Murata Power Solutions: Partial discharge characteristics of different transformer insulation systems in DC-DC converters
Even the lowest power isolated DC-DC converters may need to have the highest insulation ratings to protect against electric shock. Measurement of ‘partial discharge’ characteristics of the insulation provides a means to assess long term reliability of the barrier, especially under the condition of repeated stress. This article presents some findings and comments about the relative merits of the different barrier construction methods for best partial discharge performance. Read More
Some applications of DC-DC converters require them to provide ‘galvanic isolation’ – a barrier between the primary and secondary circuits. Isolation may be used simply to prevent a ground loop by breaking the primary and secondary return connections which may actually join elsewhere. The other extreme is where the barrier forms part of a safety isolation system to prevent electric shock. In any case the DC-DC barrier needs to be specified correctly to prevent malfunction or injury.
There are many international standards agencies such as UL and IEC that prescribe the construction of the converter barrier for different market applications and levels of isolation. The common situation is normally assumed of a relatively low ‘normal’ or ‘working’ voltage with occasional higher voltage transients depending on an assumed ‘over-voltage category’ of the end system. To ensure the integrity of the barrier, a one-off ‘hi-pot’ test is performed by the component manufacturer at a high voltage, typically 3 kV ac for IT systems or 4 kV ac for patient-connected medical systems. During this test, ‘breakdown’ must not occur but ‘partial discharge’ may be occurring un-noticed.
So what is partial discharge (PD) and what does it mean when it occurs? Any insulation consisting of material in one or more layers has air gaps of some size either as micro voids in the material or between layers. PD occurs when the voids locally break down under high voltage field stress effectively shortening the total insulation thickness but not causing breakdown across the total thickness. It is accepted that the effect is not damaging for low repetition rates. In systems prone to repeated high voltage transients however, carbonisation can occur. Voids then cumulatively become short circuits leading to higher electric field stress across the remaining insulation which eventually fails completely. In the power grid this is a real concern and much work is done to measure, characterise and minimise PD.
In a typical application and in the relatively benign environment around a low power DC-DC converter, you could expect that high voltage transients would be rare. However, there is an application where a continuous, high voltage, high frequency stress is present. This is in the so called ‘high side’ drive circuit where a device such as an IGBT is in a bridge configuration where its gate drive and associated DC-DC power supply is not referenced to ground but to the IGBT emitter. In applications such as motor drives, inverters, welders etc. this reference point can be a voltage switched at tens of kHz to thousands of volts with very fast edges (dV/dt). Under these conditions, the DC-DC converter sees the stress across its barrier and PD is a real possibility leading to eventual failure.
All is not bad news however. Low values of measured PD at high stress voltages using specialised test equipment can be a used as an indicator of the long reliability of the insulation. This test is non-destructive whereas the alternative of a ‘hi-pot’ test looks only for gross failure and is not very useful as a reliability indicator.
The designer of an insulation barrier, whether in a DC-DC converter or elsewhere, has a choice of methods of mechanically achieving the barrier: a single solid insulator, multiple layers of thin insulation or by physical separation through air. Practical considerations often come first. For example in a miniature part, the typical air separation required by safety agencies may be impossible to achieve. Even if space is not a problem, electrical issues may then come into play; isolation transformers with large separation between windings have high ‘leakage’ inductance leading to lowered efficiency, voltage spikes and often poor regulation. Solid or multilayer insulation is therefore often preferred, an example being the popular ‘triple insulated wire’ which has three separate wraps of insulating film over the conductor. This forms a barrier that may be less than 0.1 mm (4 mil) in total thickness but meets agency requirements for reinforced insulation at 250 V ac working voltage.
These different systems however have very different partial discharge characteristics. Air separation obviously is the best and breakdown is only limited by flashover across the total air gap or perhaps along any surface bridging the barrier. Solid insulation can be excellent but must have a controlled level of voids whereas multi-layer insulation clearly has a risk of air spaces between layers. To decide which of single-solid or multilayer insulation is acceptable, an understanding of the relationship between void size and partial discharge inception voltage is needed. The relationship is not linear and is described by the ‘Paschen’ curve after Friedrich Paschen, a nineteenth century scientist. The curve (Figure 1) shows that as void size decreases, the voltage across the void to cause void breakdown decreases but at a slower rate with a minimum at about 10 µm. As an example, at 100 µm size, about 1000 V causes PD inception. However at ten times that void size, 1000 µm, it takes only five times the voltage, 5000 V, for PD inception. Of course, if the total stress voltage across the whole material is fixed and the electric field is uniform across the material, the voltage across any void decreases anyway with decreasing void size.
Some practical examples now show the real choice available. Solid insulation of 0.4 mm is common to meet reinforced agency-rated isolation with a ‘hi-pot’ test voltage of 3000 V ac. or 4242 V peak. Let’s assume that this test voltage is representative of real transients that might produce PD. This gives a field strength of 10.6 kV/mm across 0.4 mm producing 106 V across 10 µm, 1060 V across 100 µm and 2120 V across 200 µm, plotted as points A, B and C in Figure 1 (shown above). Points above the curve represent breakdown, implying that void sizes should be controlled to be significantly less than 100 µm which is realistically achievable.
If however, another common insulation system of 3 layers of 0.05 mm polyester tape is used, the same test voltage produces 28.3 kV/mm across the total thickness of 0.15 mm or 283 V across 10 µm and 2830 V across 100 µm shown as points D and E. Point E is quite certain to produce partial discharge and this gap size of 100 µm (0.4 mil) is easily possible between tape layers. Layered insulation can also be affected by environmental contamination and humidity.
Triple insulated wire is less likely to have significant voids between layers due to the tightly controlled manufacturing process but the risk is still there. Note that the Paschen curve describes the PD inception voltage. Once discharge occurs, the voltage must be lowered significantly to extinguish the local breakdown in the void.
It is clear that if insulation cannot be achieved in total by air separation, solid material with a known low level and low size of voiding is best. In any case, increasing the material thickness beyond the agency-specified minimum reduces the field strength across each void and likelihood of partial discharge. Once a barrier is designed such that low levels of partial discharge are expected under stress test conditions, the actual measured value of PD, typically in pico-Coulombs, can be used as a long term reliability indicator.
Disclaimer: by clicking on this button, you accept that your data might be communicated to this company. If you do not want us to communicate your data, please update your details on your profileDownload White Paper