Hi, We had a Roach1 that had been used for a long time but only on designs that did NOT use any of the CX4 ports. When it came time to finally use it to ship around 10 GbE traffic we found something amiss.
Pin 1 of U21 wasn't well soldered to the PCB pad so the logic to apply power to the active CX4 transceiver would work correctly. It might work for a moment just as the transceiver was connected to the 4port Fujitsu module but would eventually loose power. I think that pin was slightly bent and not in the same plane as the pins on the IC so there was a gap between the pin and PCB pad and thus solder didn't make a good connection during assembly. The fix was a trivial manual resoldering. This functionality, powering active CX4 links, isn't tested during the standard test suite at Digicom otherwise it would have been found. Has anyone else experienced similar behavior ? If this is common enough we could supply Digicom w/ a couple of active CX4 links and then have them run a couple of additional tests: test 1) try the 10GbE tests with the active CX4 links as well as the the passive cables already used. test 2) leave the active CX4 links in during the long memory tests that are run prior to shipping and see if the CX4 links stay operational throughout. thanks, Matt Dexter
