In the last week a few tests have been performed with TCS chillers to investigate better and find possible mitigations. Mitigation attempts are up to now not effective. Although some strategies are still to be attempted at the chiller side, we will try to investigate also the the possibility to mitigate the coupling. Here a summary of the latest attempts and possibile further strategies:

1. on Feb 9 the heater of the WI chiller has been disconnected. Of course the noise disappeared, but in spite of many trials modifying control parameters, the temperature stability could not be recovered. One limitation is that the chiller firmware does not allow to completely disable the heating loop. The WI chiller heater was re-connected on Feb 13 but the WI chiller was kept off till Feb 16.
2. on Feb 14 the NI chiller has been switched on alone (i.e. while the WI chiller was off). All along this time the noise never appeared and as it was confirmed by the chiller on-board display, the heater never turned on (note that the chiller does not provide remote information on the heating status, the information is provided only at the on-board display). It was also tried to increase the T set point from 19°C  to 22°C: the heater turned just for a few minutes during the transition, than it kept silent again. See Figures 1 (UPS spectrogram 24h Feb 14), 2 (UPS spectrogram 1hour during T set change) and 3 (chiller temperature and T set). We have to conclude that "something is different with the NI chiller" which makes it never turn on the heater. (for conclusions see point 2. below)
3. on Feb 16 the WI heater was plugged on IPS (ENEL) instead than on UPS as usual, and switched on. We observed that the 1Hz noise at the level of INJ signals somehow reduced (a factor 2 or 3) but it was still present. As expected, the 1Hz comb noise disappeared from UPS and appeared on IPS, the noise is roughly the same at the magnetometer location: Figure 4 (shows noise with WI chiller on IPS compared to chiller OFF) and Figure 5 (compares noise with WI chiller on IPS and on UPS). (for conclusions see point 3 below)

Conclusions and further steps:

1. operating the chiller with the heater off seems not feasible.
2. we will further investigate for possible differences between chiller NI (not noisy) and WI (noisy). One possiblity is that there is a difference at the level of the firmware or the settings (indeed the Chiller technician we spoke with mentioned that up to his knowledge the heater PWM cycle is 3Hz, while in order to explain our observations it should be 1Hz cycle) . Another possibility is that the "magic" condition of the NI chiller is due to the difference in thermal load for example because of the less heat dissipation in the shorter water pipes (also to be checked if there is a different regulation of the water flow).
3. The fact that the noise does not disappear when the chiller is powered on ENEL, makes us suspect  that even plugging it on a dedicated UPS would not solve the problem.
4. One possible mitigation strategy is to replace the chiller with that of Virgo (which was only cooling) and its custom control (however we know from data that the old chiller performance was at the level of 0.1°C at the most, while the new chillers performs at 0.01°C, and seems that the new CO2 lasers need this level of T stab.)
5. Another strategy that we will start pursuing is to investigate further the coupling path and act on it.