#2. UHF Partial Discharge Monitoring - Let’s be honest!

In the first article of this series, we looked at how it is invalid to scale UHF PD measurements in units of pC. Now let’s consider the matter of frequency ranges for UHF PD sensors.

Badly specified frequency ranges for UHF PD sensors.

A common error is to demand a wider bandwidth from the sensor than will ever be needed in practice. In the 1990s, I was involved with research into GIS PD monitoring with National Grid in the UK. This resulted in a proposed operating frequency range of 500 - 1,500 MHz, which was perfectly suitable for GIS. Nothing has altered about the physics of PD in SF6, despite the passing of 25 years. Furthermore, in power transformer applications it is becoming increasingly evident that an upper frequency of 1,000 MHz is likely to be more than sufficient to detect PD. Yet I have seen documents that require UHF sensors for power transformers to be specified over 300 - 3,000 MHz.

You may say, “surely 300 - 3,000 MHz is better than 500 - 1,500 MHz”. Unfortunately, lack of clarity is a common feature of specifications for UHF sensors. For example, we normally regard bandwidth as being defined by -3 dB points, with a flat region in between. However, I have tested sensors of various kinds claimed to operate over a band from X to Y yet the variation in sensitivity over that band might be 20 dB. It’s important to be clear that this does not mean that the claim of operating over the range X - Y is necessarily invalid, because effective detection of PD is not governed by absolute sensitivity but by the signal-to-noise ratio (SNR) in the frequency range that is used.

UHF sensors are not ‘spot frequency’ sensors - a certain finite bandwidth is always required if the detected signal energy is to be non-zero. In most cases, for general PD monitoring purposes they are used in wideband mode (although certain communications signal bands may be filtered out to reduce interference). Even so-called ‘zero span’ measurements with a spectrum analyser select a filter bandwidth such as 1 MHz. Whatever operating frequency range is used, if noise and interference was completely absent, sensitivity could be increased indefinitely by adding gain. It is noise that puts a limit on the minimum PD signal that can be detected. The SNR is governed by factors that include not just the sensor performance, but also the monitoring hardware itself and the general EMI environment in which the PD monitoring system operates. To be fair to all suppliers of UHF sensors, specifications should include not only (i) the frequency range but should also (ii) define the units of sensitivity measurement, (iii) the flatness of response that defines the boundaries of the claimed frequency range, and (iv) most importantly, define the test configuration in which the response is measured.

To conclude, a comment on point (iv): I believe that one technique sometimes used to generate an impressive response for UHF sensors over a wide frequency range is to test only the ‘antenna’ part in isolation, instead of ensuring that it is properly mounted as it will be once installed on the HV equipment. UHF PD sensors are commonly tested for frequency response and sensitivity using an electromagnetic TEM/GTEM cell system. A good example of sensitivity ‘cheating’ can be illustrated using the example of sensors to be mounted at GIS inspection windows (sometimes present at disconnectors), such as the one shown at the top of this article. These round glass windows can be used as positions to retrofit external UHF sensors. With this kind of mounting, the ‘antenna’ is set back from the inside wall of the GIS tank by a smaller diameter tube and by the pressure window. When calibrating the sensor response, it is essential to mount it on a test plate with the same kind of window. If the procurement process for UHF PD monitoring does not insist on fair procedures for testing sensitivity, then companies who calibrate their sensors properly will lose out to competitors whose sensors will never perform as well as the sales literature may suggest.

If you need technical support with UHF PD detection technology, High Frequency Diagnostics can offer the following independent and confidential services:

• Evaluation of PD monitoring system specifications to ensure that they are realistic (for customers or vendors),
• Review of tender submissions to ensure the client knows what questions to ask and what validating evidence should be sought to confirm potentially exaggerated specification claims,
• UHF sensor design and sensitivity testing,
• Design and supply of UHF PD sensor calibration systems,
• PD emulation and ‘blind testing’ of installed PD monitoring systems to validate operation and test manufacturer’s claims,
• Preparation of evidence-based reports and literature reviews to rebut unrealistic or unscientific claims relating to UHF PD detection and monitoring.

https://www.hfde.co.uk/Contact.html

Martin Judd is Technical Director of HFDE Ltd, a Scottish company that specializes in condition monitoring and diagnostic techniques for high voltage (HV) electrical insulation systems. His role involves regular working on-site with electrical utility companies, applying advanced techniques for detecting, diagnosing and monitoring partial discharge (PD) activity in HV equipment such as power transformers, open terminal substations, switchgear and GIS. Before founding HFDE in 2014, Martin was Professor of High Voltage Technologies at the University of Strathclyde in Glasgow, where he managed the HV Laboratory and pioneered many aspects of the ultra-high frequency (UHF) technique for PD detection. He has authored more than 250 scientific papers on PD and related topics and is a member of the IEEE DEIS Diagnostics Technical Committee. Dr Judd currently serves as UK member of two CIGRE working groups, “Improvement to PD measurements for factory and site acceptance tests of power transformers” (A2/D1.51) and “PD measurement on insulation systems stressed from HV power electronics” (D1.74).