Cyclone 5 recovered from scrap boards

[Miti]

Look what I recovered from some scrapped boards, together with the 128M configuration flash and 25MHz clock! And I think there are more in the bin.
I simply couldn't let them go in the garbage even though I know they're hard to reball, install, etc.
Do you think is worth trying to use them in some TBD projects in the future? I used some very small pitch parts in some of my projects but I haven't tried BGAs yet.
I thought about going to the next level for awhile, my toaster reflow oven would love it, but I just didn't find the project and the time.

[james_s]

I would have tried to work out how to use the original boards they were on, that's usually a lot easier than trying to re-ball parts. You'll need a multilayer PCB to install those on.

[Miti]

I would have tried to work out how to use the original boards they were on, that's usually a lot easier than trying to re-ball parts. You'll need a multilayer PCB to install those on.

Nope, can’t do that, part of the NDA. I can’t take customer’s boards or schematics out of the plant.

[Scrts]

Afaik these are fairly big in terms of LUTs and also have a hard memory controller.

[OwO]

I've successfully soldered salvaged BGAs of that size to a board without reballing (LGA style). It was an Artix XC7A100T (676 ball) and a Zynq XC7Z010 (400 ball). They both work fine and the Zynq runs Linux and passes memtest without issues. I've also used salvaged DDR3 chips by reballing the board rather than the chip.

I could design a PCB for the Cyclone V on 4 layers, but I don't have the inclination to do so because I don't plan on using Altera parts in the future, but if you can get the PCB designed I can help prototype it. PM me or add me on discord (username in signature).

[Miti]

I've also used salvaged DDR3 chips by reballing the board rather than the chip.

Yes, I was thinking about reb.... well, balling actually, the PCB rather than the BGA. Did you use real balls or you just used lots of flux and moved a ball of solder over the pads to form bumps? I would think the bumps would be too uneven to solder reliably all the pads on the BGA.

[OwO]

I used a BGA stencil (the thick type) and applied paste over the footprint. Then without removing the stencil I used hot air to melt the paste into balls. (if you remove the stencil and reflow, the balls will merge because the volume of paste is larger than usual.) As it cools down you can remove the stencil (try to do this when the solder is solid but the flux is still liquid).

[Psi]

I have a stack of these boards with 4 DSP and 1 FPGA on them.
If they're worth the postage to someone, let me know.
I'm in New Zealand

Some of them are still soldered on, some have been unsoldered. 
All of the DCDC 10A voltage regs have been pulled off the boards (was the reason i got them). So boards are no longer that useful.

[BrianHG]

Look at what I found, DIY reballing.  You will need a dummy stencil for reballing, solder paste & a hot air gun, bu here is someone doing it: (Begins at 5:45)



Since these stencils are cheap, then it is worth it if you have everything else:
Reballing Stencil

[Bassman59]

Look at what I found, DIY reballing.  You will need a dummy stencil for reballing, solder paste & a hot air gun, bu here is someone doing it:

We (day job) have used similar kits to re-ball parts that we needed immediately but for whatever reason were not available. We took known-good parts off of boards which were obsoleted/superseded (previous rev boards). Of course there is always the danger in damaging parts in the removal and rework process, but I don't recall any real fall-out.

The OP got these parts but he has no way of knowing whether they're actually any good.

[Miti]

The OP got these parts but he has no way of knowing whether they're actually any good.

Oh, I know they are good. They come from brand new boards that were scrapped at customer’s requests. Layout errors.

[OwO]

[OwO]

Here is a Zynq soldered LGA style (no reballing). The chip and PCB are first tinned by dragging with a knife tip, with the iron at a consistent angle (so that the amount of solder on each pad is mostly consistent). Then during reflow it's mandatory to press down on the chip to ensure all pads make contact. For larger packages you have to press down at several points on the chip because of board flex.

[OwO]

LGA soldered Artix-7 (XC7A100T, FGG676):

[asmi]

I have my doubts about long term reliability of such sloppy soldering jobs as repeated heat cycles (these Zynqs heat up quite a bit, and so do memory modules) can cause joint failures.
It should be OK for prototypes though, unless these prototypes are actually going to be used for reliability testing.

[Miti]

I have my doubts about long term reliability of such sloppy soldering jobs as repeated heat cycles (these Zynqs heat up quite a bit, and so do memory modules) can cause joint failures.
It should be OK for prototypes though, unless these prototypes are actually going to be used for reliability testing.

I don’t think anybody would use this technique for production. This is only for quick and dirty prototyping with some parts that you salvage from other boards , like Bassman59 said, or like I did. These Cyclone chips are over $300 CAD a piece at DK and quite capable.
I did try moving a ball of molten solder in a pool of flux on top of the pads on the board and I was amazed how even the solder bumps are. I think I’ll give it a try one day when my work schedule goes back to something less inhuman and I can find the time to make a board for them.

BTW, what would you like to see around this chip on a prototyping board?

[soFPG]

Then during reflow it's mandatory to press down on the chip to ensure all pads make contact.

So how did you achieve this? Put on a little weight on the chip? Because I imagine opening my oven during reflow would drop the temperature drastically and with a conveyor belt system it seems impossible to access the PCB during a reflow run.

[OwO]

Was using a hot plate. You can use scrap PCB pieces to press the chip with one hand and hold the hot air gun with the other.

[Miti]

I could design a PCB for the Cyclone V on 4 layers, but I don't have the inclination to do so because I don't plan on using Altera parts in the future, but if you can get the PCB designed I can help prototype it. PM me or add me on discord (username in signature).

If you did, what would be your stack-up for such a board?

[OwO]

top - traces
in1 - solid ground plane
in2 - traces
bottom - ground plane

Power would be usually routed on top or in2, but in areas where non-critical traces are on in2, power/other signals can be routed on the bottom layer too.

[Miti]

What power modules do you recommend for Cyclone V? I recovered the integrated power modules from the original boards as well but the footprint is not really... “hobbyist friendly”.

[OwO]

That footprint does not look too bad. You can probably solder it by tinning the pads on the board and reflowing.
I usually use standard sot23 buck converters (AP3429, FP6373, etc) for loads up to 2A, and something like JW5068A for higher current.

[Miti]

Is there any difference between the dedicated clock inputs? If I read correctly the pinout (I'm looking at U484 in Pin Information for the Cyclone® V 5CEFA7 Device), CLK0p goes all the way to P9, which is almost in the middle of the chip. I guess there's a reason for that, equal paths to the edges of the silicon???
There is CLK2p at U13, which is closer to the edge.

[OwO]

I'm pretty sure all clock inputs first get routed to the middle of the chip, at least this is what the xilinx device view shows. All clocks must go through a global clock buffer at the center of the chip, that then drives a big fanout network.

[Daixiwen]

The Cyclone 5 FPGA has 2 or 3 clock modules placed on the periphery of the chip. Clock placement can be important if you have very tight timing requirements, but for regular use any clock will just be connected to a global clock resource and its physical position does not really matter.