I started shopping out prices for those re-balling jigs you see in dosdude1 clips and am starting to wonder whether I’m going to, foolishly, level up to buying a board pre-heating station and mountable heat gun, with intent to just start swapping out BGA CPUs, PCHs, and GPUs on early Intel Macs which are either faulty (or which can be bumped up to something even nicer).
When I first saw this post, I was skeptical as to how feasible it would be to perform this kind of rework without a full lab with well-characterized tools and magnification. Hopefully my weekend's project serves to provide some inspiration for the scrappier among us.
Over the holidays, I was given an 13" A1466 MacBook Air, which was suffering a little with a dead battery, the base 4GB of RAM, and what looked like the remnants of a coffee spill. After a quick look
elsewhere on these forums and then on Aliexpress, I picked up a $24 set of four Samsung K4EBE304EB-EGCG 32Gb chips with the intent of upgrading this A1466 to the maximum 16GB of RAM, as
@dosdude1 has done before and documented well.
I also picked up an 8208 hand-held hot air gun that purports to have temperature control as an upgrade over my discount Harbor Freight paint stripper that I've mainly used for heatshrink since I'm not brave enough to point it at anything more expensive. Once I had this hot air gun in hand, though, it occurred to me that I could probably pull off a RAM swap using my home lab rather than borrowing a more professional setup like I did in
my previous attempt at a RAM upgrade.
I wound up using one of those cheap 60W USB Type-C PD hotplates as a board pre-heater, which was only feasible because this A1466's 820-3437-B logic board is completely flat underneath the RAM chips. Combining that hot plate (eventually set to 200ºC, but its thermocouple is only loosely attached to its heating element) with the hot air from above (set to 350ºC), I was able to pick the chips off with a very finely-pointed pair of tweezers.
Unfortunately, I was a little too impatient with the first RAM chip I removed, gouging the board with my tweezers and tearing a few of the no-connect (N/C) pads off the board. This is a mechanical shame but thankfully not an electrical problem. I was more careful with the other three and extracted them without any issue.
In dosdude1 style, I tinned the pads with leaded solder and then used solder wick to get the pads flat enough to receive the new chips. I was able to do this with no problems using a classic T18-style iron. One hiccup (not pictured) is that the new Samsung chips, despite having four times the memory capacity of their Elpida predecessors, were narrower and so the alignment markings I'd made on the board to mark where the corners had been were not entirely helpful.
I was able to align the new chips by eyeballing the center point and also by comparing against a different A1466's flooded logic board which happened to have Hynix chips in the same package.
Refitting was the reverse of removal, with the hotplate again serving as a preheater and the hot air being applied from above. Thanks to the surface tension of molten solder, the classic "poke the chip and see if it wiggles" test worked out perfectly for all four chips.
I reassembled the Mac just enough to boot it to a no-hard-drive-detected flashing question mark and confirm both that I hadn't hopelessly shorted anything and that the RAM was nominally working.
At this point, the SPD data needed to be corrected so that the Mac would know that it now had 16GB of RAM available. Since there's no dedicated SPD EEPROM on the board, this data is stored in the SPI flash that holds the UEFI configuration, specifically in the UEFI section with UUID D357D673-4816-486C-9982-A8F6B7E0F569.
While most people simply desolder the SPI flash chip, I decided to try my luck at reprogramming it in-circuit with a CH341A programmer and some 34AWG fly-wires. This worked out perfectly, and I was able to patch the SPD data as per dosdude1's guidance, using the "old engine" version of UEFITool which allows for patching.
(Apparently sometimes you can't just back-feed power to the board like I'm doing here, but I figured it was easy enough to solder to the SPI flash test points that it wouldn't hurt to give it a try before reflowing the chip off the board.)
After booting into a fresh install of Mavericks, the Mac is reading my patched SPD data correctly and detects all 16GB! (the Logitech receiver there is because
somebody didn't plug the trackpad/keyboard cable all the way in...)
So in conclusion, it's totally possible to do BGA rework at home without any magnification or especially fancy tooling. Probably the most premium soldering equipment in this project was my SMD291 flux, which is admittedly expensive - but good flux (of whatever brand) makes life so much better than trying to work without it.
This is obviously smaller-scale than the CPU/PCH/GPU swaps
@B S Magnet was considering in the quoted post, but for anybody contemplating the dive into BGA madness, the barrier to entry is not as steep as it may appear. The difference with that sort of scale is that you'd really want a larger preheater and possibly a larger nozzle on the hot air gun so that the board and chip don't warp due to heat...
One other classic BGA challenge not addressed here is reballing; at the ball sizes used here, it is likely feasible to reball with the naked eye, but newer components get increasingly dense and would definitely push the limits of my vision.
Now to find a battery for this Mac and try not to succumb to the siren song of refurbished stereo microscopes...
