Hi, > I have performed the following tests on PCBAs #2, #3, #5 and #6: > * power up and check current, all boards around 0.6A
The results is same as mine here. Means they are all consistency. Good. > All the boards passed the tests except PCBA #3 which could not netboot. The > Ethernet carrier was detected and the autonegociation worked, but the board > failed to send packets. After inspection, I found out that pin 18 (TXD1) of U2 > was not correctly soldered. I manually re-soldered it, and netboot and Linux > worked. Did pin 18 short with adjacent pins? If yes, means AOI process doesn't inspect well. If not, that's soldering joint problem. Then human visual inspection might not easily check out. But I need to feedback to smt factory surely. > A similar error (not covered by the tests above) was found on PCBA #5 where > pin 4 (D) of U4 was not properly soldered either. We might need to review the > SMT soldering process to reduce the occurence of these problems. Or are we > within the normal failure rate of SMT lines? Not about within failure rate. Quality about solderability needs good material quality(no oxidation, etc.), good temperature control profile, good human management process. From the functional success of fpga, the temperature control of profile is surely good under bga soldering during smt IR-reflowing . So far this time the parts prepared are most bulk ones not reel. A moisture control monitoring is hard to face those bulk ones if I checked them when I received. The spartan-6 chips I didn't take apart well-sealed bag which I tried to take risk not take apart to check if chip package is correct. So U2/U4 was yes that I took apart and probably got oxidation caused they were not soldered well. So I'll notice this. > About the substitution parts listed on [1]: > * both Ethernet connectors (J4) are OK > * both JTAG headers (J6) are OK > * both power jacks (J11) are OK > * both VGA connectors (J17) are OK Sounds good. Means sourcing can be more varied. > * for the memory card connector (J19) the Molex device seems better. Although > others can be mounted and probably made to work, their metal latch sometimes > gets stuck and sometimes falls on the components creating opportunities for > short-circuits if the board is powered up. The Molex device opens/closes very > smoothly and the metal latch stays in place and does not fall on the board. Yes, the other two here samples exactly have the conditions on latch and smooth open/close. So let us choose Molex firstly. > We may also want to design the case so it blocks the latch and prevents it > from reaching the components. This can also be useful with the Molex > connector, so the latch cannot be manually pushed to create short-circuits. Yes, agreed. A case block that prevents it from touching components when opening. Thanks for this report which is good for next sourcing and run. Adam _______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkym...@freenode Webchat: www.milkymist.org/irc.html Wiki: www.milkymist.org/wiki
