Paul said:
VanguardLH said:
daytripper said:
[...]
The mobo maker could just make a plastic frame to hold the chip in place
(both for position along with affixing to the mobo via spring clip) and
the user would use a soldering iron with a tip designed for the BGA grid
pattern. The user would buy the mobo they want, the CPU they want, and
then do a one-time solder of the CPU onto the mobo.
[...]
That's some funny stuff right there.
Unless you're serious, of course...
Cheers!
I was serious. You do know what "ball" means in BGA, right? It's a
ball of solder. So why can't the chip, even a CPU, come prepped with
the balls of solder on its pads, the mobo come with balls of solder on
its grid and using feedthroughs so the solder is reached from the
backside of the board, and all you have to do is keep the chip pressed
against the grid, keep it aligned, heat up the solder gun with a
matching grid tip, and just melt all the solder to weld the chip to the
grid?
You've never applied new solder to the underside of a PCB so it heats
the solder on the other side through a feedthrough to use solder wick on
the other side when you cannot otherwise reach the other side with a
soldering iron? Heat travels.
Of course, we're talking about DIY'ers that know how to solder and that
it flows towards the heat source and what level of heat to apply and not
the boobs that barely know how to push down the level for a ZIF socket.
Not having sockets doesn't mean you can't DIY. It means the DIY'er will
need better skills than pushing stuff into a socket or slot.
You at least want to solder all the balls at the same time.
Hence why I said a soldering iron tip that matches the grid you want to
solder but assumes you can get heat flow through feedthroughs in the PCB
(but that'll make the PCB more expensive so this probably won't happen).
I'm not sure a cheap soldering iron (with an array of different tips,
like one for BGA wiper tips) is going to work well. It won't have the
temperature regulation of a professional soldering station where you can
dial in the target temperature and it maintains it during thermal load.
I've seen machines that runs the PCB through a hot tin bath (or "wave"
that flows over a roller to create a reverse trough). The balls get
melted in a row, not the entire grid at once, as the PCB rolls over the
wave of solder. You need to preheat the PCB and chips before hitting
the solder wave; else, the uneven heating can result in "popcorning" of
the chips (their tops pop off). Obviously extremely few home users are
going to get a wave soldering machine. The point I'm making is not all
of the balls have to melted all at once (but they should be preheated
all at once).
There is a magic alignment effect, where the wetted contacts
tend to "pull" the chip into alignment, such that the chip
rotates to the grid of contacts underneath. You want the
solder to fill the pads properly, which is going to
happen if all the balls melt at the same time
and the chip settles into place.
But if the chip slides into a carrier to force alignment, something like
a modified PLCC socket but made for alignment purposes instead of
providing for contacts:
http://www.hermann-uwe.de/files/images/bios_chip_plcc_socket.preview.jpg
then alignment doesn't rely on wet solder flow pulling the chip in
place. Yes, this is a socket but that doesn't preclude designing a new
socket (which would be much thinner than this) and whose only function
is to force alignment of a unsoldered BGA chip along with a temp clip to
apply downward pressure (until after soldering is complete then remove
the clip and attach the HSF probably using PCB mountings).
I'm saying something could be figured out to provide alignment and
pressure so a home user with decent soldering skills could install a BGA
chip on a prepped grid on a PCB. That doesn't mean it will happen since
manufacturers are trying to save on tenths of a cent so adding such an
alignment and soldering helper device probably won't happen (unless it
comes as an extra you buy separately to use with the BGA chip but then
the mobo would have to be designed to allow the space for it). There
are a lot of HSFs that are designed for what's there rather than the
design of the mobo planning for them.
If you were a home user, and desperate for adventure, you could try a
toaster oven. That's the closest thing to IR reflow you can arrange
for real cheap.
The closest I've seen to multiple contact soldering and for cheap is
using a pancake griddle on which to lay the prepped PCB with BGA chip
atop. Obviously the pancake griddle method is applying heat on the
opposite side of the PCB than where the BGA chip gets soldered, so
off-side BGA [de]soldering is possible.
This next guy uses a pancake griddle but elevates the PCB off of it.
That's because this next guy is reflowing the solder rather than
replacing the chip. He uses the griddle to up the temperature but a
heat gun to elevate to the melting temperature:
(looks like the guy is in a cramped dorm, ignore the background "noise")
Here's another guy using a griddle to preheat, foil tape for a heat
guard, and a heat gun to remove the BGA chip:
So pancake griddle, soldering iron with grid tip, or heat gun are all
methods to melt the solder balls. Because these methods may not keep
the chip always aligned because they may liquify the solder balls
unevenly. That is why I thought of some carrier or holder in which the
chip sits and perhaps a spring for a little downward pressure would
ensure alignment. A CPU as a BGA chip is going to have a *lot* of
contacts versus the smaller BGA chips you encounter. I'm assuming a CPU
is going to be a much larger BGA chip with a lot more pins thus with
more mass and more susceptible to sliding around if the PCB is unlevel.
The "carrier" could be something really simple, like just 8 posts, 2 to
snug in each corner or the chip, that either come in the mobo or can be
inserted and removed in mobo holes when needed. I figure something to
help alignment rather than rely on the wetting pull might reduce the
number of bogus "faulty" returns to mobo vendors or botched jobs.
Somehow I suspect mobo vendors aren't going to allow any soldering and
doing such will void any warranty - but then how often do users change
their CPUs? By the time the user wants to upgrade or replace a faulty
CPU is probably after the mobo warranty has expired.
