To help with the alignment, I place small corner marks on opposite corners of the BGA package in copper. Remove the soldermask over these marks. This eliminates alignment errors due to silkscreen-to-copper misalignment (which can be quite large). One additional step I use is that I rub (figure 8 pattern) the BGA package, balls down, on a clean sheet of paper (placed on a flat surface). Keep doing this until ALL of the balls are showing a small amount of flattening. This ensures that all balls make contact with the PCB since they are now all planar.
I like your idea of rubbing the balls on a sheet of paper, makes sense to try it at some point. And yes placing any kind of marks in the copper layer would mean perfect alignment with the BGA pads, however in practice I have not had errors with just the silkscreen marks so for me it's really not worth the trouble.
@@voltlog I've seen significant misalignment between silk and copper for both domestic and China PCB manufacturers. Not every time, but frequently enough for it to be a problem. Marks on the copper will always be aligned to the copper...the only downside is that being on the copper layer, they can block some routing paths.
The ESD protection diode test is creative, it would work on all digital IO, but a lot of analog IO doesn't have them as it would affect their performance. Things like ADC inputs and voltage references.
I was able to successfully recreate this on the second try 🥳 Thank you so much for providing your gerber files. I was able to even get away with the cheapest process at JLCPCB (my go-to PCB manufacturer for years now, sorry PCBway) which is leaded HASL surface finish. I only needed to rework it a bit with flux and desoldering braid so the pads were nice and even. The first try I messed up by pushing on the eMMC, so I had to remove and reball the chip, which I surprisingly managed to do with my regular solder paste and a cheap stencil off Amazon. On the second try I just placed it very carefully within the silkscreen boundaries and heated it up with 290°C hot air from all sides and decided it should have melted by now at some point. And tadaa, it worked perfectly.
I found another useful thing about measuring the internal ESD clamp diodes a while ago... I had a board with some BGA SDRAM connected to a GPU (Dreamcast repro), but it had graphical glitching when powered up. It looked like most of the soldering was probably OK, but the graphics had a tell-tale sign that maybe one of the SDRAM address pins/balls wasn't soldered right. When I measured between Gnd or VCC, and the vias on the underside of the board (meter in Diode mode), I got similar readings for most of the signals. It showed a similar value for most of them, but one end of each trace is connected to the GPU as well. All of the address signals showed very close values for the voltage drop, like 0.487V or something, except ONE, which showed maybe 0.528V. Sure enough, I could just about see with a USB microscope that one of the solder balls near the end of one SDRAM chip wasn't quite reflowed. I heated it for a minute or so with the hot air station set to 360C, and saw the solder ball visibly change shape. It fixed the problem. ;) So, the relative voltage drop of each signal might also help if you have a board with multiple chips already connected together. (Some IO pins might measure a different average voltage drop, depending on how many chips are connected to that same signal, or just the type of IO pins on the chip etc., but it's a good rule-of-thumb to check for soldering issues.) It can also be handy to only used tented vias underneath BGAs when doing a prototype, so you have other surrounding vias that aren't covered by soldermask, which lets you shove a multimeter probe on the vias without needing to carefully scrape the soldermask away.
On a slightly related note: I bought the motherboard from an Xbox One S about a year ago, just to have a play with it, because it was cheap. It was listed as having a "very loose HDMI port", but the soldering tested OK there. I tried hooking up a fan, PSU, and Blu Ray board, but of course got no output. It's likely the board is dead anyway, or not worth the effort to repair. I couldn't figure out how the Jaguar CPU/GPU first booted, as there was no obvious SPI Flash on the board. It probably has a small boot ROM in the custom Southbridge, but also the large-ish eMMC Flash chip. What I didn't realize is that you can connect most eMMC flash chips to an SD card slot (on a PC) using only the 1-bit SPI mode (CMD, D0, CLK, and power), and it will often still work. :o (I thought they might have ditched supporrt for 1-bit SPI mode on most of these chips now, but apparently not.) So I might try hooking that up soon, and finding a new image to write to the chip. A board like yours could be ideal for recovering data from eMMC chips using only a few pins, assuming you can buy a BGA test socket for not too much money.
I soldered a 128GB eMMC chip to an Atomic Pi SBC, after desoldering the included 16GB chip and cleaning the pads. Been working for 1.5 years no problem.
Seriously, you're making it so simple and sweet such that I want to give you a hug. I have a part that is .5mm pitch with solder balls 0.35 mm just like yours. Do you think having via on pad still work with your technique? Thanks.
Vias tend to be problematic because they can suck the solder into the hole, creating a void between pin and pcb but results may vary. There is also the option to have plugged vias from the manufacturer.
@@voltlog Is there enough solder ball to make a connection despite of leaked solder into the hole? If tape the other side of the via with tape, would it work?
@@BS-my2ky Results will vary, you will have to test and inspect the vias from the other side. If you see solder wicked in there, you can be sure there isn't enough left on top for a reliable connection.
