Tool/software:
Dear Support Team,
Sorry for creating a new thread, but the old thread is already locked.
We have thoroughly tested the new PCBs made on a true high-speed laminate and with small improvements in power decoupling and layout. They perform much better, but still not perfectly. Depending on the combination of a PC port and a connecting cable, the problems appear or not. However, even when the problems do appear, the periods of undisturbed operation are now minutes or even tens of minutes, while the previous PCBs exhibited link losses within seconds. Same as with the old PCBs, disabling U1/U2 support in TUSB8020B through an EEPROM helps, the problems disappear completely then.
We have tested cables of different lengths from different vendors. Some PC ports performed generally well, while some were sensitive to the cable selection. Shorter cables often (but not always) perform better. We have even dissected the cables around their plugs and have seen very different levels of manufacturing quality with respect to shielding near the plugs and keeping lengths in differential pairs. Funny that the some of the best performing cables looked the worst, and vice versa.
Interestingly, our oldest prototype (7 years old) -- not that different in PCB trace layout and lacking many improvents, made on a mediocre laminate -- is still unbeatable with respect to the link stability in various port/cable combinations (with standard U1/U2 enable settings). This is the oldest of our TUSB8020B chips. But it is the same silicon revision as the fresh ones so it should not matter, should it?
Regards,
Michal
