For CERN RF, fixed latency is used (the variant with the wr ref clk used for application):

• we measure max latency at 7.5us level
• the fixed-latency is set for 8us
• we can observer late frames cnt increase The above seems to be a bug.

Moreover, the latency of nearly 8us over a single switch and 1m cables... it is too much. Looking at different systems in operation, this seems consistant (i.e. ~6-8us over 1 wr switch measured using WR Streamers). When another switch is added, the latency is at 11-12us level (increase of 4us). It is a bit too much to have the overhead of up to 4 us. It is possible that this can be introduced inside the streamers module itself... and could be optimized.