In the last few years, I have come across with some works in which the acquisition of PD signals is done by oscilloscopes. The fact that caught my attention is that often the scope is set to record data continuously during tens of milliseconds (several cycles of power frequency). We can call this configuration “time-based acquisition”. When the acquisition is completed, all the PD pulses can be observed in relation to the test voltage phase.
This is handy and can also provide a quick insight about the PD source and so on. However, it might be difficult for detailed analysis purposes. Let’s summarize some disadvantages of the time-based acquisition.
- Inefficient usage of acquisition memory. In this mode, the scope is recording actual PD pulses AND anything else between pulses (noise, disturbances, etc.).
- Difficult to extract the PD pulses from the data vector. Finding local maximum might consume a lot of processing power.
- Compromise between pulse resolution and length of data vector. High sampling rate to record very fast details in our waveforms can eat up all the memory in a short acquisition length.
Fortunately, this is not only an issue concerning the acquisition of PD signals. This is a very frequent issue that scope users face every day and as a result, now more manufacturers are adding a feature called “Segmented Memory” or FastFrame acquisition mode to their scopes. This is simply a mode in which you focus all the resources of the scope to record a signal only when a trigger condition is met. According to the Application Note from Tektronix “oscilloscopes with FastFrame Segmented Memory allow you to partition the available acquisition memory into frames (memory segments) of hundreds of thousands of samples. This capability facilitates a burst trigger rate of 310,000 frames per second (acquisitions/sec)”. A nice description of the principles behind the Segmented Memory feature can be seen in this video.
Here is when the biggest difference arises: the acquisition memory is only used to save actual and meaningful signals.
Let’s see a comparison example. In the screenshots below, the blue trace to the left shows the acquisition of corona discharges during 100 ms. As it is known, they occur near the negative peak of the test voltage, but in this time-based acquisition also the background noise between pulses was recorded. Conversely, the blue trace at the right shows one out of N corona pulses recorded in FastFrame mode. Note that for this particular example, each pulse lasts only 2 us.
FastFrame mode can help make better use of memory and increase resolution of data, but then what about the phase information needed when creating the well-known phase resolved PD patterns? In other words, in PD measurements we want to measure the PD pulse itself and also the phase angle of the test voltage at which the pulse occurred. To take the advantages of FastFrame mode and still being able to retrieve the phase information, here at the High Voltage Laboratory of TU Delft we have developed an analog synchronization circuit that outstands for its simplicity. Thus, one channel of the scope is dedicated to the acquisition of the PD pulse and a second channel to the acquisition of the synchronization signal.
The synchronization circuit implements a zero-crossing detector and a saw-tooth generator in such a way that the zero crossing point of the test voltage triggers the saw-tooth generator. The period of the saw-tooth signal is the same as the test voltage as can be noticed from traces yellow and dotted red in the left screenshot above. Therefore there is a direct relation between the phase angle of the test voltage and the instantaneous value of the saw-tooth signal.
Since the period of the saw-tooth is 20ms (50Hz), then during the 2 us acquisition the saw-tooth signal is almost constant as shown by the yellow trace in the right screenshot above.
The analog synchronization circuit does not produce a perfect saw-tooth signal. However, this is not a problem because the signal can be easily curve-fitted by cftool from Matlab to determine the value of the phase angle ø as a function of the instantaneous value of the saw-tooth signal.
Schematics and details of the synchronization circuit are available is our open access paper [3]. All of you are invite to test this approach and we are pleased to help you in the process. Just contact us!