Open bapivee opened 4 years ago
Update to focus on data transmission.
Should we optionally allow the use of the RX clock to generate the TX clock in the other direction. Either end of a connection should have the option of being the clock master.
Grab the text from the old version of the spec. Review at the next meeting.
Good impact above the stack for the elastic buffer crossing clock domains.
Not sure what is the need for a master slave configuration? Each D2D is independent Training and calibration happens on the RX Clock Data needs to flow independent of the TX/RX pair
https://github.com/opencomputeproject/ODSA-BoW/blob/009cd9eda3e79b7c0bf5f1e7520417e303ed1640/spec/bow_specification/bow_specification.mdk#L376
How does each side process the received clock - in data transmission and at initialization. Could be the same as leader/follower. Should we copy from the AIB?