*3.3. Measurement Setup*

Since the expected PD current consisted of a pulseless current with superimposed impulse currents, the magnitude and frequency content of the whole signal were very broad. Hence, it was reasonable to use two different measurement setups.

#### 3.3.1. Measurement of Impulse Currents

The measurements were performed using a high bandwidth oscilloscope-type Teledyne LeCroy WavePro 735 ZiA with an analogue bandwidth of 3.5 GHz, a maximum sample rate of 20 GS/s (using four channels), and a memory of 128 MSa per channel to allow long-term, high bandwidth measurements.

In order to ge<sup>t</sup> a deeper understanding of the partial discharge peak current ˆ *I*PD and the time differences between subsequent impulses, a high frequency current measurement was necessary. Hence, the needle electrode was directly connected to the oscilloscope using a low loss coaxial cable and a terminating resistor *R*meas = 50 Ω in parallel to the input impedance of the oscilloscope used *R*osci = 1 MΩ || *C*osci = 16 pF. Even though the used measurement device had a high analogue bandwidth, the measurement circuit (Figure 5a) led to a frequency dependent measurement impedance, especially due to the capacitance *C*cable of the coaxial cable used (Figure 5b).

**Figure 5.** Frequency dependence of the measurement circuit used.

Since the partial discharge current was calculated using Equation (6), the frequency dependence of the measurement circuit was not taken into account. This led to an amplitude error in the presented partial discharge peak currents, if their frequency content exceeded several megahertz. Due to the unknown frequency content of the partial discharge current, this approach seemed appropriate and was already used in other different publications [34,36]. Furthermore, one has to mention that the electrode arrangemen<sup>t</sup> influenced the measured PD current as well, mainly due to the stray capacitances between the protruding needle, the ground/high-voltage electrode, and the encapsulation of the test vessel [37]. Because the influence of these capacitances could not be easily quantified, they were neglected in this investigation.

Each occurring PD impulse triggered a 1 μs long sequence. The minimum time difference between two subsequent sequences was Δ*t*min = 1 μs according to the manufacturer. If a PD occurred in the dead time between two sequence recordings, it could not be detected.

$$I\_{\rm PD} = \frac{\mathcal{U}\_{\rm osci}}{50 \,\Omega} \tag{6}$$

Not only the partial discharge current amplitudes were investigated, but also the time differences between subsequent impulses. Since it was expected that they could occur with a rather high time difference of several milliseconds to seconds, the sequence mode of the oscilloscope was used (Figure 6) [38].

**Figure 6.** Sequence sampling mode.

#### 3.3.2. Measurement of Pulseless Currents

Even though the pulseless PD current could be measured with the oscilloscope as well, it was, in terms of the expected high amplitude differences between the impulse current and the pulseless current (Figure 2), reasonable to measure the smaller pulseless currents separately. Therefore, a transimpedance amplifier was connected to the needle electrode instead of the oscilloscope. Two low-pass filters with a cut-off frequency of approximately 4 Hz and two amplifiers allowed precise measurements of the pulseless PD current with a high SNR in the range from 0.1 nA to 100 μA. The data logger was connected via Bluetooth to the measurement computer and transmitted the mean value of the measured direct current every second. The operational readiness of this measurement device was already proven by other investigations [3,39].

#### **4. Test Execution and Data Evaluation**

## *4.1. Test Execution*

In order to understand the partial discharge behavior in dependence of the electric field strength, the investigations were performed at different voltage levels, starting from the inception voltage. Due to the missing general accepted definition of the inception voltage [38,40], the voltage with the first measurable partial discharge current impulse was determined as inception voltage *U*i in this investigation (cf. Section 5.1). The investigations were performed at multiples of this voltage at absolute gas pressures of *p*SF6 = 0.1 MPa, 0.5 MPa and 0.7 MPa. The dew point of the SF6 used was lower than −35 °C with a purity of at least 99%. The experiments in synthetic air were carried out at an absolute gas pressure of *<sup>p</sup>*syn. air = 0.5 MPa. The gas consisted of 20.5% oxygen (O2) in nitrogen (N2) according to the manufacturer. Its moisture content was less than 2.0 ppmmol.

Studying the impulse currents, the number of recorded sequences (each with a duration of 1 μs) depended on the time difference between subsequent impulses. For the measurements at inception voltage, five-hundred sequences were recorded and analyzed; if the applied voltage was higher than the inception voltage, 3000 to 4000 sequences were examined. An exception was the investigations at positive polarities of the protrusion at inception voltage: at higher gas pressures of *p*SF6 = 0.5 MPa and *p*SF6 = 0.7 MPa, only ten sequences could be recorded due to the long time difference between two impulses.

The studies of the pulseless partial discharge current were carried out with a stepped voltage rising test. At each step, the voltage was kept constant for 5 min, and the mean value of the PD current measured within this period was evaluated.

The needle electrodes used were changed after every test execution in order to avoid the influence of changing tip radii on the PD behavior. More precisely, one needle was used for the investigations at one polarity and one pressure value, but different voltage levels.
