Evaluación de los aprendizajes.

As alluded to above, microphonic noise is caused by mechanical vibration of electronic circuit components with respect to surfaces at different potentials. Aside from the microphonics due to LN2 fills, we observe two major classes of microphonics in MALBEK:

• HV micro-discharges: Reverse polarity pulses related to HV micro-discharges [121,

158]. The following sections outline the DSP methods used to remove these events. • Ringing: We also observe waveforms that are triggered by electronic ringing. Again,

the following sections outline the DSP methods used to remove these events.

To eliminate ringing and HV micro-discharge induced signals, a suite of three cuts has been developed. These cuts are applied in a tiered approach, where the first tier removes waveforms that are relatively easy to identify as microphonics and the second and third remove the remaining events. The first cut eliminates microphonic waveforms based on the derivative

amplitude to energy ratio (A/E). This cut is used by the GERDA collaboration [72] to discriminate between single- and multi-site events. This metric can also be used to identify non-physics signals. Figure 5.25a shows the A/E value versus energy with only the timing cuts applied. An example of a waveform removed by this cut is also shown in Figure 5.25b. To avoid confusion with a multi-site event cut, this cut will be referred to as the max-min cuthereafter. The second cut implemented is based on the technique developed in Ref. [159], which analyzes the energy calculated with two separate shaping times. This is done off-line by performing two separate trapezoidal filters with differing peaking times (11µs and 5µs). This cut will be referred to as the microphonics cut hereafter3. Figure 5.26a shows the microphonics cut, where the max-min cut has already been applied illustrating that non- physics related backgrounds remain following the max-min cut. The curves shown have been calculated in the same manner as the wpar and t10−90 acceptance curves, injecting known

waveforms into the test input of the preamplifier and calculating acceptance curves such that 99% of pulser signals fall between them. The third cut applied to remove microphonics/noise is based upon the integral, or sum of the ADC values, of the waveform and will be referred to as theintegral cuthereafter. Figure 5.27a shows the integral cut, where again the max-min cut as already been applied to the data in Figure 5.27a. The curves have been generated in the same manner as the microphonics cut. It should be noted that the integral and microphonics cuts both remove a small fraction of slow-signals (<3% of the total slow-signal population) from the energy spectrum.

In addition to the max-min, integral and microphonics cuts, a cut based on waveform health has been implemented. The slope of each waveform baseline is calculated and the resulting distribution is fit with a Gaussian function of width σ for both DS3a and DS3b. Events with a baseline slope outside a±3σ window are removed. This will be referred to as the baseline slope cut hereafter. This cut removes pile-up events, or events which occur close in time to a previous trigger. The baseline slope distributions for DS3a and DS3b are shown in Figure 5.28 – both fits have P-values of 0.998. Additionally, an example of a pile-up waveform is shown in Figure 5.29. This cut removes <0.05% of all events.

3

For historical reasons only - Ref. [76] used the same cut and referred to it as his Microphonics Cut.

(a)

(b)

Figure 5.25: (a) The derivative maximum divided by the energy of the waveform versus energy for DS3a is shown here. All events which fall above the blue line are microphonics. The location of the blue line was chosen such that the two classes of events here are completely separated for energies greater than 600 eV. (b) An example of a ringing waveform that is removed with this cut.

(a)

(b)

Figure 5.26: (a) The microphonics cut for DS3a is shown here. 99% of events fall between the curves as trained with a pulser. The max-min cut has already been applied to these data in advance. (b) An example of a HV micro-discharge waveform that is removed with this cut.

(a)

(b)

Figure 5.27: (a) The integral cut for DS3a is shown here. 99% of events fall between the curves as trained with a pulser. The max-min cut has already been applied to these data in advance. (b) An example HV micro-discharge waveform that is removed with this cut.

Baseline Slope (ADC/ns) -0.01 -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 0.01 Counts 0 5000 10000 15000 20000 25000

(a) DS3a Baseline Slope

Baseline Slope (ADC/ns) -0.01 -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 0.01 Counts 0 5000 10000 15000 20000 25000 (b) DS3b Baseline Slope

Figure 5.28: The baseline slope distributions are shown here for DS3a and DS3b. For both data sets, a Gaussian function (red) was used to fit the distribution. The vertical red dashed lines indicate the±3σ boundaries.

Time (ns) 0 10000 20000 30000 40000 50000 60000 70000 80000 Vo lt a g e ( a rb .) 7000 7200 7400 7600 7800 8000 =37.33 ns 10-90

Run=19417, Time=2557099916490, Energy=1.46 keV, t

Figure 5.29: A typical pile-up event removed with the baseline slope cut.