Replace FPGA? [success!]

njcnjc
edited November 2020 in Problems Now Fixed
I'm sad to say I had a "zorching" incident where 5V was applied to one of the I/O pins on my kiwi/beagle. It took out both the kiwi and the beagle. I measure ~1 ohm on the 3.3 line of P9 and also across C29. The 3.3A, 3.3G, 1.8, 1V all look good if I apply 5V power. It looks like 3.3 is coming from the Beagle. If I connect the Beagle, it immediately shuts down.

So my questions are:
  • It seems likely that the FPGA is the damaged component, is there a way to verify this?
  • Which is the correct speed grade to replace it with, "1" or "2"?
    • XC7A35T-1FTG256I
    • XC7A35T-2FTG256C
I'm reasonably confident I can replace the BGA. I did it successfully on a Red Pitaya several months ago. The part is about $40 on Digikey, so it's worth a try.

I'm just looking for any advice to maximize my chances of success.

Thanks,
Nick W1NJC

Comments

  • jksjks
    edited October 2020
    It's great that you have the equipment and skills to replace BGAs with hot air. I wish I did. All I can do with my hot air machine is blow 0603 components off the board and into the carpet where they are never seen again.

    The lowest speed grade is fine (-1 I think). Nothing in the FPGA (currently anyway) runs faster than the 66 MHz ADC clock.

    One thing to definitely do after you remove the FPGA is measure the 3V3 rail again to see if one of the ceramic caps is shorted. The larger values (4.7uF, 100uF) are only rated at 6.3V which should have survived 5V, but you never know.

    You might be able to check for shorted caps before taking the FPGA off by using an IR camera to look for overheating. Or use Louis Rossmann's trick of covering the caps with IPA and seeing if one (or more) evaporate the IPA faster than the others.
  • njcnjc
    edited October 2020
    Thanks for the tips, John. Yeah the hot air station I have at work is pretty decent. When I did the Red Pitaya I used kapton tape over everything I didn't want to blow away and it seemed to work. That was the first time I'd ever attempted something like this and that part is a 400-ball 0.8mm pitch BGA. It took three attempts but I learned some things in the process, and this part is "only" 256 balls at 1.0mm so I'm hopeful that I'll be successful. In case anyone is interested, here are some things I learned last time:
    • Use a high enough temp. I found I needed about 500°F (260°C) to get things flowing quickly and evenly.
    • Use a tip that is about the size of the part. My first attempt was with a smaller tip and I was moving it in circles - this was not successful.
    • Be liberal with flux. I found that the part was adhered but didn't work. I applied flux to the perimeter of the chip and tried again. Capillary action must have sucked in the flux and helped things flow because it worked after that.
    • This goes without saying, but you must have perfect alignment. I used a high-res closeup picture of the original part before I removed it and a microscope when I placed the new part. Basically I looked for any features on the board that are useful for alignment: vias, traces, silkscreen, etc.
    Yes, I had planned to measure for shorts on the voltage rails after the FPGA is removed. I suppose I could connect a current-limited supply to the 3.3 and look for hot spots. I do have a FLIR I could use but the IPA or cold spray trick works too. At this point I'm feeling pretty sure that at least the FPGA is bad.

    OK on the -1 speed rating. I see now there is a "1C" printed on the part. The 1I (industrial temp rated -40°C to + 100°C) is available so that's what I'm going to get. $50 w/ tax+shipping.
    I'm really glad you responded because I wanted to make sure there wasn't something else I might have been missing. Now I feel comfortable proceeding.

    I'll post here with my results.

    Thanks,
    Nick W1NJC
  • jksjks
    Yeah, Rossmann is the king of "can never use too much flux" in his videos, lol.

    You might want to edit your post to say 1.0mm for the Kiwi BGA pitch. Almost trivial these days compared to the 0.3mm CSP packages that exist.

    Good luck!
  • njcnjc
    Good catch, thank you. I've fixed it. I have successfully removed the part and cleaned the pads. No shorts now :) I had to up the air temp to 530°F before she broke loose. I remember something similar last time too. I'm ordering the part, it probably won't arrive until next week.

