We're running into an issue where 9.7% of the time MYKONOS_waitInitCals returns the error: MYKONOS_ERR_WAIT_INITCALS_ARMERROR. The data we measured was over a span of 1300+ configurations of the chip across a few unique mykonos profiles. This specific error is split evenly across profiles that target RX and TX only configurations. To be clear, the profilesValid mask is set to 0x7, which means the TX,RX, and ORX profiles are valid, however when we run a "TX only" configuration we only run initial calibrations for the TX side (including the QEC calibration). Same goes for RX only configuration.
Essentially, 5% of the time our configuration sequence times out for initial calibrations when we used TX calibration flags, and 5% of the time it happens with RX.
The kicker to all of this is, 90% of the time we don't experience these timeouts on the same set of unique configurations. As I mentioned, we logged this data over 1300+ configuration sequences, but there's a small set of unique configurations that are getting repeated. So the question is, why does the MYKONOS_waitInitCals timeout so inconsistently? Why does this number drop to zero when QEC calibration flags are commented out?
I tried looking at similar posts to this forum about this very issue but it doesn't seem to be very clear. One thing mentioned is making sure the ports are impedance matched, but these configurations are all happening in the same environment so why would it fail 10% of the time?