Hmm, I don't think your diode check is complete. It does correctly detect broken connections and shorts from IO to ground/supply pins. However, I would also check each combination of IO to IO, to confirm there are no shorts there. (Even if two IO are shorted, the diode to Vcc/ground would still show.)
Theoretically, yes! copper will have perfect alignment with the BGA pad. However in practice I have not encountered errors in the silkscreen alignment that would cause problems during soldering with this kind of pitch.
Very nice tip for checking the pin solder; in your experience would the datasheet mention the protection diodes ? Or we can always the pins before assembly of course !
Even if the protection diodes are not explicitly mentioned, you can guess that they are there if the "Absolute Maximum Input Voltage" is 0.6V or 0.7V above the power rail voltage. (And vice versa for the abs min input voltage.)
并不是中国的助焊剂便宜不好用，而是在ailibaba上找到一款合适的有英文说明的助焊剂对你来说有些难，其实中国还是有很多便宜好用的助焊剂， you can input the key words in Chinese "金鸡牌" `in Alibaba ，I use this almost in 10 years .
Translation: "It is not that Chinese flux is cheap and not easy to use, but it is difficult for you to find a suitable flux with English instructions on ailibaba. In fact, there are still many cheap and easy-to-use fluxes in China." ... "Golden Rooster brand."
The 60 fps makes the video unfortunately not watchable in 1080p, and even in 720p it is not as smooth as 1080p in 30 fps :-( Thanks for sharing you BGA experience :-)
Thank you for the follow-up video! It was helpful to see your techniques. One thing I would add is to check that the silk screen is properly aligned before using it to align a small pitch BGA. It is possible to have the silk misaligned by more than one ball pitch, even on advanced PCBs. This particular BGA package makes it difficult, but looking at other BGA packages from a shallow angle under a microscope can be used to check ball alignment.
If the balls are closer to the edge of the BGA package it certainly works to look at it from the side but in this case you can't see anything. I have never seen silkscreen alignment errors from PCBWay, their process seems to be pretty good and I order PCBs quite often. A bulletproof alternative would be to add some corner marks in the copper layer but I think that's not necessary in practice.
Unfortunately I've had it a couple times with them, but it's not much of an issue because I can usually see the balls from the sides. That's a great idea about adding corner marks in the copper.
Excellent video as usual. I'm curious as to the reason for the kapton tape on your hotplate - is it something you always use or has it some specific benefit for this type of component?
Sometimes I use it to solder PCBs that have thermal vias in pads which get solder paste. This mean that solder/flux can leak trough the via onto the hot plate causing a sticky mess. The kapton tape is there to protect the plate, I just replace it when it gets messy.
7:10 - if you have a working board, you can use any device that can chart I(V) characteristic chart and measure specific pins (for reference). Then you can quickly check any PCB and locate non-soldered (or damaged) components that need to be fixed. This is the most common and quite cheap technique to be achieved with a current sense amplifier (INA186A5 for example) and inverter to 20Vpp 200mA sine wave (depending on circuit capabilities it can be lowered to 3.3Vpk or 5Vpk, however, most devices will survive even those 10Vpk). Of course, you also need an internal series resistance to make your device safe from shorts (recommended is to switch them using relays). The benefit is that this technique can find exact damaged/unsoldered components (especially if you reflow BGA and accidentally desolder some 0201 which you forgot to fix). The disadvantage is that you need a known good circuit to make reference. It is also not perfect, but can solve 95% of problems. And for reference: checking a circuit with several hundred pins takes about 10-15 minutes ;) (depending on amount of pins in that you are referencing).
Louis Rossmann has some pretty interesting BGA soldering techniques with just a hot air gun and a ton of flux when he does macbook motherboard repairs.
This method is ok but temperature control is lacking and also you need a proper reflow profile for repeatability. A mini reflow oven is ideal for this kind of rework. Also try to brush the flux on in an even layer as too much will create waves and bubbles like you mentioned. At work i use the Martin Minioven 05 and for big pcb work a PDR focused infrared rework station.
thanks for this. i've been struggling with failed reballs on those ssd adapters, so the tip about testing the solder job with a meter has finally allowed me to get a successful read!
Marco Reps uses a homebrew vapour phase in a recent video. Just need a hotplate, metal container, glass lid and of course the fluid (whatever it's called) to make the vapour. Maybe you can give that method a try.
That sounds like something Marco Reps would do, but I looked for such a video and didn't find it. Do you recall what the main topic of the video was? Are you sure it was Marco Reps?
Great video thank you. Will you be getting any more of the boards in stock? I want to practice before attempting to replace the bga on my old tiny laptop
Temperature of the hot plate needs to correspond to the type of solder you are using. BGA packages tend to have lead-free solder balls so you should aim for 240-250 deg C
I'm thinking for BGA's you would want to follow a controlled temperature profile over time dictated by the solder ball type and flux constituent make-up. Led free seems more problematic. I guessing it would be quite specific to each case.
You can basically emulate a temperature profile with a hot air gun, start off far away slowly heating the board then move the hot air closer until it melts and then slowly move it away. Having a hot plate at a base temperature would help too.
