Left – variation with time of the power dissipated by the PID regulating the temperature of the bolometer plate (red curve; same as the green curve in Fig. 29). Subtracting the part that is correlated with the cosmic ray heat input as measured by the SREM gives the green curve. Subtracting the part that is correlated with the PID on the dilution stage gives the blue curve. The constancy of the blue curve shows that cosmic rays and variations in the dilution stage account for all of the very low frequency variation in heat input to the bolometer plate. Right – frequency domain versions of the same curves.