New results for the problem I have (double byte sent instead of one) :
It seems that there is no link with the fifo behaviour, contrary to what I said, but only with CTS signal behaviour of the UART !!!
When there is no dataflow management by the bluetooth tranceiver (UART's CTS never goes high), there is no problem.
Different lenght of Fifo and circular buffer (Fifo lenght > or < circular buffer) could modify CTS frequency. And CTS frequency is linked to the appearance frequency of the double byte.
I'll export Fifo's signal to determine if there is a link with full state fifo frequency and the double byte transfer.
I'll tell U