec on bow and twist for your PCB? How will you keep your PCB from bending when you "press" the FPGA?

Why are you resisting use of a socket?

 

[dick_freebird]

With care in board design you might find pogo sockets
that can hit the land array for soldering and only have
a bit of mechanical overhang, letting a product board
also serve validation / characterization and even low rate
IQC.

But always a purpose built test board will be better for
repeatability and longevity if you make the investment,
and maybe let you get at things an application board
would hide.

 

[aminpix]

"Press the FPGA on its spot"? How? You expect to be able to get hundreds of balls to make good contact by pressing them onto a PCB? What's the spec on bow and twist for your PCB? How will you keep your PCB from bending when you "press" the FPGA?

Why are you resisting use of a socket?

Thank you for your advice. I found SG-BGA-6049 matches my FPGA. where can I buy it?
I usually design my boards and send it to China to manufacture and put components on it. JLCPCB don't have SG-BGA-6049 in his inventory. Since it is a BGA component it is cheaper to ask Chinese manufacturers put it on the board. Do you know any manufacturer to put SG-BGA-6049 on my board?

 

[barry]

Seriously?
It looks like that is an Ironwood part. (It sure would have been helpful if you had posted that information).
But maybe you can get the part from, oh, I don’t know, Ironwood??? Or, look on the manufacturer’s website for distributors?

There are thousands of companies in the US who can solder this part. If you want to use a Chinese vendor, I guess you’ll have to ask them their capabilities.