Is the retimer's output physical connection length declared for the D-PHY 2nsec limit independently of input side connection length i.e.: the repeater becomes a defined link source and the receiver on the other end the link sink per D-PHY definition?
As an application example, a receiving (sink) SoC device is located on a PCB and the required signal breakout prevents locating a CSI-2 cable socket nearby. The full 2nsec budget on the input side of the retimer is needed to support a 25cm cable to an off-board camera sensor. Since the DPHY440SS is a retimer, and I believe the D-PHY flight-time applies to the wire length entirely for SI reasons (differential and inter-lane skew delay, signal loss, etc.), the signal and timing cleanup provided by the retimer effectively resets the total distance to it's output port, meaning that a full 50cm link (input + output) can be achieved as long as the retimer is located half way. Since each connection segment of the link stays within the 2nsec limit, the link meets the spec.
If the spec actually applied to total absolute link-layer delay, solutions that convert CSI-2 lanes to LVDS and back for long reach links would not be compliant and would therefore not exist. Am I missing something? I would expect TI to have had many customers with the same or similar question and be in a position to publish a clear application guideline and design example.