The times are correlated with changes in the demodulation phase of B1_d3 photo-diode. It is however not connected to the phase being tuned or mis-tuned, but apparently only on the absolute value of the phase shifting asked to B1_Lo3. Before August 11 and after August 18 the dephasing was between 0 and -20 and the noise was present. In between these two times the dephasing was around -180 and the noise was not there.
In conclusion, it seems that the noise coupling depend on the dephasing asked to B1_Lo3 board. I do not understand how this can happen, unless there is some strange malfunctioning of the board itself.
We might try to make a short experiment rotating the phase and see if we can switch on or off the noise.