Some people used the toaster oven method, to fix Nvidia GPU solder
joints. But I would still put this idea in the "repugnant" category.
You have absolutely no control of the temperature profile that way,
and the toaster oven is going to be heating all sorts of stuff you
don't want heated (think "burned plastic").
That's why even with the pancake griddle method I would think you would
want to deflect heat away from the parts you are not soldering. In a
couple videos, I see them using foil tape to deflect heat. Another idea
would be to use a metal block on the griddle along with some heatsink
paste to get the griddle's heat to only apply to the underside of the
PCB where you are reflowing the solder under the BGA chip.
(Voids caused by excessive moisture level on the solder balls.)
I think the moisture problem is caused by using water-based rosin. I
haven't looked into this source of soldering defect. I don't know if
the use of directed heated air (e.g., heat gun) eliminates the moisture
issue. From what little that I read, the moisture problem is not due to
ambient moisture in the air but of moisture *inside* the BGA package
getting baked out. I'm not sure why a chip maker would permit an
atmosphere inside their chip that contained moisture. Seems a defect in
or a poor process during manufacturer, and the maker is that slack in
building the chip then how much better quality would be the rest of the
chip.
http://www.intel.com/content/dam/ww...g-databooks/packaging-chapter-14-databook.pdf
Section 14.7: Moisture Sensitivity
It looks to be a manufacturing problem, not actually a soldering
problem. It isn't ambient/outside moisture causing the problem (so it
isn't a problem of how you perform the soldering). It's a problem with
moisture *inside* the chip that expands and pops off the top of the
chip. Whether soldering, applying heated air, or for whatever reason
the chip gets exposed to high temperature heat, a moist chip can pop.
You could, for example, be using a heat gun on some other component but
heat up the bad one to pop its lid. To prevent popcorning, it appears
the user of the chip (home users or fab plants) are expected to leave
the chip inside the dessicant bag until just prior to use. If the
problem isn't of poor manufacture of having excessive moisture inside
the chip then it appears the popcorning problem is related to poor
handling by the consumer/user of the chip.
http://glenbrook.webworksnow3.com/blog/wp-content/uploads/2011/03/bga_fig4.gif
With care, I'm told you can get defectivity down to around
1 ball in 100,000. That means, if you solder down a hundred
chips each having 1000 balls underneath, one of the chips
will have a single bad solder joint. It would take a little
effort and expense though, to get that good at it. The
results of home users doing such soldering, isn't going to be
that good.
The typical DIY'er doesn't have $200+ temperature-controlled soldering
stations or even the soldering skills for just splicing wires together.
I don't remember how many times I've helped with replumbing a house to
watch someone apply heat at the joint and think that's where they heat
up the solder instead of letting it flow towards the heat by applying to
torch past the joint overlap. There's a reason why DIY hardware didn't
involve soldering. Hell, I remember working on the test floor where we
tested the mainframes and watching some so-called engineers raping a PCB
trying unsolder a part and then gobbing on the solder to replace it. I
still have memories of one guy that would poke the tip of a soldering
iron into a blivet and circle the iron in a conical pattern trying to
heat up the solder inside to then suck it out all the while hearing the
squeaking of the tip against the dry blivet. I've seen some folks using
an soldering gun instead of a soldering iron so they end up using way
too much heat in too large an area. Soldering seems almost an inate
skill, like whether you can play a piano or just bang your fingers
against the keys.
Probably a better setup is ordering both the mobo and CPU (BGA) together
and having the vendor do the initial soldering for a fee. You get to
choose what you want on the mobo but someone with proper gear and
expertise does the solder work. Then sometime later when you want to
upgrade the CPU, and probably after the mobo warranty has expired, you
can try to remove and replace the BGA chip yourself. You can do it
hoping the PCB is level and that wetting gives you proper alignment but
I would think something to aid alignment would reduce the number of
botched jobs by amateurs, like me.
I think someone already mentioned using a converter card where the BGA
style CPU would get soldered onto a card and the card shoved into a
slot. This is reminscent of the Slot 1/2 CPUs. Of course, it means the
CPU manufacturers would be producing BGA style chips while someone else
was manufacturing the cards and doing the soldering for you so you end
up buying the BGA CPU already on a card. Mobo makers have to follow by
providing CPU slots.
Of course, all of this "what do we do know" conjecture is based on
someone's hypothesis of what Intel /might/ do. The computer industry is
rife with technologies or schemes that dead ended. Remember the
microchannel bus? Was it the EISA hosts where if you lost the floppy
then you might not be able to boot the computer? I was surprised at the
3 years the SECC/SEPP form (Slot 1) for CPUs lasted considering how
often the mobo slot got cracked which resulted in less pressure on the
fingers against the card's pins. Several times I had to use cable ties
looped together across the top of a Slot 1/2 CPU to hold it down because
the mount brackets cracked or broke, or having to scavenge from failed
Slot 1 CPUs to dismantle the shell and replace the fan. How long did
Microsoft's BOB last?
I'm pretty sure if Intel decides to direct its manufacturing to
producing only BGA style CPUs that something else will arise to meet the
challenge of installing and replacing those CPUs outside the mainline of
computer parts makers. Or DIY'ers will have to improve their soldering
skills and perhaps their tools from those that don't want to learn and
practice getting weed out. Look at what happens with cars. Some folks
simply don't want to learn beyond the aspirated engines they grew up
with and/or don't want to buy the tools to keep up with the changes.
Having to solder or reflow BGA parts will become a differentiating
factor in typing the DIY'ers.