    Stay tuned!
  • n6gnn6gn
    When chasing low impedances or shorts on a board like this, before the most drastic steps I often use a current-limited supply set to something reasonable such as perhaps only a few hundred mA, and then use a millivoltmeter across PC traces to follow the current.
    Had the problem been associated with a shorted capacitor you could likely have found it and avoided removing a potentially still good FPGA. It sounds like in this case current following like this would have led to the FPGA anyway but it's a useful trick.
  • jksjks
    They probably use lead-free solder which is maybe why you had to go to 530F ? Not sure though.
  • njcnjc
    Success!

    The BGA operation worked and the kiwi is functional again using a BBB that I had.

    Now, I've been trying very hard to get the MT7601 USB wifi dongle working and it's proving just as difficult as last time. I have managed to finally get it working but via connman this time. It seems like it takes a while for the kiwi service to respond on the wifi address. The address displayed in the kiwi admin page always shows the eth0 address. This thread is not really the place to talk about this but I wanted to give you an update.

    Nick
    cathalferrisHB9TMCChrisSmolinskiPowernumptydl7awl
  • jksjks
    Wow, congrats on that!

    A while ago someone succeeded in replacing the GPS front-end chip (24 QFN with thermal pad). I thought that was difficult enough. Never expected to see someone attempt the BGA.
  • njcnjc

    So the kiwi has been working great...until Sunday! After some debug, I found that the 1.0V rail was shorted to ground. I eliminated components by removing the diode, and then the inductor on the power supply. Using a FLIR device and running some current I determined the problem was under the FPGA. So I figured what the hell, kapton taped up the board again and fired up the heat gun to reflow the BGA. It worked!

    Nick

  • jksjks

    Lol -- that's just awesome. When you first replaced the BGA how did you prep the pads after removing the bad FPGA? Did you do anything? Wick the existing solder paste? Flow with leaded solder and wick that? Hail Mary?

  • njcnjc

    I used the really small and thin solder wick and an iron, being careful not to rip anything up or scrape any solder mask. It took patience. I used a microscope to inspect and went over any pads that I wasn't satisfied with. Then I cleaned them with isopropyl alcohol and some other board cleaner stuff we had at work. Then I used "BGA repair tacky flux" on the board before placing the part (which is pre-balled) and heating it up. I didn't have a stencil so I just slathered it on there, fairly liberally. As mentioned above, it took a few tries and additional flux around the perimeter before it worked.

    I'm not sure what happened to cause the short - especially 6 months later. I guess it was some misplaced solder or a fragment of something under there. It was a dead short too, like 1-ohm or less. I'm glad the reflow worked because I wasn't too thrilled at the idea of replacing it again...plus with the chip shortages they seem to be hard to come by.

  • dl7awldl7awl

    "I'm not sure what happened to cause the short - especially 6 months later."

    Maybe there had been some whisker growth - I have already seen some cases like that.

    BTW, respect and congrats!!!

    73, Manfred

  • PowernumptyPowernumpty

    I wonder if it possible to flush any residual flux out from under the BGA afterwards?

    https://www.dfrsolutions.com/hubfs/DfR_Solutions_Website/Resources-Archived/White-Papers/Reliability/Review-of-Models-for-Time-to-Failure-Due-to-Metallic-Migration-Mechanisms.pdf

    njc
  • njcnjc

    Thanks, guys. It's likely I used too much flux, but it seemed necessary. Maybe I should hold the board vertical, heat the chip (not to solder melt temps but enough to liquify the flux) , and apply IPA to the top and let it flow down and wash out. Technically it's a "no-clean" flux but perhaps it incentivizes whisker growth. Then again, having some flux left over under there makes a subsequent reflow operation more likely to work. It's a little early for a white paper, but maybe after some more coffee!

    73,

    Nick

  • PowernumptyPowernumpty
    edited April 2021

    No worries on the white paper I only posted it to make it look like I can read.

    (my dog told me "no clean flux" is in there though)

    I did see metal migration in sata power cable adapters once, blew me away, looked like it had been hit by lightning but must have been cooking up for weeks/months before eventually setting fire.

Sign In or Register to comment.