Within the present work measurements done in different setups are compared. It is observed that the noise contributions in the CRESST main setup and a test setup significantly vary.
4.4.1 Baseline Noise
The relevant quantity that is typically determined to describe the noise contribution for CRESST detectors is the baseline noise. Within the present work the latter is determined from empty baselines, i.e. records triggered randomly by the DAQ. To determine the baseline noise a template pulse of fixed height is superimposed to these empty baselines and the resulting artificial pulse is fitted with the template [74]. The widthσ of the resulting Gaussian peak in the amplitude distribution is determined. It equals the baseline noise, since only the baseline noise contributes to the uncertainty of the pulse height.
Typically the absolute measured baseline-noise level in terms of voltage is observed to be similar for all detectors operated in the CRESST setup. This is expected as the parameters defining the noise sources discussed before are similar or equal for all detectors.
For example the phonon detector with the lowest energy threshold in CRESST-II phase 2 named Lise has a baseline noise of σPU,0 = 0.9 mV. The phonon detector with the best overall performance in CRESST-II phase 2 named TUM40 has a baseline noise of σPU,0 = 0.7 mV. In CRESST-II phase 2 the baseline noise of the different phonon detectors varies by a factor of∼3 between the best and the worst detector. For light detectors the baseline noise is in average slightly larger compared to phonon detectors.
The lowest baseline noise for a standard light detector measured in CRESST-II phase 2 (with the detector named Leon) isσL,0U = 1.4 mV. Also for the light detectors the noise varies by a factor of ∼3 between the best and the worst detector.
However, it is observed that the noise level measured in the test setup in Munich is considerably larger by a factor of up to∼10 (for examples see chapter 7).
4.4.2 Measured Noise Spectra
The noise spectra measured in the different setups are compared in the following. To determine the noise spectra empty baselines are exploited, as they are a sample of the
(a) TUM40
Figure 4.7: Noise spectrum of the phonon detector TUM40 (a). The frequency spectrum of empty baselines measured in the main CRESST setup is shown in blue, while the expected noise spectrum is shown in black. The latter is the sum of the individual noise contributions due to thermal noise (green dotted), Johnson-Nyquist noise of the thermometer and the shunt (blue dotted), SQUID noise (red dotted) and 1/f noise (magenta dotted). The individual contributions are calculated with the parameter values determined for TUM40 as given in (b).
The amplitude of the1/f noise is varied to obtain best matching with the observed spectrum.
noise. The frequency spectra of empty baselines measured with TUM40 in the main CRESST setup in Gran Sasso is shown in figure 4.7a in blue. In this spectrum many sharp lines are visible. They can be attributed to electric interferences (e.g. at 50 Hz) or to mechanical vibrations induced by the cryogenic facility11. At high frequencies the spectral density drops due to an anti-aliasing filter with a cut-off frequency set at 10 kHz during CRESST-II phase 2.
The expected noise spectrum (black line) is the sum of the individual noise contribu-tions due to thermal noise (green dotted), Johnson-Nyquist noise of the thermometer and the shunt (blue dotted), SQUID noise (red dotted) and 1/f noise (magenta dotted).
They are calculated with the parameters given in figure 4.7b according to the equations given in section 4.3. The parameters are determined from the pulse shape of TUM40 (τin, Geb) or from a measurement of its superconducting transition (RT, Te, m). The bias currentIB is chosen for the detector at the beginning of the measurement and the temperatures of the SQUIDTS and the heat bathTb are given by the setup. The only parameter that is not known a priori is the amplitude of the 1/f noise (∆RRT
T )1/f. The latter was varied to obtain best matching with the observed spectrum. Over a wide frequency-range the expected and the observed noise spectrum match well. Only at low
11In a dilution refrigerator, as used in CRESST, a precooling is achieved by pumping on a volume of liquid helium. In this way, this volume is cooled to a temperature of 1 K and the helium becomes superfluid. The continuous refilling from a bath with liquid helium can induce a noise contribution at discrete frequencies in the high-frequency range similar to the one observed in figure 4.7a [116].
The test setup in Munich was modified similar to the solution proposed in [116]. This might explain why fewer of such noise peaks are observed in the test setup (see figure 4.8).
(a)TUM26-b
Figure 4.8:Noise spectrum of the phonon detector TUM26-b (a).The frequency spectrum of empty baselines measured in the a test setup is shown in blue, while the expected spectrum is shown in black. The latter is the sum of the individual noise contributions due to thermal noise (green dotted), Johnson-Nyquist noise of the thermometer and the shunt (blue dotted), SQUID noise (red dotted) and1/f noise (magenta dotted). The individual contributions are calculated with the parameter values determined for TUM26-b as given in (b). The amplitude of the1/f noise is varied to obtain best matching with the observed spectrum.
frequencies (.10 Hz) a small additional contribution is observed.
A noise spectrum measured in a test setup in Munich with the detector TUM26-b is shown in figure 4.8a (blue line). Compared to the spectrum measured in the main CRESST setup only a few less pronounced lines at large frequencies are visible. This indicates that the detectors in the test setup are less influenced by electric interferences and mechanical vibrations.
Also for this detector the expected noise contributions (dotted lines) are calculated with the parameters given in figure 4.8b. Due to the measurement conditions the pa-rameters determined in the transition measurement exhibit large uncertainties. Details on this detector and its parameters can be found in chapter 7. Similar to the noise spectrum of TUM40 the amplitude of the 1/f noise (∆RRT
T )1/f was varied to obtain best matching of the total expected spectrum (black) with the observed spectrum. In this case the amplitude of the 1/f noise is higher by a factor of 10 compared to the amplitude in the detector TUM40. The contribution from the Johnson-Nyquist noise as well as the SQUID noise are similar. Due to the relatively high bias current of 22µA together with the low thermal coupling Geb the amplitude of the thermal noise is significantly higher.
For high frequencies &200 Hz the measured noise spectrum is well described by the expectations. The slightly higher expectations in the kHz region is due to a filter, which is not modeled in the expected spectrum. However, for small frequencies below
∼ 200 Hz a large additional noise contribution is observed. Hence, there must be an additional noise source that causes a difference in the low-frequency range where also
the signal contribution is expected. This additional contribution is more pronounced in the test setup compared to the main CRESST setup. The test setup is located above ground and, thus, the observed event rate is significantly higher compared to the main setup at Gran Sasso. A high rate leads to events where the detector has not completely relaxed back to its operating point. Therefore, the resulting record features a pulse located on top of a non-flat baseline. Records with a non-flat baseline reduce the quality of the template fit and, thus, introduce larger uncertainties in the determination of the energy12. In the case of empty baselines, the resulting records feature a signal component and appear in the low-frequency range of the noise spectrum just as the signal.
This can explain the rise of the spectrum determined from empty baselines measured in the test setup at low frequencies. It also indicates, that the higher noise contribution is mainly determined by the experimental site. Thus, the noise is expected to be smaller in the main CRESST setup at Gran Sasso.